The G3PIY Homebrew Power Supply for Valve Equipments

A word of warning.. high voltage power supplies like this can be very dangerous

even when switched off

 The following details are not intended as a guide to constructing the power supply exactly as described below but to cover some ground to give you ideas to build your own. The key item will be the mains transformer and you can use one similar to mine, or even two transformers, especially if you aim to use the PSU with the later model of the 19 set which has special biasing arrangements. You'll see below how the initial design changed as tests were made and new features were included.

I already have a power supply which delivers a single variable high voltage but, it's not ideal to drive the 19 set with a single HT supply as combining HT1 and HT2 is really only acceptable in testing rather than on the air, where it's desirable to get the best power out of the set. After initially testing my 19 set I decided to build a dual HT power supply (and at the same time make it flexible enough for other things as well). I suppose the purist will frown and suggest a proper contemporary power supply should be used (and I actually have one) but, for ordinary use of the 19 set in a typical ham shack, a mains powered unit will be more satisfactory. At this point I should outline the HT requirements between different versions of the 19 set. All require a standard HT of 275 volts for the receiver and transceiver sections but, as far as the elevated HT of 500 volts for the 807 power amplifier stage is concerned, newer versions of the set have a slightly different requirement concerned with biasing arrangements.

My older set has a standard HT requirement for both voltages; that is each has a common ground connection for zero volts. Later versions of the 19 set require the zero volts connection of the 807 HT to be isolated from the chassis. The older sets have a smaller, 6-pin power connector, compared with the newer MkIII which uses two larger connectors. You can easily make a power supply suitable for the MkIII set but you may need to use two HT transformers unless your junk box is very big and well stocked.

The HT voltages for the 19 set are not critical being nominally 275 and 500 volts, being derived in a normal installation from a power supply unit containing a pair of rotary transformers driven from 12 volts. Having the ability to crank up the two voltage rails can significantly improve the RF output.

The main requirement for the new power supply is a suitable transformer and I found, in a box in my workshop, an old Admiralty transformer that looks to be eminently suitable.

Read this note about transformers

The windings are all clearly marked as follows:

620v-550v-375v-0-375v-550v-620v and two windings at 5v 3A (originally used for rectifiers)

The HT winding is marked 250mA for the centre winding and 200mA for the extra windings so the overall rating is 315.5VA. A pretty big transformer!

I decided to start with the 12 volt DC needed for the 19 Set valve heaters. Connecting the 5 volt windings in series was straightforward as the sense of the outputs are marked on the transformer top. As the thing is marked 230 volts it should produce around 14.4 volts DC, dropping under load to around 13 volts which is ideal. I wasn't too happy with this arrangement... read on.

A pair of 6.3 volt windings or a 5 volt plus 6 volt could have been used but as the peak output of the former might be a little high, somewhat depending on the winding resistance, you could for example add a string of suitable diodes in series with the output (each will give you around 0.7 volt).

The lowest HT output may be miles too high for the 19 set but I have in mind adding a voltage stabiliser and a switch for varying the output, say from 250 volts to 450 volts or so (this will be HT1). The 550 volt winding again is miles too high, but a variable stabilised output is the aim, which should allow say 450 volts to 750 volts (this will be HT2).

Because the transformer is pretty heavy (25lbs, or more than 11Kgm) and because of the high voltages involved I decided to construct a case from MDF. This is very easy to work, offers good insulation and anyway I have several narrow cupboard doors bought from Ikea's bargain basement for £1 each. One piece gives me front and bottom panels, and two sides.

The MDF is not far short of an inch thick so the control switches will be countersunk into the inner surface of the front panel. I'll also fit a voltmeter and an ammeter.

Here's some pictures showing construction progress...

I cut the countersunk holes using a circular cutter then a small chisel to lift out the centre. The holes for the meters were too big for my cutters so I marked a circle and used a small drill. The finned heatsink mounts a full wave bridge rectifier for the 12 volt supply and the long tagstrip mounts diodes for the two full wave rectified high voltage outputs. The upper tagstrip will be used to mount voltmeter dropper resistors.

 Front view. The meters were from the junk box. The voltmeter has a 1mA movement so for 500 volts fsd the dropper resistor is 500Kohms or for 1000 volts fsd 1 Mohm.

I always use at least two resistors in series to handle the resistor voltage rating. Be warned; in this age of transistors some resistors will have a low voltage rating and a set of these will need to be mounted in series to provide a decent rating.

 The long tagstrip originally mounted six diodes per rectifier. I used 1N5404 diodes because I had a lot of these. As they are only 400 volt rated I used three in series for each of the two half-wave legs. The RMS voltage for the HT1 supply is 375 volts and the peak is 530 volts. Full wave rectifier diodes should be rated at double the peak = 1060 volts, so 3 diodes will give 1200 volts. HT2 has an RMS voltage of 550 and a peak of 777 volts and double this is 1555 volts so ideally 4 in each leg would be best so I added extra diodes, one in each phase.

The mains input socket is an IEC connector cut from a piece of computer power supply case. The 12 volt bridge rectifier is an SB256 (now coded SB2506) rated at 25 Amps and 420 volts RMS because I have several of these and its easy to keep cool.

