| Created: 18/10/06 | Last update: 2021/05/18
<<< OzValveAmps

AVA100 Power Supply Units

High voltage and heater power supplies for valve guitar amplifiers to 50+ watts output.
Discussion and examples

Update: 23/7/09

Caution Danger high voltage
Caution 1:
All of these power supplies are potentially DEADLY, even the smallest. Take note of all warnings, cautions, and high-voltage proceedure. Be particularly careful of charged power supply caps - these can remain significantly charged for days if not bled down by a lamp load or similar.

Caution 1a:
The cans of the capacitors used in the voltage multiplier must be assumed to be live, and should be suitably insulated against accidental touching or contact with each other. With the centre-fed multiplier the transformer secondaries are also “live”.

Caution Work in progress
Caution 2:
The AVA100-Series is a work in progress and as such may contain significant errors.


Power supplies

A valve amp requires a power supply which delivers a few different voltages.

The basic supplies required are for the heaters and the high tension HT supply. Additional supplies may also be required to provide negative bias for the output stage, and low voltage for any solid-state sections.


The easy one to dispose of first is the heater supply.

For example, 6GW8's require 6.3 volts AC at 660mA (0.66A) each, so a minimum of 1.2 amps is required.

6BQ5's requires 750mA each, 1.5A total, and still enough spare to run some 12AX7's at 300mA at 6.3Vac each (parallel heaters).

It is fairly easy to provide a DC supply for the heaters if desired, but the simplest solution is a transformer that delivers 6.3Vac, and the trusty 2155-line will deliver 2.36 amps at 6.3 volts. This actually leaves 1.76 amps spare, so adding in a twin triode (0.3A) or five is possible.

Altronics type M2155L is identical to the Jaycar MM2002 and costs the same.

There are trannies with higher current ratings but they are listed as having a 6 volt output. 6.0 volts is actually at the lower tolerance limit (-5%) for 6.3 volt heater valves, so the 6.3 volt option is to be prefered.


Update: 10/6/08

Another feature of the 2155-series is that 6.3Vac centre-tapped is also available if that form of hum-dinging appeals. To do this, ground the tap marked “9.1V”, and take 6.3V between tap “6.3V” and tap “12.6V”, (no other connections to be made to the secondary). This should still supply (a now balanced) 6.3Vac at 2.36A for no extra components.

HT - High Tension

The preamp stages are normally powered at minimal current (1 or 2mA per stage) drain from the main HT supply that feeds the output stage, through suitable resistor-electrolytic decoupling networks. So the requirements of the output stage determine the HT supply design.

This HT is distributed to the different stages via a decoupling network of resistors in series, with big electrolytic caps to ground. The idea being to stop unwanted feedback, positive or negative, between stages via the HT line. Note that these front-end decoupling caps may normally operate at a lowish HT voltage but they will be subject to the full unloaded output voltage of the HT supply during warm-up before the valves start drawing current, and must be voltage rated accordingly.

Generally the main HT voltage is between about 250 and 500 volts, and at currents up to several hundred milliamps in a big amplifier.

For the 6GW8, 6BM8 and similar the HT will be between 250 and 300-odd volts at up to about 100mA for a simple amp, to perhaps 250mA for a twin combo fully optioned with multiple preamps, tremolo and reverb at full boogie.

Note: actually the supply of heater current is the key limitation to adding stages.

HT Trannies

In Ye Olde Days we would have simply bought a suitable tranny with a 250-0-250 or 285-0-285 volt winding (with two or three brutal heater windings), add a suitable valve rectifier, and we would be away. But apart from scrounge and specials (made to order), you can't get trannies like this any more. My own Playmaster 117 uses a tranny recovered from a dead B&W 21-inch TV set which is still more than equal to the task, but they haven't built telly's like that for decades.

The whole world has gone solid-state and low voltage/high current, so there is quite a range of such trannies. But as solid-state amps progress, so the required rail voltages have been rising, as high as +/-70 volts in some cases.

