Last page update: 12 October 2017

red-blue line


Note: this page contains about 165 photos and diagrams! It may take a while before they have all been downloaded. Please be patient!

"Bernhard" is the German codename for the ground-station ("Stellung", "Anlage") of the "Bernhard/Bernhardine" radio-navigation system that was used by the Luftwaffe during part of WW2. The beacon ground-station is a complete radio transmitter installation - including the antennas. A complete installation is a "Funk Sende-Anlage", abbreviated  "FuSAn". The FuSAn developed for the "Bernhard" system was FuSAn 724. Its two transmitters had an output power of 500 W. The FuG 120 "Bernhardine" is the airborne counterpart of the "Bernhard" beacon. It comprises a Hellschreiber-printer and control electronics. It printed the bearing data transmitted by the selected "Bernhard" beacon. Note that over a dozen different fixed and mobile FuSAn types were developed for various radar and radio-navigation systems of the Luftwaffe, primarily by Telefunken, Lorenz, and the DVL (Deutsche Versuchsanstalt für Luftfahrt), ref. 141.

The "Bernhard" FuSAn comprised a large rotating antenna system (ca. 25 x 35 meter). The main sub-systems of this rotating navigation beacon are (see Figure 1):

  • The upper structure ("Gerüst"), consisting of:
  • a large antenna system that comprises three antenna arrays,
  • a cabin ("mitdrehender Geräteraum") with the transmitters,
  • a small square "shed" near each of the four corners of the cabin.
  • A large concrete ring, with
  • a circular rail track, and
  • four locomotives for rotating the upper structure.
  • A small round central-support and equipment building ("feststehender Geräteraum") in the middle of the ring.
  • A remote antenna mast, for monitoring the signals transmitted by the beacon.
  • Sources of electrical power.
Berhard station

Fig. 1: The "Bernhard" ground-station of the "Bernhard/Bernhardine" radio-navigation system

(click here to get full size)

Basic characteristics of the "Bernhard/Bernhardine" system are:

  • Frequency: 30 - 33.1 MHz.
  • Transmitter power: 2 × 500 watt (FuSAn 724)
  • The available literature often refers to "Bernhard" as FuSAn 724/725. The 725-version was intended to have more powerful transmitters: 5000 W each. However, there is no evidence that these transmitters were ever entered into service, were even developed or available off-the-shelf. Ref. 20 and 21 state that they were planned only. The exact reasons for the power increase is unknown, but would typically be extended range and improved immunity against interference and jamming.
  • The wiring list of the "Bernhard" station contains several items with two gauge specifications: one for the 500 W transmitters, an a much heavier gauge for 4000 W transmitters (i.e., not 5000 W, cables nr. 6-8, 33, and 34, in ref. 189)
  • Antenna system dimensions: ≈28 x 35 m (HxW, 92x115 ft).
  • Antenna system track diameter: 22.5 m (≈74 ft).
  • The specified average track radius (mid-point between the inner and outer rail) was 10.5 m, and the concrete ring has a width of 1.5 m: 10.5+10.5+1.5=22.5 m.
  • Antenna system weight: 120 tons (265000 lbs), ref. 21; some literature states the weight as 102 tons (ca. 256000 lbs), or 100 tons (ref. 20).
  • Antenna rotational speed: 12 degrees per second (2 revolutions per minute, see the "rotational speed" section below).
  • This means that the small locomotives that turned this enormous antenna installation, moved at a respectable linear speed of about 8 km per hour (5 mph).
  • The speed was kept constant to within about ±0.2-0.3 % (!) Note that by design, the printer in the aircraft can not work with a beacon that turns at a different speed.
  • Accuracy: initially ±1°, then improved to ±0.5°, finally reduced to ±4° by using a single-trace ( = simpler) printer system and a single transmitter (though unclear if this was ever operational).
  • Operational range: 150-500 km (80-270 nm), depending on aircraft altitude with respect to the "Bernhard" antenna (p. 22 in ref. 15).


The base of the entire rotating "Bernhard" superstructure is a large concrete ring. The ring has a circular rail track on top of it, for the four locomotives that actually rotated the system. The superstructure (equipment cabin + antenna systems) is basically a large turntable. Standard width of the ring is 1.5 m (5 ft), as measured at several of the still-existing rings. The specified diameter of the center of the circular rail track ( = midway between the two rails) was 2x10.55 = 21.1 m (ref. 193). Hence, the outside diameter of the ring was about 21.1 + (1.5 / 2) = 21.9 m.


Fig.2: Concrete ring with the circular rail track


Fig. 3: Satellite image of the "Bernhard" site at Arcachon/France - overhead view (ca. 2013)

(source: http://www.geoportail.gouv.fr/)

A rail track comprises two main elements:

  • the track structure:
  • steel rails
  • fastening system. This typically takes the form of ties (crossties; UK: sleeper, D: Querschwelle), and fasteners to fix the rails to the ties. Their purpose is to maintain the correct distance between the rails ( = gauge), hold the rails upright, and transfer the loads from the rails to the underlying track bed and ground. Fasteners take the form of rail anchors, tie-plates, base-plates, sole-plates, rail-chairs (D: "Schienenstuhl"), bolts and nuts.
  • the track bed or foundation: typically layers of ballast, sub-ballast, and subgrade (layers of crushed rock, gravel, and sand) on top of the natural ground. For applications with very high loading ("weight on wheels"), the track bed may be "ballastless": a continuous slab of reinforced concrete on top of a subgrade. The latter is the case of the "Bernhard" track. This has a major advantage compared to the traditional track structure: no need for regular heavy maintenance (e.g., tamping the ballast and associated re-aligning of the rails) to restore the desired track geometry and smoothness.

The standard rail profile of the Reichsbahn (and the Bundesbahn until 1963) was S 49 (weight = 49 kg / m). It was, and still is, also used for narrow gauge tracks, tramway and subway tracks. It is unknown which rail profile was used for the "Bernhard" track, but it would have made sense to use the readily available national standard profile.

Bernhard rail track

Fig. 4: "Bernhard" track and Reichsbahn rail profile "S 49"

(source "Bernhard" track image: adapted from Fig. 1 in ref. 193)

Fig. 4 above shows the curved rails, fastened to I-beam ties (UK: "sleepers") with standard clamps. Fig. 4 also shows a "rail gauge rod" (a.k.a. "gauge tie rod" and "gauge tie bar") between the rails. There were 80 such "Spurstangen", evenly distributed between the 120 cross-ties.  A rail gauge rod is a member bar that is specially designed to joint two steel rails at the rail bottom. Their purpose is to protect the rails from tilting, and to hold the track to gauge (prevent the rails from spreading). A distinction is made between single-ended and double-end rods, depending on one or both ends being adjustable. Here, a simple steel rod is used, with both ends threaded.

The "Bernhard" track is "narrow gauge": the distance between the rail-heads is less than local standard gauge. Standard gauge in Germany is 1435 mm ( = 56½ inches). It is the same as in North America, most of Europe and the Middle East, in China and Australia. For this standard gauge, narrow gauge is defined as 500-1435 mm (18-56½ inch). To ensure the accuracy of the "Bernhard/Bernhardine" system, stable and accurate rotation of the antenna system was required. This translated to very tight tolerances for the rail track. They were specified by Telefunken (Dept. V/Mo in Berlin-Zehlendorf) and the Hein, Lehmann & Co. company - Telefunken's standard manufacturer of the antenna installations.

Per the 1942 adjustment and verification instructions for the "Bernhard" rail track, the nominal dimensions and tolerances are as follows (ref. 193, Fig. 5 below):

  • Gauge ("Spurweite", distance between the inside of the rail heads): 842 mm.
  • On-center distance between the rail heads: 900 mm (≈ 3 ft), acceptable tolerance: ±1 mm ( = 5/128 inch).
  • The top of the inside rail is higher than the outside rail by 24 mm (nearly 1 inch). Acceptable tolerance: ±1 mm ( = 5/128 inch).
  • The top of the rails must be at the same height, all the way around the track. This is verified for 20 evenly distributed points around inside rail of the track. The height of these 20 points must be within ±2 mm ( = 5/128 inch) of each other.
  • Average radius of the track (mid-point between inside and outside rail): 10548 mm (10.5 m = 16 ft 4 inch). Allowable tolerance: ±15 mm (0.6 inch).
  • There is a large ball bearing at the center of the roof of the round building in the middle of the concrete ring. The top flange of the ball bearing raceway must be higher than  inside rail by 1297 mm ( = 4 ft 3 inch). Allowable tolerance: ±30 mm (1.2 inch). This is not an extremely critical dimension: it can be compensated when installing the locomotives, by placing shims between superstructure frame and the ball-joint ("Kugelzapfen") on top of the locomotives.

After installing the rails, all specified dimensions had to be checked against the associated tolerances. All measured values were recorded on a special form, and sent to department V/Mo of Telefunken in Berlin-Zehlendorf. This enabled determining trends during regular re-verification. For some measurements, a reference point had to be defined. It had to be marked clearly and permanently. Ref. 193. 


Fig. 5: cross section with nominal dimensions and tolerances

(source of dimensions and tolerances: ref. 193)

The documented "Bernhard" installations have at least six different fastening systems. In all cases, the rails are fixed in place either with 120 ties, or with 120 sets of anchors in the concrete.

  • A: the concrete ring has 120 pairs of rectangular vertical holes, with a square dimple half way between them. Example: Be-15 Szymbark/Bytów. Possibly pre-fab ties were used that had a rectangular vertical post at both ends, and those ends were inserted into the vertical holes. The purpose of the dimples is unknown.
  • B: the concrete ring has 120 pairs of rectangular vertical holes, but there is only a dimple for every third pair of those holes. Example: Be-14 Aidlingen/Venusberg. There, the holes measure about 11x22 cm and are about 50 cm deep; the dimples are 17x17 cm. A pair of steel rods is anchored in the bottom of each rectangular hole. The part of the rods that sticks out above the ring is threaded (M20). The holes are not placed very accurately. However, as the upper part of the steel rods can be moved around, this is not an issue when installing the rail fasteners.
  • C: the concrete ring has 120 pairs of rectangular vertical holes, but there are no dimples. There are no steel rods anchored in the holes. Examples: Be-12 Nevid/Plzň, Be-16 Sonnenberg/Hornstein.


Fig. 6: The various forms of rail fastening and associated features in the concrete ring

  • D: the concrete ring has no holes, but there are 120 sets of 2+2 threaded rods sticking out where there are vertical holes in versions A-C. This requires more accurate placement of the rods than in cases A-C. Examples: Be-4 La Pernelle, Be-11 Trzebnica/Trebnitz.
  • E: the ties are sections of I-beam that are embedded into the top of the concrete ring; 2+2 vertical bolts are welded onto each end of the I-beams.
  • concrete ring with a flat top. Examples: Be-10 Hundborg, Be-3 Le-Bois-Julien, Be-6 Marlemont, Be-2 Mt.-St.-Michel-de-Brasparts, Be-7 Arcachon. At Archachon, the I-beams are 15½ cm wide and 17 cm tall ( = standard I-beam per DIN 1025), and are embedded in a concrete layer that was poured separately.
  • concrete ring with a rounded top. Example: Be-8 Schoorl/Bergen, where the ties are 1.3 meter long and the bolts are placed at 13 and 27 cm from each end of the ties. Rounding the concrete top must have required additional effort. It is unclear why it was done.
  • F: two sets of 120 ties, one set is narrower than the ties at all other sites. Example: Be-0 Trebbin which was also a "Bernhard" test site. The narrow ties have a width of 7 cm, and may have been from an initial version of the track. However, both sets of ties are embedded into the same layer of concrete, i.e., both sets were already installed when the concrete of the top layer was poured. The bolts on the narrow ties appear to have been removed with a grinding tool, suggesting that they predate the wide ties.


Fig. 7: The various forms of rail fastening and associated features in the concrete ring

Berhard station

Fig. 8: Pairs of threaded rods

I visited the remains of the "Bernhard" ring of Buke in April of 2015. This ring was destroyed by the British after the war, exposing the cross-section of the concrete. It suggests that the concrete was cast with as many as six distinct parts, see the photo below. The top layer is about 7 cm (≈3 inch) thick.

Bernhard station

Fig. 9: Cross-section of the ring at Buke

A number of the "Bernhard" rings have rail-ties (sections of I-beam) that are embedded into a separate layer of concrete on top of the ring:

Bernhard station

Fig. 10: Top layer of concrete, with embedded I-beam rail-tie at Arcachon

The inside rail of the track has a radius of less than 10 meters. This is very small for a rail track, and causes a large  difference (≈5%) in the speed between the inside and outside wheels of the locomotive bogies. This "slippage" causes problems with normal bogies that have rigid axles, with wheels that cannot turn independently. It also causes problems with traction, required locomotive tractive effort, and wear of the wheel flanges and the rails. One way to solve this, is to use wheels with a smaller diameter (here: 5%) on the inside rail of the track. In turn, this requires that the inside rail be raised slightly (here: 2.4 cm):


Fig. 11: Raising the inside rail with a block on the ties

The "Bernhard" installations at, e.g., Be-3 Le-Bois-Julien, Be-6 Marlemont, and Be-10 Hundborg, had a block on all ties. At Be-5 Mt.-St.-Michel-Mt.-Mercure, there is only a block on every fourth tie.

Berhard station

Fig. 12: A block on the inside end of all ties of Be-6 at Marlemont

(source: unknown)

Berhard station

Fig. 13: A block on the mounting plate on the inside end of all ties of Be-10 at Hundborg

(source: Hundborg Lokalhistoriske Arkiv; used with permission)

 A jig was used to check the gauge and the height difference between the inside and the outside rail. Gauge and height verification is repeated at each of the 80 rail gauge rods:

Bernhard rail track

Fig. 14: Measuring the gauge and the height difference between inside & outside rail height

(source: adapted from Fig. 2 in ref. 193)

The bottom of the jig has a notch at each end. A spirit level (a.k.a. mason's level; D: "Wasserwaage", F: niveau à bulle) is placed on top of the jig. The notches are sized such that for nominal dimensions, the spirit level shows "horizontal" and the gap between the jig and the rail heads is 3 mm. The latter gap is verified with calibrated shims. The shims are also used to raise the jig on the side of the inside or the outside rail, until the spirit level shows "horizontal":

Bernhard rail track

Fig. 15: Measuring the height difference between inside & outside rail height

(source: adapted from Fig. 4 in ref. 193)

Generally, the aim is for trains to run without any contact between the flange of the wheels and the rail head. Such contact causes stress on the frame of the bogie, and wear on the wheels and rails. This is why wheel treads normally have a conical shape that widens towards the flange of the wheel. On a circular track with a very small radius, it may be necessary to align the bogie axles radially with the curvature of the track. I.e., the axles always point at the center of the circular track, perpendicular to the rails. No information is available about the bogie design of the Bernhard locomotives. So it is unknown if this approach was actually used.

Berhard station

Fig. 16: Bogie axles angled towards the center of the circular track

The relative height of points around the track, and the relative height of the top of the round building is verified with a two-tube water level (D: Schlauchwaage, F: niveau à eau). This level works on the principle of "communicating vessels". It consists of a sufficiently long flexible tube that is filled with water. There is a glass tube at both ends of the tube. The tubes have an adjustable scale. First, a reference point ("datum", D: Normalpunkt) is chosen. As many points around the track must be measured, it is most convenient to use a reference point at the round building in the middle of the circular track. This will also be used to measure the height of the building. A two meter long surveyor's pole is pounded into the soil, and against the overhanging roof of the round building. With a sprit level, the height of the top of the ball bearing in the roof is marked on the pole:

Bernhard rail track

Fig. 17: Measuring relative height around the track and of the top of the round building

(source: adapted from Fig. 6 in ref. 193)

With a two-tube water level, the height of the top of the inside rail is measured at rail gauge rod number 1. This height is also marked on the pole. The relative height between top of the rail and ball bearing flange can now be determined with a ruler. The height of the inside rail is then measured at every fourth rail gauge rod (i.e., at 20 points in total).

Obviously the circular rail track was not transported to the "Bernhard" sites in one piece. Most likely, they were delivered as sections of curved rail. It is unknown which type of rail joints were used:

  • Conventional joints with a small expansion gap (D: "Stoßlücke"). The rail sections are connected with so-called "fishplate" joint-bars (D: "Schienenlasche") and bolts through the rail. The "fishing" is the vertical web between rail head and rail base. The gaps cause the familiar clickety-clack noise when the train wheels bump over them. There is a tie underneath both of the joining rail ends. None of the "Bernhard" tracks have pairs of ties that are placed right next to each other. So it is unlikely that this type of joint was used.
  • Expansion joints (a.k.a. "breather switch", "adjustment switch", D: "Dehnungsfuge", "Schienenauszug"). The ends of the mating rail sections are tapered diagonally, and are not bolted together with fishplates. This type of joint allows for smoother transitions ( = reduced noise and vibration) than conventional joints.
  • Fusion-welded seamless joints. The welding is done on-site with the "Thermit" process that is described below. The result is a continuous welded rail (CWR, D: "durchgehen geschweißtes Gleis", "lückenloses Gleis"). Such rails need very solid anchoring, in order to to avoid warping of the tracks due to thermal contraction/expansion (e.g., "sun kink").
  • The "Bernhard track has a length of π x track diameter = π x 22 ≈ 69 m. The expansion of steel rails is ca. 12 mm per °C per km. Concrete has a expansion coefficient of 10 mm per °C per km. Assuming a very moderate summer-winter difference in rail temperature of 50 °C, the relative expansion would have been ca. 7 mm (≈1/4 inch). Note that this is a steady-state value, and steel and concrete have different thermal inertia.
  • For CWR, the Deutsche Bahn installs a heavy concrete rail-tie every 60 cm. To minimize expansion/contraction forces, the rails are installed when the temperature is around 20 °C. The ties of the "Bernhard" track are spaced by... 60 cm!

Berhard station

Fig. 18: Three standard types of rail joints

(left-to-right: conventional joint, expansion joint, seamless joint)

The "Thermit" process was discovered by the German chemist Hans Goldschmidt around 1890. He was a student of the German chemist Prof. Robert Bunsen (yes, of the famous "Bunsen burner" - actually the "Bunsen & Desaga" burner). Goldschmidt patented the process (originally intended for purifying metals) in Germany 1895, and named it "Thermit". It is a simple but intense exothermal reaction: a mixture of powdered iron oxide (commonly known as "rust") and aluminum is converted into iron and aluminum oxide, plus enough heat - over 2400 °C (4300 °F) - to fully melt the mixture in a matter of seconds! The reaction is typically started by igniting a magnesium "sparkler" stick or ribbon that is put into the mixture. Unlike blast furnace smelters, basically no external heat needs to be applied. The aluminum slugs will float on top of the melted iron. The mix normally contains additives, to obtain the desired steel alloy. The process can be used for welding large steel parts, such as shafts, cables, pipes, and rails. It can also be used for casting of parts, and for under water welding. The first commercial welding application were tramway rail projects in Essen/Germany in 1899 and Berlin in 1901. The deutsche Reichsbahn followed in 1928. The process was patented in the USA in 1928, by the Metal & Thermit Company (named Goldschmidt Detinning Company until the first world war). Ever since the 1920s, it is the standard process world-wide for welding rail tracks. The intense pyrotechnic process has also found military applications (grenades, incendiary devices, etc.). The aluminothermic Thermit rail-welding process is as follows:

  • The ends of two rail sections that are to be welded, are brought together with 2-3 cm spacing (1 inch) and are aligned.
  • The gap is clamped with a two-part ceramic shell that has the same shape as the cross-section of the rail. This is needed to avoid the molten steel from running off. The sides of the clamp are sealed with special molding sand.
  • The ends of the two rail sections are pre-heated to ca. 1000 °C with a blow torch.
  • A ceramic "funnel" pot is placed on top of the molding clamp, and is filled with Thermit-mixture.
  • The mixture is ignited with magnesium ribbon, and then boils. The entire process only takes about 25 sec.
  • After cooling off, the pot, clamp, and slugs are removed.
  • The weld is ground, to get a smooth rail head.

Berhard station

Fig. 19: Thermit-welding of tramway tracks in the city center of Bremen/Germany around 1900

(source unknown)

Rails that are dirty (grease, decomposing leaves, rain, snow, and ice) significantly reduce traction of locomotives. Traction heavily depends on the adhesion coefficient μ between the steel wheels and the steel rails (ref. 158):

μ ≈ 30% for clean, dry rails.
μ ≈ 20% for rails that are clean but wet or icy, and for greasy rails.
μ ≈ 5-10% for rails covered with (decomposing) plant leaves.

This is why the final version of the "Bernhard" stations had a track-cover that moved with the rotating installation (p. 20 in ref. 183). Not only to maintain traction and avoid disturbance of the constant speed ("Gleichlauf"), but also to avoid accidents.

Based on available photos, such a cover was installed at least at the "Bernhard" stations Be-4 at La Pernelle (but not yet in March of 1943), Be-8 at Schoorl/Bergen, Be-9 at Bredstedt, and Be-10 Hundborg. The "Bernhard" station Be-0 at Trebbin did not have such a cover.

Berhard station

Fig. 20: A steel track-cover moves with the rotating platform

(station Be-4 at La Pernelle, France, July 1944)

Berhard station

Fig. 21: Cross-section of the track cover - made of sheet metal and support braces

Berhard station

Fig. 22: The cover has a "box" around each pair of support wheels

(Station Be-10 at Hundborg/Denmark)

Berhard station

Fig. 23: Side-view of the track cover - "box" around each pair of support wheels

Berhard station

Fig. 24: Cross-section of the track cover - "box" around each pair of support wheels

The small round building below the rotating cabin contained electronic equipment. Operators had to get to an from that building, even when the station is rotating. So they had to cross the rail track. To facilitate this, two sets of stairs were integrated into the moving rail cover, placed at opposite sides of cabin (p. 20 item 6 in ref. 183, though some photo material suggest only one such set of stairs...).

Berhard station

Fig. 25: One of the two sets of stairs for crossing the covered track

(Station Be-4 at La Pernelle/France)

Berhard station

Fig. 26: The destroyed "Bernhard" Be-8 at Bergen/Schoorl - upside-down section of track cover

(source: ref. 127)

Berhard station

Fig. 27: Staircase section of the track cover

(photo - turned right-side-up: ref. 127)

The wooden cabin on the turntable has an entry door at each end. Stairs are attached to the turntable, to get in and out of the cabin, while rotating or stationary (p. 20 item 4 in ref. 183).

Berhard station

Fig. 28: The stairs attached to the turntable, providing access to both ends of the cabin

(Station Be-9 at Bredstedt)


This section describes the interconnections within and between the main elements of the "Bernhard" ground station:

  • the rotating superstructure, comprising the
  • the antenna system, and
  • the equipment cabin
  • the stationary support & equipment building below the rotating superstructure
  • the electrical power distribution & conversion building
  • the remote monitoring antenna mast and radio receiver.

The interconnections between the dipoles of the antenna system, and the connection between the transmitters and the antenna system are discussed in detail in the next section ("The antenna system").

The large cabin that rotates with the antenna system contains the two transmitters and the associated power supplies (motor-alternators). These power supplies are powered by regular 50 Hz 3-phase AC voltage ("Drehstrom"). The cabin also contains an electrical power distribution panel and the control panels for the four locomotives. All four locomotives are powered by DC-voltage. One of the locomotives also has a 3-phase synchronous AC motor, powered by a fixed-frequency 3-phase AC voltage ("geregelter Drehstrom").  Lighting and heating appliances in the cabin are powered by single-phase AC voltage ("Wechselstrom").

