Fullscreen HTML5Med Res (??MB)HTML4

Vintage Electronics Braybon Bros Automatic Voltage Regulator I have personal experience with this fascinating device that started life in the 1940s but was manufactured into the 1960s. It was used to control alternators to produce a more-or-less constant AC output in the face of varying loads. A production Braybon T3 AVR, serial number 4777. By Fred Lever B efore the Second World War, AC mains power was spreading around Australia. In the main cities, 240/415V three-phase was reticulated with homes utilising 240V single-phase. Appliances were being made to comply with this, using the 3-pin plug and socket system still in use today. Outside of the big cities, factories and farm properties installed their own generators using low-voltage DC or medium-voltage AC. The DC systems usually ran at 32V. Some installations had storage batteries charged by a generator (petrol, diesel or wind-powered). 240V AC systems started displacing DC systems as they could utilise the ever-growing commercial range of appliances and electric motors. Smaller systems were usually under 10kW with a single phase, while more ambitious 10-50kW systems were 3-phase types, allowing larger motors in workshops and homes. Townships installed generator systems large enough to power a local network, including street lighting, shop premises and homes (or domiciles). Examples of a farm/factory set and a town powerhouse installation are shown in Photos 1 & 2. 90 Silicon Chip The typical farm/factory set shown in Photo 1 comprises a Lister two-­cylinder diesel engine coupled to an alternator to produce AC. The alternator set was made by the Sydney firm Braybon Bros, providing around 6kW at 240V AC. Braybon and many other small manufacturers found a growing market to supply generating sets, alternators and control equipment for defence requirements during wartime, and the subsequent post-war boom into the 1960s. The typical town powerhouse shown in Photo 2 has two identical sets with six-cylinder diesel engines, possibly from Blackstone, coupled to brush alternators of around 200kW capacity. The power generated was 3-phase 415V AC. This was sufficient for a small town to run the main street lighting, shop and house power. You can identify the engine prime movers easily. Each drives an alternator that in turn drives a smaller DC generator called an exciter. The exciter and alternator are electrically connected in a series configuration, allowing the AC alternator to have its high-power voltage level controlled by the lower-power DC exciter using Australia's electronics magazine an automatic voltage regulator (AVR). The generating set in Photo 1 has the AVR control system mounted proud at the end. That is the two metal boxes, the top one being the hand rheostat (HR), and the lower one is the AVR. These items are production examples of a Braybon Type B rheostat and a Braybon Type 3 AVR. Out of sight is a gauge to show the output voltage, plus some operator controls, including the AVR hand/auto switch. Braybon Bros invented this type of AVR in 1940 to fill a wartime lack of supply of imported AVRs. The new AVR was very successful, despite its utter simplicity and modest cost in comparison to what had been available. I decided to make a replica of the design to investigate why it worked so well despite its apparent simplicity. AVR operation Fig.1 shows a basic electrical circuit of a generating set with an alternator, exciter and an AVR. The exciter is depicted on the left, with shunt field XF and armature XA. The armature supplies excitation power to the alternator field, AF. The output windings deliver three-phase AC power to the load. siliconchip.com.au Photo 1: a typical alternator that would have been used to power a farm or factory. The exciter generates a DC voltage, adjusted by resistors LR and HR. LR is set to provide a maximum limit and HR, the hand rheostat, is the manual voltage control giving a range of 50% to 150% of the rated voltage. The switch allows either manual control of the voltage by rheostat HR or, when switched to auto, the AVR is introduced to the field circuit. The AVR senses the AC voltage from the alternator and automatically adjusts its internal resistance. When properly set up for automatic operation, the hand rheostat provides an initial voltage of about 70% of the rated voltage on no load. The AVR is then switched on, taking the voltage up to 100%. The AVR then tries to maintain this despite a varying load. Fig.2 includes a typical excitation curve of an alternator, showing the change in AC voltage against DC excitation. With no load, the AVR is adjusted to settle at point A on the curve, the