When we look for higher voltage transformers we find a rather odd problem; these are also the trannies with the highest power ratings. In most cases they are much more than required, 160 watts and up, and therefore heavy and fairly expensive.

Altronics and Jaycar types. (VA = volt-amps = watts)
M4925, 5025, 5125, and MT2144 are torroids

Back-to-back (serial)

One answer is to take two more sensibly-rated trannies and connect them back-to-back, producing an isolated 250 volt winding. In the SC Mudlark this was done using a couple of torroids.

This has a couple of potential advantages if multi-tapped transformers are used, allowing some adjustment of the final output such as driving a nominally 9 or 15 volts winding from 12 volts for more or less output voltage.

In theory transformers are bi-lateral devices, that is they work the same both ways.

This may be truer of torroids, but most available low-power E-I transformers do not work that well in reverse, in fact they tend to be pretty lossy forwards too but we don't notice.

Having two trannies in cascade doubles the effective series resistance of the supply, reducing its voltage regulation (having more “sag”). Double the weight is also a consideration in a portable amplifier the Mudlark can ignore.

A particular disadvantage is that the entire power for the amplifer has to pass through both, which means for a 60W amp consuming 120W you need two 120VA cores, a total of 240VA of core, copper, and weight for only 60W output.

SMPS - Switch-Mode Power Supplies

At least one article has been published (in Silicon Chimp) on converting a computer power supply to deliver high voltage by rewinding the transformer potcore.

Regulated High-Voltage Supply For Valve Amplifiers by Leonid Lerner
How to modify a surplus PC power supply to produce a 700V or 400V high-voltage rail.
Silicon Chip, Issue: 190 Published: 28 July, 2004

This is a major project in its own right.

It has the advantages of cheapness, low weight, and you can have exactly whatever voltage you desire - regulated.

For my part I think it's simpler to buy a bobbin-wound low voltage tranny of suitable core wattage rating, strip the secondary, and wind my own. It amounts to much the same thing without the complexity (and potential noise radiation) of the SMPS. With fewer bits it may also be more reliable. ;))

Series connection

Well, if we are going to have two trannies for the HT supply, what happens if we use available low-voltage trannies with their secondaries in series to double the initial AC voltage?

For our 60W amp we only need two 60W cores, half the VA of serial or back-to-back connection, because the cores now share the load between them. At least 30% lighter for a better regulated supply.

Voltage multipliers

Another option is to use a voltage-multiplying rectifier. These take the form of capacitor and diode ladders that act as charge pumps. These can be voltage doublers, triplers, quadruplers, and so on. Multiplier-rectifiers have a bad reputation for sagging badly under load, but as we will see, that is undeserved.

The advent of demanding SMPS (above) has produced a new generation of electrolytic capacitors of high voltage, high capacitance, small size, and low ESR or high ripple current rating. These are particularly suited to voltage multiplier applications. And if you scrounge power supplies or disposable cameras they're Free!

Voltage multipliers have been built using valve rectifiers, despite the difficulty obtaining isolated heater supplies, but there are many suitable silicon rectifiers available that allow us to avoid this problem. In fact 1000 volt 1 amp diodes are so cheap they are almost free. If you use scrounged bridges they are Free!

Tie the “AC” leads of a (recovered from dead PC supplies) bridge rectifier together, and call it two diodes in series. But first - test all diodes with your trusty little yellow DVM for about 700 millivolts forward (not shorted) and no reverse leakage at all.

Notional valve amp HT supply
(actual-build details, below)

Don't Panic! This is only a simple half-wave rectifier times four, but drawn somewhat differently to what you may be used to. It's two doublers stacked on top of each other and driven in the middle to make a voltage quadrupler. The series resistor is switch-on surge limiting (and some sag), the shunt resistors a safety bleeder. You'll find half of this in many Moody's, and in the Playmaster 102.


On one half-cycle the first caps in series (left hand) charge up to the transformer peak voltage. On the other half-cycle this pre-charged cap is added to the tranny peak voltage, so charges the second (right-hand) reservoir caps to twice the peak voltage on the tranny winding.