Bernhard wiring

Fig. 29: Electrical power and signal interconnections of the rotating part of the system

(source: derived from ref. 189 and 190)

The AC and DC voltages and tone-modulation signals for the transmitters are transferred to the rotating cabin via a slip-ring assembly that is attached to the ceiling of the small round brick building below the rotating turntable. Slip-rings allow electrical lines to traverse continuously rotating mechanical joints. The shaft of the slip-ring assembly is extended to the heart of the beacon installation: the optical encoder disk. The three light-projectors of the encoder disk are AC-powered. Two of the three associated photo-cells are used to key the modulation tone of the compass-scale transmitter on/off in the rhythm of the symbology to be transmitted. The third photo-cell generates a tachometer signal for monitoring the rotational speed. The signals transmitted via the two antenna arrays are received via a radio at a nearby mast, and printed with two Hellschreiber-printers (the same HS 120 printers as installed in the aircraft). All AC and DC power is passed through a power  distribution & control panel to the slip-ring assembly and to local equipment.

Bernhard wiring

Fig. 30: Electrical power and signal interconnections of the round equipment room below the rotating superstructure

(source: derived from ref. 189 and 190)

The DC and two 3-phase AC voltages are provided by a nearby power generation & conversion building. This building receives 3-phase AC power from the public power grid or from a local diesel generator. The selected AC power is rectified with a mercury arc rectifier (MAR), to power the DC-motors of the locomotives. Part of the DC-power is converted to 3-phase fixed-frequency AC-power, for the synchronous AC-motor in one of the four locomotives. This DC-AC conversion is done with a Conz-converter.

Bernhard wiring

Fig. 31: Electrical power and signal interconnections of the round equipment room below the rotating superstructure

(source: derived from ref. 93A, 189, and 190)

Several electrical power and signal cable types are used throughout the system (ref. 189):

  • NGA ("Normen-Gummi-Ader für feste verlegung in Schutzrohr") rubber insulated wire for use in metal conduits).
  • NPA (metallumflochtener "Normen-Panzer-Ader", metal-braided (armored) cable; ref. 197).
  • NMH ("Normen-Gummischlauchleitung in mittlerer Ausführung, Handgeräteleitung", medium rubber-conduit cable for hand-held equipment),
  • "R.L.M. Erdkabel" (military cable for in-ground use).
  • "Juro" (???) cable. I only have this designator handwritten in Sütterlin-script in ref. 189. If you know anything about this type of cable, please contact me!
Bernhard wiring

Cable gauges vary from 1.15 to 12.5 mm diameter ( = 1 to 125 mm2 cross section). All gauges are specified for aluminium wire! The current rating of these cables varies from 1 to 250 amps. This is about 20% less than for copper wiring of the same gauge.


The "Bernhard" ground station is the rotating radio-navigation beacon of the "Bernhard/Bernhardine" system. This means that its transmission is not simultaneously in all directions (i.e., omni-directional), but it sweeps the horizon with its directional radio beams. Secondly, it is a 2-channel transmission: one channel is used to transmit the momentary azimuth value of the beam-direction. This is done in Hellschreiber-format. The second channel is used to transmit a signal that disappears ("null") when the associated beam is pointing straight at the antenna of the navigation-radio receiver in the aircraft.

So, how is this all done? Clearly, we need two antenna sub-systems: one for each transmission channel. The antenna sub-system for the azimuth-data must create a single-beam radiation pattern. The antenna sub-system for the pointer-channel must create a twin-beam radiation pattern, with a sharp null between the two beams. That is, two narrow beams, that point slightly to the left and right of the centerline of the single-beam. To get the sweeping effect, the two antenna systems must be aligned (pointing in the same direction), and be continuously rotated together.

Let's look at the actual antenna systems to see how all of this was implemented. Figure 32 below shows three development stages of the antenna system of the initial UHF Bernhard system. It operated at a frequency of 300 MHz, which is equivalent to a wavelength of 1 meter. The fourth development form (1940) used a parabolic antenna with an aperture of 6 λ (ref. 3, 181), which translates to a beamwidth of almost 10°. This antenna is shown in the far right photo in the next figure.

Bernhard antenna system

Fig. 32: Three antenna systems developed for the UHF-version of "Bernhard"

(left: 1935 (twin-beam only), center: 1936 , right: 1940; source: ref. 3)

The antenna system in the center photo of Figure 32 most clearly shows the arrangement of a large number of vertically-oriented antenna elements:

Bernhard antenna system

Fig. 33: The three antenna sub-systems and the resulting radiation pattern (top view)

(dipole arrays diagram: adapted from Reichspatent 767532)

The antenna sub-system at the top (in the green box) shows a group of five identical, parallel antennas that are interconnected. These antennas are simple dipoles. As the name suggest, a dipole has two identical "poles", typically straight metal rods or wires. The two halves of the dipole are connected to the transmitter via a 2-wire feedline. The next figure illustrates the radiation pattern of a single, vertically oriented dipole antenna in free-space (i.e., sufficiently far from other objects and ground). It is a doughnut-shaped  omni-directional pattern - which is not what we want! 

Bernhard antenna system

Fig. 34: The omni-directional radiation pattern of a single vertical dipole antenna in free-space

(left: 3D pattern of the strength of the radiated signal; center: vertical cross-section of the pattern, right: top view)

So how do we get a directional beam-pattern? This requires that we concentrate the transmitted energy in the desired direction(s), and radiate less energy in all other directions. This can be done in (at least) three ways (ref. 139K-a):

  • By placing a reflector surface behind the (dipole) radiator.
  • By using a "longwire" antenna (such as a Rhombus antenna), with a radiator length of several wavelengths. This is not very practical for a rotary antenna system that operates at a wavelength of ca. 10 m.
  • By combining the radiation pattern of multiple dipoles.

or a combination thereof.

Let's keep it simple and use several identical dipoles that are placed in parallel, and arranged on a straight line ( = linear). As all dipoles lie in the same plane, this is called a planar array. Let's use the same spacing ( = equi-distant) between adjacent dipoles: a linear array. We feed all dipoles of this array with the same radio-frequency current (= same amplitude and same phase). I.e., a uniform distribution of the dipole excitation. The in-phase (co-phase) feeding makes the array concentrate the radiated energy in the direction that is perpendicular ( = broadside) to the plane that is made up by the parallel dipoles. We now have a uniform linear broadside array. The actual pattern depends on the distance between the dipoles. Figures 35 and 36 below show this dependency for an array of 3 and of 4 parallel dipoles, respectively. This is beginning to look like what we want!

Bernhard antenna system

Fig. 35: Top view of radiation patterns of a uniform 3-dipole broad-side array (dipoles spaced by  0.2 - 1 λ)

Bernhard antenna system

Fig. 36: Top-view radiation patterns of a uniform 4-dipole broad-side array (dipoles spaced by 0.1 - 2 λ)

We can make several general observations:

  • The direction with maximum radiation intensity, is independent of the number of parallel dipoles and their spacing.
  • For spacing larger than 0.1 λ, all patterns have one or more clear side-lobes, in addition to the "forward" radiation  of the main beam. This is not desirable.
  • For spacing equal to, or larger than, half a wavelength (0.5 λ), some side-lobes become very significant and approach the strength of the main beam. These are called grating lobes. This is also not desirable. Note: maximum gain for a given number of dipoles is typically obtained for a spacing of 0.5 to 0.7 λ.
  • The main beam becomes narrower ( = directivity) and stronger ( = gain), as the number of dipoles and/or spacing increases. This is desirable. However, each additional dipole adds less gain than the previous addition ("law of diminishing returns"). The latter is illustrated in Figure 37 below. Also, if the number of dipoles is increased without increasing the overall size (span) of the array, then the directivity is reduced (main beam becomes wider), ref. 139J.
  • As, for a given number n of dipoles,  the spacing increases, the number of (equal strength) maxima increases. Between these maxima, there are n - 2 smaller lobes.ref. 139J.
Bernhard antenna system

Fig. 37 The radiation pattern of uniform broadside arrays of 1-8 parallel dipoles (spacing < 0.5 λ)

(source: Figure 3.46 in ref. 139A)

Bernhard antenna system

Fig. 38: Radiation pattern of uniform broadside array of 3 parallel dipoles (spacing = 0.5 λ) - without reflector screen or dipoles

(left: array configuration; center: 3D radiation pattern; right: top view of the radiation pattern; my 4NEC2 model is here)

How get can we reduce, or even eliminate, the side-lobes and rear-lobe? There are several techniques, with varying degrees of complexity. The basic parameters that we can adjust, are amplitude and phase of the dipole current, spacing between the dipoles, and the use of reflectors (ref. 139A-139H): 

  • One way is to not feed all dipoles with the same current amplitude. That is, non-uniform excitation instead of uniform. Instead, use amplitudes that taper off towards the dipoles at both ends of the array. This is called current "grading" or "tapering". The amplitude coefficients should be equal to those of a binomial series. For dipole spacing less than 0.5 λ, this eliminates all side-lobes. Current-grading coefficients may also take the form of a Fourier-series distribution, Power-of-Cosine distribution, Taylor-series distribution, etc. However, this approach very significantly complicates the system for feeding the dipoles. Also, the uniform distribution produces the highest directivity (see p. 518 in ref. 139H), though the first side-lobe is at best only about 13 dB down from the main-lobe.
  • Side-lobes can be reduced by  putting a reflector surface behind (and parallel to) the plane of the dipoles. This is what we see in the three "Bernhard" antenna systems of Figure 32 above. For UHF frequencies (λ ≤ 1 m), constructing such a conductive "mirror" surface. Likewise for small VHF arrays (2-3 dipoles). Reflectors are typically placed at a distance of 0.25 λ from the dipoles (1939 Telefunken/Lohmann patent 767531, lines 114-116; also p. 18 in ref. 138).
  • For true broadband operation (i.e., a bandwidth of at least +/- 10%  about the center frequency, whereas "Bernhard" is "only" +/- 5%), a reflector distance of  0.34 λ actually provides much less variation in feedpoint impedance, with that impedance closer to being purely resistive (ref. 139K-b).
  • Instead of a solid metal surface, as wire mesh may be used, as long as the size of the openings is less than 0.1 λ. Instead of a mesh, it is also possible to use a sufficiently dense "curtain" of un-tuned wires that are parallel to the dipoles.
  • To get the desired effect, the planar surface or mesh may have to be extended around the array. I.e., the array is placed inside a shallow metal box or wire basket, that is open in the broadside / forward direction of the array.
  • Instead of a reflector surface or mesh, a reflector-dipole can be placed behind (and parallel with) each primary dipole, e.g., at a distance of 0.25 λ (e.g., p. 18 in ref. 138). The reflector-dipoles may be passive ("parasitic"). In this case, they are not connected to the transmitter, but receive radiation from the other dipoles and re-radiate it. The reflector-dipoles may also be actively powered by the transmitter. The result of the latter method is illustrated in Figure 39 below.
Bernhard antenna system

Fig. 39: The radiation pattern of uniform broadside arrays with identical array placed 0.25 λ behind it, fed with 90° phase difference

(source: Figure 3.42 in ref. 139A)

Figure 40 shows my simulation of the effect of placing a reflector mesh-screen (0.1 λ mesh openings) at 0.25 λ behind the 3-dipole array: a significant reduction (14.4 dB) of the side-lobes (compared to Fig. 38 above). In the actual "Bernhard" antenna system, there is a mesh-screen above, below, and to the sides of the array. I.e., a mesh cage that is open in the forward direction. This further reduces the side-lobes. I did not expand my simulation model this way.

Bernhard antenna system

Fig. 40: Radiation pattern of 3-dipole broadside array with a reflector screen (not a cage)

(left: array configuration; center: 3D radiation pattern; right: top view of the radiation pattern; my 4NEC2 model is here)

Radiation pattern of the 3-dipole array with reflector screen - rotating at 2 rpm

(©2016 F. Dörenberg; simulation in free-space, i.e., effect of the earth's surface not taken into account)

So, now we basically know how to create a single-beam pattern with acceptably small side-lobes. But how do we get the required twin-beam pattern? How about using two separate broadside uniform arrays that are placed side-by-side? Then we "just" have to create a small angular difference between the two main beams. This can be done both mechanically and electrically. The mechanical approach was used in the small version of the "Knickebein" beacon system. The left-hand and right-hand dipole arrays are clearly placed at an angle, and the two beams cross-over in front of the antenna system. See Figure 41. Looking down onto the antenna system, the two sides form a shallow "V". Hence the name "Knickebein" ("crooked leg" = "dog leg").

Knickebein large & small

Fig. 41A: Knickebein ground-station - large (left) and small (right)

(source: Fig. 36 & 37 in ref. 181; red circle shows size of a man)

Knickebein pattern

Fig. 41B: Radiation pattern of the Knickebein beacon

The electrical method for pointing the main beam of a single broadside-array in a direction other than  perpendicular, is called "beam slewing" or "electronic beam steering" (see Section 3.14 / pp. 271 in ref. 139A). Here, the dipole-current in the dipoles of one half of the array, is given a phase-angle difference with respect to the dipoles in the other half of the array. The size of the phase-offset determines the amount of beam-slew or beam-tilt. To get a twin-beam pattern, one could use two such split-arrays and put them side-by-side.

However, this is unnecessarily complicated for what we are trying to achieve! All we need to do, is put two uniform linear broadside arrays side-by-side, and simply feed them in an anti-phase manner (180° phase difference between the two arrays). See p. 72 in ref. 137. The resulting radiation pattern has two main-beams, symmetrically with respect to the normal (perpendicular) direction, with a small angle and a sharp, deep null between the two beam directions. See Figure 42 for a 4+4 dipole array configuration. Other than the rear-lobes, this is what we want!

Bernhard antenna system

Fig. 42: Radiation pattern of two side-by-side arrays that are fed anti-phase (without reflector screen or dipoles)

(left: array configuration; center: 3D radiation pattern; right: top view of the radiation pattern; my 4NEC2 model is here)

Before discussing the VHF-version of "Bernhard", there are two aspects of the UHF antenna system that need to be mentioned:

  • The configuration of the twin-beam dipole array.
  • Feeding a rotating antenna system from stationary transmitters.

The twin-beam UHF antenna system comprised two side-by-side arrays (see the magenta boxes in Figure 33 above). The photo shows that each array actually comprised two sub-arrays of five dipoles each. These sub-arrays are placed one above the other (a so-called "stacked" array). This was either done as part of side-lobe suppression (see Section 3.16 (p. 276) in ref. 139A), or to increase gain. However, the latter is less likely. The system used two identical transmitters (i.e., same output power). The power of one of these transmitters was split in two, for the two halves of the twin-beam array. Increasing the gain of the twin-beams would increase the operational range of the beacon, but it would not make sense for the twin-beam system to have a significantly larger (or smaller) range than the single-beam system above it.

The VHF "Bernhard" system operated at frequencies in the 30-33.1 MHz band instead of 300 MHz. That is, a nominal wavelength of about 10 m instead of 1 m. This basically means that the antenna system of the VHF "Bernhard" is about ten times as big as its UHF predecessor. Figure 43 below shows the large antenna system. It measured about 35x25 m (WxH, ≈82x115 ft). We recognize the upper dipole array (green box) for the single-beam transmission, above the two side-by-side arrays (magenta boxes) for the twin-beam transmission. 

Bernhard antenna system

Fig. 43: The two antenna sub-systems of the Be-10 "Bernhard" station at Thisted/Denmark

Note that the dipole-array configurations are not the same as those of the UHF "Bernhard" system:

  • The single-beam array now comprises three parallel dipoles instead of five. Clearly, having used arrays of five dipoles would have made the entire antenna system much larger.
  • There is no solid reflector surface behind the top-array. Instead, there is a wire-mesh reflector behind and around the array. Obviously the mesh is much lighter, and also has less wind resistance. It is placed at a distance of 0.25 λ from the dipoles.
  • The twin-beam array now comprises two side-by-side sub-arrays of four dipoles each, instead of five. These two sub-arrays are fed in an anti-phase manner (180° phase difference between the two arrays).
  • There is no reflector surface behind the bottom-array. Instead, there are reflector-dipoles. They are placed at 0.25 λ behind the primary dipoles. The reflector-dipoles are "active", i.e., powered by the transmitter. Active reflectors are more effective for side-lobe reduction than passive/parasitic reflector dipoles or rods  (see p. 71 in ref. 137). The phase angle of the excitation of the reflector-dipoles leads that of the main dipoles in front of them by 90°.
  • There are two different configurations of reflector-dipoles, see Figure 15A/B:
  • aa reflector-dipole behind each primary dipole (Figure 44A).
  • a reflector-dipole only behind the two primary dipoles of each sub-array that are closest to the centerline of the antenna system (Figure 43 and 44B).

Bernhard antenna system

Fig. 44A: Top view of the 2x(4+4) array configuration of the twin-beam antenna

Bernhard antenna system

Fig. 44B: Top view of the 2x(4+2) array configuration of the twin-beam antenna

Bernhard antenna system

Fig. 45: Radiation pattern of two side-by-side arrays that are fed anti-phase (with active reflector dipoles at 0.25 λ)

(left: array configuration; center: 3D radiation pattern; right: top view of the radiation pattern; my 4NEC2 model is here)

Radiation pattern of the 2x(4+4) arrays - rotating at ≈2 rpm

(©2016 F. Dörenberg; simulation in free-space, i.e., effect of the earth's surface not taken into account)

Below are illustrations of the radiation pattern of the "Bernhard" antenna system. They are from pre-WW2 Telefunken/Lohmann patents, and therefore relate to the early UHF-version of "Bernhard". However, they are similar to the final VHF-version of "Bernhard" (see p. 95 in ref. 3). Note that the pattern has significant side-lobes and rear-lobes. This is confirmed by the official manual of the FuG 120 "Bernhardine" bearing-printer system that was used in the aircraft (p. 6 in ref. 15).

Bernhard antenna system

Fig. 46: Single-beam and twin-beam radiation patterns of the UHF "Bernhard" antenna system

(left: Figure 2 in the 1936 patent 767354; center: Figure 1 in 1938 patent 767523)

In the Cartesian plot (right-hand plot in Fig. 46), the side lobes of the twin-beam curve are a factor 24.4 : 5.2 = 4.7 below the main lobes. This voltage attenuation factor is equivalent to a power attenuation of 10log(4.7x4.7) = 13.4 dB. The next figure shows a Cartesian plot of the the VHF "Bernhard":

Bernhard antenna system

Fig. 47: Cartesian plot of the radiation pattern of the VHF "Bernhard" antenna system

(source: adapted from page 6 in ref. 194, 1944)

My simulation model generates similar patterns: 

Bernhard antenna system

Fig. 48: Cartesian plot of the radiation pattern, based on my antenna simulation model for the 2x(4+4) arrays

(both curves converted from power gains; the grayed area corresponds to the -50° to +50° range of the plot in Fig. 46 above)

Based on available photos,  the 2x(4+4) dipole array configuration was used at the "Bernhard" installation of Trebbin (Be-0), Mt.-St.-Michel-de-Brasparts (Be-2), and La Pernelle (Be-4). The 2x(4+2) configuration was used at the installation of Bredsted (Be-9), Thisted (Be-10), and Arcachon (Be-7). This suggests that the transition between the two configurations was made around station Be-8 in Schoorl. Why remove the outermost reflector-dipoles? Figure 51 shows the difference in the 3D radiation pattern of the two configurations. Without the outermost reflector dipoles, there are more (and slightly stronger) rear-lobes. However, this may not necessarily have had a negative impacted on the performance of the "Bernhard/Bernhardine" system, based on the time-constants of the automatic receiver-gain control of the SV 120 printer-amplifier in the aircraft.

Bernhard antenna system

Fig. 49: Radiation pattern of the 2x(4+4) array configuration (left) and of the 2x(4+2) configuration

(My 4NEC2 models for these two configurations are here and here)

Transmitter power output (TPO) is the radio-frequency (RF) power that the transmitter produces at its output. This is not the same as the effective radiated power (ERP) of the antenna system. ERP is basically the product of transmitter output power, losses in the feedline system (and radiation by that system) including splitters, and gain of the antenna system in the direction of maximum gain. ERP is referenced to the same TPO being input to a single dipole. The "Bernhard" dipole arrays concentrate the RF energy into the main lobes of the radiation pattern. This provides a gain with respect to a omni-directional reference antenna. Based on my simulations of the antenna arrays, the main lobes of the twin-beam pattern had a gain of 13.81 dBi. That is: 13.81 dB compared to an isotropic radiator (i.e., 13.81 dBi). ERP is referenced to a dipole, which has a gain of 2.15 dBi. Hence, the twin-beams had an estimated gain of 13.81 - 2.15 = 11.66 dBd. With a 500 watt "Bernhard" FuSAn 724 transmitter, this would have resulted in an ERP of 7.3 kW - far from the unrealistic 5 MW ( = 40 dB) that is sometimes suggested in literature.

Figure 50 below illustrates the length, thickness, and spacing of the dipoles, as well as the spacing between the dipoles and the reflector-cage. The size of the two men in the photo can be used as a reference. Note that the VHF "Bernhard" operated at a frequency of 31.55 MHz ±1.55 MHz. Hence, the nominal wavelength λ is 9.5 m.

Bernhard antenna system

Fig. 50: Dimensions and spacing of the dipoles and the wire-mesh reflector cage

The table below shows the dipole and spacing dimensions for three different assumptions regarding the height of the men in the photo above:

Bernhard antenna system

Fig. 51: Estimated dipole and spacing dimensions, based on photogrammetric analysis

I have run antenna simulations for about 30 different combinations of parameters in Figure 51, for the array of 2x(4+4) dipoles. A table with the results is here. General conclusions are consistent with antenna theory: 

  • When the dipole length is decreased, gain of the main beams decreases (not desirable), the front-to-back ratio decreases (not desirable), and the strength of the first left & right side-lobe increases relative to the two main beams (not desirable)
  • The gain of the two main beams is relatively insensitive to spacing between the front dipoles and the reflector dipoles.
  • The strength of the first side-lobes left & right of the two main beams, is relatively insensitive to the variation of the parameters. The relative strength is -9 to -12 dB.
  • The strength of the other side- and rear-lobes is sensitive to spacing between the front dipoles and the reflector dipoles, and to the dipole diameter.
  • The width of the two main beams is relatively insensitive to array dimensions, other than the number of dipoles. The beam-width is 10°-12°.
  • The angle between the two main beams is relatively insensitive to array dimensions, other than the number of dipoles. The angle is 20°-24°.
  • The total number of pattern-lobes increases when lateral spacing between the dipoles is increased. For the evaluated configurations, this even number varied from 10 to 16.

As in the UHF "Bernhard" system, the dipoles are "full-wavelength": they have a nominal length of 1 λ. Note that this is contrary to the original 1936 Telefunken/Lohmann Reichspatent nr. 767531 and 767532, in which all dipoles are "half-wave" (½ λ). Why use dipoles 1 λ dipoles instead of ½ λ dipoles?