Mains wiring in place. I've used a double-insulated grey cable in most places because I have a large drum of it.

 Some more wiring in place. The front tag strip is for meter dropper resistors. At this stage I tested the 12 volt output. A full wave bridge is bolted to the finned heatsink. Note the smoothing capacitor behind the heatsink. This is mounted on the rectifier and is initially 2200uF 50vw. The output with no load was 13.5 volts. When I test under load I might have to use a larger capacitor.

Later I load tested the 12 volt supply and found it dropped to 11 volts with a load resistor of 10 ohms. Adding additional reservoir capacitance bumped this up to about 11.8 volts (which is still too low) so I decided to fit a second transformer. As space was a bit limited I fitted a small toroidal type marked 0-6volts twice at 5 Amps. This produced over 18 volts with no load. In theory it shouldn't have exceeded 17.7 volts at 240 volts but I'm considering fitting a series pass regulator for this supply as I don't wish to damage any valves.

The regulator shouldn't get very hot as the waste voltage will be quite small because of circuit resistances coming into play at a couple of amps. A second option is to use a three-terminal TO3 style regulator. I have a box of 12 volt versions and adding a diode in the ground pin connection will increase the output a little, however these turned out to be obsolete type which despite their size are only rated at one amp. I also have some 5 volt 3-terminal regulators which are heavier and might be pushed up to 3 amps. Adding a 7.5 volt zener diode in its ground lead should do the trick. Alternatively I could use a TO220 NPN transistor with a 12 volt zener diode stabilising its base, however this might be awkward to design as the input voltage would vary quite a bit, thus affecting the zener current.

After puzzling over the best design of the 12 volt supply I decided to get a 7.5 Amp regulator. This is an LT1083CP which can be programmed via two resistors to deliver a suitable voltage. I used a 1Kohm pull-up resistor and an 8.2Kohm fixed resistor plus a 1Kohm pot and adjusted this initially to deliver exactly 13.2 volts. Placing a pair of 6.8 ohm resistors in parallel across the output (to draw just under 4 Amps) showed the ouput voltage didn't budge. The LT1083 is screwed directly to the heatsink holding the bridge rectifier placing the heatsink at 13.2 volts as the output connects to the centre pin and hence the exposed metal of the device. This doesn't matter as the bridge rectifier metal is isolated. If I'd used an insulating pad the power dissipation figures would have been degraded.

As I hadn't so far included measurement of the 12 volt output I decided to dispense with the voltmeter toggle switch and intead use a rotary switch to connect the meter. The most suitable I had was a 3-pole 4-way switch so this provides a spare setting if I decide to make further design changes. One option is to add a stabilised 6 volt output using the now redundant pair of 5 volt transformer windings. I may use the toggle switch to place the ammeter in either HT output.

 The next picture shows the HT wiring in place. Preliminary tests showed I'd got around 370 volts and 500 volts or so without reservoir/smoothing capacitors.

I added some surge limiter resistors. These are 40 ohms for the lower HT supply and 58 ohms for the higher voltage and are not at all critical.

I have lots of old circuit boards from lift drive units. Most units are driven from 3-phase mains and develop a high voltage DC supply using a 3-phase bridge rectifier and lots of smoothing capacitors. I used two capacitors in series for the lower voltage supply and three in series for the higher voltage. The capacitors are 1200uF rated at 400 vw so two in series provide 600uF smoothing and three give 400uF smoothing. To provide a discharge leak and a means of balancing the capacitor terminal voltages I fitted more parts salvaged from the drive unit. These are 100Kohm wirewound resistors and one is fitted across each capacitor. It takes about 5 minutes after switching off the mains for these resistors to discharge the capacitors.

 After fitting the smoothing capacitors more tests revealed the results below. The first shows a reading of 3.95 volts representing 790 volts output. The second indicates 2.8 volts representing 560 volts output.

These correlate with the expected peak voltages from the 550v and 375v windings and will vary a little depending on mains voltage (236V RMS during this test).

View showing progress so far. The meter indicates the slowly discharging high voltage.

 More work is required to add switchable variation of the two high voltages, fixing various parts to the case, adding output connectors etc etc

The simplest way of arranging a switchable output voltage was to use a series pass transistor. I selected a BUH515 left over from CRT monitor repairs. This has a max Vce of 700 volts and a max Vcb of 1000 volts and can pass several amps as long as it doesn't get too hot. I decided to set the minimum HT at 183 volts and add 33 volts per switch setting as I have a large bag full of 33 volt zeners, type BZX55-C33. These have a max dissipation of 500mW. The minimum voltage was determined by a 1.5KE150A TVS diode and by soldering 33 volt zeners around the periphery of a 12-way rotary switch I could vary the output voltage from 183 to the max value of 550. This method is not ideal because my rotary switch is a break before make type and the voltage can flip up to maximum between switch positions. I might try a capacitor from the transistor base to ground to delay any rise in voltage between switch settings.