Exactly the same thing is happening below, but in reverse making a minus supply which adds to the output because it it stacked in series with the upper section, thus giving four times the peak winding voltage.

How much charge gets pumped?

Since the pumping rate is fixed by the mains frequency the amount of power this supply can deliver and it's voltage regulation depend a lot on the size of the caps used. I could buy big high-voltage caps, but the caps I've already recovered are 330uF. Will they be big enough?

The charge (Q, in Couloumb) on a cap is;

Q = C x V

...and we are lucky that charge is also simply defined as Amp/seconds past a point;

Q = I x t

[these two relationships are well worth pasting in your hat]

I hope it is obvious therefore that;

C x V = I x t

So we can now relate capacitance, voltage, and current over time. This is bound to be useful.

Voltage Regulation

The regulation of a simple supply is quoted in percentage voltage drop between no-load and full rated load (although the range 10% to 90% load might be more practically meaningful). Most of the small trannies being considered here have a quoted regulation of 5%. In practice this means the rated voltage at the rated current, so the no-load voltage will be somewhat higher, here about 5%.

Voltage multipliers don't have a great reputation for regulation. One of the reasons were the limited capacities and ripple currents available, but modern electrolytic caps have changed that.

Using a 25+25 volt tranny as an example; when un-loaded the first cap will charge to the peak output voltage of the tranny.

This is root(2)=1.414 times the rms value, or around 70 volts. We must remember to subtract the diode voltage drop (~0.7V), but since the output voltage is quoted for full load we can expect a nominal 50 volt tranny to be producing 5% more or about 52.5 volts unloaded, so this covers any diode drops.

On the other half-cycle this stored 70 volts is added to another 70 volts from the winding, so the output cap is charged to 140 volts.

The same process is going on in the lower half producing -140 volts, so we have 280 volts across the whole supply. Unloaded.

So what happens when we load it? Ah... well, the HT voltage sags some and the superimposed hum voltage goes up from virtually nothing.

Elsewhere we have worked out that we need about 100mA of HT at full crank.

Applying Q = I t above, this is 100 milli Couloumb per second.

The caps recovered from a couple of ATX supplies are 330uF/200V. So across the output we have two in series (330/2=165)uF charged to 280 volts. Applying Q = C V gives 165e-6 x 280 = 46.2 mQ of charge stored; enough to keep the amp running for about half a second.

Since we don't really want to know the exact value, only if it is going to be enough, we can make a number of simplifications that avoid having to deal with exponential functions.

We assume that the amount of discharge between mains cycles is small compared to the overall time constant, and so what is really an exponential decay can be (very closely) approximated to a straight line; and that the change in current drawn by the amp won't fall significantly from the 100mA assumed.

The supply is recharging every 20mS and the amount of charge used by the load in that time is 100/50 = 2mQ per 50Hz cycle. So the cap charge will fall from 46.2 to 44.2mQ.

Q = C V [transpose, divide b.s. by C for V alone]

V = Q/C = 44.2mQ/165uF = 267.9 volts minimum.

The ripple voltage will therefore be 280 - 268 = 12 volts peak-to-peak or about 8.5 volts rms. Average is (268 + 280)/2 = 272.5 volts. So ripple should be a pretty reasonable 3% at full load.

Strictly the relationship Vpk = root(2) Vrms is only true for a sine wave, and the ripple is a rounded triangular waveform, but again this approximation will satisfy our needs here.

The tranny sags, the multiplier sags, but mostly the surge-limiting resistor looks like being the major source of sag in this supply, and this is totally under our control.

The supply can be made stiffer, have better regulation, by using a bigger tranny, and by using more capacitance such as paralleling caps to get say 660uF or more each. And of course tweeking the surge limiting resistor, maybe finding a suitable power thermistor to replace it.

Ideally this supply would be followed by a suitable iron-core smoothing choke, as used in Goldentone amps, but such an animal is a bit rare these days, so for most builders a wire-wound resistor will have to serve.


Even ignoring the price, recovered caps still look like the best choice on technical grounds.