rated voltage. When the Photo 2: large alternators like these were used to power small towns, including street lighting. maximum load is applied, losses increase in the alternator, so higher excitation is required to advance to point B, by AVR action, to restore the voltage to normal. The rise in DC excitation from no load to full load in these alternators was in the range of 2-4 times, depending on many factors. Making a Braybon Type 3 AVR That brings us to the subject of this article: making a replica of the Type 3 vibrating armature point AVR (AVR T3) to glean some insight into how it worked. This AVR type was the third arrangement by the designer, S. C. (Stan) Braybon. Earlier types were the Type 1 solenoid carbon pile and Type 2 oscillating rotor point. There were some variations in the initial units, but once into production, the configuration stabilised and was virtually the same throughout the 40-odd-year life of the design. Fig.1: the basic principle of controlling alternator output by varying the exciter voltage. siliconchip.com.au Australia's electronics magazine The author has the unique position of having worked for Braybon Bros, both making and servicing the Braybon AVRs, as well as developing the subsequent solid-state AVRs for the Braybon set range. Back in the 1970s, I had a toolbox holding all sorts of spares used on service calls. When no longer used, I parked it under a bench in my workshop. It lay almost untouched for decades, to be opened only when searching for some piece of gear related to 1960s gensets, like carbon slip ring brushes or field rectifiers. For this article, tipping all the contents out and sorting through the junk at the bottom revealed some truly unobtainable, crucial parts for a T3, not the least being a couple of sets of tungsten vibrating points and some badly machined metal parts. They could at least be re-machined and put back to use! This enthused me, so I set to work using photographs of a stock AVR and Fig.2: the relationship between the DC excitation voltage and alternator AC output voltage is not completely linear. September 2026  91 a factory AVR test unit in my collection as a guide to make new parts. A production unit In the lead photo we have an AVR T3, like the one mounted on the genset in Photo 1. The top cover is removed, revealing not that much in the way of parts underneath! Nothing is missing. By referring to Fig.3, you can see almost all the active items. Coil M is the red bobbin at the rear of the lead photo, rectifier RB is the selenium plate device and transformer TX the grey object. The ballast resistor (BR) and capacitor (C) are out of sight underneath. The rocking armature sits on the frame of coil M, with the points (PP) on the nose, and the reference spring (S) just visible halfway between the points and the centre pivot. The spring reaches through the baseplate and hooks onto a clever right-­ angle spindle mechanism that appears at the front as the round control knob. This AVR is fitted with the point-­reversing switch, RS, the toggle facing the front of the photo. With reference to Fig.3, a simple explanation of how it works is as follows. Accept that if the points (PP) are closed by the pull of spring S, a field resistor such as HR in Fig.1 connected to the REG terminals will be shorted out, and the AC voltage will rise. When the pull of coil M due to the rising AC voltage exceeds the pull of the spring, the points open and the AC voltage will fall. The points will ‘hunt’ for a balance point, and so the AVR will hunt around a voltage. On M, winding SC creates the pull-in proportional to the AC voltage delivered by TX, BR and RB. Coil CC is connected in anti-phase to SC, and the changing field current tends Photo 3: I started by recreating the magnetic assembly. 92 Silicon Chip Fig.3: the configuration of the voltage regulator. The magnetic field generated from the AC voltage to be regulated is opposed by spring tension to control a set of points. to damp any change of state, adding an ‘anti-hunt’ action. At first sight, that is all you need to know. Like anything else, the actual way it works is more complex. Building one and testing it was the only way to discover more. To do this, all I had to do was draw up a list of parts required, refurbish or make each part, tick them off the list and put it together. Simple! The magnetic path and machined a thread on one end to secure it to the Bakelite base plate. I similarly made the brass post to carry the lower stationary contact from a length of 12.7mm (half-inch) diameter brass rod with one end tapered and tapped for 3mm. The other end was threaded to secure it to the baseplate. I made these