After testing the HT output at all its twelve settings with a load of 3.6Kohms, I found an average voltage drop across the smoothing capacitors of about 7%. The output was typically about 350 volts at 100mA.

The BUH515 transistor can dissipate a fair amount of heat when appreciable current is drawn so I fitted it to a heatsink. The specific type of transistor I used is an insulated version which doesn't need any special mounting hardware at some expense of its maximum power rating. At 180 volts output and 200mA the BUH515 will have up to 400 volts across it and at 200mA this will result in 74 watts which is a little high (I didn't test at this level). At 300 volts output and 100mA current the dissipation will be 25 watts. Both figures assume the voltage across the smoothing capacitors is unchanged, but in reality will be reduced. I'll carry out some tests to see what the actual numbers are.

Recapping. I tested with a 3.6Kohm load so the load power was 9W to 84W (183V to 550V) and the BUH515 dissipated V x I or (550-183) x (183/3600) = 18.6W.

Above is the regulator for the lower HT voltage

 Having got the lower of the two HV supplies running successfully I tackled the higher of the two. Again, using a BUH515 series pass transistor I wired the 12-way switch with a set of 33 volt zener diodes. This time, instead of a single 150 volt TVS diode I used a single 150 and two 120 volt TVS diodes in series. This gave me 390 volts plus 11 x 33v or a total of 753 volts worth of zeners. After wiring it and checking I turned on the supply and monitored the two high voltages. The higher of the pair read progressively lower as the switch was rotated anti-clockwise and less zeners were selected, reaching a minimum of 400 volts output. The highest voltage was 760.

Above the eleven 33V zeners including one wired back-to-front (which I corrected)

 OK so far. I then added a load of four 1.8kohm wirewound resistors in series. This would draw about 55mA at the minimum voltage and 105mA at the highest voltage, assuming the latter didn't drop too much under load, or 22 watts and around 80 watts respectively. The BUH515 would not be under much stress at the higher power output, as it's Vce is lowest, but would dissipate about 21 watts of heat at the lowest output voltage. I chose a 50kohm wirewound resistor for feeding the chain of regulating zeners, giving a maximum zener current of around 8mA.

All went well for 5 minutes then there was a loud bang and the output voltage read zero. One of the TVS diodes split into two, the other two measured short-circuit when checked. All the 33 volt zeners were OK, but the BUH515 had burst into smithereeens. Back to the drawing board.... I know the BUH515 Vce rating could be exceeded under fault conditions as the collector voltage was 795 and the minimum Vce max rating 700 volts at 100mA. Was this the problem? My junk box appears to have a 2SD1441 whose safe operating graph indicates it can run pass 2 amps with a Vce of 800 volts if I need something better or was it one of the TVS diodes that failed first or something else?

Thinking about it. The maximum voltage at the collector is 795. The minimum voltage at the emitter is governed by the zener diode stack voltage which (at the lowest output voltage) is the sum of just the TVS diodes, which is 150 + 120 + 120 = 390V. Therefore the maximum value of Vce is 795 - 390 = 405V which is miles lower than the permitted Vce of 700V.

For the transistor to self destruct, due to this rating being exceeded, its Vce must have reached more than 700V and I can't see how this could have happened.

Above the blown-up regulator circuit. 1.5KE150A in half and BUH515 missing its lid.

 One thing I've neglected to consider is that the BUH515 is a bipolar transistor so it will draw base current which will be dependent on its gain (these big transistors can often have a very low gain). If the base circuit limits this current what will happen? I assumed the collector current would be limited and/or the extra current would be drawn through the base circuit. In my case the three TVS diodes and the 12 x 33v zener stack are fed through a 50Kohm resistor and at the maximum output voltage setting the whole stack of zeners is in circuit. The 50K resistor is designed to provide a zener current plus base current of 795/50mA or circa 16mA. As far as the 33 volt zeners are concerned this represents around 528mW. They're rated at 500mW which is not quite good enough if the base current was zero, but they were still intact. The TVS diodes are rated at 5W and these see between 2 and 2.4 watts dissipation.

The TVS diodes are connected between the base of the BUH515 and the 11 x 33v zeners and as the transistor draws base current this is added to the flow through the 50K resistor. This extra flow produces an extra voltage drop across the resistor which could (if too high) cut off the clamping effect of the zeners, reducing the voltage output of the power supply. Lets say at an output through the load of the four 1.8K resistors of 110mA (at 795 volts) the gain of the BUH515 is 10. This will result in a base current of 11mA and 11mA through the 50K zener resistor reduces the zener current to 5mA which is still OK. The 33V zeners are working at 165mW well within their rating now that I've included typical base current in the calculations.

The end result was this. One of the TVS diodes had blown into two and the others had failed short-circuit. All the 33 volt zeners were OK, but the BUH515 had blown up.

What else could explain the problem? Could it have been due to excessive heat? The transistor is an insulated type having a much lower power rating than its metal tabbed equivalents, but still some 60 watts.