More Power

Igor - throw the switch!” (heh heh heh)

As I've said, one of the problems is that the available trannies tend to be excessive in terms of power capacity for the needs of a small amp.

A supply to drive a pair of EL34/6CA7's, 6L6GC's, or even KT88's seems quite reasonable using this approach. But you will naturally have to beef up the cap values and perhaps voltage ratings.

Don't forget capacitor ripple current rating. The currents in the series caps are double the output side currents, so if you want to pull 250mA DC these caps will have to cope with half an amp. But when you can get them free you can simply parallel them up.

While I have included tripler voltages above (Vp3) there is a potential problem with this arrangement in that it produces un-equal half-cycle currents in the tranny winding, possibly leading to core saturation and overheating.

Actual Builds!

Update: 10/6/08

Okay, so much for the joys of theory. How does the actuality compare? The following supplies have now been prototyped and tested.

AVA101-1PSU Phase 1

Grant Wills claimed serial number 1 with his Lamington build, and demonstrated that the proposed voltage quadrupler is a practical and cheap HT supply.

A Phase 1 PSU, The Lamington power supply circuit.

This is the supply used for The Lamington and drives it to 15 watts output.

With 220uF's pumping 470uF's this follows the convention that if you have different value caps available you use the lower values in series and the higher values in shunt.

The HT decoupling network is also shown. The actual value of HT decoupling electros for the preamp and driver stages is unimportant as long as it is “enough”, and in cases where the off-load HT is 330V or less scrounged disposable camera flash caps will be ideal. All voltages are representative at idle.

Addendum: It is strongly suggested that builders fit a 150k bleeder resistor across each 470uF, and include a neon tell-tail within the supply (see AVA105, below).

Note: if you resort to connecting identical caps in series for increased voltage rating you must fit a bleeder across each cap to keep the voltages balanced. If you connect identical units in parallel for greater current you must fit a current-balancing resistor, around 47 ohms, in series with each. If you go series-parallel (say for your hex of KT88's) then you must fit both.

When I asked Grant about the supply voltage regulation he replied “about 5%, pretty much the same as the transformers themselves”. Very satisfactory. 310V out on idle.

AVA101-2PSU Phase 1 - “The Flea”

ex-flash caps under load
(l-r) flash-cap quadrupler, load lamps, AC from 2x 2A trannies

Update: 12/6/08

Test results for quadrupler built from scrounged disposable camera flash units, available Free! from your local film processor.

These flash units contain an electrolytic cap of between 80uF and 160uF, depending on the make, at 330 volts working, plus a high voltage diode.

Caution!Shock hazard! Caution! when disassembling the disposable cameras be sure to discharge the cap - some have been found with significant charge remaining days after the last use.

I wanted to see how this quadrupler would behave when fed with different AC voltages by tapping down on the trannies (both 2 amp units in this case).

Four diodes and two 80uF for C1 & 2, and two 120uF for C3 & 4 were recovered. A 220k/1W was recovered from a dead TV and used as the fixed bleeder across the output. The load consisted of three 40W globes in series with a 27 ohm resistor in the earthy end to measure the actual current (light globes are very non-linear as resistors, the current depending heavily on the temperature of the lamp filament).

Vac nominalVac actual VdcoVdcl@ Idc mAWatts
12+24 (or 17.5+17.5)40.122118280.0014.6
24+30 (or 27.5+27.5)60.5333295100.0029.5

Vac nominal - voltage marked on tapping
Vac actual - actual voltage applied to multiplier
Vdco - DC output voltage, no load
Vdcl - DC output voltage, loaded
Idc mA - loaded output current
Watts - computed power output

The caps themselves got warm to the touch after a couple of hours at the highest power, so I would suggest a nominal output limit of 100mA continuous with recovered flash caps (but certainly parallel them if needed, or use larger ones from ratted computer PSU's, which we will come to, below).

While I used a mixture of tappings for these tests I suggest you stick to using the same tap on both trannies to balance loading. The actual tappings available are nominally 12, 17.5, 20, 24, 27.5 and 30 volts; so there are intermediate combinations not shown here.