  • The radiation pattern of a 1 λ dipole has slightly more gain and slightly narrower beams than that of a ½ λ dipole.
  • 1 λ dipoles are more suitable for broad(er)band operation than antenna systems with ½ λ dipoles
  • A 1 λ dipole may be harder to properly match to a modern solid-state transmitter via a 50 Ω coax cable than ½ λ dipole (at resonance), but that is not a real argument against 1 λ dipoles when using a transmitter with tube amplifiers in combination with an open-wire feedline.
  • The mid-point of each "leg" of a 1 λ dipole in principle has (near) zero voltage. This makes it a convenient point for attaching the dipole to a support structure. This is discussed further below. The neutral point of a ½ λ dipole is the feedpoint, which is not convenient for attachment.
  • Implementing a uniform array is easier with 1 λ dipoles. They have high feedpoint impedance. Therefore, they are voltage-fed instead of current-fed. A feed-system for supplying all dipoles of an array with same amplitude and phase, is easier with voltage than with current. The dipoles need to be fed in-phase ( = co-phase). That is, the phase difference between all dipoles within a sub-array must be 0° = n x 360°, where n is an integer number. Half of the 360° is simply obtained by having a feedline length of ½ λ = 180° between adjacent dipoles. Obviously, this is easy to do: just space the dipoles ½ λ (with a small adjustment for the velocity factor of the wire). A tuned ½ λ section of feedline also has the advantage that it does not act as an impedance transformer: the impedance at one end, is transferred 1:1 to the other end. In the Bernhard antenna arrays, the feedline between adjacent dipoles is a balanced open-wire transmission line (TL). I.e., two parallel wires that are suspended in the air. The second half of the required 360° is simply obtained by connecting the top element of the even dipoles and the bottom element of the odd dipoles to the same feedline wire, and vice versa. I.e., switching  polarity at each dipole. The latter switch-over can be implemented two ways, see Figure 53:
  • By crossing the feedline wires between dipoles. This (standard) approach is illustrated in the 1936 Telefunken/Lohmann Reichspatent nr. 767531 and 767532, However, with this method, the feedline wires are not perfectly parallel. This may disturb the characteristic impedance Z0 of the feedline.
  • Without crossing the feedline wires between the dipoles. Instead, the cross-over is made at the feedpoint of every other dipole (see p. 72 in ref. 137).

Bernhard antenna system

Fig. 52: In-phase voltage feeding of a uniform array of 1λ dipoles

The top and bottom antenna arrays are each is connected to a transmitter in the large equipment cabin that rotates with the antenna system. This connection ("Energieleitung") consists of a shielded two-wire transmission line (a.k.a. shielded two-wire TL, shielded balanced TL, shielded pair; D: "(ab)geschirmte symmetrische Bandleitung", "(ab)geschirmte Zweidrahtleitung", "(ab)geschirmte symmetrische Doppelleitung"), see p. 110 and 111 in ref. 181. This is basically two parallel wires in metal tubing or braiding, with a dielectric material between the wires and the round or oblong metal conduit. The conduit is attached to the steel trusses.

Bernhard wiring

Fig. 53: Antenna system interconnections

(source: derived from ref. 181)

The null-direction of the bottom array must calibrated ("Nullverstellung", cable item 56 in ref. 189) so as to be exactly aligned with the value of the bearing angle is transmitted via the top array. This null-direction is not only affected by the accuracy of the construction of the antenna system, but also by the phase and amplitude equality between left-hand and right-hand sub-arrays of the bottom antenna. As shown in Figure 53 above, there was a means to adjust the amplitude of the signal amplitude fed the left-hand and to the right-hand sub-array. It was located at the transition from the shielded 2-wire cables from the transmitters, and the open-wire line from there to the dipoles. Note that there was no means to adjust the phase. Per ref. 181 (p. 111), the amplitude adjustment was mechanical, and the amplitude adjustment did not cause a phase shift. This suggests that the amplitude adjustment was implemented by selecting a tap of a transformer. The 1935 Telefunken/Runge/Krügel/Grammelsdorff patent nr. 737102 proposes using a fixed-location remote receiver to check the direction of the beam-null, as measuring and balancing antenna feed-currents does not guarantee its correctness. Antenna feed-currents could then be adjusted with variable capacitors across the feed-lines or adjustment of the coupling at the transmitter or the feed-point.

The following photo shows two antenna cables emanating from the bottom of a hexagonal box that is mounted to the right of the top of the door, and a ventilation screen below it, at floor level. This suggests that the transmitters were located in the cabin section behind it. This is consistent with the layout drawing of Bernhard station Be-0 at Trebbin.


Fig. 54: The entrance end of the cabin at Bredsted

(source: Australian War Memorial photo SUK14634; also part of photo on page 5 in ref. 5)

Available documentation does not mention the spacing between the feedline wires and the diameter of those wires. It can also not be determined accurately enough from the photos. Hence, the characteristic impedance of the feedline cannot be calculated. However, it was probably at least several 100 Ω. As shown in Figure 53 above, the feedline from the transmitter was attached to the end of each 4-dipole sub-array. The 3-dipole top array was fed at the center dipole. This provided a sufficiently broadband feed system for the "Bernhard" operating frequency range of 31.5 MHz +/- 5%. See Fig. 2 in ref. 139K-c. Note that open-wire feed-lines do have some disadvantages (ref. 139K-d):

  • the wires always radiate to some extent, which may affect the radiation pattern of the antenna system
  • snow and rime accretions affects the impedance characteristics
  • depending on their placement, capacitance between insulators (see Fig. 55 below) affects the impedance characteristics.

Thick dipole radiators have a large cross-section. At the dipole feedpoint these large cross-sections would be facing each other and form a significant capacitance. This is undesirable. Furthermore, the large cross-section of the radiators must be connected to the feedline wires. The feedline wires have a much smaller diameter, typically no more than several mm. A large step-transition in conductor diameter is also undesirable in antenna systems. A standard solution is to give the feedpoint-end of the dipole radiators a pointed shape, like the tip of a sharp pencil. See p. 70/71 in ref. 135. See Figure 55. The pointed tip can also easily be adapted to implement a connection to the feedline wires without having to cross those over between adjacent dipoles.

Bernhard antenna system

Fig. 55: Conical feedpoint-end of the dipole radiators

Bernhard antenna system

Fig. 56: Pointed radiator-feedpoint tips of several "Bernhard" installations and of a small "Knickebein"

The dipole radiators were made of large diameter tubing (note: "tube" is specified by outside diameter, "pipe" by inside diameter).  Ref. 13. A solid rod would have been much heavier, cost a lot more material, and not perform any better as an antenna: only the "skin" radiates. It is unknown what material they were made of: steel, copper, brass,... Copper would have had less losses, but steel pipe would have been much more easily available, stronger, and could easily be welded to an attachment arm.

Figure 57 below shows the standard textbook diagram of the sinusoidal distribution of current and voltage along the radiators of a full-wave dipole. The curves are only valid for "vanishingly thin" radiators.

Bernhard antenna system

Fig. 57 current & voltage distribution of a full-wave dipole in free-space

(note: only valid in free space (far enough away from ground and objects), and for very thin radiators)

When the radiators have a diameter that is not infinitesimally small, the current distribution is no longer purely sinusoidal - see Figure 58 below. The current at the feedpoint becomes significant. This explains the reduction in feedpoint resistance when the diameter of the radiators is increased (i.e., the λ/d ratio is decreased).

Bernhard antenna system

Fig. 58: Current distribution of a thick full-wave dipole in free-space

(dashed line: Fig. 4.9 in ref. 135; solid line: my 4NEC2 model)

The feedpoint resistance and resonance length of a dipole depend on the ratio of the wavelength λ, and the diameter of the dipole radiators. The diagram below shows this dependence for a single dipole. For the same relative thickness, a full-wavelength dipole has bandwidth that is about 1.3x larger than that of a half-wave dipole (see Section 4.3 in ref. 135). Note that each "Bernhard" system operated at a fixed frequency in the range of 31.5 MHz ±5%. It was obviously highly desirable to be able to use the same dipoles at all "Bernhard" installations. This required a relatively broadband antenna system, which was facilitated by using full-wave dipoles.

Bernhard antenna system

Fig. 59: Feedpoint resistance and resonance length as function of the diameter of the dipole radiator

(source: figure 4.7 in ref. 135)

Photogrammetric analysis of the available photos suggests that the "Bernhard" dipole radiators were quite thick: a diameter of about 9 cm (3½ inch) for a wavelength of about 9.5 m. I.e., a ratio λ/d ≈ 100. Based on this λ/d ratio and the graph above, the dipoles should have a length of ≈0.87 λ, not 1 λ. This is consistent with the photogrammetric estimate of the actual dipole length. The bandwidth of a dipole not only depends on its length, but also on its diameter. A dipole has a feedpoint impedance that consists of resistance and reactance. The resistance is relatively insensitive to the dipole diameter, but the reactance (capacitive or inductive) is not! The thinner the dipole, the more the reactance changes for a given frequency change away from resonance. In other words: the smaller the bandwidth. So: thicker is better!

The voltage distribution in Figure 57 above shows that the voltage is zero at a distance of ¼ λ from the feedpoint. That is: at the mid-point of each dipole radiator "leg". This means that this "neutral" point can be connected to ground, without affecting the performance of the antenna. This makes it a convenient point for attaching a dipole leg to the structure of the antenna system, without needing some form of insulation. However, this applies only to a single dipole in free-space. In a real dipole, the voltage and current distribution are distorted due to coupling with nearby objects (in particular other dipoles in the antenna system and the support structure) and ground (esp. with vertically oriented dipoles). Assuming that the same dipole length was used at all "Bernhard" stations, the position of the neutral point would also depend on the operating frequency of the particular station (one of 32 channel-frequencies in the 30-33.1 MHz range). From the available photos of "Bernhard" installations, it cannot be determined whether or not an insulated attachment was used. There is another advantage of attaching the dipole legs at mid-point: the high-voltage parts of the antenna (the feedpoint and the tips) can be kept away from the antenna support structure. Hence, the effect of loading by stray capacitance ( = loss and pattern distortion) between the dipoles and the structure is minimized.

Bernhard antenna system

Fig. 60: Attachment of dipole neutral-points to the support structure of the antenna system

(side view of the antenna array; center: Be-9 at Bredstedt, ref. 5; right: similar dipole attachment used in the small "Knickebein", ref. 13)

The same neutral-point attachment method was used in several dipole-array antenna systems, such as that of the FuG200 "Hohentwiel" ship-detection radar (ref. 136, multiple arrays of four 1λ dipoles with passive reflector dipoles, operating at around 550 MHz, i.e., λ = 55 cm), the DMG-3 microwave link (ref. 138), and the "Knickebein" beacons.

FuG 200 antenna system

Fig. 61: Junkers Ju 188 with 4 arrays of 4 driven + 4 passive dipoles of the FuG 200 "Hohentwiel" radar

(right-hand image clearly shows crossed-over feedline wires between the dipoles )

Small Knickebein antenna system

Fig. 62: Antenna sub-systems of a small "Knickebein" beacon - 2x(4+4) driven full-wave dipoles

(source left-hand photo: ref. 33; right-hand: France 1941, Bundesarchiv Bild 101I-228-0322-04)

Hein Lehmann logo

The antenna systems were built by Hein, Lehmann & Co., Eisenkonstruktionen, Brücken- und Signalbau of Berlin-Reinickendorf (moved to Berlin-Tempelhof after the war). The company was incorporated in 1888, and was active in sheet metal, steel constructions, bridges, railway signals, hangars for "Zeppelin" dirigibles, etc. Ref. 140, 177A/B/C.

The company had a department ("Abt. Funkbau") that constructed and installed (very) large antenna masts and towers ("Funkmaste, Funktürme") for Telefunken (incl. the latter's "Funknachrichten und Navigation" dept.). Examples are the famous Funkturm (radio tower) in Berlin-Charlottenburg (1926, still standing tall to this date (2017)), and antennas for the Langwellensender around the world, notably at Lahti/Finland (1928), Nauen/Germany, Kootwijk/The Netherlands, and Sidney/Australia. Also the huge antennas of the gigantic 1 megawatt Goliath VLF transmitter station of the Kriegsmarine. It was used for worldwide communication to (submerged) submarines - including broadcasts in Hellschreiber format.

On 31 July 1941, Telefunken placed an order (purchase order nr. 253/12129) for 12 "Bernhard" antenna systems. They cost 60810 Reichsmark (RM) each, excluding in situ installation (ref. 177A-177C). This is equivalent to roughly US$380,000 and €352.000. These prices are estimated for the end of 2016, based on general inflation data (ref. 178A-178C). Note that Consumer Price Index (CPI) inflation data does not necessarily apply to specific products (such as antenna masts, electronics) or services. At some point in time, the antenna manufacturer reduced the price by 5.7%.

The final assembly and installation at the Bernhard-sites was billed separately. For Be-2 through Be-7 (France, 1942), Be-8 (The Netherlands, 1942), and Be-12 (Czechoslovakia, 1944), Hein, Lehmann & Co. charged an installation cost that ranged from 20300 RM to 31000 RM. This is estimated to be equivalent to about US$113,700 – US$173,600 or €105.400 – €160.900 (again, at the end of 2016).

Telefunken also placed an order (purchase order nr. 253/40567) for antenna systems for stations Be-13 through Be-22. The antenna system of Be-13 (Buke/Germany) was delivered and installed. The antenna system of Be-15 (Szymbark/Poland) was delivered to the station site, but never installed. The antenna systems for Be-14 (Aidlingen/Germany) and Be-16 (Hornstein/Austria) were manufactured, but never delivered. Manufacturing of the antennas for Be-17 through Be-22 was only about 50% completed by the end of the war. So, a total of 22 "Bernhard" stations was planned!

In 1941, Telefunken also placed an order for six “Diode masts” (steel lattice masts for the remote monitoring antenna), for 2020 RM each. This is estimated to be equivalent to about US$12,600 or €11.700 (end of 2016).

Furthermore, Telefunken ordered "Panzerung von Holzhäusern" (purchase order nr. 253/33163) for 12 Be-stations, at a cost of 4167 RM each. That is, sheet-metal protection of the wooden cabin below the Bernhard’s antenna system. “Panzerholz” is plywood that is covered with sheet metal armoring on one side or on both sides. The order was probably placed in 1943. Based on that, the estimated equivalent 2016 cost is US$21,900 or €20,250 each. The metal protection was only installed at five stations (Be-2, Be-3, Be4, Be-8, and Be-10). Possibly it took the form of large panels that could be slid in front of the cabin windows.

Hein Lehmann logo

The antenna structure of the "Bernhard" station Be-4 at La Pernelle/France is reported as having grey-black camouflage colors (ref. 128).

Bernhard system

Fig. 63: British intelligence compiled by RAF Photographic Reconnaissance Units (P.R.U.) and "agents" (e.g., members of the resistance movement in various countries, such as Yves Rocard in France, ref. 128)

(source: plate 15 in ref. 34, reproduced in ref. 91)

Bernhard system

Fig. 64: RAF oblique aerial photo (3-March-1943) of "Bernhard" station at La Pernelle  and drawing derived by A.D.I. Science

(source: ref. 172A)

Bernhard system

Fig. 65: ADI Science drawing derived from aerial photos by RAF P.R.U.

(source: ref. 172A)

Not surprisingly, given its size and shape, the Bernhard ground station was easily mistaken for a German radar installation of the era:

GAF radar antenna systems

Fig. 66: The "Bernhard" antenna system was easily (and often) mistaken for a German radar installation

(source: Figure 26b in ref. 34, p. I-30 in ref. 13)


The "Bernhard" beacon FuSAn 724 comprised two identical transmitters, each with an output power of 500 watt:

  • One connected to the upper antenna array (to the center dipole, from where it was distributed to the outer two dipoles). This transmitter is referred to as the Kennzeichensender ("Kz-Sender"), as it transmits the compass scale with integrated station-identifier ("Kennzeichen").
  • One connected to the dipole-arrays of the lower antenna. The transmitter output was split between the four sub-arrays (front-left, front-right, rear-left, rear-right). This transmitter is referred to as the Leitstrahlsender ("Ls-Sender", lit. guide-beam transmitter), and transmits a constant-tone signal.

This 500 watt transmitter was a standard aerodrome approach-beacon transmitter of the Lorenz company (ref 195, 196): "Anflugführungssender" model AS 4:

Bernhard transmitters

Fig. 67: The Lorenz AS 4 beacon transmitter

(source: Fig. 3 in ref. 143)

Characteristics of the AS 4:

  • Output power: 500 watt at nominal primary supply voltage (220/380 ´volt AC, 3-phase, 50 Hz), 300 watt for primary supply voltage 10% below nominal.
  • Modulation:
  • Amplitude modulation (AM).
  • Modulation index: 90% ( = modulation depth = ratio of modulation signal amplitude and carrier amplitude), adjustable.
  • 5 HF stages: crystal oscillator, first frequency multiplier (2x), second frequency multiplier (2x), first push-pull amplifier, and the final push-pull  power amplifier.
  • Internal power supply (with forced-air cooling):
  • Input: 24 volt DC from the external power supply NA 500 (which also provides four anode and bias voltages)
  • Output: 4, 8, and 23 volt DC (heater voltages for tubes/valves and oscillator crystal oven)
  • Built-in 1150 Hz tone generator (not used in the "Bernhard" application)
  • Crystal oscillator with oven (+58 °C).
  • Power-up time:
  • 70 sec (time delay relay), to ensure that the cathode of each tubes is sufficiently heated to produce full electron emission, prior to applying the anode voltage.
  • An additional ≈3 minutes for achieving stable frequency.
  • The transmitter channel-frequency could be changed in 3-5 min, depending on the number of operators and their qualifications (ref. 183). This requires adjustment of five anode- and antenna-currents, and eliminating keying-clicks by adjusting the tone-pulse shape (Section III in ref. 143).
  • Size: 114x121x70.6 cm (WxHxD, without the feet; 4x4x2.3 ft)
  • Weight: 402 kg (900 lbs)
  • Housing: "Silumin" die-cast. Silumin was an aluminium-silicon alloy of the Metallbank und Metallurgische Gesellschaft in Frankfurt/Main; marketed in the USA as "Alpax", dating back to the early 1920s.

The transmitter has two frequency doublers after the crystal oscillator. Hence, the crystal oscillator operated at 30-33.3 / (2x2) = 7.5 - 8.33 MHz. It is unknown if the oscillator operated at (or near) the fundamental crystal resonance frequency, or at an overtone frequency (i.e., near an odd integer multiple (typ. 3, 5, or 7) of the fundamental). The crystals of the standard AS 4 transmitter were made by the Loewe company (probably the Radio Frequenz G.m.b.H subsidiary of Loewe-Opta in Leipzig, frmr. "Radio A.G. D.S. Loewe"). The crystal was placed in a small enclosure with a heating element. This is referred to as a "crystal oven". Its purpose is to keep the temperature of the crystal near the point where the slope of the crystal's frequency vs. temperature curve is zero. The Loewe crystal oven had a temperature setpoint of 58 °C (136 °F), ref. 143. In the FuSAn 724, the Loewe crystal and oven were replaced with a module from Telefunken. First of all, the transmitter had to operate at a carrier frequency that was different from the standard AS 4 frequencies. With the Telefunken crystal module, the carrier frequency had an accuracy of ±1 kHz over the normal operating conditions. I.e., around ±30 ppm (±30 x 10-6) at 30-33.3 MHz. It is unclear if the Telefunken module was also needed to improve the frequency accuracy.

In the FuSAn 724, other modifications were made to the AS 4 (ref. 195). For example, the AS 4 has a built-in generator for a constant 1150 Hz tone, and a modulator stage that is tuned to that frequency. However, one of the FuSAn 724 transmitters used a constant 1800 Hz tone, and the other a keyed 2600 Hz tone. In both cases, the built-in 1150 Hz tone generator was not used. Instead, these tone signals were generated externally. Hence, several filter capacitors in the modulator stage had to be replaced. Also, the modulation index (a.k.a. modulation depth) was made easily adjustable with a potentiometer, instead of the fixed-value resistor.

Standard equipment of the "Bernhard" station included a set of spares tubes (valves) for all equipment - about 60 tubes in total (per sheet 18 & 19 in ref. 189). For the two transmitters combined, the following tubes were kitted: 2x AF7, 12x RS289, 4x RS282, 7x RS329, all made by Telefunken.

The external power supply of the AS 4 is "Netzanschlußgerät" model NA 500 (ref. 143), with the following characteristics:

  • Input power: separate inputs for 220 and 380 volt 50 Hz 3-phase AC ("Drehstrom"), 5 kVA.
  • Output power: -24, -100, +400, +1000, and +2000 volt DC; 20, 220, and 4 volt 50 Hz AC.
  • The input step-down transformer was connected to either 220 and 380 volt 50 Hz 3-phase AC.
  • The voltage of one of the three output phases of this transformer was regulated with an 8 amp "carbon pile regulator" (a.k.a. "Kohledruckregler", "Pintsch-Regler"). This is a fast electro-mechanical voltage regulator, comprising a stack of several dozen carbon discs or rings ("Kohlescheibensäule"). Ref. 208A, 208B. The resistance of the carbon stack depends on the pressure that is applied to it. This pressure is applied by an electromagnet, whose DC control current is derived from the controlled voltage with a separate transformer-rectifier. This control current depends on the load, as well as on the primary 3-phase voltage. This closed-loop control keeps the regulated voltage constant to within ±3% for ±10% input voltage variation, In case of over-voltage of the primary 3-phase power, the regulator mechanism reaches its extreme position, This actuates a contact that shuts down the anode voltages, after a persistence delay of about 10 sec.
  • The input transformer fed three separate single-phase transformers, each followed by a selenium rectifier bridge and a filter, to generate -24, +400, and +1000 volt DC. The 24 volt DC was reduced to 4, 8, and 23 volt DC in the internal power supply of the transmitter.
  • The -100 and +200 volt DC were generated with a similar scheme, but with two separate 3-phase transformers.
  • The 20 volt AC was used by a motor in the AS 4 that continuously turned a shaft with four notched disks, associated with switching the antennas that were normally connected to the AS 4 (but not in the "Bernhard" application)
  • Cooling: forced-air (fan).
  • Size: 183x67x50 cm (WXHXD, 6x2.2x1.6 ft).
  • Weight: 340 kg (752 lbs).

Bernhard transmitters

Fig. 68: The AN 500 power supply of the AS 4 beacon transmitter

(source: Fig. 7 in ref. 143)

Bernhard transmitters

Fig. 69: The AS 4 transmitter and NA 500 power supply

(source: Fig. 4 in ref. 143)

The planned 5000 watt transmitters of the never-realized FuSAn 725 may have consisted of simply adding a final-amplifier stage and associated power-supply to each of the FuSAn 724 transmitters. I.e., using the latter as powerful modulators. This was not an uncommon approach. E.g., the 5 kW Lorenz Lo5000 "Ehrenmahl" transmitter used an 800 W "Ehrenmal" transmitter with its 3 kV power-supply, and added a 5 kW final-amplifier with a 7.5 kV power-supply.

Bernhard transmitters

Fig. 70: 3 kV power supply (left), 800 W transmitter (2 columns), 5 kW amplifier and 7.5 kV power supply of the Lo5000

(source: J. Bieschke, used with permission of seefunknetz.de)

Note that the wiring list of the "Bernhard" station contains several items with two wire gauge specifications: one for the 500 W transmitters, an a much heavier gauge for 4000 W transmitters (i.e., not 5000 W; cables nr. 6-8, 33, and 34, in ref. 189).