At an output of 795 volts the Vce would be small but at the lowest voltage (where none of the zeners are in circuit, just the TVS diodes) of 400 volts the value of Vce was 395 volts. Output current was 55mA. Power dissipation in the BUH515 was therefore around 22 watts, well within its rating of 60 watts, unless the heatsink wasn't adequate? With transistors, it's not the power rating that's important per se it's the temperature of the chip inside the case. As the BUH515 is designed for TV applications it would ordinarily see only short pulses, not a steady state DC current as is the case here. My guess is the chip got too hot and a short developed between the collector and base. This would have placed 795 volts on the base across the TVS diodes. The three diodes rated between them at 490 volts would have seen 795 volts and they are designed to go short-circuit at this level of overload. All three would have failed short circuit and, as there's a considerable charge in the smoothing capacitor, 795 volts across a short circuit would have resulted in a loud bang (as it did) and blown up one of them. At more or less the same instant the transistor exploded.

Another point to consider is this. With no load the power supply was fine, so the fault probably occurred due to current flow through the BUH515, and it didn't explode instantly, it took 5 minutes... increasing chip temperature might explain this? Many devices are designed for specific purposes and their data sheets don't always cover operation at low currents for example.

Anyway, it's academic as I've decided to change the design and use a 1200 Volt 15 Amp IGBT in place of the bipolar transistor. This should be better as it will draw negligible gate current and hence preserve the zener stabilizing voltage. I'll also use a string of 33 volt zeners in place of the TVS diodes and apply the full HT across the whole array instead of switching some in and out of circuit, thus making the zener current and dissipation constant and so the rotary switches will now carry three wires instead of two (a pair carrying the high voltage from the string of 33 volt zeners and ground, plus one to select the final voltage). To protect the IGBT I'll consider fitting a 15 Volt zener across the gate to source. . I may also add a 500mA fuse in each HT feed to the IGBTs.

It's funny how device codes have grown as their size diminishes. The IGBT is an NGTB15N120IHLWG, a real mouthful. Click it to see the spec.

Above is the Mk2 circuit now superseded by Mk3 being tested. See the text for descriptions of the various parts


 The Mk3 circuit has an extra diode in each leg of the HT2 feeds; the ammeter is switchable between HT1 and HT2; the LT circuit now uses a separate transformer and a pre-set regulator for 13.2 volts; a small 12 volt cooling fan, plus ballast resistors in the transistor collectors.

The new transistors arrived and these are insulated gate transistors not FETs so should work in pretty much the same way as say 2N3055 types except base current will essentially be absent and base voltage will be the governing factor. I've decided to use a simple series pass regulator circuit and as base current is not important I'll fit a 47kohm resistor in series with the control voltage circuit to prevent damage in the event of something nasty happening.

Construction was easy as the mechanical bits were already in place. As the new transistors have exposed collectors I used high voltage sticky pads as insulation. These are cut out of sheet material and I rubbed the heatsink with fine emery cloth and applied a slight smear of heatsink grease to the back of the transistors before fitting the insulating pads and screwing down the transistors. This type has a mounting hole insulated from the collector so you don't need messy mounting washers.

Once construction was temporarily complete I plugged the power supply into a variac just in case I'd done something silly. Cranking the voltage up proved all was well however and I was able to make some measurements.

The lower HT supply was variable from 140 volts to 512 volts on the panel meter and checking this showed the latter was 504 volts and the collector voltage 538.volts.

The higher HT supply was variable from 284 volts to 680 volts (=687 volts) with the collector at 795 volts. Note that I didn't want too high a maximum voltage so added a string of 33 volt zeners to limit the output.

I've taken the precaution of mounting the zener diode strings which determine the minimum voltage on tagstrips mounted adjacent to the transistors so that if I decide to change the voltage settings this will be easy. There are some complications however. As the high voltage rectifiers and transformer windings have some resistance the collector voltages will be somewhat dependant on power output. I measured about 7% voltage drop running 100mA output so I guess a 10% drop in output voltages will be typical under load (both supplies share some of the same secondary winding). Because the collector voltages will drop I had to take account of this when choosing the size of the zener string, hence the slight loss of maximum output voltage under no load conditions. In fact I might have to make a modification or two once load testing is completed.

The new design has a 13kohm 2watt resistor feeding HT1 zener chain (11 switchable and 4 fixed) and a 12kohm 2watt resistor feeding the HT2 zener chain (11 switchable and 8 fixed). The 4 zeners fixed dictate HT1 minimum output of 140 volts and the total of 15 zeners dictate the maximum output of 512 volts. HT2 has 8 fixed zeners resulting in 284 volts output and a total of 19 zeners dictating a maximum output of 680 volts. The 33 volt zeners vary slightly from nominal which explains the 8% discrepancy. The design zener current is 5mA which gives a dissipation of only 165mW and feed resistor wattages of only 300mW. These figures are miles better than those for the original bipolar power transistor design.

Below the redesigned regulator circuit before neatening up and fixing into place.

To the right of each transistor is the set of zeners (4 & 8) for determining the maximum output voltage

Below: the front panel

I might remove the two meters and change their scale markings later. The voltmeter reads 1000 volts full scale and the milliammeter 250mA full scale.