Here is the main reason for using multi-tapped trannies - they allow tweeking of the HT supply voltage to “heat up” or “cool down” the output stage as desired or needed. In this respect this supply arrangement is much more flexable than more traditional supplies and allows variation of the HT alone without changing the heater voltage.

The trannies I used for these tests were the 2 amp version, much higher than required here, because they were to hand in the Phase 5 PSU (see below) intended to eventually drive a pair of 6L6's.

Whatever sort of valve build you are considering, a sample studio tonebox, a Watkins or Vox AC30 clone, or something much bigger, this quadrupler supply arrangement has shown it will deliver the goods and is very flexable and practical at minimal cost.

Where high voltage is required at low current, such as only a few milliamps for a stand-alone preamp, trem, or reverb, there are smaller 24Vac and 30Vac at 150mA trannies avaliable. The internal resistance of this supply is around 400 ohms and we must expect that these lighter transformers will produce a higher internal resistance and more sag under load, but for light loading this should be insignificant.

Again the reminder that even “small” power supplies like these are easily capable of delivering a lethal current. Sure 400 ohms is perhaps a hundred times higher than the mains, but it only takes a mere 10mA to kill and these supplies won't even notice if you happen to get connected across the output. The Earth Leakage Breaker or “Safety Switch” on your switchboard only protects you against direct connection with the mains and gives no protection at all the other side of a transformer.

AVA105-1PSU Phase 5

A most serious supply - not for newbies.

Phase 5 PSU closed

The second build is my own Phase 5 PSU.

Inspired by Grant's progress, I built this power supply.

The circuit...
Phase 5 PSU circuit
(not shown is the Mains line filter already fitted to the ATX-PSU case used)

Internally ...
Phase 5 PSU internal
(l-r) ex-PSU cap quadrupler, heater tranny, line filter (above), 2x 2A 30V multitap HT

Test results

VdcmAdc Watts

AC input: two by 30V 2Amp in series

The upper voltage limit is set by C3&4 at 400V max. Caps from disposable cameras have a lower capacity but operate at 330 volts, lifting this limit to 660Vdc.

The upper current limit is set at one-quarter of the transformer current rating, or the maximum ripple current rating of the caps used, whichever is the lesser. Flash caps can be expected to have a low ESR and good ripple current rating.

To test this supply I used some lightglobes of different wattage as a dummy load connected in series across the output (don't use CFL's as a load; but they are another potential source of suitable diodes). In the earthy end I also placed a 27r/1W resistor (also from the dead supplies) so I could monitor both voltage and current-as-a-voltage.

375V no-load, 350V from idle current to 80 watt level plus. Regulation is around 6% at 35 watts, and 12% at 80 watts, but basically 350V from 10% to 90%. It delivers 250mA continuous at 349V and is obviously capable of even more.

I haven't tested it to extremes yet, but this Phase 5 supply should be capable of supplying at least 500mA at 250V at the maximum 120W capacity. I'm looking towards 6L6's, which are available new.

Having these taps allows us to tune the supply voltage for normal or “hot” running.


Update: 17/8/08

Loadlines for the two multiplier arrangements, ex-computer caps in red above, ex-photoflash caps in blue/green below. The AC source in all cases is a pair of 30 volt (multi-tap) 2 amp trannies.

y=Volts, x=mA
The coloured dots are test points
Performance above test points implied, but not tested

Plotted in WinPlot.

New: 17/12/10

This is a layout that shows how most of the supply could be built on a Printed Circuit BoardPCB, and you could indeed etch your own board drawn with waterproof marker, or adapt it to Vero/strip board, or just hard wire it on blank matrix board.

0.1-inch dot pitch

Andrew_k's Flea PSU build

New: 21/10/09

Andrew_k is a member of the AGGH forum and posted these photos of his Flea PSU build, with the addition of regulated 12.6V, 14H/60mA choke and multiple RC smoothing filters.

psu-boards-akhp.jpg psu-guts1-akhp.jpg psu-guts2-akhp.jpg psu-finished-akcp.jpg psu-rear-akhp.jpg

More Tests - Hextupler, Octupler

Thanks for the following to Ralph Parkhurst who is working up his Quad-6BQ5 build.