The AM transmitter for the twin-lobe beam was modulated with a constant 1800 Hz tone (emission designator A2N). Ref. 15, p. 81 in ref. 181. Hence, the spectrum of the transmitted signal consisted of the carrier frequency, and a sideband line at 1800 Hz distance on both sides of this carrier. See Figure 71. For a discussion of AM modulation, associated spectra, and demodulation, see ref. 211.

Bernhard transmitters

Fig. 71: RF-spectrum of an AM transmitter modulated with a constant 1800 Hz tone

The carrier for the mono-beam was modulated with a 2600 Hz tone. Hence, the spectrum of the transmitted AM signal consisted of the carrier (10 kHz away from the other carrier), and a sideband at 2600 Hz on both sides of the carrier. This is similar to the spectrum of the carrier modulated with the constant 1800 Hz tone. However, the 2600 Hz signal was not a constant tone! It was keyed on/off with the pixel-pulses in Hellschreiber-format, to represent the symbology of the compass-rose scale that was to be transmitted. I.e., On-Off Keying (OOK) modulation, which is the simplest form of Amplitude Shift Keying modulation (ASK), emission designator A2A.

The spectrum of a tone with frequency f0 that is keyed on & off with a square wave with period 2T, consists of a spectral line at f0, and an infinite Fourier-series of sideband lines at odd multiples N of 1/(2T) on both sides of f0. The height of the sideband lines decreases with N, via a (1/Nπ)2 relationship. See the left-hand image in Figure 72. Keying a tone on/off with a square wave signal is not particularly useful for conveying information. If the tone is keyed with a (quasi) random sequence of on & off pulses with duration T, then the sideband spectrum lines are "smeared" into half sine-wave envelopes. See the right-hand image in Figure 72.

Bernhard transmitters

Fig. 72: spectrum of a constant tone frequency that is modulated with fixed-length on & off pulses

(left: keying with a square wave with on & off pulses with duration T; right: keying with random sequence of such pulses)

The shortest Hellschreiber-pulses of the "Bernhard" system (one black or white pixel) had a duration T = 2.3 msec (ref. 15). This implies a shortest pixel cycle ( = 1 black pixel, followed by 1 white pixel) of 2 x 2.3 = 4.6 msec. A stream of such pixel cycles is a square wave with a period 2T = 4.6 msec.  Hence, the spectral lines were spaced by 1000 / 4.6 = 217 Hz. Obviously, the stream of Hellschreiber pixels did not consist of alternating black and white pixels, all with the same 2.3 msec duration. The black pixel pulses were 1 to 10 pixels long, the white pulses 1 to 14. So the spectrum was as illustrated in the right-hand image of Fig. 72.

The tone-pulse transmitter limited the bandwidth of the sidebands with a 400 Hz filter (ref. 15). I.e., to twice the 217 Hz pixel rate. Note that in 1935, Hellschreiber manufacturer Siemens-Halske, the British Cable & Wireless company (actually, the Communications Division of its subsidiary Marconi's Wireless Telegraph Company Ltd.), and the Reichspostzentralamt (central office of the German national postal authority), recommended to limit the transmitted bandwidth to 1.6 times the pixel-rate (here: 217 Hz, resulting in 1.6 x 217 = 350 Hz). Ref. 28A, 29, 142. Also see the "Hellschreiber bandwidth" page. This was deemed the minimum to ensure legible printing. I.e., not high-quality printing.

The crystal-controlled carrier frequencies of the two transmitters were spaced by 10 kHz (ref. 24, 181 (p. 81), 183). The frequencies were placed -5 kHz and +5 kHz with respect to the center of the channel bandwidth of the EBl 3 receiver used by the FuG120 "Bernhardine" printer system in the aircraft (p. 78 in ref. 181, p. 16 in ref. 183). This receiver had 34 fixed channels in the 30-33.3 MHz band, spaced 100 kHz. The lowest 32 channels were allocated to the Bernhard/Bernhardine system. There was a +/- 0.5 kHz tolerance on each of the two carrier frequencies (p. 81 in ref. 181). I.e., the carrier spacing varied from 9 to 11 kHz. The nominal bandwidth of the transmitted signal was 10 kHz + 1800 Hz + 2600 Hz + 2x 217 Hz = 14834 kHz. For the given tolerance of the carrier frequencies, the bandwidth was roughly between 14 and 16 kHz. Figure 73 shows what the nominal combined RF-spectrum of the two transmitters looked like.

RF and audio spectra

Fig. 73: Nominal RF spectrum of the combined "Bernhard" transmitter outputs

(source: frequencies taken from ref. 15 and ref. 181)

Note that this spectrum comprises two complete AM signals simultaneously. This is very (!) different from a single AM signal that is modulated with two separate baseband signals (e.g., a constant 1800 Hz tone and 2600 Hz tone-pulses), ref. 211.

It is unclear if the frequency of carrier-2 was 10 kHz above carrier-1, or vice versa. The FuG 120 manual shows it one way (Fig. 6 in ref. 15), but ref. 181 (Fig. 41) and ref. 183 (Fig. 4) show it the opposite way... The information available on the associated SG 120 tone filter also does not indicate the relative position of the constant-tone and Hellschreiber tone pulses.

The signals from the "Bernhard" beacon were received with a standard FuBl 2 "Funk-Blind-Landeanlage" approach-beacon receiver system in the aircraft. It comprised an EBL2 and an EBL3 beacon receiver:


Fig. 74: Block diagram of the EBL3 and EBL2 receiver

(source: adapted from ref. 72)

The "Bernhard" signals were received with the EBL3. This is a superheterodyne ("superhet") AM receiver. This means that it de-modulated the amplitude modulated signals with an asynchronous product-detector: first the received signals are heterodyned, i.e., multiplied (mixed) with the sinusoidal signal from a local oscillator. The output signal from the mixer is filtered, and finally passed through an envelope detector ( = diode rectifier + smoothing filter + DC-blocking).

In the "Bernhard" application, the EBL3 received two complete AM signals simultaneously. For a discussion of AM modulation and demodulation, see ref. 211. The audio output of the EBL3 consisted of the constant 1800 Hz tone from the "Bernhard" beacon's twin-beam antenna, and the 2600 Hz Hellschreiber tone pulses that represent the azimuth symbology. See Figure 75.

RF and audio spectra

Fig. 75: Audio spectrum of the EBl 3 receiver output

The EBL3 had an unusually large bandwidth of about 16 kHz. As derived above, the worst-case bandwidth of the "Bernhard" signal spectrum could be nearly 17 kHz. So it was very important to perform careful calibration of the EBL3 (p. 9 in ref. 15), and precise tuning. Otherwise, either the constant tone signal or the Hellschreiber tone-pulses could end up on the upper or lower flank of the bandpass, and be significantly attenuated.

To the best of my knowledge, no original "Bernhard" audio recordings exist today. So I have simulated the sound of a transmission. The recording below comprises two beam-passages: once without the azimuth information in Hellschreiber-format (i.e., only the constant 1800 Hz tone), and once the constant tone plus the constant stream of 2600 Hz Hellschreiber-pulses. The simulated beam-passage duration is about 3½ seconds. Actual beam-passage duration was 3-5 sec, depending on distance from the beacon.

Bernhard sound

Simulated sound of two "Bernhard" beam-passages - without & with Hellschreiber tone pulses


As stated before, the "Bernhard" beacon continuously transmits the momentary pointing-direction (bearing, azimuth) of the antenna system in a pictorial format, the equivalent of a compass scale strip:

Patin remote compass

Fig. 76A: Remote compass, as used in Fw190 etc.

(made by the "A. Patin & Co. G.m.b.H." company of Berlin)

Bernhard compass card

Fig. 76B: Rasterized compass scale of the Bernhard system

(sources: patent 767524 (top), ref. 15 & 181 (bottom))

Bernhard compass card

Fig. 76C: Rasterized compass scale of the Bernhard system

(source: Fig. 48 in ref. 181)

":Bernhardine" hell printout

Fig. 77: Signal-strength bar graph, azimuth data and station identifier "R", as printed with a "Bernhardine" Hellschreiber

(source: ref. 3; the strip indicates that the receiver is on bearing of 239 degrees from ground station "R")

The 360° compass-scale strip is rasterized into pixels, and continuously transmitted pixel by pixel, as the Bernhard beacon turns. Pixel columns are "scanned" and transmitted top to bottom. At the bottom of each pixel column, the pixel stream continues seamlessly at the top of the column to the right of it. Each pixel is transmitted as a tone pulse. This is known as the Hellschreiber transmission system, invented and patented by Rudolf Hell in 1929. See the "How it works" page. On the receiver side, the stream of received tone pulses is printed directly by a Hellschreiber printer, to recreate the compass scale strip on a paper tape.

The 1938 Telefunken patent 767524, the 1942/43 Bernhard/Bernhardine system descriptions (ref. 181 (Fig. 43 & 48), ref. 183), and the 1944 manual of the "Bernhardine" Hellschreiber printer (ref. 15) show the following pixel format (see Fig. 76B and 76C above):

  • Columns of 12 pixels.
  • 3 pixel-columns per degree:
    • 3 columns/deg x 360 deg = 1080 pixel-columns in total.
    • 1080 pixel-columns x 12 pixels/column = 12960 black & white pixels total.
  • "Degree" tick-marks, 1 pixel wide:
    • 10-degree tick marks: 10 pixels high.
    • 5-degree tick marks: 5 pixels high.
    • 1-degree tick marks: 3 pixels high.
  • Azimuth numbers: a character-matrix of 5x5 pixels (WxH, per patent 767524), 5x6 pixels per ref. 15 &181.
  • Beacon identification letter (e.g., "M" and "R" in Figure 76B above): 5x5 pixels

At first sight, Figure 76B and 76C above indeed appear to show a column height of 12 pixels. However, when looking closely, all alphanumeric characters clearly show "half pixels". The text of the patent and ref. 181 (p. 98) also references ½-pixels. But also note that there are no single black or white ½-pixels! The smallest black or white image element ("kleinstmögliche Bildpunktlänge") is two ½-pixels, i.e., one "full" pixel. This not only applies within each column, but also at the transition from the bottom of one column to the top of the next column. With this clever scheme (part of the Rudolf Hell's patent), any image element can start at any ½-pixel boundary (for improved legibility), but the required signalling bandwidth is related to the duration of a "full" pixel. I.e., half the bandwidth that would be required if the shortest image element duration would also be a ½-pixel.

Note that the 360° compass scale shows a one- or two-digit value for every ten degrees of azimuth. That is, the numbers 1 - 36. This is also the standard for identifying the (magnetic) heading of runways at aerodromes. Note: there is no runway nr. 0, or a runway with a number larger than 36 (other than in stupid movies). The azimuth (bearing from the beacon) was referenced to True North ( = QTE), see §10 in ref. 14 and ref. 15. These days, aeronautical radio-navigation beacons, like the runways, are referenced to Magnetic North ( = QDM; exception: Canada's Northern Domestic Airspace, a polar region).

The complete compass-scale comprises 12960 "full" pixels. The Bernhard beacon makes a 360° revolution in 30 sec. Hence, 1 full pixel has a duration of 30 sec/12960 ≈ 2.3 msec ("kleinstmögliche Bildpunktlänge", p. 9 in ref. 15). The shortest pulse cycle (i.e., 1 black pulse + 1 white pulse) is 2 x 2.3 = 4.6 msec. This is equivalent to a maximum pixel-rate of 1000 / 4.6 = 217 Hz. As the pixels are binary (black & white), the telegraphy speed is 217 baud. This is close to the telegraphy speed of the standard Presse Hellschreiber system (225 Bd), and almost twice the speed of the military Feld-Hellschreiber system (122.5 Bd). High-quality printing of Hellschreiber pixel pulses requires that at least the second harmonic of the pixel rate be transmitted. I.e., a transmitter and channel bandwidth of at least a little over 400 Hz. This is why the "Bernhard" compass-scale transmitter has a bandwidth corner-frequency that is limited to that value (p. 9 in ref. 15).

The compass scale pixel-stream was transmitted by the "Bernhard" beacon, and was printed by the Hellschreiber printer of the FuG 120 "Bernhardine" system in the aircraft. In this particular application, the Hellschreiber printer was synchronized to the "degree" tick-marks of the pixel stream. I.e., the the "degree" tick-marks are used as synchronization pulses. The motor of the "Bernhardine" printer spindle runs a little faster than the nominal speed that is required to print the three pixel-columns that make up one degree of the compass scale in exactly 30 sec / 360° ≈ 83.33 msec. The motor runs fast by 1-2% (p. 91 in ref. 181), or 1.5% nominally (p. 18 in ref. 183). Note that "teletype" (telex) teleprinters are also synchronized to the transmitting station. They use an explicit start-bit at the beginning of each 5-bit letter code. The motor of such teleprinters typically also runs 1-2% faster than the nominal transmission speed.

The synchronization mechanism in the "Bernhardine" Hellschreiber printer uses an electro-mechanical "catch and release" mechanism, see the synchronization section of the "Bernhardine" page and Fig. 77. The printer spindle makes exactly one revolution per "degree", and has a small hook-shaped notch. The "catch" hook of the lever must be in the "catch" position, just before the notch arrives at the end of each spindle rotation. I.e., at the bottom of the third ( = last) pixel column of each degree. The notch is briefly held ( = spindle stops turning) until the first pixel of the "degree" tick-mark of the next "degree" column is received.

Bernhardine sync mechanism

Fig. 77: Complete cycle of the printer synchronization process

To make this synchronization scheme work, at least the 1.5 pixels preceding each "degree" tick-mark (i.e., at the bottom of the preceding pixel column) must be white. This minimum pause is equal to 1.5 pixel / (3 x 12 pixels) = 4% of the transmission duration of 1 degree. This is just large enough to accomodate several factors:

  • Tolerances in the rotational speed of the "Bernhard" beacon: it was regulated to within ±0.2-0.3 % of exactly 2 rpm.
  • Tolerances in the printer motor speed: it was regulated to within ±0.5 % with a centrifugal regulator.
  • The reaction speed of the spring-loaded "catch and release" synchronization electro-magnet.
  • The 2400 Hz sinusoidal modulation tone has a cycle time of 1000 / 2400 ≈ 0.4 msec. The pulse-detection circuitry that drives the electro-magnet always loses part of one tone-cycle, both at the start and at the end of each pulse.

The entire 12960 pixel sequence of the rasterized compass scale must be transmitted every 30 sec ( = one revolution of the beacon). The original 1936 Telefunken patent (767354) mentions the possibility of transmitting the compass data via the principles of facsimile ("fax")  or TV-video ("Bildfunk oder Fernsehprinzip"). Other early patents for rotating beacons propose using a fax system with an electro-chemical (moist impregnated paper) drum printer (1929, J. Robinson, patent nr. 562307 ), or a Nipkow scanning-disk (UK/US: Nipkov) to transmit the information as low-resolution video to an unspecified compatible imaging device (again J. Robinson's 1929 patent, and also the 1933 patent nr. 620828 of the Marconi Co.). These solutions require continuous optical scanning of a drum that has a printed compass scale strip on it. The drum rotates at exactly the same speed as the beacon, and is aligned with it. An other approach is to scan the rasterized compass scale strip only once, and record the result on a gramophone record or magnetic wire or tape recorder (Marconi's 1933 patent nr. 620828). The recording is then played continuously at the right speed, isosynchronized ( = speed & phase) to the beacon rotation. One of the disadvantages of the latter approach is that the gramophone record and wire or tape wear out rather quickly.

However, as suggested by the 1936 Telefunken patent (767354), the "Bernhard" pixel sequence was stored as a track at the circumference of an optical encoder disk ("optischer Zeichengeber", "Impulsgeberscheibe", "Kennzeichenscheibe"). The disk was mounted on the central shaft of the rotating antenna system of the "Bernhard" beacon. The pixel track passed between a light source and a photocell. The output of the photocell was used for on/off keying of the 2600 Hz modulation tone of one of the two "Bernhard" transmitters.

The disk was made of glass (p. 62 in ref. 20). Patent 767524 proposes to implement the pixels as radial lines, engraved into a disk with a blackened surface. However, the same patent also proposes a "negative" master disk, from which a copy can be made for each beacon station. This suggests a photo-chemical process, rather than mechanical engraving. Engraving (as proposed in ref. 20, and p. 124 of ref. 21) would have been an extremely (and unnecessarily) laborious process, given the number of pixels and the required very high accuracy of the pixel pattern on the disk: "1/2-pixels" have an angular width of only 360° / 25920 < 0.014° = 50 arcsec (p. 91 in ref. 181)! An optical disk has important advantages (p. 98 in ref. 181):

  • Reproduction of a "master" disk via a photochemical contact-print method is relatively easy and inexpensive. It is also easy to verify the quality and accuracy of the copies.
  • Allows direct-drive connection to the antenna system of the rotating beacon (no gearing with inherent inaccuracies).
  • The non-contact "light source plus photo cell" arrangement causes no wear of the disk, which would otherwise cause inaccuracies.
  • Embodies the primary source of inaccuracy into a single, stable device. This disk is the heart of the beacon system!

Telefunken patents 767524 and 767354 also consider other implementations of the disk: a pixel track implemented as interconnected metal patches that are scanned with a slip contact (similar to the character drum of the Feld-Hellschreiber), or an optical pixel track with holes through the disk.

The Bernhard navigation system comprised a chain of ground stations. So the station had to be identified by its transmission. This was done by adding a station-identifier letters (callsign) between the 10-degree numbers of the transmitted compass scale, see the letters "M" and "R" in Fig. 76B. The associated pixels could be implemented in several ways (patents 767524 and 767528):

  • on the same disk as the compass card data, integrated with the compass data track.
  • on the same disk as the compass card data, but as a separate track.
  • as a dedicated track on a separate disk, mechanically aligned and synchronized with the compass card disk. This has the advantage that the same compass card disk could be replicated for all ground stations, and the station identifier be put on a "personalized" disk. In case of a separate track (on the same disk or on a separate disk), the output of the associated photocell is simply combined (logical "OR") with the output of the photocell of the azimuth track.

Patent 767937 proposes an optical disk with multiple (e.g., 8) concentric tracks. This would enable the quick change of the station's identifier letter, and/or expansion to a 2-letter callsign, to be able to distinguish more than 26 beacons.

The actual implementation of the "Bernhard" disk is slightly different from the patents, see Figure 78:

Optical disk mount

Fig. 78: The optical encoder disk of the "Bernhard" beacon

(source: adapted from Figures 45-47 in ref. 181)

The actual optical encoder disk has three types of track:

  • compass scale symbology,
  • tachometer,
  • station identifier letters. Each letter is repeated 36 times (i.e., every 10 degrees of the compass scale), evenly spaced around its track.

All tracks of the disk are concentric. The tachometer encoder track was used for monitoring the rotational speed of the beacon inside the round equipment building below the rotating cabin. The encoder track generated a square wave signal, with a pulse frequency of 36 Hz: 3 pulses per degree (see Fig. 78), 2x360/60 = 12 degrees rotation per sec, hence 3 x 12 = 36 Hz. The frequency was measured with an accuracy of ±0.1 % (p. 80 ref. 181).

Patent 767528 includes the cross-section diagram of what a 2-disk attachment might look like. One disk with the compass scale, and a second disk for the beacon's station-identifier letter ("Funkfeuer kennzeichnender Buchstabe"). There is a light source mounted on one side of the track of interest, and a photo cell on the opposite side of the disk:

Optical disk mount

Fig. 79: Two stacked optical disks with, mounted on the shaft of the rotating antenna system

(source: Figure 4 of patent 767528)

Patent 767354 proposes that the light source be a gas discharge lamp, powered with an AC signal. This makes the output signal of the photocells also AC, which is easy to amplify. Patent 767529 discusses shaping the output signal of the photocell by amplification and clipping, to obtain rectangular keying pulses for the transmitter.

Again, the actual implementation is slightly different from the patents:

Optical disk mount

Fig. 80: Diagrammatic cross-section of the encoder disk assembly

(source: adapted from Figure 44 in ref. 181 )

Optical disk mount

Fig. 81: The optical disk assembly of the "Bernhard" beacon (mounted on a test stand)

(source: adapted from Figures 49-50 in ref. 181 )

The glass disk was very flat (no distorting and light-scattering wobbles), and very accurately coaxially aligned with the shaft. The shaft of the disk was connected to the centering spigot ("Zentrierzapfen") of the beacons rotating superstructure via torsionally stiff but axially flexible couplers (p. 80 in ref. 181).

The light source is a so-called slit projector ("Spaltprojektor"). It comprises a high-intensity lamp, a slit diaphragm, and a lens system. This arrangement is primarily used in ophthalmic and other medical slit lamps, and dates back to the 1850s (Hermann von Helmholtz). The "Bernhard" slit projector focuses a very narrow, homogeneous, well-defined line of light onto the track of the optical disk. The line had a width of only 30 μm (1.2 thou, 1.2 mil). This is less than half the width of an average human hair (75 μm). The disk had a diameter of 40 cm (15¾ inch), ref. 181. I.e., slightly larger than an old 78 rpm phonograph record. Hence, the circumference of the compass-scale track was about π x 40 ≈ 126 cm (= 4+ ft) for 12960 "full" pixels. I.e., a single pixel-line had a width of about 1260 / 12960 ≈ 0.097 mm = 97 μm, slightly wider than an average human hair.

Optical disk mount

Fig. 82: Close-up of the compass scale track of the optical encoder disk

(source: adapted from Figure 47 in ref. 181)

Figure 78 above clearly shows that there were only 24 letter-tracks! However, ref. 181 (p. 100) mentions tracks for "all letters of the alphabet". The duration of an individual pixel has to be independent of the distance from the letter-track to the center of the disk. All points on the disk have the same rotational speed. But tracks closer to the center of the disk, have a smaller track length. Hence, they have a lower linear speed, and the pixel-width must be proportionally smaller (so as to have the same pixel-duration in all letters). The slit-projector was placed directly above the track, so the narrow light beam had the same width at all positions. Possibly, the slit-projector and photocell for the letter-tracks could not be placed close enough to the shaft of the disk for a 25th and 26th letter. Or, the pixel lines would have been too narrow for a 25th and 26th letter. Note: in total, only 22 "Bernhard" stations were planned (item 12 on p. 3 in ref. 177B).


The "Bernhard/Bernhardine" system is simply a "UKW-Richtstrahl-Drehfunkfeuer und Empfangszusatz mit Kommandoübertragung". That is, a  rotating VHF directional beacon system, with command-uplink capability.

Obviously, the primary purpose of a beacon is to be a navigational aid. With a single beacon, only the relative bearing ( = direction) to/from that particular station can be determined (unless the beacon somehow allows the slant range ( = distance) between beacon and aircraft to be determined).  I.e., only a position line ("Standlinie") from the beacon (with known location) can be determined, and neither distance from the station, nor a position point. Position determination is done by combining the bearing from at least two beacons with known location. I.e., by means of conventional triangulation ("Kreuzpeilung").


Fig. 83: Determining bearing angle with a single beacon, and position via triangulation with two beacons

(the bearing angle is measured clockwise from North)

The "Bernhard" beacons were used by fighter aircraft that were engaged in intercepting inbound enemy aircraft, primarily bombers. It was the task of regional fighter control stations ("Jägerleitstellungen", "Jägerleitstände") to guide fighter aircraft to their target. The required guidance, instructions, and information was continuously provided via HF and VHF voice radio (radio telephony, R/T). A standardized short message format was used for the broadcast stream of the so-called "Running Commentary" ("Laufende Reportage"). This was used in German fighter control systems such as "Zahme Sau" ("Tame Boar") and "Wilde Sau" ("Wild Boar"). See §22-25 in ref. 6, ref. 25. In particular, later during WW2, with increasingly frequent Allied bombing raids on Germany, the fighter control voice frequencies became saturated. On top of that, the voice frequencies were also subject to Allied jamming. As a backup, the same  short messages could be broadcast in spoken form via certain radio navigation beacons, and also via some Morse beacons (§59 in ref. 25).