I might add another switch and arrange the circuitry to indicate either HT1 or HT2 current.

 The next step is to test the power supply performance (progressively) under load, then I'll try running each supply at 100mA then the pair together and see the effects on output and collector voltages. The design should enable each output voltage setting to remain more or less fixed as output current is varied.

HT1 load resistors at min and max settings should be (140/0.1) ohms = 1.4Kohms @ 14watts and (512/0.1) ohms = 5.2Kohms @ 51 watts

This corresponds to transistor dissipations of (538-140) x 0.1 = 40 watts and (538-512) x 0.1 = 2.6 watts respectively, and total powers of around 54 watts.

HT2 load resistors at min and max settings should be (284/0.1) ohms = 2.8 Kohms @ 28 watts and (680/0.1) ohms = 6.8 Kohms @ 68 watts

This corresponds to transistor dissipations of (795-284) x 0.1 = 51 watts and (795-680) x 0.1 = 11.5 watts respectively, and total powers of around 79 watts.

What do these transistor dissipations mean? I must admit I'm a little rusty here but something like this might be about right..

There are several factors to consider

The actual transistor construction leading to its thermal contact characteristic = 0.68 degrees C per watt

The insulating pad and its efficiency. Mine is rated at 73 degrees C x cm squared per watt which with a pad thickness of 0.15mm = 0.016 degrees C per watt

The performance of the heatsink. Mine is unpainted finned aluminium in a restricted airflow so let's say has a rating of 1 degrees C per watt = 1.0

The maximum transistor junction temperature from the above figures will be of the order of 51 watts x (0.68 + 0.016 + 1) + ambient (say 25 degrees C) = 111 degees C.

The key figure is the heatsink, so by adding a small computer style fan perhaps running at a reduced voltage all will be well.

However, from these figures it would look better if I were to increase the minimum HT2 voltage by around 33 volts to reduce the transistor dissipation as the max junction temperature of the IGBT is quoted as 150 degrees C and, as I'm sharing the heatsink between HT1 and HT2, I suppose this factor will effectively raise the ambient temperature perhaps adding a further 25 degrees C to the final figure?

Reducing the highest dissipation of 51 watts can be done by removing one more 33 volt zener.

HT2 would then have a range of 317 to 713 volts and dissipation about 47 watts. Two fewer zeners reduce this to 44 watts and 99 degrees C for the max junction temperature.

Another option is to add a ballast resistor before the HT2 transistor collector to use up some of the waste power, perhaps 500 ohms which would drop 50 volts @ 100mA and reduce the transistor dissipation by 5 watts. A resistor of 1 Kohm might affect the regulation.. the reason being that the voltage feeding the string of zener dodes drops below the level needed to pass current through them (ie. below 512 or 680 volts for HT1 and HT2).

I decided to remove two 33 volt zener diodes making the maximum transistor dissipation 44 watts.

Semiconductors can safely get very hot but you have to keep an eye on the data sheet for the particular device. Some quote enormous power dissipations at 25 degrees C but a closer look at the figures might say zero watts at 100 degrees C. I've seen dozens of really high power devices that have blown up because a fan failed or a lift motor seized and thermal runaway just destroyed the thing. Fortunately, the transistor I've selected specifies 250 watts dissipation at 25 degrees C and, even at 100 degrees C can manage 50 watts, so it should be OK.

More flexibility can gained because it's quite feasible to add a switch or two and change either or both outputs to be continuously variable. This can be done by adding a potentiometer across the raw HT to ground and connecting the wiper to the transistor gate. This design would have a lower limit set by a fixed resistor in the earthy end of the pot. For example a 1 Mohm pot with a 470Kohm fixed resistor would give me around 250 to 800 volts output.

I tested HT1 using a load of 3.6Kohms. The indicated voltages dropped by only a couple of volts under load so regulation is not bad.

The following results were obtained.

HT1 switch positions: 518, 480, 448, 418, 380, 344, 312, 278, 240, 204, 176, 140 volts

HT2 switch positions: 724, 700, 660, 628, 592, 560, 522, 492, 458, 420, 384, 352 volts

Using a load of 3.6Kohms on the HT1 output I got from 38 to 143mA output. The heatsink started to get slightly warm at the 140 volt setting representing a dissipation of 15 watts.

I also need to change the two rotary switches to make before break types as some circuits might not work too well when the HT suddenly shoots up to maximum. The option which I vaguely considered is adding a capacitor or pull up resistor. As I'm now using a gate resistor of 47Kohms a 1uF capacitor will give me about 50mSecs and a 10uF about half a second or so switching delay. The easiest thing was to fit a couple of cheap 12 way rotary switches type CK1034 which I ordered and fitted. Oddly, although they're described as 12-way they have only 11 switch positions. Presumably one position is lost when the make-before-break option is chosen? After the new switches had been fitted the HT now moves smoothly up and down without the glitches to maximum voltage.

Below is the state of the design with the pair of IGBTs and LT toroidal transformer fitted plus the new rotary meter switch fitted in place of the original toggle switch, but yet to have the HT current switch fitted.