parts and confirmed that the dimensions suited each other by a rough mock-up, shown in Photo 3. I made the steel parts for the magnet coil assembly first. I had the subspec armature and new points from my trove of toolbox parts, but needed to machine the L-shaped bracket and the coil core. The easy way to achieve the L shape was to part off a piece of 3 × 2.5 × ¼-inch mild steel angle and shape the top of the L to a 45° chamfer so the corresponding pivot milling in the armature fitted freely. I made the coil core next, machining it from 25.4mm (one inch) diameter mild steel rod. I cut it to length The Bakelite baseboard Photo 4: the voltage adjustment screw mounted on a Bakelite baseboard. Photo 5: the magnetic assembly and spring are mounted on the baseboard. Australia's electronics magazine For the baseboard, I cut a 200 × 200mm piece from 9.5mm (3/8-inch) thick Bakelite sheet. The voltage-­ control spindle mechanism casting mounts on a centreline, and this determines the exact position of the coil assembly as the reference spring hooks vertically from the casting arm up to the armature. As shown in Photo 4, I refurbished the casting, positioned it on a centreline and fitted it. That allowed me to drill the holes for the magnet assembly siliconchip.com.au Photo 6: metal screws are used to make connections between the two sides of the Bakelite baseboard, making it an early type of doublesided circuit board. and do a trial fit, shown in Photo 5. Note the vertical position of the spring, determining where all else fits. Transformer and magnet coil To produce a transformer, I found a junk unit of the same core size as the original with a good 240V primary. I rewound the secondary to 10V AC and fitted the frame with angle-mounting foot brackets. The magnet coil needed a bobbin former to slide onto the steel core. I used a piece of 25.4mm inner diameter plastic conduit tube and for the end flanges cut the ends off a Jaycar hookup wire spool. I used plastic glue to fix the pieces into a functional bobbin, then wound on an estimated 800 turns of 0.7mm (0.028-inch) diameter Lewmex wire for the shunt winding plus a series winding tapped at 40, 80, 160 and 400 turns. The ballast resistor and modern rectifier came from stock. Next, I drilled all the holes needed in the baseboard for the parts, terminals and lead-through screws to make connections from one side to the other – see Photo 6. With the parts on hand and the board made, I could then assemble the unit and prepare it for testing. Photos 7 & 8 show the assembly in progress. Note the simplicity of the device. The casting turning the axial drive of the voltage control knob into a vertical direction to adjust the spring tension on the armature is a clever piece of design. Once all the parts were assembled, I applied a static test by driving the AVR AC input from a variac and terminating the REG terminals into a 3A 32V DC supply with a resistance load. That allowed me to adjust the vibrator running of the armature to a smooth action and to graph the output current change of the points with varying AC input voltage. Fig.4 shows the AVR static test circuit and the response curve. The DC load current varied between 0.7A and 1.7A from 255V down to 220V. As the AC voltage falls, the AVR increases the current through DC load XF. That is exactly the logic required for an exciter field load correction. When 60 turns was selected on the series winding, the slope of the response became shallower, indicating negative feedback lowering the gain of the AVR. siliconchip.com.au Photo 7: most of the parts are now mounted on the top of the board. Photo 8: the wiring on the underside of the Bakelite board. Australia's electronics magazine September 2026  93 Scope 1 shows the voltage to the actuating coil M being a 100Hz pulsing wave. The armature balances between the spring and magnetic pull sources like a see-saw and also vibrates at 100Hz, providing a ‘chopping’ action. If the AC voltage is low, the spring pull is dominant and the point duty cycle is high, as in Scope 2, and near maximum field current flows. If the AC voltage is high, the magnetic pull is dominant, the duty cycle is low (Scope 3) and the AVR has little effect on the field current, so the AC voltage drops. Some observations With the AVR set to hover at 240V, I measured the spring tension and found it close to 1200g. The spring wire is about 0.035-inches in diameter (~0.9mm) and operates with about a 10 thou (0.25mm) gap between turns. That suggests the spring is working in a linear part of its range. The magnetic path is about 20 × 20mm through the core but only 9.5 × 9.5mm through the frame parts. With a measured 0.6A DC