The 1938 Telefunken patent 767512 already addressed using the "Bernhard" beacons to broadcast the relative direction for intercepting enemy aircraft. It proposes to transmit an extra heavy tick-mark in the compass scale, see Figure 84B. This could be used to indicate a target azimuth (the radial from the ground station) that is to be intercepted. The patent suggests implementation with a fixed notch (cam) on the edge of the encoder disk, and a contact that is adjustable to the desired target azimuth (Fig. 84A). The notch would actuate a switch that was simply connected in parallel with the output of the photocell of the optical disk. However, this target-azimuth marker method would not have been very practical: it could only be received by aircraft that are already flying on nearly the same bearing from the beacon station as that target. Also, the command/guidance would be limited to conveying a target azimuth.

Notched optical disk

Fig. 84A: Notched optical disk

(source: Figure 2 in patent 767512)

Tarfet azimuth bar

Fig. 84B: Target azimuth bar superimposed on azimuth data

(source: Figure 3 in patent 767512)

However, the "Bernhard" beacon already broadcasts information in textual form: the fixed numbers of the compass scale and the station identification letter. Why not use the same Hellschreiber system for transmitting short, programmable text messages? Excellent idea! But the beacon rotates, and the beam is only received during about 3+ to 5 sec out of every 30 sec rotation (depending on distance from the beacon). At 12°/sec rotation, about a 40-60 degree section of the compass scale is printed each time. Clearly, it must be guaranteed that the complete message is printed during a single beam passage, independent of the bearing from the station on which the aircraft is flying within the operating range of the system. This limits the number of text characters that can be put into a single message.

Let's assume that transmission of the message is continuously repeated as the beacon rotates. To always guarantee that a complete message is received, not one, but two back-to-back copies of the message must fit within a single beam passage! If not, the end of one message is printed, immediately followed by the beginning of the next copy of the same message. This by itself is bad enough, but it is also not possible to determine if all characters of the message have been received. This factor of two further limits the number of text characters that can be put into a single message...

Figures 76B and 76C above show that a traditional Hellschreiber font with five pixel-columns per character is used for the compass scale. Also as standard for Hellschreiber fonts, each character has a leading and a trailing pixel-column that is blank. This is for character spacing. The compass scale has three pixel-columns per degree. Each character spans 5 + 2 = 7 pixel-columns. Hence, at least about 20 characters can be received per beam passage (also per item 9 on p. 10 ( = Blatt 11) of ref. 198A). Therefore, the length of one complete message is limited to about 20 / 2 = 10 characters. To mark the beginning of each message, a delimiter symbol must be used. The start-delimiter of the next message-copy automatically marks the end of the preceding one, which allows confirmation that a complete message has been received. This leaves up to 9 characters of actual fighter guidance information. The messages had the same content about an inbound enemy bomber formation as the "Running Commentary" guidance and instructions, but captured as simple coded groups of letters and numbers. The following format was used (p. 275 in ref. 5B):

  • The "+" symbol as a message-start delimiter.
  • Altitude (in 100s of meters) of the lead-aircraft of the enemy bomber formation.
  • A two-digit identifier of the "Jägergitter" air defense "box" in which the lead-aircraft is currently located (e.g., "QR" for the box around the city of Mainz).
  • The two-digit course of the bomber group (in 10s of degrees).
  • Size of the group (estimated number of aircraft).

In some literature, allied bomber formations are generically referred to as "bomber streams". However, that term only refers to a specific form of sequencing (night) bombers, used by the RAF from the end of May 1942 until the end of the war. The purpose of this tactic was to create a string of bombers (with designated altitude bands and time slots), that would pass through the narrow German (night) air defense system via a minimum number of "boxes". This defense system, established in 1940 by then-colonel Josef Kammhuber, comprised a chain of rectangular airspace zones ("boxes"). The chain eventually reached from Denmark to the north of France, and was referred to by the British as the "Kammhuber Line". Ref. 27. The zones had search and tracking radars (first Freya and Würzburg radar systems, later also Würzburg Riese) and groups of search lights (some radar controlled). Funneling all bombers through one or a few boxes, quickly overloaded the defense capability of the boxes (two night fighters per box, an estimated 6 intercepts per hour).

Depending on the message content, 8-10 message characters were sent (including the delimiter). The following hand-drawn figure from 1945 illustrates the message "+40KA27100". Unfortunately, it not only (incorrectly) suggests that the station identifier was sent every 20° instead of every 10°, but also (incorrectly) suggests that the messages were sent in addition to the compass scale. Note that these command messages were transmitted instead of the compass scale, as there was simply no third printer-track.

Bernhard reportage track

Fig. 85: "Bernhardine" print-out with Reportage track at the bottom

(source: ref. 6 (1945), very similar image in ref. 5)

In the compass scale format, there is one tick-mark for every degree. These tick-marks are also used to synchronize the compass scale Hell-printer to the Hell-format transmission (as explained in the "Optical Encoder Disk" section above). As the same printer channel is used for the command messages, it must now also be synchronized to the command message transmission. This requires tick-marks. Here, a tick-mark is implemented at the top of every single pixel-column, instead of every third column (p. 4 ( = Blatt 3) of ref. 198A). Retaining the tick-mark at the top of every third pixel-column could only have worked with a Hell-character font that is 6 pixel-columns wide (incl. blank columns for character spacing). With a tick-mark at the start of each pixel-column, there is no such limitation. Also, having more sync tick-marks than once per degree does not upset the sync mechanism: the synchronization electro-magnet of the compass scale printer-channel is activated by all black pixels, whether tick-mark or other part of the transmitted symbology.

The upper track of the "Bernhardine" printer is used to plot the signal-strength curve. The curve has a sharp V-shaped dip in the middle, which is used as a pointer for the compass scale that is normally printed below it. Clearly, the pointer-curve is not used in combination with the command messages. However, the signal-strength printer channel is not turned off, as it is linked to the automatic motor start/stop function.

The next figure illustrates what a real print-out with a command message would have looked like:

Bernhard reportage track

Fig. 86: re-created Bernhardine print-out with command-uplink message

As stated above, the "Reportage" messages were sent instead of the compass scale data from the optical encoder disk. So, somewhere, these messages were converted from a text-string input to a Hellschreiber pixel stream. This could have been done with a keyboard and tape-puncher, combined with a "punch tape to Hellschreiber pulse-sequence converter". This was the normal way with the "Presse Hellschreiber" system. The tape could be looped through the tape reader to repeat the message. Of course, speed and text font would have had to be adapted. However, p. 87 in ref. 2 and p. 392 in ref. 7B suggest that a different method was used to program the text character sequence: inserting jumpers ("Stöpsel") into a patch-board ("Stecktafel"). Mid-2015, I finally obtained confirmation of this, by the photo shown below (ref. 93A). It was taken inside the cabin below the rotating superstructure of the Bernhard installation Be-10 at Hundborg/Denmark. The photo shows two transmitter-modulators, two monitor Hellschreiber printers (printing signals from a nearby remote monitoring receiver and antenna), and a patch-board with patch cords. There are 9 jacks for each of up to 9 selectable characters. With some difficulty, one can see that the left-hand column is labeled A-Z, and the right-hand column 1-9, 0, +, ... So, the conversion from text strings to Hellschreiber pixel streams was done at the Bernhard station, based on telephone or teleprinter messages from the regional fighter command & control center.

Reportage - Kontrollpult

Fig. 87A: Lower left-hand corner - patch-board & patch-cords for selecting command-message text string

(source: Figure 30 in ref. 93A; photo taken at Be-10 Hundborg/Denmark)

The actual patch-board and patch-cords appear to be very similar to what was used during the 1930s in small standard telephone switchboards:

Reportage - Kontrollpult

Fig. 87B: Patch-board section ("Stecktafel") of a "Klappenschrank" telephone swithboard model OB-14

(source: Fernmeldemuseum Dresden; "OB" in "OB-14" stands for "Ortsbatterie", meaning local battery operation)

This patch-board method is similar to what was used in the mechanical Siemens-Hellschreiber-sender model 44 in the 1960s. This sender has a character-drum with 19 notched disks and associated slip-contacts: seven disks to generate the pixel sequence for the characters A - G, ten for the figures 0 - 9, and one notched disk for the character "-". This machine sent a string of eight characters, based on a discrete code at its inputs (representing status and self-test results from a telephone exchange system).

Reportage - Kontrollpult

Fig. 88: The inside of a Siemens-Hell model 44E with a stack of notched character-generator disks at the center

A similar method with a stack of notched disks may have been used for generating Bernhard message strings in Hellschreiber format. When the messages had to be sent, the output of the photocell of the optical azimuth disk was simply disconnected, and the pixel stream of the text message generator was used instead.

Note that the command-uplink capability was not a standard feature of the "Bernhard" stations. Implementation of this data link system required a modification to the beacon, and was only implemented at two or three stations by the end of the war (including station Be-10 at Hundborg in Denmark).

Apparently, late 1943 / early 1944, the Lorenz company also experimented with an expanded Hellschreiber-based command data-link system, referred to as "Sägezahn" ("sawtooth", ref. 26A).

The Bernhard/Bernhardine system was the first and only operational ground-to-air data-link system of the second World War that had freely formattable messages! Since about the year 2000, the same concept has been introduced to "modern" civil aviation: Controller-Pilot Data Link Communications (CPDLC); since 2015 its usage is mandatory in European airspace above 28500 ft. CPDLC is for up-linking routine ( = non-time-critical) air traffic control instructions and clearances to aircraft via digital radio. Purpose: reduce the significant time that air traffic controllers spend on routine communications over VHF voice links, and help reduce miscommunications as well as "stuck microphone" issues (which block the radio channel). However, contrary to the Bernhard system, the pilot can now respond to messages, request clearances and information, and declare an emergency.


The official German Bernhard/Bernhardine system description documents (ref. 181, 183) clearly state that the Bernhard-beacon rotated once every 30 sec, i.e., at 2 rpm. The specified diameter of the center of the circular rail track ( = midway between the two rails) was 2x10.55 = 21.1 m (ref. 193). Hence, the track length was 66.28 m. This means that at 2 rpm = 120 rph, the small locomotives that turned this enormous antenna installation, moved at a respectable speed of 8 km per hour (5 mph). As described in the "Locomotive system" section below, the rotational speed of the system was determined by a single synchronous 3-phase AC motor in one of the four locomotives. The 3-phase AC power was provided by a DC-AC inverter that had a fixed reference frequency. Hence this motor could only turn at the reference speed, and the speed was monitored very accurately. The rotational speed of the beacon was kept constant to within -0.2 to +0.3% ! See p. 80 in ref. 181 and p. 8 & 18 in ref. 183.

The official German system descriptions and the manual of the FuG 120 "Bernhardine" printer system also state the same 2 rpm speed (top of Section B on p. 5 in ref. 15, sheet 8 & 10 of ref. 198A). Moreover, the "Bernhardine" printers were simply not compatible with antenna rotational speeds that deviated more than a couple of percent from that nominal speed! As with other types of synchronized teleprinter systems, the motor of the Bernhardine-printer had to turn 1-2% faster than nominal system speed (p. 18 in ref. 183).

There are some persistent statements that the "Bernhard" beacons rotated with a period other than 30 sec. Some of the sources for these statements are the following:

  • The French resistance explicitly reported in 1943 that this station Be-3 at Le-Bois-Julien rotated once per minute (p. 62 in ref. 91).
  • A British 10 cm [ = 300 MHz] radar station at Fairlight (on the East Sussex coast, east of Hastings, 87 km [54 miles] northwest of Be-3 at Le-Bois-Julien) concluded that the station appeared to rotate once a minute (ref. 173B, December 1943). Likewise, extensive radar measurements early-June 1944 also concluded that the rotation period was 52-60 sec (ref. 173A).
  • Luftwaffe POWs in the UK reported 1 rpm for the Bernhard system in general (§10, 18, 19 in ref. 6)
  • A "reliable informant" who saw the Marlemont station, reported to the British that he was told [correctly] that it rotated with a speed of 8 km/h. Based on the wrong British photometric assumption that the Bernhard-ring had a diameter of 82 ft [25 m], a rotational period of 36 sec. was estimated. Ref. 173E.
  • A 1946 US air force survey of German electronics development, stating that "... information is printed once per minute" (ref. 93B). This may have been based on war-time "intelligence" from other sources.

The 1936 Lohmann/Telefunken patent 767528 states that the limiting factors for the upper limit of the antenna's rotational speed, are the printing speed of the Hellschreiber and the required pixel resolution of the printed information. Given the large size and weight of the antenna system, there are obviously also mechanical considerations for the upper speed limit. The patent proposes to resolve this, by quadrupling the number of antenna beams, spaced at 90º intervals. Each optical encoder disk would simply have four "light source plus photocell" pairs (two pairs shown in the diagram above), that could be adjusted to account for angular offsets between the beam centerlines.


Below the antenna installation, and rotating with it, is a steel truss bridge. The rectangular bottom frame of the bridge is made of heavy I-beams. It is supported by the four locomotives and by the round building at the center of the concrete ring. The upper frame of the bridge was suspended from the large lattice truss-joist of the lower antenna system.


Figure 89: Cabin of Be-4 at La Pernelle/France

Inside the bridge is a long wooden cabin ("mitdrehendes Holzhaus"). It measured about 20x4x3m (LxWxH, ≈66x13x10 ft), based on p. 20 in ref. 183, as confirmed by photometric analysis of available photos. Its length is close to the inside diameter of the ring (≈20.5 m). The cabin was made of heavy wooden planks. There is a entrance door at both ends of the cabin.

Ca. 1943, Telefunken contracted its standard antenna structure supplier, Hein, Lehmann & Co., to provide "Panzerung von Holzhäusern" for 12 "Bernhard" stations (purchase order nr. 253/33163, ref. 177C). That is, for sheet-metal protection of the wooden cabin. "Panzerholz" is plywood that is covered with sheet metal armoring on one side or on both sides. The price was 4167 Reichsmark per BE-station. Based on general inflation data, this is equivalent to ca. US$21,900 or €20,250 (early 2017, ref. 177). The metal protection was only installed at five stations (Be-2, Be-3, Be-4, Be-8, and Be-10), before the course of the war intervened. The protective panels probably took the form of large panels that could be slid in front of the cabin windows, see Fig. 91 below. The wooden part of the panels involved Fa. Rostock in Trebbin (ref. 176A). This company operated three owned or leased sawmills in Trebbin (close to Be-0) since the 1930s, and supplied wooden construction materials for a number of "radar" installations and other Wehrmacht constructions.


Fig. 90: The cabin at Be-10 Hundborg/Denmark - with protective siding panels

At some "Bernhard" sites, these sliding covers are on the outside of the bridge frame that suspends the cabin from the truss-joist. At other sites, these panels slide between the cabin and that frame.

The cabin contained the two transmitters, AC/DC electrical power distribution and controls for the locomotive motors, and for cabin heating & lighting. See the "Electrical & signal distribution" section. Ref. 13 (p. 4.09) suggests that the cabin was divided into three sections. The section on the right (looking at the front of the antenna system) contained the transmitter equipment. The center section housed the controls for the four electric locomotives. The section on the left was a workspace. A plaque at La Pernelle (Be-4) states that it had one or more beds in it.

The photo below shows the power distribution and control panel ("Schaltwandtafel") inside the rotating cabin of Be-10 at Hundborg. The spoked handwheel in the lower right-hand corner (sheet 15/20 in ref. 189) belongs to the circuitry for bringing the locomotives up to speed from standstill, and for slowing down to standstill ("Kontroller und Anlaßwiderstand"). Also see the "Electrical & signal distribution" section.

Berhard station

Fig. 91: German engineer describing controls in the cabin of Be-10 at Bredstedt to a member of the RAF-ADW

(source: Australian War Memorial photo SUK14636, public domain; ca. August 1945)


At each of the four corners of the cabin, there is a "closet" or square silo of about 1½x1½x2 m (WxDxH; ≈5x5x6½ ft). They are mounted on a cantilever, away form the main cabin and above the rear bogie of the locomotive underneath. There are no openings for ventilation in the walls or the roof. No conduits appear to emanate from the bottom. From the available photos it is clear that there are no windows, and it appears that there is no door. Whatever was in there, apparently did not require access! So it was not electrical or mechanical equipment, nor a container with brake sand for the locomotives (for which there would have been a tube descending in front of the powered wheels). Hence, the purpose of these sheds remains entirely unclear.


Fig. 92: One of the cantilevered corner sheds at La Pernelle

(source: 1946 film clip Cinémathèque de Normandie)

The next photo shows that the sheds were empty at some point in time:


Figure 93: Low-altitude oblique RAF aerial photo of the La Pernelle site

(source: ref. 172A; photo by G.R. Crankenthorp, taken on 3 March 1943)

Possibly they contained dead weight (e.g., stone, sand, concrete, lead), to get more weight on the traction wheels of the locomotives - even though the entire structure carried by the locomotives was already quite heavy by itself. No such blocks have been found at "Bernhard" sites were there still are visible remains... However, the blocks must have had considerable weight: a triangular stabilizing arm was installed between the cantilever supporting each block, and the bottom frame of the bridge:


Figure 94: One of the four stabilizing arms of the turntable at Be-10 in Hundborg

At some sites, these sheds have a pointed, four-sided roof (e.g., Be-4 at La Pernelle, Be-7 at Arcachon). At others, the roof is flat (e.g., Be-9 at Bredstedt, Be-10 at Hundborg).


Figure 95: Corner-sheds with a pointed roof (Be-4 at La Pernelle/France)


Figure 96: Corner-sheds with a flat roof (Be-10 at Hundborg/Denmark)

Unlike the main cabin, the walls are not made of wooden planks: there are no seams. In available post-war photos of "Bernhard" installations (La Pernelle, Arcachon), the main cabin is completely stripped of its materials - but not the four corner-sheds! Either the material was hard to remove and carry away, not valuable enough, or not usable for some other reason.


Fig. 97: Post-war photo of Be-7 at Arcachon - installation dismantled and stripped, except for the corner sheds


Fig. 98: Post-war photo of Be-4 at La Pernelle - installation dismantled and stripped, except for the corner sheds


The superstructure of the "Bernhard" (i.e., the antenna system and bridge with cabin) was rotated by four electrically powered locomotives on the circular rail track. Each locomotive had two bogies (US: trucks). The photos below shows that each bogie had two axle-boxes and leaf-spring suspension. Such axle-boxes typically have greased sliding bearings (a.k.a., journal bearings, not ball bearings). Each of the locomotives had 2 x (2+2) = 8 wheels, so the four locomotives had 32 wheels in total. The weight of the superstructure was carried by the four locomotives and the round central support building below it. Let's assume that the weight was evenly distributed. So, the combined locomotives carried 4/5 x 120 = 96 metric tons. Hence, each wheel carried a weight of 96 / 32 = 3 metric tons. This is well below the standard railway limit of about 16 tons/wheel, for the load at which both the rail head and full-size wheels are damaged.

Berhard station

Fig. 99: Each locomotive has two bogies (Be-10 at Hundborg) - 32 wheels in total

(source: www.gyges.dk, used with permission)

Based on photometric analysis, the locomotives measured about 4x1.2 m (LxH, 13x4 ft), and the small wheels had a diameter of about 50 cm (20 inch).

Berhard station

Fig. 100: The side of the locomotive on the outside of the track - direction of motion is to the left

The next photos show that the locomotives had a large round access hole at both ends, normally covered with a rectangular cover plate. The right-hand photo shows that at least one of the locomotives had external down-gearing between a motor and the forward bogie. The down-gearing ratio is small: about 1.6:1. The gearing is covered, so it is not known if it was a belt or a chain.

Berhard station

Fig. 101: Close-up of a locomotives and external gearing of one of the locomotives (right)

(source: 1946 film clip of Be-4 at La Pernelle, Cinémathèque de Normandie)

As explained further below, each locomotive was driven by two motors. It is unclear if both bogies of each locomotive were driven, or only the forward bogie. It is also unclear if both axles of each driven bogie were driven, or only one per bogie. Dividing motor power evenly between multiple axles optimizes the use of the available traction, but complicates the construction.

The photos in Fig. 101 and 102 show that the locomotives supported the weight of the superstructure at the point halfway between the front and rear bogies - which makes sense. There was ball socket ("Kugelpfanne") at this point on top of the locomotive (p. 8 in ref. 193). The socket held a large downward-pointing ball-stud ("Kugelzapfen") that was mounted underneath the I-beam frame of the superstructure. Between the superstructure and the locomotives, there are only conduits for electrical power cables to the motors (and a tachometer signal, but only at locomotive nr. 4, see Fig. 103, 111).

Berhard station

Fig. 102: Ball joint on a locomotive of Be-12 at Nevid and electrical conduits to the motors (magenta circle)

(source: brdy.org)

There were three motor types used in the locomotives (ref. 190):

  • "Hauptantrieb": main drive DC-motor. All four locomotives had such a motor.
  • "Nebenantrieb": auxiliary drive DC-motor, with separately controlled field winding. Only locomotives nr. 1-3 had such a motor.
  • "Synchronantrieb": 3-phase AC synchronous motor drive. The frequency of the AC power was accurately regulated. I.e., it was not taken from the variable 50 Hz public power grid or a local generator. Only locomotive nr. 4 had such a motor.

Bernhard wiring

Fig. 103: Motorization of the four locomotives

(source: derived from ref. 189, 190)

That is, each locomotive had two motors! The DC motors provided gradual acceleration from standstill, and the majority of the drive power that was required during normal operation. As explained further below, the required highly accurate speed control was provided by the single synchronous 3-phase AC motor in locomotive nr. 4.


How powerful did the locomotives actually have to be? Let's do a simplistic reasonableness check, using the definition of "horsepower". On a level track( = horizontal), the required locomotive horsepower HPloc is (ref. 156, 157):

HPloc = W x T x S / 375


W = total gross train weight in tons (1000 lbs)
T = total Train Resistance (a,k,a, Starting Resistance) per ton = 8 lbs/ton. Note: this is a standard value used in the railway industry. Modern rail systems have a lower resistance.
S = speed in mph
375 is a constant that assumes that no HP is used for driving accessories (gearing, compressor, alternator, ...)