 Above is a view of the low voltage regulator temporarily wired in place and a small fan for IGBT cooling not yet fitted properly. The pale brown choc block is for the 12 volt output.

Below shows the new HT output current switch in place.

 You can see the circuitry for monitoring HT current (on the upper tag strip). I used four resistors in total. One each of 0.5 ohm is in series with each high voltage output (=current sensing resistor). Another resistor of about 0.5 ohm is directly across the milliammeter (=safety resistor) and a fourth resistor was selected to make the meter reading exactly right. To check this I connected a variable low voltage power unit (in series with a 10 ohm current limiting resistor) across each sensing resistor and set the voltage to produce 200mA in the circuit (as shown by the low voltage power supply output meter). The 50mA milliammeter in the homebrew power supply indicated 42mA. I then tried adding various resistors across the milliammeter until the reading dropped to exactly 40mA. The final choice was 3.3ohms and this was soldered across the safety resistor. The milliammeter is now set for 250mA full scale. Before this I'd measured the resistance of the 50mA milliammeter and found it was about 2 ohms. This points towards the resistor values used above. I say "points towards" because shunting a milliammeter is not an exact science. Exact shunt resistors were found by the trial and error method given above, having first established rough resistor values required to make the scale read 250mA.

The two sensing resistors are connected to a 2-pole toggle switch connected to the milliammeter. DOWN selects the higher HT rail and UP selects the lower HT rail.

The rotary switch (right of centre front panel) connects the voltmeter to the 12 volt output with fsd set to 25 volts and the two high voltage supplies which have 500 and 1000 volts fsd.

The wiring really needs tidying up now. The sides will be fitted soon and the fan will be fitted in a suitable position. A top cover will be added also.

Anyone reading this far will have realised (I hope) that I have included some detailed calculations because the design is critically dependent on the HT transformer, voltmeter and ammeter. Turning to the semiconductors: the key problem is temperature. Before selecting specific devices it's vital you read and understand the detailed specs because most have headline ratings which assume only nominal temperatures and infinite perfect heatsinks. At least in this day and age you don't have to rely on dodgy germanium semiconductors with thermal runaway and self-destruction, as was the case when I started in industry.

 Although I initially designed the power supply for an early 19 set, small changes will make it suitable for later models where the HT negative supply is not grounded.

In addition, because of the ratings of the transformer I selected, the power supply could be used for a T1154 transmitter, hence I'm carrying out some modifications. One is to increase the maximum value of the HT voltage and the second is to add a 6 volt DC output for the T1154 valve heaters.

You'll note that there are three sets of taps on the HT secondary winding viz. 375, 550 and 620 volts. These are RMS values and with a full wave rectifier can produce up to 530, 777 and 876 volts DC respectively. Currently I'm not using the 620 volt taps.

The low voltage for the 19 set is around 13.6 volts DC and for the T1154 about 6 volts. The low voltage LT stabiliser could be switched to provide either voltage.

EXPERIMENTS WITH THE LT1083 (see the spec)

At this point I tested the 13.6v output. The transformer feeding it via a bridge rectifier has a pair of outputs each marked 0-6V 5A After connecting four 1ohm power resistors in series across the output terminals I obtained the following readings:

 AC input off load  13.8 V RMS  DC output  13.6 V DC  Output current  zero
 AC input full load  12.8 V RMS  DC output  9.6 V DC  Output current  2.4 A

Clearly something amiss. Looking at the various losses you'll find the transformer and secondary wiring resistance is about 0.4 ohm (1V/2.4A). The LT1083 will have a guaranteed drop of 1.5 V at 7.5 A. The rectifier spec declares a loss of 8 W at 5A representing a total voltage drop of 1.6 V or for each diode 0.4 V. Therefore to achieve the 13.6 V output at between 2.4 A and 5 A will require a bridge output of about 15.5 V compared with the measured voltage of some 10 V across the smoothing capacitor. The peak voltage was measured as 17.2 V so the ripple should be less than 17.2 V - 15.5 V or 1.7 V.

Looking at the existing reservoir/smoothing capacitor. This is 2200uF and this value at 2.4 A load will develop only around 10 volts. What is required then is a much larger capacitor, or an increase in the AC voltage input to the bridge. In theory, increasing the smoothing capacitor to about 30,000uF should result in 1.7 volts ripple which subtracted from the peak of around 17 volts should allow the desired output of 13.6 V to be developed. Strange to say, I added a 30,000uF capacitor and the stabilsed voltage was 13.6 V with no load and dropped to exactly 9.6 V with a 4 ohm, 3 ohm 2 ohm and a 1 ohm load. Time to look more carefully at the spec of the LT1083 perhaps? The answer appears to be given by the clue that the output voltage sticks at 9.6 V at any load current other than zero. According to its spec the LT1083 has a bug (my paraphrase) such that two different stabilised outputs can be obtained because of instability. This instability is caused by two factors, firstly I omitted to fit a stabilising capacitor which is supposed to be circa 25uF and secondly the no load output must be at least around 10mA. This current should be drawn by the output voltage potentiometer (or a permanent load resistor). In my design the potentiometer resistors are 1Kohm and 5.6Kohm. The latter has a 500 om pot wired in series to adjust the output voltage. Originally, I'd fitted the stabilising capacitor so my initial low current tests were satisfactory, but when I modified the circuitry to provide switchable 6 volts and 13.6 volts I forgot to put it back. Putting back the capacitor should solve the problem.... well it didn't.