coil current, the amp-turns of the coil is 480At (0.6A × 800 turns). The flux density in the frame is low enough that the iron path is not saturated, so it is linear except for the air gap. The pole air gap is about 3mm and the flux is concentrated in this gap. The pole face attracts the back end of the armature against the spring tension. Since the applied current is not a square wave but a sinusoidal halfwave, the effect of the vibration is to chop the point current into square waves as in Scopes 2 & 3. Back at Braybon, when I ran an AVR on a genset, the overall set noise tended to mask the buzzing. Now, when run on the bench, the noise is moderate, similar to a soft-spoken voice. You can tell how the AC voltage is going by the pitch and noise level of the points. This AVR speaks to you! Construction methods I tried to stick to old-school methods as much as possible. There are no crimped wire connections; all the lug shank connections are soldered. All joints on a 1940s production AVR were soldered with a wall-gas heated copper iron, solder stick and flux paste. AVRs used either lacquer or cotton-covered wire for the coils. Today, I used 1960s Lewmex high-temperature motor winding wire for the coils. The coil former in mass production was moulded Bakelite; mine used glued PVC sections. I had to use modern 1960s push-on quickchange terminals to make changes easy around the rectifier and series resistor. The rest is just how it was: drilled holes in the Bakelite plate, brass screws everywhere with spring or fan lock washers. I did use the flexible PVC-covered wire of the 1960s. Production AVRs had a lot of bare 16-gauge tinned copper wire connections. When I made them with the point changeover switch, you insulated the bare wires with spaghetti sleeving or used fabric-insulated wire. I painted my bare steel parts with etch primer and a light coat of hammer Fig.4: some experimental results obtained while tuning my replica AVR. 94 Silicon Chip Australia's electronics magazine silver. During 1960s production, we plated all the parts with cadmium. The shop had a plating bath with a cyanide acid solution and cadmium plates. OH&S, eat your heart out! That would not be allowed now. I did wash my hands after each use of the plating bath. Photos 9 & 10 were taken near completion of the project. Dynamic AC testing Once assembled to the point shown in those photos, I could test the AVR in closed-loop mode connected to an alternator. Getting a suitable alternator was a bit of a problem! I have a 10kVA generating set with a 1960s alternator in mothballs. Instead of the trouble of bringing that back into service, I decided to assemble a bench simulator that would provide 240V AC in response to a DC control voltage. That was accomplished using a Lucas (the prince of darkness) 12V DC car generator belt driven by an AC motor to provide a real-world ‘exciter shunt field’. I coupled that in series to a magnetic amplifier that had an AC output in proportion to the DC armature voltage input. That mimicked what a small exciter-alternator pair would look like to an AVR. The Lucas ‘exciter’ has inductance and a time constant. The ‘alternator’ has a time constant inherent in the magnetic cores plus added lag capacitance across the DC coil circuits if needed. In practice, the time constant obtained was about one second or slightly more. A drawback of this simulator was that the wave shape had a high harmonic content (see Scopes 4 & 5); however, this was not too dissimilar to some of the dreadful alternators I fitted AVRs to in real life! I could have added a harmonic filter to remove some of the bumps, but decided to leave that complication unless the wave shape created caused insurmountable testing problems. It turns out that this type of AVR does not like the distorted wave shapes one bit! As the smooth operation of the armature is rather dependent on the wave shape, the narrow high harmonic wave with bumps tends to make the points ‘chatter’ erratically. However, the wave shape was good enough to obtain meaningful tests. Scope 6 shows the exciter shunt field voltage delivered by the AVR, responding to conditions of no load on the left, siliconchip.com.au Scope 1: the voltage applied to the coil during operation. Scope 2: the field voltage waveform when the alternator output is 220V AC. Scope 3: the field voltage waveform when the alternator output is 260V AC. Scope 4: my test setup AC output voltage with no load. Scope 5: the AC output voltage onload. Scope 6: the exciter field with a load applied after two seconds and then removed after eight seconds. Scope 7: the alternator field voltage during the test shown in Scope 6. Scope 8: the