Converting this to metric units, we get:

HPloc = W x T x S / 271


W = total gross train weight in metric tons (1000 kg)
T = total Train Resistance per metric ton = 8 kg/ton
S = speed in km/h = mph / 1.609

Total weight of the rotating superstructure was 120 metric tons, distributed among the four locomotives and the central support at the center of the concrete ring. The locomotives carried 4/5 x 120 = 96 metric tons. Wind load would increase this value. The specified diameter of the center of the circular rail track ( = midway between the two rails) was 2x10.55 = 21.1 m (ref. 193). Hence, the track length was  π x 21.1 ≈ 66.3 m. This means that at 2 rpm = 120 rph, the small locomotives moved at a respectable speed of 120 x 66.3 = 8 km per hour (5 mph). Hence, the required total locomotive horse power is 96 x 8 x 8 / 271 = 22.7 HP at the traction wheels. There is down-gearing between the motor and the driven wheels. Let's assume a reasonable transmission efficiency of 85%. For the required total motor horsepower we now get:

HPmotor-total = HPloc / 0.85 = 22.7 / 0.85 = 26.7 HP

The "Bernhard" system used four locomotives. So, the required motor horsepower per locomotive is:

HPmotor = HPmotor-total / 4 = 26.7 / 4 ≈ 6.7 HP

The rolling resistance of a railway vehicle (which is a science all by itself) is the sum of all forces acting through the wheels and the axles, that oppose motion of that vehicle. There are many sources of rolling resistance. Some vary only with weight (e.g., journal bearing resistance, rolling friction, track resistance), some are linearly proportional to speed (e.g., wheel-flange contact, wheel-rail interface, lateral and vertical movement), some depend on the square of the speed (e.g., aerodynamic), and some on the fourth time-derivative of displacement. Examples:

  • Bearing friction.
  • Elastic deformation of the wheel-tread and of the rail, in the wheel-rail contact area. Note that the contact surface between a wheel and the rail is a very small elliptical area, called the "contact patch". It is typically only about 15 mm (0.6 inch) across! The weight of the wheel and the load that it carries, makes a "dent" in the surface of the rail head. This dent moves with the wheel, and is like a very small bow wave. So, trains actually always go uphill, even if the track is horizontal!
  • Losses due to wheel creep (during accelerations and decelerations, the elastic deformations cause the actual wheel displacement to be be different from its rolling distance).
  • Losses due to grinding of wheel flanges against the rail head, and "hunting" (horizontal back-and-forth waving movement of the bogies on the track).
  • Wheel noise (vibration and resonances in the wheels).
  • Suspension "jounce" (the fourth time-derivative of displacement), due to impacts on rail joints (if any), and associated rebounds. Bumps and bounces convert horizontal momentum into vertical momentum; the associated energy is dissipated in the suspension.
  • Track deformation (Rayleigh waves).
  • Aerodynamic drag that acts on exposed wheels and on the body of the locomotive. At low speed, this is  normally quite small, if not negligible. However, in the case of "Bernhard", there is also drag of the large antenna system due to system movement and wind load. The antenna system is symmetrical with respect to the vertical axis of rotation. So there is a "push & pull" effect, depending on whether the movement is upwind or downwind. Wind will change the apparent weight on the locomotives.
  • Curve resistance, due to the radius of the curvature of the track. Note that regular "1 meter" gauge track  typically has a minimum curve radius of 45 - 60 meters, about 4 - 6 times that of the circular "Bernhard" track! Without special measures, the curve resistance would have been quite high.
  • The "Bernhard" track had a gauge ("Spurweite", distance between the inside of the rail heads) of 842 mm. The on-center distance between the rail heads was 900 mm (≈ 3 ft). Ref. 193.

In normal rail applications, total resistance at low speed (less than about 15 km/h, 10 mph) is dominated by friction of the axle bearings (ref. 159). Note that a train with steel wheels on steel rails has a friction factor that is about 80% lower than a truck (UK: lorry) with rubber tires on pavement! Also note that the central support below the rotating superstructure had a large ball bearing (diameter ≈40 cm ≈16 inch). Clearly, it caused rotational resistance.


At least some of the DC motors of the electro-locomotives of the Be-13 station at Buke were built by Ziehl-Abegg Elektrizitätsgesellschaft m.b.H., of Berlin-Weißensee (ref. 99). Their power was 10 kW (13.6 metric horsepower, 13.4 US hp), with a large down-gearing ratio. It is unclear if these were the main drive DC motors, or the auxiliary drive DC motors. The motors at (some) other Bernhard stations may have been built by Siemens (e.g., ref. 103). This was probably Siemens-Schuckert, who also manufactured electric locomotives.

Ziehl-Abegg is a company specialized in electric motors. It was founded in 1910 by Emil Ziehl and the Swedish investor Eduard Abegg as Ziehl-Abegg Elektrizitäts-Gesellschaft m.b.H. Abegg dropped out of the partnership the same year, as he could not come up with the required funds, and the patent that he brought into the deal (ref. 214) proved useless. However, Abegg's initial "A" was retained (as a solid triangle) in the the "Z-A" company logo. Ref. 154. In 1897, Emil Ziehl invented the external rotor motor ("Außenläufermotor", outrunner motor: stator inside the rotor - very compact and excellent weight balancing). In 1904, he invented electrically powered gyroscopes with gimballed suspension. Prior to 1910, Emil Ziehl had developed electric motors and tested generators at AEG, and developed gyro-compasses at Berliner Maschinenbau AG (BEMAG, frmr. Eisengießerei und Maschinen-Fabrik von L. Schwartzkopff). BEMAG was a manufacturer of locomotives powered by steam, compressed air, and electricity. Ziehl-Abegg made DC-DC converters (DC-motor + generator) for Zeppelin airships and airplanes. Telefunken was a major customer. They also made electro-mechanical transverters ("Drehstrom-Gleichstrom-Umformer", i.e., AC-motor + DC-generator, as in Ward-Leonard Drive Systems) for elevators and generation of anode voltage of large transmitters, transformers for directional-gyros (e.g., SAM-LKu4), motor-generators such as the U 4a, and the motor-generator-alternator of the U 120 of the "Bernhardine" system. After the war, the production facilities were carried off to the Soviet Union. In 1947, the company restarted, this time in the south of Germany. These days, Ziehl-Abegg AG is a manufacturer of electric motors for elevators, ventilation and air-conditioning systems.


Fig. 104: 1912 advertizing poster, listing in 1943 Berlin phonebook, wall plaque on the building in Fig. 106

(source of poster: ref. 155)


Fig. 105: 1938 Ziehl-Abegg postmark on an envelope

(source: www.briefmarken12.de)


Fig. 106: Company buildings of Ziehl-Abegg Elektrizitätsgesellschaft m.b.H in Berlin-Weißensee

Electrically powered rail vehicles (electric and diesel-electric train locomotives, streetcars/tramways, subways) traditionally used DC traction motors. This remained the case until well after the advent of solid-state power electronics, in particular gate turn-off (GTO) thyristors, in the early 1960s. DC traction motors where primarily of the series-wound brushed type. I.e., with a commutator, and the field windings in series with the motor's armature windings. Note that brushless DC-motors only date back to the late 1950s, ref. 160.

Series DC-motors can produce their highest torque at low speed: as much as 3-8 times the full-load torque at nominal speed. This is ideal for traction applications. For a given field flux, DC motor speed is determined by the armature voltage, whereas the delivered torque is driven by the armature current.

Berhard station

Fig. 107: Basic types of wound Direct Current (DC) motors - classified by placement of the field winding

Berhard station

Fig. 108: Basic characteristics of series, shunt, and compound DC motors

As stated above, the speed of a DC motor depends on the voltage across the motor's armature and the field flux. Standard methods to vary the armature voltage of a series motor are:

  • An adjustable resistance placed in series with the motor's armature.
  • Ward-Leonard drive system.
  • Rectified adjustable AC-voltage (ref. 161, 163G).

There are many other flavors of motor speed control (variable AC frequency, Pulse Width Modulation, ...). They are generally beyond the scope of this discussion or of the technology available at the time.

The adjustable series resistance (rheostat) method has a major disadvantage: poor speed regulation ( = speed is highly load-dependent). There are also very high losses ( = heat dissipation) in the series resistance at low speed. That is, during speed-up from, and slow-down to stand-still. Some speed-control methods for series DC-motors are illustrated in Figure 106 and 107.

Berhard station

Fig. 109: Speed control of a series DC-motor via field or armature diverter, tapped field-winding, variable series resistance

Berhard station

Fig. 110: Speed control of a series DC-motor via series-parallel reconfiguration of split field-winding and multiple motors

The Ward-Leonard Drive System (D: "Leonardsatz", "Ward-Leonard-Umformer", "Leonard Doppel-Umformer") is basically an electro-mechanical way of generating a variable DC-voltage to control the speed of a DC-motor. Ref. 162A-162G. It was invented by Harry Ward Leonard in 1892. For about 75 years, there were few practical alternatives to this system - until the advent of solid-state power-electronics such as thyristors in the 1960s. Worldwide, this was the normal way to provide smooth, step-less control of the speed of high-power DC-motors, from zero to full speed. It has been - and still is - used in many applications, such as cannon/gun- and turret-aiming, elevators, rolling mills, cranes, hoists, mining (colliery) winders, diesel-electric propulsion of locomotives and of special ships, strip-mining shovels, and heavy radar antennas. German WW2 radar antenna systems with a Ward-Leonard drive include the "Wassermann S" (FuMG 42; see figure 2 in ref. 162G; 36-60 m tall / 4 m diameter column, weight up to 60 tons) and the AEG-Telefunken "Würzburg Riese" (FuSE 65) with its large dish antenna (7.5 m diameter, 9.5 tons; ref. 162F). Allied radar system also used Ward-Leonard drives. E.g., the 20 ton antenna of the British "Marconi Type 7" was rotated with a 15 HP DC-motor, controlled with a Ward-Leonard set comprising a 24 HP 3-phase motor, a main DC-generator, and a small DC exciter generator. Ref. 162H.

The Ward-Leonard Drive System consists of a Ward-Leonard Drive Unit and a shunt-wound DC motor. The Drive Unit consists of a motor-generator. The motor (referred to as the "prime mover") has a near-constant speed. This can be a 3-phase or single-phase synchronous AC motor, or a combustion engine (diesel, gasoline/petrol) with a speed governor. The output shaft of the motor is coupled (direct-drive) to the input shaft of a DC-generator. The output voltage of this DC-generator is connected to the armature of the DC-motor that drives the load. The DC-motor need not be located near the motor-generator. The shunt-field of the DC-motor is connected to a constant voltage source; hence, the motor's excitation field (flux per motor-pole) is constant, and the torque only depends on the armature current - independent of the motor speed. The shunt-field of the DC-generator is connected to that same constant-voltage source, though via a rheostat (large variable resistor).

Berhard station

Fig. 111: Ward-Leonard Drive System

The generator's output voltage is varied by changing the generator's field current with the rheostat. In turn, this changes the DC-motor's armature voltage, and hence its speed. The constant voltage source may be a rectified AC voltage (if the prime mover is an AC-motor). It can also be generated with a small DC exciter-generator ("selbsterregter Erregergenerator"), that is also driven by the prime mover, and has its shunt-field connected to its own armature (hence, "self-excited").

The Ward-Leonard drive unit is an electro-mechanical multi-kilowatt amplifier: a small change in the input current (generator field) results in a large change at the output (generator armature voltage and current). However, in its basic form, it is an open-loop control system: the rotational speed of the load is not measured and fed back in order to adjust the generator's field current. Hence, that speed is not regulated with the high precision required in the "Bernhard" application. Note that it is possible to expand the basic Ward-Leonard system with such a feedback loop.

Another drawback is that both the AC-motor (or the engine) and the DC-generator must be dimensioned for the full and peak power of the load-driving DC-motor(s) and system inefficiencies.

System efficiency is driven by the product of the efficiency of the three machines (AC-motor, DC-generator. DC-motor), and typically lower than that of rheostat control and field control methods. A single Ward Leonard drive unit can control multiple load-sharing DC motors in parallel ("group control"). Variations of the Ward-Leonard drive system are electro-mechanical amplifiers such as the Metadyne (1930s) and the Amplidyne (1940s).

As elegant and effective as the Ward-Leonard drive system may be, it was not what was used in the "Bernhard" to provide a variable DC-voltage to the locomotive drive system! The DC motors were regulated with series rheostat for speed up from standstill and down to standstill.


There are two basic types of AC motors: asynchronous motors and synchronous motors. Both have two main parts: a stator and a rotor. Both motor types have the same stator. In a 3-phase motor, the stator basically consists of three pairs of formed coils of wire. Each coil is mounted in the slots of a laminated steel core. The coil pairs are spaced evenly around the stator. See Fig. 112. The coils of each pair are connected in series.

Berhard station

Fig. 112: Simplified cut-away view of the stator of a 3-phase AC motor (synchronous or asynchronous)

Each coil-pair is energized by one phase of the 3-phase AC electrical power. Each energized coil-pair forms a pair of magnetic poles. The resulting magnetic field extends into the air gap between the stator and rotor, and into the rotor. The magnetic field strength and polarity of each pole-pair changes cyclically, as the AC excitation is sinusoidal. When all three phases are connected, the stator generates a rotating magnetic field (RMF). This field has a constant amplitude and rotates with the same speed as the 3-phase excitation. E.g., for an excitation frequency of 50 Hz = 50 cycles/sec, the RMF rotates at 50/sec x 60 sec/min = 3000 rpm, and at 3600 rpm for 60 Hz excitation.

Berhard station

Fig. 113: Concept of how an RMF is generated by a 3-phase stator when excited with 3-phase AC power

We all know that unlike magnetic poles (North-South) attract each other, and like magnetic poles (North-North, South-South) repel each other. The motor's rotor can turn freely inside the stator. Let's take a rotor that consists of one or more magnetic pole-pairs. The stator generates a rotating magnetic field, so the magnetic rotor poles will try to remain aligned with that field: the rotor turns. There are several ways to make a stator pole-pair, see Fig. 114:

  • A permanent magnet (bar magnet).
  • A coil with the ends short-circuited. By itself, such a coil does not generate a magnetic field. However, if a varying magnetic flux is induced in this coil, a current will circulate in the coil. The direction of this current is such that it opposes its cause (Lenz's Law). The cause is the varying induced flux. With the RMF of the stator, the induced flux in the rotor winding only varies if the rotor does not turn at the same speed as that RMF. This induced varying flux, combined with the induced current, generates an electro-magnetic force (EMF, Faraday's Law) that acts on the coil conductors. The magnitude of the torque ( = rotating force) is proportional to the relative rotational speed of the rotor, compared to the synchronous speed ( = the speed of the RMF). The speed difference is called "slip". The direction of the torque is such that torque is reduced (remember: Lenz' Law): in other words, such that the speed difference is reduced.
  • Important: there is no rotational force if the rotor turns at the synchronous speed! Hence, the rotor never turns at the synchronous speed, but always slower. The amount of slip depends on the mechanical load that is driven by the motor. The heavier the load, the larger the slip ( = lower motor speed). If the load varies, the speed varies. An AC motor with such a rotor is called an asynchronous motor.  As it works on the principle of induction, it is also called an induction motor. Low-power asynchronous motors can have a slip of 5-10%, whereas asynchronous motors with a higher power rating have approx. 2-5% slip.
  • The rotor coils can be implemented as actual coiled wires (in which case the rotor windings are typically made accessible via slip rings), or simply be implemented as so-called "squirrel cage" (two parallel metal rings with a number of evenly spaced metal bars between them, often at a skew angle).
  • A coil that is energized by a DC voltage. This is equivalent to a permanent magnet. As the rotor has to turn, the DC power is supplied via slip rings. When the rotor is at standstill or at low speed, the alternating polarity of the stator's rotating magnetic field (RMF) sweeps by the poles of the rotor relatively fast. Each rotor pole is cyclically briefly pulled into one direction (without producing sufficient starting-torque), and than briefly in the opposite direction. The rotor may vibrate but will not turn.
  • Important: a motor with such a stator is inherently not self-starting! The motor needs a supplemental drive mechanism (e.g., a starter winding incorporated into the motor, or another motor) to first be accelerated to 90-95% of the synchronous speed (i.e., < 5-10% slip) - without energizing the rotor. At that point, the rotor is energized and automatically pulls into synchronism (a.k.a. "in step") with the RMF: the rotor poles are locked to the RMF and the rotor turns at synchronous speed. This is great for a constant speed drive application: no need for a closed-loop speed control system! For obvious reasons, a motor with such a rotor is called a synchronous motor.
  • Important: a synchronous motor turns at synchronous speed, from no-load to full-load!
  • Contrary to the asynchronous motor, the motor torque is generated as a result of the physical angle between the stator and rotor. This phase angle is called the load angle, coupling angle, or torque angle. An increase in mechanical load causes this angle to also increase - but synchronous speed is maintained! If the mechanical load ever exceeds the motor's maximum torque, the rotor completely loses synchronism (drops "out of step") with the stator's RMF, and the motor comes to rest. The same happens if the rotor supply voltage or the stator supply voltage is reduced excessively.
  • For a constant load, the motor's EM torque is equal to the load torque and the torque angle is a non-zero constant. A sudden change in load will upset this steady state. The locking between the rotor and the RMF is not rigid! A sudden increase in load causes a temporary slow down of the rotor, which simultaneously increases the torque angle and the EM torque. This accelerates the rotor back to the synchronous speed. As the rotor reaches synchronous speed again, the torque angle is larger than needed and the rotor speed overshoots the synchronous speed. This reduces the torque angle, and the EMF torque drops simultaneously. The rotor decelerates and the rotor speed now undershoots the synchronous speed, etc. I.e., the rotor speed oscillates around the sync speed. This phenomenon is known as "hunting" and "phase swinging". Under certain conditions, these oscillation may diverge ( = exponentially increase in amplitude), even to destructive levels. The oscillation can be reduced by adding damping windings (a.k.a. amortisseur windings) to the rotor, and by large load inertia (e.g., a heavy flywheel).

Berhard station

Fig. 114: Possible configurations for a pole pair of an AC motor

The following graphs show the torque-versus-speed characteristics of an asynchronous and a synchronous AC motor:

Berhard station

Fig. 115: Torque-vs-speed characteristic of a typ. asynchronous AC motor (left) and of a synchronous AC motor

The synchronous rotor speed of an AC motor is clearly proportional to the RMF that is generated by the stator, i.e., to the frequency of the AC excitation of the stator. However, it also inversely proportional to the number pole-pairs of the rotor. The simple formula is given in the figure below. For instance, for a 50 Hz excitation, the synchronous rotor speed Ns is 3000, 1500, 1000, and 750 rpm, for 1, 2, 3, and 4 pole-pairs, respectively. The number of pole-pairs is an integer value, so it obviously cannot be chosen as freely as an excitation frequency. Compared to asynchronous motors of equal power and speed, synchronous motors are attractive for low-speed ( < 300 rpm) and ultra low-speed drive applications: their efficiency is high, their power factor can always be adjusted to 1 (via field current adjustment), and they are less costly.

Berhard station

Fig. 116: Synchronous motor rotors with 1, 2, & 3 pole-pairs and slip-rings for DC power

(rotors shown with salient ( = protruding) poles, typ. for low speed applications, rather than cylindrical rotor with distributed windings)

The characteristics of the synchronous motor of "Bernhard" locomotive nr. 4 are not known: neither the frequency of the 3-phase AC, nor the number of rotor poles, nor the torque rating. All that is known is the gauge of the aluminium wiring of the 3-phase AC power (ref. 189): 16 mm2 (≈ 5 mm Ø, equivalent to about 5 AWG). Modern 4-conductor insulated aluminium cable of this gauge (e.g., NAYY-J) has a current rating of 50 amps. The aluminium wiring of the AC motor's DC field excitation (and of the DC motors) was 2.5 mm2 (≈ 5 mm Ø, equivalent to about 13 AWG). Modern 2-conductor cable of this gauge has a current rating of about 18 amps. Aluminium has 61% of the conductivity of copper.


The task of the locomotive motor control system is to smoothly increase the rotational speed of the "Bernhard" beacon from stand-still to exactly 2 rpm, and to accurately maintain that speed. The signals transmitted by the "Bernhard" beacon were printed aboard the aircraft with the "Bernhardine" Hellschreiber-printer. The compass-scale channel of this printer was synchronized to pulses transmitted by the beacon. As explained in the "Optical Encoder Disk" section (after Fig. 77), to make this synchronization scheme work, the allowed tolerance on the 2 rpm beacon speed was only ±0.2-0.3 % (p. 80 in ref. 181 and p. 8 & 18 in ref. 183). This small tolerance had to be met, independent of variations in the motor load (e.g., rail resistance around the circular track, wind load), and independent of amplitude and frequency variations of the 3-phase 50 Hz primary AC power. Note that towards the end of the war, the minimum frequency of the 50 Hz power grid was reduced to 43.3 Hz in the Central German block, and to 41 Hz in the Western German block (ref. 14).

One standard way to control and regulate motor speed is with a closed-loop control system. This requires tachometer feedback of the momentary speed, for comparison against the speed set-point. The amount of speed error (and possibly one or more of its time derivatives, and its integral) is then used to command the motor to speed up or slow down. If the control system is properly configured and dimensioned ( = control laws/algorithms), and it has sufficient control authority ( = "power"), then the torque-vs-speed curve of such a drive system can be made to approach that of a synchronous motor (see Fig. 115 above, ref. 215). So, when constant speed is required - as is the case here - then why not go straight to an inherently synchronous AC motor drive and use an AC power source that has a sufficiently constant frequency? Yes, indeed, why not! Doesn't this basically just move the control system from the motor to the AC generator? Yes, indeed. But there it is easier to implement, as we shall see.

On the one hand, the locomotive drive system must operate with 3-phase primary AC power that has varying frequency and amplitude. On the other hand, DC power ( = rectified AC power) is required for several reasons. First of all, as explained in the "synchronous AC motors" section above, a synchronous motor is not self-starting. It must be brought close to synchronous speed by other means. Here: with DC motors. Also, the rotor of a synchronous AC motor is DC-powered. So, a 3-phase AC rectifier is required. And to complete the AC drive system, we need a "DC to fixed-frequency 3-phase AC" converter.

The next diagram illustrates the three main blocks of the "Bernhard" locomotive drive system:

Bernhard wiring

Fig. 117: Top-level block diagram of the "Bernhard" motor drive system

(source: derived from ref. 189, 190)

The "Electrical power control, conversion & distribution" block has the following functions and associated control panels:

  • Selection of the source of 3-phase 50 Hz (nominal) AC main power: the public power grid, or the local generator of the “Bernhard” station.
  • Protection against over-voltage / over-current conditions of the primary AC power, and emergency shutdown. For the latter, there was a shutdown button located in the rotating cabin, in the round building below it, and outside the concrete ring.
  • Conversion of the selected 3-phase 50 Hz AC power to DC power. This was done with a 6-anode Mercury Arc Rectifier described further below.
  • Conversion of DC power into 3-phase AC power that has a constant frequency (unlike the primary AC power). This conversion was done with an electro-mechanical DC-AC inverter; in this case, a so-called "Conz" converter as described further below.
  • Distribution of the AC and DC power. The selected 3-phase 50 Hz AC main power is distributed to the rectifier unit and to the rotating cabin (via a slip ring assembly in the round building below that cabin). The DC power from the rectifier unit is distributed to the DC-AC inverter and to the rotating cabin, Also see the "Electrical & signal distribution" section. 

The motor drive control panel was located in the rotating cabin. It had separate controls for the three motor types (main drive DC, auxiliary drive DC with separate field winding, and asynchronous AC). The block diagram in the next figure illustrates the power conversion and motor drive control functions with more detail:

Bernhard wiring

Fig. 118: Electrical power and signal interconnections of the locomotive system

(source: derived from ref. 93, 189, 190)

The DC motors provided gradual acceleration from standstill to close to the nominal speed. They continued to provide the largest part of the drive power during normal operation, while the synchronous AC motor in locomotive nr. 4 provided the very precise speed regulation. It is unclear why the DC drives were implemented with two DC motors per locomotive, and why they had different control methods. The four main-drive DC motors were controlled via a large rheostat arrangement in the rotating cabin (see the hand wheel on the "Main Drive" control panel in Fig. 119 below). The three auxiliary-drive DC motors had a separate field winding that was wired to the "Auxiliary Drive" control panel (but apparently no control wheel).