The answer was the two resistors forming the potentiometer were drawing too little current ie 2mA. The spec suggests 10mA minimum load so although with no load the output was OK, suddenly drawing several amps flipped the LT1083 to a second stable configuration where it stabilsed at 9.6 volts.

I swapped the potentiometer resistors for a 100 ohm and a 560 ohm. The 500 ohm adjusting pot was then set to provide 13.6 volts output. Adding a 3 Amp load showed the output dropped to 12 volts. Checking the LT1083 input showed thios was around 13.8 volts, not quite the 1.5 volts required for proper operation.

The large transformer has a couple of 5 volt 3A windings so I added one of these which raised the off load LT1083 input to 25 volts.

Now, adding a 3A and a 4A load didn't shift the output from its setting of 13.6 volts. I can also add the second 5 volt winding (in the correct phase) to increase the output current to at least 5 Amps. I tried this and found I could get exactly 5Amps at 12.5 volts without any obvious shift in voltage from no-load. I adjusted the output to 12.5 volts as the fsd of my meter was 25 volts and 12.5 represented half scale. I'll select a suitable TVS diode and fuse to protect against circuit failure.

The next step is to provide a stabilised DC output of 6 volts. Because of the danger of damaging valve filaments I'll need to add a TVS diode and fuse in each of the two outputs.


The power supply has two high voltage outputs. Testing was accomplished by connecting a suitable resistance across each output and measuring the current at each setting of the selector switch. I connected in series four 10 watt 1.8Kohm power resistors and checked the higher voltage output. The meter reads 1000V fsd and is presently calibrated 0-5V. At a reading of 2.0 I measured about 55mA. As the voltmeter failed to move when the load resistors were connected I was surprised to see the current flowing. At the maximum output where the meter read 3.2V representing 650 volts I measured 90mA. Again the voltage didn't vary from no-load. When I disconnected the load the croc-clip drew a large arc. Power outputs were 22 watts and 58 watts respectively.

Turning to the lower HT voltage, I set the meter to about 300 volts and connected the load resistors. The current measured a few mA and the output dropped close to zero. Clearly something is wrong. Either the transistor has failed in some way or the grid bias resistor is open circuit or gone very high? My guess is the latter so that the collector current isn't being supported by the output from the zener chain. I'll check tomorrow. Well, I checked and found the transistor had degraded having a leak between its gate and emitter. Reading the transistor spec gave plus/minus 20 volts as the max permitted gate emitter voltage. In my application the collector is say 500 volts, the gate say 300 volts and off load the emitter will follow the gate and settle at say 299 volts, however the instantaneous voltage at the emitter when switching on with a load resistor to ground will make the gate emitter votage equal 300 volts, so I guess that's the problem? I should perhaps add a zener diode or a TVS diode (uni or bi-directional) to prevent excessive volts appearing between gate and emitter? This should have a rating of less than say 15 volts. A TVS diode will be best because these are rated at very high instantaneous powers. Looking in my stock of parts an SMBJ5.0CA should do the trick.

Later, I fitted a new transistor (type GT15Q102) for the failed one. This has a very similar spec. and a comparable body shape.

I fitted the 5 volt TVS diode as planned and the high voltage circuit worked properly. I also added a 5 volt TVS diode to the other high voltage transistor.

During testing I checked the panel mounted milliammeter. 55ma and 91.6mA on my multimeter read 12 and 18 on the latter which has a design fsd of 250mA and a scale marking of 50mA. This means that true current = meter reading times 5, so 12 represents 60mA and 18 represents 90mA so the shunt I've added isn't too bad.


The table below gives the test results for the lower of the two HV supplies (load = 7.2Kohm)

 Switch Setting  Meter  Voltage V  Current mA  Power W
























































and the higher of the two... (load= 7.2Kohm)

 Switch Setting  Meter  Voltage V  Current mA  Power W
























































Provision for 6 volt DC output

 Having acquired a T1154 I decided to make some changes to the power supply. The first is to add a 6 volt output for the valve heaters. Because I do not intend to simultaneously power up both a 19 set and the T1154 I used the same transformer winding that provides the 12 volt output to provide the new 6 volt output.

The solution is straightforward and I decided to use a standard 6 volt regulator in combination with a series pass power transistor to give me several amps of DC. I also added a second transistor to provide a short-circuit protection plus a TVS diode to prevent over-voltage. The 12 volt supply uses a pair of 6 volt windings plus an additional 5 volts from the HT transformer but for 6 volts output I may not include the extra 5 volt winding. A spare position on the voltmeter selector will provide monitoring for the voltage.