voltage across the points with no load. Scope 9: the voltage across the points with the maximum load. load on in the centre and load off at the right, with about 12 seconds across the screen, or one second per division. Starting from the left, after approximately three seconds, the load is applied. Note how the AVR points react rapidly and overshoot the field voltage. The armature bounces a few times until stable, balanced vibrating is reached. On the load release, at around the eight-second mark, the armature tips toward the core, opening the points with the voltage diving low. It then bounces a few times before settling in balance again in the no-load condition. Scope 7 shows the exciter armature output for the same conditions of load. Here, the rotor mass absorbs most of the spiky switching of the field but follows the outline of the response, with a couple of bounces either way on load application and shed. Given the one-second-per-division scans, the time constant of the AVR and alternator from disturbance to settling is again about one second. Scope 8 shows the switching voltage across the points with no load. Here, both the duty cycle and amplitude are low. In Scope 9, on full load, the duty cycle and amplitude have risen. The point chopping contains the underlying 100Hz vibration, but this is modified by the armature rocking about trying to follow the coil’s pull. Neither condition shows a regular rhythm, as the AC voltage is always moving about and the AVR hunts slightly trying to maintain balance. Compare these with Scopes 2 and 3, which were taken while running openloop with no AVR action. would be 216V AC. Full load on this simulator was decided to be when the excitation had increased to three times that of no load. In real life, one would set the voltage high on no load, say to 250V, and live with 226V at full load. If that range sounds a bit wide, back in the day, the mains could vary by 20%, usually downwards, so a 10% drop was quite acceptable. Note also the AVR’s response may not be ‘RMS responding’. In particular, a change in wave shape could skew the magnetic pull one way or the other from RMS. Figuring out if such an AVR is responding to the RMS, peak, average of the voltage or something else would be an interesting mathematical exercise, but far beyond my capabilities! In practice, the AVR would be set with damping turns selected so it was not too unstable in response to load changes, and the resulting voltage regulation would be accepted. siliconchip.com.au Voltage regulation With this setup, the AC voltage regulation was about 10%; that is, if set to 240V at no load, the on-load voltage Australia's electronics magazine September 2026  95 Photos 9 & 10: the completed unit, ready for testing, and a custom made cover for it – like the original units had. For a three-phase alternator, Braybon always offered just single-phase sensing; this was found to be ‘good enough’. It was only when alternators were required to give closer regulation, approaching ±1% in later years, that we developed solid-state AVRs with more gain and more powerful stability circuits. Still, that is a story for another time. An early form of op amp The operation of the magnetic path and vibrating armature is more complex than at first sight. What struck me when contemplating this is how close the magnetic circuit is to being an operational amplifier. The armature pivots like a see-saw depending on the difference in the pull of the inputs, just like an op amp. The output of this ‘op amp’ is the points mounted on the end of the armature. I can liken the design to an op amp run from a chopper power supply driving a power FET to control an external DC current. If this sounds farfetched, consider Fig.5. The AC input supplies both the rail and the level-sensing applied to the ‘op amp’. The sensing level through SC drives the op amp’s negative input, and the reference spring S drives the op amp’s positive input. You can consider the spring as a zener diode. The output of the op amp chops in response to the unfiltered supply rail, and its DC level depends on the difference between the +S and -SC inputs. 