Berhard station

Fig. 119: German engineer showing the locomotive control panel of Be-9 at Bredstedt to a member of the RAF-ADW

(source: Australian War Memorial photo SUK14636, public domain; ca. August 1945)

Each locomotive had two motors and two bogies. Most likely (but unconfirmed), each bogie had one motor. Each bogie had two axes. It is unclear if only one axis of each powered bogie was driven, or both axes (better traction, but mechanically more involved). At least one bogie of one locomotive had external gearing, with a belt or chain:

Bernhard wiring

Fig. 120: External gearing on a locomotive of Be-4 at La Pernelle

(source: 1946 film clip of Be-4 at La Pernelle, Cinémathèque de Normandie)

Since the AC synchronous motor provided inherent accurate speed control, there was no need for a closed-loop speed control system with a speed sensor. But there were two speed sensors: a tachometer in locomotive nr. 4 (with the synchronous AC motor), and a tachometer track on the optical encoder disk (measurement accuracy 0.1%) in the round building below the rotating cabin and superstructure. However, they were there for speed monitoring and alerting purposes only.


A rectifier is an electrical device that converts alternating current (AC) to direct current (DC). The device achieves this by allowing electrical current to flow through it in one direction only. A half-wave rectifier only passes either the positive or the negative half of a full AC voltage cycle. A full-wave rectifier passes the positive half cycle directly, and the negative half cycle with inversed polarity:

Berhard station

Fig. 121: half-wave and full-wave rectification of single-phase and 3-phase sinusoidal AC voltages

(assumes ideal rectifiers/diodes (no forward voltage drop, etc.), no source reactance (typ. inductive), no output smoothing, and no load)

By 1930, the Mercury Arc Rectifier (MAR, a.k.a. Mercury Vapor Rectifier; D: "Quecksilberdampfgleichrichter", ref. 163A-163N) had become the best method for rectifying high power AC voltage in industrial applications and electrification of light and heavy railroad. The discovery of the unidirectional current-flow of an atmospheric arc between a mercury pool and a carbon electrode, goes back to 1882 (Jules-Célestin Jamin and his co-worker Georges Maneuvrier, ref. 163H). The MAR was invented around 1900, and a glass-envelope MAR was patented by P. Cooper-Hewitt in 1902 (ref. 163J), based on his mercury vapor lamp. He marketed a metal-envelope MAR in 1908. In 1914, Irving Langmuir patented the concept of using a control-grid between the anode and the mercury-pool cathode (ref. 163K). This made it possible to arbitrarily choose the actual moment of arc initiation ( = switch-on via phase angle control, instead of it being determined by the primary power), and thereby vary the DC output. MARs can also be configured as an inverter instead of a rectifier, i.e., as a DC-to-AC converter.

MARs are a form of cold-cathode gas discharge tube. The rectifier consists of a glass or stainless steel vessel. The vessel is evacuated, or filled with inert gas. There is a pool of liquid mercury at the bottom of the vessel. This is the cathode. The vessel has one or more upward arms with a graphite anode. Clearly, a MAR is a static rectifier, as opposed to the mechanical rotary converters that preceded the MAR.

Full-wave rectification of a single-phase AC voltage requires two anodes. See Figure 122. For rectifying 3-phase AC power, the MAR must have at a multiple of three anodes.

Berhard station

Fig. 122: left: Simplified principle diagram with a 2-anode MAR (left) and an active 3-anode MAR

Like a fluorescent lamp, a MAR must be started. Conduction is initiated by dipping the starting (igniting) electrode into the mercury pool, passing a high current, and retracting the electrode. This locally heats up the mercury (the "cathode spot" or "emission spot") and vaporizes it. This starts abundant emission of electrons by the mercury cathode. The mercury vapor is ionized by the stream of electrons that flows to the anode, and causes plasma discharge (arc) between the anode and the cathode. The mercury ions emit both visible blue-violet light, and a large amount of ultra-violet radiation. The light may have another color when the vessel is filled with an inert gas, e.g., pink as in Fig. 122 (probably argon). Evaporated mercury condenses on the cool wall of the vessel (hence the large bulbous form), and returns to the mercury pool at the bottom of the device. The plasma discharge stops as soon as the anode voltage drops below a certain level, or anode current is interrupted. Hence, for rectification of an AC voltage, ignition must be synchronized with that voltage. Alternatively, excitation electrodes may be used to maintain the plasma. The anode material does not emit electrons, so electrons can only flow from the cathode to the anode. I.e., current can only flow from the anode to the cathode. The ripple on the DC output current is smoothed with a series-inductance ("choke coil").

The heat of the mercury vapor must dissipate through the glass envelope in order to condense. To help keep the glass envelope cool enough, an electric fan is typically installed below the MAR (as clearly visible in Fig. 125 and the the right-hand photo of Fig. 126 below).

Berhard station

Fig. 123: Effect of cooling on voltage vs. current curve ( = losses & efficiency) of a glass MAR

(source: Fig. 66 in ref. 192)

For operating temperatures below 10 °C (50 °F), special measures must be taken to protect the MAR against damage from instable operation, and the attached transformers against current surges. This may be done with surge diverters and cathode heating. Ref. 161. Note that mercury freezes around -39 °C (-39 °F).

The MAR anodes are connected to AC power via a transformer. Each phase of the secondary side of this transformer has an inductance: inductance of the secondary winding itself, and transformed inductance of the primary transformer windings and the AC power line. Inductance prevents current (here: anode current) from varying instantly. Hence, when one anode becomes conductive and its current is building up, the current of the adjoining previously conductive anode anode is still dying down: for a short time, both anode arcs are active simultaneously! This cyclic "overlap" phenomenon effectively short-circuits the main transformer's secondary phases that are associated with these anodes. For rectifier circuits, the "overlap angle" (a.k.a. commutating angle) is the commutation time interval when when both devices conduct. This causes the rectifier's DC output wave to temporarily drop to the average of the overlapping sinusoidal transformer phase voltages, which significantly distorts the ripple (Fig. 25, 29, 33, 37 in ref. 163C).

Berhard station

Fig. 124: The effect of phase-overlap on the DC voltage wave

(overlap due to source reactance or load; source: Fig. 25 & 29 in ref. 163C)

The voltage drop during the overlap period is proportional to the output current, and is also a function of the number of anodes and of the transformer inductance. As load current is increased, the operating time (duty cycle) of each rectifier phase is increased, and more phases will overlap. In case of a short-circuit load, the significant voltage drop across the rectifier arcs and resistive losses in the transformer (and other parts of the rectifier circuit) prevent full-time overlap of all phases.

MARs can be constructed for hundreds of kilovolts and tens of thousands of amps. They have been used, and sometimes still are (!), as rectifiers for locomotives, radio transmitters, control of industrial motors, welding equipment, aluminum smelters, high-voltage DC power transmission, etc. Glass-bulb MAR designs are typically limited to 250 kW (500 volt, 500 amps). For higher power levels, a steel-tank version was developed around 1908. Siemens-Schuckert developed a compact double-wall water-cooled tank rectifier around 1920. Through the 1960s, high power (up to gigawatts, ref. 163L) high-voltage DC (HVDC) transmission line systems were designed with MAR rectifiers and inverters. MAR technology was succeeded by ignitrons and thyratrons (ref. 163F), and then solid-state Gate Turn-Off devices (GTOs, e.g., thyristors).

In May of 2015, I obtained the black & white photo shown below. It shows the large 6-anode MAR of the Be-10 "Bernhard" at Hundborg/Denmark. The MAR does not appear to have control-grid electrodes. It was installed in a typical MAR-cubicle. The required primary transformers were located in the adjacent MAR-control cabinet. Likewise, the choke-coil, though the DC-motors may have had enough inductance so as not to require such a DC-current smoothing coil.

Berhard station

Fig. 125: The rectifier system of Be-10 at Hundborg/Denmark

(sources: (left) US NARA image nr. 111 SC 269041; (right) ref. 93; the thick "disk" below the MAR is actually the spinning cooling fan)

The "Bernhard" MAR was made by the Gleichrichter Gesellschaft m.b.H company in Berlin, manufacturer of rectifiers since 1919. They were acquired by the Swiss company Brown-Boveri & Cie. (BBC) in 1921 (ref. 191). BBC became ABB (ASEA Brown Boveri), after a merger between BBC and ASEA AB of Sweden in 1988. The "Bernhard" MAR was a model S 18 T "Glasgleichrichterkolben" (glass bulb rectifier; pdf page 20 in ref. 189). The complete rectifier cubicle with all the equipment and controls was model DRA 300A / 220 V, also of the Gleichrichter G.m.b.H. (pdf page 20 in ref. 189). The model designator suggests that the MAR had a rating of 330 amps DC at 220 volt AC. The cubicle included the standard cooling fan as well as a bulb heater. The circuitry around this MAR included 17 fuses! Standard equipment of each "Bernhard" station included one spare MAR (pdf page 21 in ref.189).

Berhard station

Fig. 126: Label of a BBC MAR-cubicle with a MAR built by Gleichrichter G.m.b.H. in Berlin

(source: H.-T. Schmidt homepage; MAR for 220 volt 3-phase AC @ 75 amps, 100/140 volt DC @ 150 amps, 140/165 volt DC @ 65 amps)

Other German MAR manufacturers of the era included Siemens-Schuckert Werke, several German subsidiaries of BBC, and the AEG company Apparate-Werke Berlin-Treptow (AT) that was founded in 1928. A 6-anode MAR with a height of 90 cm (3 ft, about the size of the "Bernhard"-MAR) can typically handle as much as 350 amps at 650 volts. The photo on the left in Fig. 127 below shows a small MAR, rated for only 220 volt / 100 amps (22 kW), together with its transformers. This MAR was manufactured in the 1950s by Elektro-Apparate-Werke J.W. Stalin in Berlin-Treptow. This was the post-war continuation of AEG-AT in the Soviet-occupied part of Germany. The 6-anode MAR in the photo on the right is about 60 cm (2 ft) tall. It is part an elevator (lift) system in a defunct very large WW2 air-raid shelter 140 ft (43 m) below Belsize Park in London. This MAR is still operational in modern days (at least through the year 2014). The cubicle is similar to the "Bernhard" cubicle in Fig. 125.

Berhard station

Fig. 127: TITLE

(sources: collection of Technical University Freiberg(left); ©2000 Nick Catford Subterranea Brittanica; both used with permission)

The "Bernhard" MAR shown in Figure 125 has six anodes. For obvious reasons, these multi-arm MARs are sometimes referred to as "Krakengleichrichter" ("octopus-rectifiers"). Compared to a 3-arm MAR, a 6-arm MAR reduces the ripple in the rectified voltage. It also requires a more complicated main-transformer connection to AC power, e.g., a delta/double-star or star/double-star configuration. See pp. 18-24 in ref. 163B. Six-anodes is typically sufficient. Having more anodes does reduce the output ripple (which has a lowest harmonic frequency of six times the 50 or 60 Hz main power). However, the reduction when going from 6-phase to 12-phase rectification is less than half the reduction when going from 3-phase to 6-phase (Fig. 35 in ref. 163C for no-load conditions). Also, cost increases rapidly without increasing rectifier output, and the already low power factor (due to an undesirably large phase angle between AC supply voltage & current) is further reduced.

Berhard station

Fig. 127: A 6-anode Mercury Arc Rectifier with  delta/double-star main-transformer connection to 3-phase AC power

(source: Figure 7.41 in ref. 163E)

When the motor voltage is reduced to slow down the motors (e.g., to brake to standstill), the large inertia of the "Bernhard" turntable will reverse the load torque of the motors. However, current can only flow through the MAR in one direction - it is a rectifier/diode. So, regenerative braking (using the motor as a generator and feeding the generated electricity back to a power grid) is not an option, and the motor cannot exert a braking torque. As a result, the armature voltage will increase to undesirably high levels. This is typically handled by switching-in a large dummy-load resistors across the armature of the motors, and dissipating the generated power as heat.

Here is a 36 sec video clip of a 6-anode MAR in action (you may want to turn the audio volume down a bit - MAR systems are very loud!):

A six-anode Mercury Arc Rectifier in action

Source: YouTube; one of four MARs of the 750 volt DC power supply system of the tramway network in Melbourne/Australia. MARs made by Hewittic Electric Co. (frmr. Westinghouse Cooper-Hewitt Co Ltd.) in Walton-on-Thames/England, and installed in 1936.


As stated above, a synchronous AC motor was used for obtaining and maintaining the required accurate locomotive speed. The frequency of the 3-phase AC power from the public power grid and the local backup generator was not sufficiently accurate. Hence, the fluctuating primary AC power had to be converted to constant-frequency 3-phase AC. Before the days of solid-state power electronics (1960s), the required power conversion was done by electro-mechanical means: an electric motor, an AC generator, and a closed-loop control system. The motor, generally referred to as the "prime mover", could be AC or DC. In the "Bernhard" system, a special DC-to-3-phase-AC inverter (D: "Umformer für Gleichstrom-Drehstrom") was used: a model NGJV So, 5/2 T built by the Conz company (sheet 16 in ref. 189). It was located in power generator building near the "Bernhard" beacon. The associated control panel had the following fuses: two for 350 V, 160 A, and three for 500 V, 35 A (per sheet 20 of ref. 189).

The Conz Elektricitäts-Gesellschaft mbH company was founded in 1887 by Gustav Conz. It was originally located in the Spaldingstraße in the southern German city of Ulm. The company moved to Hamburg in 1890, and acquired a plot of land in Hamburg/Altona-Bahrenfeld (Gasstraße 6-10) in 1911. An office building and two factory building were constructed here in 1912. Ref. 207. In 1962, Conz became a wholly-owned subsidiary of Deutsche Maschinenbau-Aktiengesellschaft (DEMAG) in Duisburg. The Hamburg plant was closed in 1995.

Conz Hamburg

Fig. 128: advertizing of the Conz company from 1924

(source: Elektrotechnische Zeitung (ETZ), Nr. 16, 17 April 1924)

Conz Hamburg

Fig. 129: advertizing of the Conz company from 1937 & 1940

A standard AC-generator has a field winding that is fed with DC power, and an armature winding that outputs the generated AC power. That is: a "singly-fed" generator. However, at the heart of a Conz converter is a high-power "doubly-fed AC generator" (D: "Doppeltgespeister Drehstromgenerator"). This is also called a "doubly-fed induction generator (DFIG) and "slip-ring generator" (D: "Schleifringläufergenerator"). It is similar to the singly-fed generator, in that the stator outputs the generated AC power. However, now the rotor is excited with 3-phase AC power at variable frequency. Thus the term "doubly fed". The frequency of this AC excitation power is continuously adjusted to compensate for changes in the speed of the prime mover. The result is regulated 3-phase AC power (D: "geregelter Drehstrom") with a constant frequency. In modern times, DFIGs are widely used for large wind turbines, as solid-state inverters required for megawatt-scale wind turbines are larger and more expensive.

Conz Hamburg

Fig. 130: singly-fed and doubly-fed 3-phase AC power generator

The next figure shows the principle diagram of the AC-AC Conz generator, as patented in 1937:

Conz Hamburg

Fig. 131: Conz convertor - variable-to-fixed frequency AC-AC converter

(source: adapted from the 1937 Hans Gross / Conz patent nr. 692583; see ref. 188 for a list of related patents)

The green overlay in the figure above, shows the control loop for keeping the output frequency of the DFIG constant - independent of variations of the frequency and voltage of the primary AC power and the output load: the 3-phase AC output power drives a synchronous AC motor. This motor is small (low power) compared to the DFIG and the primary mover. It drives a centrifugal governor. Based on the rpm, the governor adjusts a variable resistance. The resistance is placed in series with the field of a small DC motor, so as to change its speed. The DC motor drives a small 3-phase AC generator that excites the DFIG. Note that it is also possible to reverse the rotational direction of the three phases of the output. The efficiency of a Conz generator is higher than Ward-Leonard converter, especially under partial load or no load (idling).

The Conz generator configuration above uses an AC motor as prime mover, and a DC generator. This configuration can be simplified if a high-power DC source is available. This is the case in the "Bernhard" system, for powering the DC motors in the four locomotives. The DC-to-AC converter configuration is also mentioned in the Conz patent. So, a DC motor was used as prime mover, and the small DC generator was eliminated:

Conz Hamburg

Fig. 132: Conz generator - DC-AC configuration as probably used in the "Bernhard" motor drive

(source: adapted from the 1937 Hans Gross / Conz patent nr. 69258; see ref. 188 for a list of related patents)

The photo below was taken in the power supply bunker or barrack of Be-10 at Hundborg/Denmark. The Conz converter is on the right and is about 2 m tall. It is installed upright, probably due to the centrifugal governor. The cabinets on the left house the MAR rectifier and associated circuitry and controls.

Berhard station

Fig. 133: The Be-10 Hundborg installation - cabinets with the MAR and its controls, and the "Conz" generator

(photo courtesy Mike Dean, US National Archives & Records Adm. (NARA) image nr. 111 SC 269041; US gov't = no ©)


At the center of the concrete ring of the "Bernhard" beacon, there is round brick building with a flat concrete roof:


Fig. 134: Concrete ring with central- support building


Fig. 135: Left-to-right - the round building of Be-7 at Arcachon, Be-14 at Aidlingen, and Be-3 at Le-Bois-Julien

(photo Le-Bois-Julien: ©2006 T. Oliviers, used with permission)


Fig. 136: The round building of Be-8 at Bergen/Schoorl (left), and Be-12 at Nevid/Plzeň

(sources: photo Be-8: ref. 127; Be-12: © Jacek Durych, used with permission)

This small building has two functions:

  • Central support for the heavy rotating superstructure ( = cabin and antenna systems) of the beacon.
  • Stationary equipment room (D: "feststehender Geräteraum").


The following items were installed in the equipment room below the rotating superstructure (see Figure 137):

  • A 15-ring slip-ring assembly, suspended from the ceiling of the equipment room. Slip-rings allow electrical lines to traverse continuously rotating mechanical joints. The rotor of the assembly was driven by a shaft that descended through the ceiling of the equipment room (see Fig. 152) and rotated with the superstructure. The slip-rings passed electrical power and signals between the stationary equipment room and the rotating cabin above it.
  • Optical encoder disk assembly, suspended from the slip-ring assembly and driven by the shaft of that assembly.
  • Three audio tone modulators:
  • a constant 1800 Hz audio tone for the pointer beam transmitter
  • 2600 Hz audio tone pulses, representing the compass scale in Hellschreiber format, for the compass scale transmitter.
  • 2600 Hz audio tone pulses, representing the command-message text string in Hellschreiber format, for the compass scale transmitter (when replacing transmission of the compass scale with the command-message).
  • Two Hellschreiber printers (the same HS 120 printers as used in the aircraft), for printing:
  • the signals transmitted by the beacon, as received by a remote receiver.
  • the command-message text string, to verify it before actually transmitting it (only installed at a few beacons).
  • A patch board and associated patch cord, for composing the up to 10 characters of the command-message (only installed at a few beacons).
  • A power distribution and control panel. The panel also indicated the exact rotational speed of the optical disk (and, hence, of the beacon), as measured by a tachometer track on that disk.
  • A switch for selecting the forward/reverse rotational direction of the beacon.
  • An emergency shutdown button.

Bernhard wiring

Fig. 137: The round building below the rotating superstructure - equipment and interconnections

(source: derived from ref. 189 and 190)

The photo below is the only one that I have of the inside of the equipment room:

Bernhard transmitters

Fig. 138: Equipment inside the round brick equipment room of the Bernhard installation at Hundborg

(source: Figure 30 in ref. 93A)

The 15-ring slip-ring assembly passed the following electrical power to the rotating cabin:

  • DC, for the DC motors in the four locomotives.
  • 3-phase AC, constant frequency. Used for the synchronous AC motor in locomotive nr. 4.
  • 3-phase AC, nominally 50 Hz (directly from the public power grid or a local backup generator). Used for the power supplies of the two transmitters, as well as general lighting and heating.

The slip-ring assembly passed the following other signals to and from the rotating cabin:

  • Constant-tone audio modulation for the pointer beam transmitter.
  • Hellschreiber tone-pulse audio modulation for the compass scale transmitter.
  • Quadrant-keying from a switch (actuated by a notched disk on the shaft of the slip-ring assembly) to the transmitters. Purpose unknown.
  • Switch closure of the emergency shutdown button on the superstructure.

No photos are available of the slip-ring assembly. As an example, the photo below shows the slip-ring assembly of German "Panther" and "Tiger" tanks of the same era. Their slip-rings have a diameter of 12 cm (≈5 inch). There  are four brass rings for electrical power (12 & 24 volt, 50 amps) and seven rings for communication and lighting.


Fig. 139: The slip-ring stack of Panzer V "Panther" and VI "Tiger" tanks

(source: ref. 145)

The standard "Bernhard" equipment included a full set of 52 spare tubes (valves), per sheet 19 & 20 in ref.189 (pdf pp. 22, 23):

  • for the modulators and transmitter-keying units:
  • 10x RV12P2000, 1x RG12D60, 1x AZ12, 6x RV275, 4x RV335, 4x RG62, all made by Telefunken.
  • 1x STV100/25Z and 1x STV280/80 made by Stabilivolt.
  • for the measurement/monitoring equipment:
  • 3x RV12P2000, 6x LV1, 6x RG12D60, 1x RGN4004, 3x RV275, 2x RV335, 2x RG62, all made by Telefunken.
  • 4x STV150/15 and 1x STV280/80, made by Stabilovolt.


The walls of the building are made of brick. There are four windows of 1.2x1.2 m (4x4 ft), and a door. Whatever equipment was installed inside this building, it must have fitted through the door or a window. Floor-to-ceiling height inside the building is about 3 m (10 ft), so the floor is well below the base of the concrete ring. This is why there is a small trench and steps that lead down to the door.

The three diagrams below show the cross-section (with measured dimensions) of the concrete ring and round building of the "Bernhard" installation at Aidlingen/Venusberg, Arcachon, and Hundborg:

Berhard station

Fig. 140: Cross-section of the installation at Aidlingen/Venusberg

(based on the measurements that I took in June of 2012)

Bernhard station

Fig. 141: Cross-section of the installation at Arcachon

(based on the measurements that I took in July of 2012)

Berhard station

Fig. 142: Cross-section of the Bernhard ring on Gåsbjerg hill at Hundborg

(based on ref. 115)

The concrete roof of the round building is supported by four columns, made of massive steel I-beams (H-beams, D: "Doppel T Träger"): the flanges are 30 cm (1 ft) wide and 24 mm (1") thick! The web of the beams is 32 cm (12½ inch) wide and 15 mm (0.6 inch) thick. So these columns have a cross-section of 30x37 cm. The columns are spaced evenly in the round wall. The roof-joists are made of the same heavy I-beams. Why would such a solid, heavy construction be necessary? The rotating superstructure weighed 120 metric tons (265 thousand lbs). Assuming the weight was distributed evenly between the four locomotives and the central support, the roof had to carry 120 / 5 = 24 metric tons (53 thousand lbs) statically!