Initially I'll string together some parts to establish the performance of the supply then tidy it up. As with the 12 volt supply the key parameters will be the resistance of the various parts when operating at the design output current. These include a ballast resistor used for short-circuit protection. The main components are a "PNP 3055" transistor (=MJ2955) supplied by Radiospares some decades ago, a BD244C protection transistor and a 7806 three terminal regulator. The circuit will be that suggested by the latter's manufacturer in Fig 14 of their datasheet..

I selected a pair of 0.16 ohm resistors in series giving 0.32ohms for the protection resistor which should allow up to 5 Amps to flow. The meter now shows 6 volts using a 10Kohm resistor to provide a full scale delection of 10 volts. The output voltage is fixed by the 7806 regulator but can be increased by about half a volt for each forward biased silicon diode in its ground lead. The T1154 requires 1.3A for each of its two PT15 and 0.7A for each of its ML6 valves, making a total of 4 Amps. Today, I rebuilt the breadboard LT circuit onto a tagstrip and fixed this into the PSU box. Testing revealed that I needed a pair of small capacitors at the input and output terminals of the 6 volt regulator which I'd fitted with a 1N4003 diode to increase the output from about 5.9 volts by a little to 6.02 volts. I connected the ground connection of the capacitors to the centre terminal of the regulator.

Connecting a dummy load provided about 3 Amps before the output dropped to about 3 volts. Shunting one of the two 0.16 ohm resistors improved things. I can now get over 5 Amps without the output voltage dropping below 6.02 volts which is also the no-load output. Next, I'll fit a TVS diode to stop the output voltage rising too high. This is pretty well essential because the no-load voltage at the emitter of the series pass transistor is over 20 volts and a circuit failure could lead to this reaching the collector. The TVS diode is rated at 5 volts but has a breakdown voltage of about 6.4 volts. If the output rises too high the diode will present a short-circuit and either the short-circuit protection will kick in or a fuse, which I'll fit in the AC feed to the bridge rectifier, will blow.

I've decided on another change because the voltage tends to drop below 6 volts at high currents with the series pass transistor running red hot. Although the MJE2955 is rated up to 200C it's maximum power dissipation is 117W. This means, if the temperature is kept within reason the thing will handle 5A with about 20 volts or so across it. However, I'm thinking about limiting the regulator output current and using a float charged lead acid 6 volt battery to provide running current. The battery is rated at 12Ah: I've also ordered a second LT1083.

Next... preliminary testing with the T1154.....

 Looking ahead with a view to adding a higher output voltage for the T1154, I haven't yet used the outer connections to the high voltae winding on the transformer. It's not really essential to have a stabilised supply, just something around the design voltage. I bought some additional high voltage diodes the other week and I have a collection of high voltage capacitors from scrap lift drive units so what can I expect? Given 240 volts input to the 230 volt transformer with 620-0-620 at 200mA, I have two options.

First, I can use a full wave rectifier with a centre tap to provide about 900 volts DC, or I can use a full-wave bridge and get 1800 volts. The latter is far too high so I should settle for the lower voltage.

Another option is to use the 375-0-375 at 250mA connected to a full wave bridge. That would give me about 1100 volts. Using the 550-0-550 winding I'd get about 1600 volts DC.

The most convenient solution is to just add a third set of rectifiers and utilise the 620 volt windings. That's the first option above and would give me 900 volts DC. With a rating of 200mA that provides up to 180 watts of DC power, given large enough capacitors for reservoir and smoothing.

I opted to just move the HT2 tappings to the 620 volt windings. As the dissipation of the regulator is critical you need to limit the voltage drop across it. I can do this by adding extra zener diodes into the base of the series pass transistor. Rather than just add more 33 volt zeners I decided to fit a 270 volt zener plus enough to give me a minimum output voltage of 500 volts. Testing showed I could vary the HT2 output from 500 to 800 volts. Running say 200mA output current will result in a power dissipation with the output set at 800 volts of 620 x root 2 (= peak voltage of 876) minus 800 times 200mA or about 15 watts.

If I was to draw the same current at the lowest voltage of 500 volts the dissipation would be 876 minus 500 times 200mA = 75 watts so I'll need to watch the current is kept to say 100mA (= about 38 watts) at the lower output voltages otherwise there'll be an earsplitting bang...

 Time for another mjor modification. The T1154 has a pair of PSU control voltages which I've decided to use. One turns on the low voltage and the other the high voltage. The first employs a relay switching supply so I'll add a small 12 volt mains transformer and rectifier to supply this. The output is fed to the T1154 and is returned when the mode switch is turned clockwise from its OFF setting. Once the 6 volt supply is active this voltage is routed to the PSU to turn on the high voltage supply when the mode switch is turned to one of the operating positions.

To avoid the need for a high voltage relay I'll carry out switching via the tranformer primaries. The high voltage will be operated by a 6 volt relay and the low voltage by a 12 volt relay. To enable the PSU to run normally I'll fit a toggle switch to over-ride the remote switching. With the toggle switch in it Local position mains will be applied to both the low voltage and the high voltage transformers. In the Remote position both transformers will be turned o by the T1154. The relay supply will be on when the PSU mains switch is turned on.


in progress...

Return home