96 Silicon Chip The chopped DC level then drives the output, in actuality the points. These bridge the REG terminals that are connected to the field rheostat of the exciter. The exciter current is fed back to the negative input of the ‘op amp’ by CC, tending to damp the AVR action and lower the loop gain. A copy of the Tirrell AVR? At university, the Braybon AVR came up as a topic and was dismissed as a local copy of the excellent Tirrell AVR design. Well, not really. The Tirrell is a much more complex device intended for use with large alternators of megawatt capacity with long time constants. It is also a precision device with a very high price tag, befitting its quality of build and performance. A simplified circuit diagram of the Tirrell AVR is shown in Fig.6. Comparing Fig.6 and Fig.3, the only common thing about the Braybon and the Tirrell is the use of points to control a current! The Braybon’s designer, Stan Braybon, described the AVR series in his 1940s hand-written notes I possess. Knowing the man as my employer in the 1960s, I know the inspiration for the Type 3 comes from his experience with motorcycle and road vehicle voltage regulators and ignition system magnetos. If you like, the Braybon Type 3 is a much-enlarged Bosch or Lucas 12/24V vibrating point battery charge regulator. Much enlarged means shifting sensing coil operation from 12/24V Australia's electronics magazine DC to 240V AC, points operation from 12/24V DC to 100V DC, chopping the points to give astatic operation and employing feedback to accelerate the armature response and to reduce hunting (stabilisation). His design notes show the evolution of the AVR types; the ‘bulletproof’ mechanical construction reflects the engineering experience of the man. The Braybon AVR Type 3 may have been ‘cheap and cheerful’, but it was an advanced design that was very effective in its diverse usage. Final thoughts While working for Braybon Bros as an electrical fitter, I encountered many types of voltage regulating systems fitted to many different types of generating sets. These systems ranged from open-loop compensating arrangements with shunt/compound field control, magnetic control via saturable reactors, to even having the voltage set by a manual control. Better-managed alternators used closed-loop devices that can be called AVRs, where the voltage level was sampled, compared to a reference, and the excitation level adjusted automatically to a standard. I came across many brands of AVRs when I was discarding a failed AVR and fitting a Braybon unit. Some I can remember are: Brown Boveri and Cie (Co), Metropolitan Vickers, GEC and Westinghouse. All of those were motorised rheostat types. There were others employing contact points, like the siliconchip.com.au engine-driven Tillitson and the Tirrell nodding point type. My brief at the time was to disconnect or remove the original AVR and graft in a Braybon unit. How I wished I had simply picked up and kept some of the marvels of engineering that went to the scrap heap! I never saw any other AVR brand that worked quite the same way as the Braybon, with a single balanced armature controlled by various sets of field coils. I thought I had found such a unit bolted to a competitor’s set, with the competitor’s nameplate attached. I was informed strongly that it was a Braybon unit of a very early build, with three coils and a rheostat volt control, along the lines of the NZ patent! That one I repaired with new points and a Braybon sticker attached. The point about the Braybon it that it is not what was called a ‘static regulator’ that moved from one excitation position to the next within a regulation band, like most motorised rheostats do in response to a load change. It is an ‘astatic’ type, where the control never sits still, hunts for a set value and responds in a non-linear way to load changes. Students will recognise that this is how a control system with a PID (proportional, integral and differential) feedback loop acts. The oscilloscope trace in Scope 6 shows some of this trait. On a load change, the excitation level accelerates almost instantly, then decelerates with overshoot to the next mean level. The effect on the AC level is to achieve the new level with almost the desired single over/ under shoot. From the more complex three-coil winding vibrator of early examples, the designer discovered fairly quickly that one coil winding could be dispensed with. Even with a mechanical voltage adjustment varying the reference spring, the AVR still exhibited fast response and satisfactory regulation and stability. After about 30 units were made, the design was settled, and mass production followed for defence and private purposes. In a nutshell, this archaic-looking vibrating-point magnetic field AVR embodied the classic features of a modern solid-state AVR, with switchmode power control and PID feedback, even if the designer may not have fully SC realised it. siliconchip.com.au Fig.5: you can think of the AVR a bit like an op amp since it uses negative feedback to regulate a voltage. Fig.6: the Tirrell AVR does the same job using a similar principle but with a different configuration. Australia's electronics magazine September 2026  97