The following diagrams show more details of the steel structure:

  • 4 large steel I-beam columns, with end-plates. The flange of these vertical I-beams is 30 cm (1 ft) wide, and the height of the beams is 37 cm (1 ft 3")
  • 4 large steel I-beam joist, with a joist-to-column brace and a triangular filler plate. The brace is a heavy steel plate (3 cm thick), as wide as the flanges of the columns and the joists. The joist and brace are mounted to the column with 8 bolts. The brace prevents the structure from racking ( = sideways swaying of the tops of the columns). The brace is mounted to the column at a 50° angle.
  • 4 steel doubler-plates, to better transfer vertical forces on the joist to the column, and to distribute any bending force to both flanges of the columns. The doubler-plate is mounted onto the end-plate of the column with 8 bolts, and to the joist with 8 bolts.
  • 2 identical steel octagonal plates, to interconnect the four joists. Each joist is mounted to the two octagonal plates with 6 bolts. The plates have hole at the center.
  • 4 small steel plates that form the sides of a box that is placed between the octagonal plates. The corners of the box are butted up against the web of the joists, but do not appear to be welded to them. The function of the box is unclear - possibly to prevent the center of the top octagonal plate from being pushed down, possibly they where used to pre-assemble the two octagonal plates.
  • 4 small steel I-beams, connecting the joists above the braces. This makes the structure torsionally stiff. I found no sign that the columns are interconnected at the bottom.
  • numerous steel concrete reinforcement rods/bars ("rebar"), placed radially inside the concrete of the roof. The ends of the rods are curled back.

Berhard station

Fig. 143: The major elements of the steel support structure of the round building

(based on my measurements of the "Bernhard" at Aidlingen/Venusberg)

Berhard station

Fig. 144: Dimensions of the steel "skeleton" of the round building

(based on my measurements of the "Bernhard" at Aidlingen/Venusberg)

A doubler-plate reinforces the joint of each joist and the associated column:

Berhard station

Fig. 145: Dimensions of the column-to-joist doubler plate

(based on my measurements of the "Bernhard" at Aidlingen/Venusberg)

Berhard station

Fig. 146: Details of the steel support structure of the "Bernhard" at Aidlingen/Venusberg

Two heavy octagonal plates interconnect the four large I-beams joists at the center of the roof. One plate on top, one from below. There are two brackets on the bottom plate, for suspending equipment. The space between the brackets is about 55 cm (22 inch).

Berhard station

Fig. 147: The octagonal mounting plate against the ceiling - with mounting brackets to suspend equipment

Berhard station

Fig 148: Dimensions of the octagonal mounting plates and associated brackets

(based on my measurements of the "Bernhard" at Aidlingen/Venusberg)

Four smaller I-beams make the top of the structure torsionally stiffer:

Berhard station

Fig. 149: Top view of the steel support structure

(based on my measurements of the "Bernhard" at Aidlingen/Venusberg)

The next photo shows the remains of the steel structure (upside down), after removing the walls and collapsing the roof:

Berhard station

Fig. 150A: The remains of the steel structure at Nevid, after destruction of the building in 2015

(source: © jdvlavicka)

Berhard station

Fig. 150B: The remains of the steel structure at Nevid, after destruction of the building in 2015

(source: © jdvlavicka)

A very large ball bearing was installed in the middle of the roof:

Berhard station

Fig. 151: The round building with the raceway of a large ball bearing in the middle of the roof

(Be-12 at Nevid/Plzeň; source: © Jacek Durych, used with permission)

It had an outer diameter of about 40 cm (≈16 inch). The ball bearing held a cylindrical tube, that rotated with the antenna system and the cabin. The bottom of this tube has a flange, for connecting to equipment that rotated with the tube. The tube had a diameter of about 14 cm (5½ inch).

Berhard station

Fig. 152: Tubular shaft descending through the roof of Be-6 at Marlemont and Be-10 at Hundborg

(source Hundborg photo: www.gyges.dk, used with permission; note the original wiring)

A small turntable is installed on top of the ball bearing, and the tube is attached to it from below:

Berhard station

Fig. 153: small turntable on top of the shaft

(Be-6 at Marlemont; note the large mounting bolts inserted into the turntable from below)


The signals transmitted by the "Bernhard" beacon were monitored via a remote receiver and antenna, located at a nominal distance of 500 m (ref. 183; 400 m per ref. 13) from the beacon. The far-field of the transmitting antennas started about two wavelengths from those antennas, i.e., at about 20 m. So, why locate the monitoring antenna and receiver so far away? Most likely because the Mercury Arc Rectifier of the locomotive drive system caused a lot of radio interference near the beacon, and because the null of the twin-beam radiation pattern was not sharp enough close to the beacon.

The 1935 Telefunken/Runge/Krügel/Grammelsdorff patent nr. 737102 proposes using a fixed-location remote receiver to check the direction of the beam-null, as measuring and balancing antenna feed-currents does not guarantee its correctness. I.e., using the remote receiver during antenna calibration.

The vertical antenna was installed on top of a steel truss mast (lattice mast, cage mast; D: "Eisengittermast"). The antenna has a pointed tip, just like the feedpoint of the dipoles of the Bernhard's antenna arrays. A ladder was integrated into the mast. It consists simply of horizontal sections of L-bracket, mounted between one of the mast legs, and the braces to one of the adjacent mast legs. The box on which the antenna radiator is mounted, is about 50 cm wide. The monitoring receiver was located at the base of the antenna (p. 21 (pdf p. 18) in ref. 183, sheet 8 in ref. 189). The received signals were printed with a "Bernhardine" Hellschreiber-printer in the equipment room below the beacon's rotating superstructure (see Figure 138 above).

Berhard station

Fig. 154: The monitoring antenna mast of Be-11 at Trzebnica/Trebnitz

(photo (1980s): courtesy C. Piotrowski, used with permission)

This "Kontrollmast" (monitoring mast) had a height of 20 m (≈66 ft; p. 21 in ref. 183). According to US photographic intelligence (ref. 13), the mast was about 30 m (≈100 ft) tall, and the vertical antenna (hollow pipe) on top of it about 2.4 m (8 ft). Photometric analysis of the photo above shows that the antenna radiator (on top of the box at the top of the mast) is about 2.6 m tall, assuming a 20 m tall mast. I.e., it was a standard 1/4 wavelength vertical antenna. It has a pointed tip, just like the feedpoints of the vertical dipoles of the "Bernhard" antenna arrays, so possibly it was just half of such a dipole leg. The mast was installed on a concrete foundation.

So far, I have been able to determine the location of the monitoring antenna of the following Bernhard stations:

The mast and antenna were built by Hein, Lehmann & Co., the same company that built and installed the dipole antenna arrays of the Bernhard systems for Telefunken. The tall mast was delivered pre-assembled to the "Bernhard" site - at least at Aidlingen/Venusberg (ref. 103). Telefunken placed an order for six such masts in 1941, at a price of 2020 Reichsmark (RM) each (ref. 177C). This is equivalent to roughly US$12500 and €11500 end-2016, based on general inflation data (ref. 178A-178C). Note that Consumer Price Index (CPI) inflation data does not necessarily apply to specific products (such as antenna masts, electronics) or services. The billing does not state if the price included the actual antenna rod.

The photo below shows the mast of Be-11 at Trebnica/Trebnitz in Poland. It is the only "Bernhard" mast that has survived to date (2014). It is being used for antennas of a local FM radio station. The officially registered height of this "object" is 22 m (ref. 129). 

Berhard station

Fig 155: Looking up inside the monitoring antenna mast, and the base of the mast at Be-11 Trzebnica/Trebnitz

(©2014 C. Piotrowski, used with permission)

Berhard station

Fig. 156: The dimensions of the standard concrete foundation of the monitoring masts

(data source: Czarek Piotrowski, used with permission)

Each mast was placed on a 3x3 meter concrete slab foundation. The concrete slab is at least 75 cm thick (2½ ft). The one shown below on the left, has two rectangular dimples (see arrows); their purpose is unknown. Possibly they are vertical holes in the concrete, for inserting steel posts for mounting equipment, such as shown in Fig. 158 below.

Berhard station

Fig. 157: Concrete foundations of the monitoring mast of Be-9 at Bredstedt (left) and Be-10 at Hundborg

(sources: R. Grzywatz (Be-9); Hundborg Lokalhistoriske Arkiv (Be-10); both used with permission)

The monitoring receiver was located at the bottom of the mast (D: "Empfänger am Fuß des Masts"; p. 21 (pdf p. 18) in ref. 183, sheet 8 in ref. 189):

Berhard station

Fig. 158: Monitoring mast - remote-receiver and cable installed at the bottom of the mast

(source: ref. 13; probably Be-4 at La Pernelle,, based on the other photos in ref. 13)

Figure 157 above shows two equipment boxes installed near the bottom of the mast. Field line installations often had lightning protection at both ends: fuses (D: "Blitzschutzpatronen") in a junction box ("Anschlußkasten", AK). This may account for the second box.

The receiver was remote-tuned from the "Bernhard" station (D: "fernbedienter Empfänger"; p. 21 (pdf p. 18) in ref. 183, sheet 8 in ref. 189). Each "Bernhard" beacon used one of 32 operating channels in the 30-33.3 MHz frequency band. To be able to use the same monitoring receiver at all Be-stations, it had to be tunable. Another reason remote-tuning is that the transmitting frequency could be changed for technical or tactical reasons. It took 1-2 minutes to change the receiver frequency to a new channel frequency (p. 27 (pdf p. 24) in ref. 183).

There was an underground cable between the receiver and the Hellschreiber printers & control equipment in the round room below the rotating superstructure of the beacon. The cable was of type "Erdkabel RLM", special in-ground cable with a very robust outer insulation ("Kabelmantel").  It had four conductors with a 1 mm2 diameter (line item 28a on sheet 8 in ref. 189; ≈AWG #18). The receiver audio was also fed to a monitoring loudspeaker in the guard office near the beacon (line item 29a on sheet 8 in ref. 189).

So, what kind of remotely tunable receiver was used? The aircraft that used the "Bernhard" beacons had an EBL-3 receiver on board. The "F" version of this receiver had remote-control. However, the control interface by itself already required 4 wires just for tuning (p. 87 and Fig. 17 in ref. 72). Note that the antenna mast is also referred to as a "Diodenmast" (e.g., ref. 99). This German term suggest that a "Diodenempfänger" was used (a.k.a. "Detektorempfänger", "Kristalldetektorempfänger"). This is known in English as a "crystal radio" or "crystal set". They were popular in the early days of radio, and got their name from a small piece of crystal that was used in the signal detector. In the 1930s, the inconvenient "crystal detector" was replaced with a diode. Such diode-receivers are not only simple, they are also passive. No separate source of electric power, such as a battery or DC voltage via a cable, is required: the simple circuit is powered by the received radio signals. Also, it is very easy to remote-tune a crystal radio: all that is needed is a small low-rpm DC motor that rotates the shaft of the tuning capacitor (esp. a capacitor without angular limitation). Four conductors would have sufficed for audio (2 wires) and DC power (2 wires, possibly with reversible polarity to change tuning direction).

Would the audio output of a diode-receiver have been strong enough, without active amplification? This depends on three parameters:

  • The signal level required at the input of the Hellschreiber printer-amplifier in the cabin. A Wehrmacht Hell Feldfernschreiber had a specified nominal output signal amplitude of 2.5 volt (900 Hz tone pulses). To ensure proper printing at a receiving Feld-Hell machine, the maximum allowed cable damping was 5 Neper, which is about 43 dB, or a voltage attenuation factor of about 140x. That is, the printer amplifier required an audio input signal with a minimum amplitude of 2500 / 140 ≈ 18 mV.
  • The signal attenuation (damping) of the audio signals over field telephone cable with a length of 1 km (the maximum distance between the mast and the "Bernhard" ring). According to a 1945 manual of the Hell Feldfernschreiber (ref. 146), its range over standard field telephone cable of type DL500 was 36 km (22 km when wet), 60 km over regular pupin-cables (D: "bespultes Kabel", "Pupin-Kabel"; cable with a loading coil/inductance at regular intervals, typ. 250 m for German field cable), and 160 km over special pupin-cable of type FL250. That is, worst-case 22 km for 43 dB damping, or no more than 2 dB for 1 km. The required minimum 18 mV plus 2 dB is about 23 mV. So that would have been the required minimum output signal of the diode-receiver.
  • The RF field-strength induced at the remote antenna by the "Bernhard" transmitters and antenna system. Per the definition of the ITU (ITU-R BS.561-2), the field strength of a ½λ-dipole with an effective radiated power (ERP) of 1 kW is 222 mV per meter, at a distance of 1 km. Also see ref. 150. The "Bernhard" installation had an ERP of at least several kW (my estimate). I.e., at least several 100 mV per meter at the monitoring antenna. That would have been more than enough to generate 23 mV at the receiver output!

The schematic below shows a crystal radio that can power a small loudspeaker (ref. 147), similar to a small 1-transistor radio. It has a double-tuned circuit, a full-wave diode rectifier, followed by a voltage-doubler. The output voltage across the capacitors can be connected directly to standard high-impedance headphones (4000 Ω). It can also be connected to a low-impedance load, such as a loudspeaker or a standard "600 Ω" phone line. To match that impedance, a simple output-transformer (about 1:6 to 1:10) must be used. Solid-state diodes were readily available at the time, such as "Sirutor" diodes that were used in various Hellschreiber models. These were quite suitable for crystal radios (ref. 148, 149), as they have a forward voltage (a.k.a. "knee" or "turn-on" voltage) of only 0.2-0.3 volt (Sirutor type 1b).

Berhard station

Fig. 159: Schematic of a "crystal radio" receiver, capable of driving a small loudspeaker

(source: adapted from ref. 147)

Berhard station

Fig. 160: A "crystal radio" receiver built per the schematic above

(source: ref. 147)


The "Bernhard" installation had several consumers of electrical power:

  • Four electric locomotives. Reportedly one of the motor types was a low-rpm model, rated 10 kW (ref. 99). Let's assume all eight locomotive motors (two types of DC motor and one synchronous AC motor) had this rating. This would add up to 80 kW in total.
  • Two model AS 4 transmitters, each with an output power of 500 watt, or 1 kW total. Each AS 4 had a power supply model NA 500, with separate inputs for 220 and 380 volt 50 Hz 3-phase AC ("Drehstrom"), rated at 5 kVA. Ref 143 (p. 7) states that each NA 500 was normally powered by a standard heavy motor-generator model "A" ("schwerer Maschinensatz A"), rated at 12 kVA (see Fig. 163 below), or the public power grid. Either way, the power source was loaded with 5 kW. I.e., ten times the transmitter output power, and 10 kW total for the two transmitters combined.
  • Miscellaneous items in the equipment room (three modulators, two printers, three projector lights in the optical disk assembly) - let's conservatively assume 1 kW.
  • Heating and lighting in the rotating cabin and equipment room below it, as well as in the ancillary building) - let's assume 4 kW.

Adding up the four  types of electrical loads, we arrive at an estimated total load of 80 + 10 + 5 = 85 kW. To translate this to the required power from an AC source, we have to assume a worst-case phase angle φ between the generated voltage and current. For a reasonable cos(φ) = 0.8, the AC-power source would have to be dimensioned for at least 85 / 0.8 ≈ 106 kVA.

The table below shows the specified power supply for a number of Wehrmacht transmitters of the same era, including other beacon transmitters. The power supply outputs are all dimensioned for at least four times the transmitter output power.

Berhard station

Fig. 161: Specified power supply rating for a number of Wehrmacht transmitters

(based on various data sheets)

To be able to operate independently from the local public power grid, the "Bernhard" installation had its own local power generator, driven by a combustion engine (diesel or gasoline/petrol). Ref. 183 (pdf p. 18) states that "Bernhard" backup generator was powered by a diesel engine ("Notstrom-Diesel"). Ref. 180 implies that "Bernhard" station Be-0 near Trebbin had a 120 kVA backup diesel generator. Note that this station was also a test site, and probably had more ancillary buildings (labs, offices, kitchen, living quarters) than a standard "Bernhard" station.

Based on the local situation (access to the local public power grid), the  "Bernhard" installation typically included a high-voltage bunker ("Hochspannungsbunker"), e.g., ref. 99. It contained one or more transformers, to connect to the multi-phase regional Hochspannungsnetz (high voltage public power grid, 110-220 kV in Germany), or to the local Mittelspannungsnetz (6-60 kV, typically 10 or 20 kV). The local Niederspannungsnetz (several km) carried less than 1 kV. Incidentally, towards the end of the war, the minimum frequency of the 50 Hz power grid was reduced to 43.3 Hz in the Central German block, and 41 Hz in the Western German block. Ref. 14.

Berhard station

Fig. 162: The two sources of electrical power of the "Bernhard" installation

The next table lists the horse power (HP) of the engine for a number of Wehrmacht electrical power generators ("Maschinensätze") for transmitters, various models of Flak-Scheinwerfer (search-lights for Anti Aircraft gun installations; ref. 152A, 152B) and even a field bakery. This ranges roughly from 1.5 - 3 HP/kVA. Assuming a 120 kVA generator, this implies an engine with 180 - 360 HP. The specification of the actual "Bernhard" generator and associated engine is unknown.

Berhard station

Fig. 163: Engine power for a number of Wehrmacht generators

Ref. 103 suggests that the generator(s) of the "Bernhard" station Be-14 at Aidlingen/Venusberg were powered by two engines that came from high-speed patrol boats of the French navy. However, the French navy's "vedette rapide" boats were standard British-built "Fairmile" models. The smallest model had two Hall-Scott Defender V12 gasoline (UK: petrol) engines of 650 BHP each. The larger "A" model had three 600 HP engines. Clearly a single engine would have been able to supply more power than needed for a Bernhard installation, even if power was also generated for the local FLAK unit and its search-light(s). On the other hand, this Be-14 station was built towards the end of the war, when materials other than rocks and stone were in increasingly short supply. A strong engine that cost nothing, is better than a correctly sized engine that is not available... There are two stone buildings near the ring of Be-14. In the middle of one of them, there is a rectangular concrete slab that measures 1.4 x 3.25 meters (4.6x10.7 ft). The slab has 6 pairs of shallow round dimples, 8 and 10 cm in diameter. The purpose of the slab in unknown. Possibly the dimples corresponded to mounting feet of a motor-generator.

Berhard station

Fig. 164: A concrete slab in the middle of an ancillary building at Aidlingen/Venusberg

Below one of the windows of the round building of the "Bernhard" at Arcachon, there is an old 4-prong junction box with triangular shape. Near this box, several old cables enter the building through the wall, just above floor level. This appears to be where 3-phase AC-power entered from the generator in a nearby building.

Berhard station

Fig. 165: Large 4-prong (3-phase) junction box on inside wall of the round building of Be-7 at Arcachon


Below is a listing of patents related to the "Bernhard" system.

Patent number Patent office Year Inventor(s) Patent owner(s) Title (original) Title (translated)
662457 RP 1935 W. Runge
K. Röhrich
Telefunken GmbH Antenneanordung zur Aussendung von zwei oder mehreren einseitig gerichteten Strahlen Antenna arrangement for transmission of two or more uni-directional beams
737102 RP 1935 W. Runge, L. Krügel, F. Grammelsdorff Telefunken GmbH Anordnung zur ständigen Kontrolle und zur Ein- bzw. Nachregulierung der geometrischen Lage eines Leitstrahls während des Leitvorganges Arrangement for monitoring and adjustment of the geometric direction of a directional beam [A/N & E/T beacons, via monitoring receiver]
767354 RP 1936 - Telefunken G. für drahtlose Telegraphie m.b.H. Verfahren zur Richtungsbestimmung Method for direction-finding [this is the primary "Bernhard" patent]
767512 RP 1938 A. Lohmann Telefunken G. für drahtlose Telegraphie m.b.H. Verfahren zur Richtungsbestimmung mittels rotierender Richtstrahlung Method for direction-finding with a rotating directional beam
767523  RP 1938 A. Lohmann
A. Bittighofer
Telefunken GmbH Empfangseinrichtung zur Durchführung des Verfahrens zur Richtungsbestimmung Receiver-side device for the implementation of the method for direction-finding
767524  RP 1938 A. Lohmann Telefunken GmbH Verfahren zur Richtungsbestimmung mittels rotierender Richtstrahlung Method for direction-finding with a rotating directional beam
767525 RP 1939 A. Lohmann Telefunken GmbH Einrichtung zur Speisung eines rotierenden Richtantennensystems Device for capacitive coupling of a transmitter to a rotating directional antenna system
767528 RP 1938 A. Lohmann Telefunken GmbH Verfahren zur Richtungsbestimmung Method for direction-finding
767529 RP 1938 A. Lohmann
A. Bittighofer
Telefunken GmbH Einrichtung zur Erzeugung angenähert rechteckiger, zur Modulation des Kennzeichensenders dienender Abtastimpulse bei einem Verfahren zur Richtungsbestimmung mittels Drehfunkfeuer Device for the generation of an approximately square pulse envelopes, for the direction finding method by means of a rotating beacon
767531 RP 1939 A. Lohmann Telefunken GmbH Verfahren zur Richtungsbestimmung Method for direction-finding [dipole antenna array arrangement with side-lobe suppression]
767532 RP 1939 A. Lohmann Telefunken GmbH Sendeanordnung zur Durchführung eines Verfahrens zur Richtungsbestimmung Antenna arrangement for the implementation of a method for direction finding
767937 RP 1939 A. Lohmann Telefunken GmbH Einrichtung zur Durchführung eines Verfahrens zur Richtungsbestimmung Device for implementation of a method for direction finding

Here are some ancillary patents:

Patent number Patent office Year Inventor(s) Patent owner(s) Title (original) Title (translated)
562307 RP 1929 J. Robinson J. Robinson Funkpeilverfahren Method for direction finding [transmission of course pointer, or compass scale info via Nipkow-video]
620828 RP 1933 - Marconi's Wireless Telegraph Co. Ltd. Funkpeilverfahren Method for direction finding [transmission of compass scale info via Nipkow-video]

Patent office abbreviation: RP = Reichspatentamt (Patent Office of the Reich), DP = deutsches Patentamt (German Patent Office)
Patent source: DEPATISnet


  • The purpose of FuSAn725, with 4 or 5 kW transmitters instead of 0.5 kW.
  • Open-wire feedline between the dipoles: characteristic impedance (spacing between the wires, wire diameter)
  • Continuous welded rail or jointed rails
  • Purpose/content of the four corner-sheds
  • Purpose of two different types of DC motors per locomotive
  • If both bogies of each locomotive were motor-driven, and one or both axes of each driven bogie.
  • Purpose of dimples in the concrete ring, between rail ties/shoes
  • Purpose of box with 5 kg of graphite (powder?) that was part of the standard equipment (sheet 20, ref. 189)
  • Purpose of the "Sektortastung" (quadrant keying, driven by a contact on shaft of the slip-ring assembly in the round equipment building) to both transmitters.
  • "juro" cable.
  • Why a special Telefunken crystal module was needed in the  AS 4 transmitters (improve the frequency accuracy/stability?).
  • Purpose of the remote tachometer in Locomotive nr. 4 (going to the rotating cabin), in addition to the tachometer track of the optical disk (in the round equipment building below the rotating cabin)
  • Frequency of the accurately regulated 3-phase AC (50 Hz), and the number of rotor poles of the synchronous AC motor of locomotive nr. 4.


Note 1: due to copyright reasons, this file is in a password-protected directory. Contact me if you need access for research or personal study purposes.

External links last checked: October 2015

red-blue line

©2004-2016 F. Dörenberg, unless stated otherwise. All rights reserved worldwide. No part of this publication may be used without permission from the author.