New: 17/11/08

I am in the process of building my son a valve guitar amp and would like to contribute towards the AVA100 collaborative project with the enclosed design (*), based on Grant Wills Lamington, with some changes as follows:

(* The circuit and PCB layout will be posted when finalised) “RedBoard 30”

  1. Power supply now uses a voltage octupler, thus requiring only one 30v 2A transformer. The 470uF caps (100VW) were obtained new from Futurlec for AU$0. 90ea plus $7.50 postage.
  2. Pull/Pull-Parallel configuration of 4 x 6BQ5s to provide 30w output
  3. DC heater supply to reduce any possibility of hum
  4. Uses a 40w Altronics 100v line transformer (M1130) with an 8-ohm load connected to the 4-ohm transformer tap (Primary 4K P-P)
  5. All components (incl pots, input socket and power supply) mounted on a single high quality PCB measuring 8.5 x 3.5 inches. This board has been designed with HV and heat in mind (thick 2mm FR2 substrate, heavy 2oz copper thickness with wide track spacing/thick tracks). The transformers are not mounted on the PCB, and the valves are mounted vertically on the ‘non-component’ side to reduce heat concerns. There is one single grounding point on the PCB to prevent hum induction via earth loop.

Here are the results of my tests with different configurations of multipliers. Some of these tests were done firstly using a 12.6v transformer. Test equipment used: Tek 2456DV scope with DMM option. All caps were 470uF.

TEST 1. Cockcroft-Walton “Parallel Halfwave Multiplier”, fed from a 12.6 v (nominal) 2.5A transformer.

CircuitACrms InputDCVoutI load mAWattsVpk-pk Ripple
Quadrupler (x4)13.5570.000nil
Quadrupler (x4)13.463.056(*1)3.51.2
Hextupler (x6)13.392.666(*1)6.12
Octupler (x8)13.514100nil
Octupler (x8)13.211975(*1)8.94
Octupler (x8)13.083180(*2)14.98

*1 Loaded with one 25w 240V lightbulb
*2 Loaded with one 75w 240V lightbulb

(NOTE: I could not test the Octupler at 56v as the some caps would require 350v rating which I did not have on hand)

TEST 2a. “Fullwave Quadrupler” as per Lamington design, fed from a 12.6 v (nominal) 2.5A transformer

VACrmsVDCoutI load mAWattsVp-p Ripple
not measured70.000nil
not measured64.056(*1)2.3not measured
not measured54.0150(*2)8.1not measured

TEST 2b. “Fullwave Quadrupler” as per Lamington design, fed from a 28-0-28 v (nominal) 2A transformer (DSE)

VACrmsVDCoutI load mAWattsVp-p Ripple

*3 Loaded with one 40w 240V lightbulb
*4 Loaded with one 25w and one 40w 240V lightbulb
*5 Loaded with one 40w and one 75w 240V lightbulb

The 28-0-28 transformers are on sale at DSE for only $9.98 [Nov '08] so I think this path provides far better regulation for a Quad x 6BQ5 design.

DSE 28-0-28 (blue) compared to Phase-5 (red)


Some applications require a high-ish voltage, 100 volts or more, but at only a few watts of power or milliamps of current. These include stand-alone preamps, trem's and reverb's, valve mike pre's, electrostatic headphones and speakers.

Mainly this is a matter of using lighter power trannies to save cost and weight (since any of the caps above should suit), just reset the bleeders to 1k per volt.

Suitable trannies include 24Vac and 30Vac at 150mA, giving up to 40mA out.


Okay, so I built eight of these disposable camera flash modules into a monster parallel ringflash for better-lit macro-photography.

And it naturally occured to me, as I'm certain if will occur to some of the more astute readers, that as they have individual batteries you could stack these up in series - say 10 at 120uF charged to 330V per stage. You know, 12uF charged to 3,300 volts!

And yes, you can simply step and repeat the cap-diode in these supplies, adding stages for more voltage (but less current) if you want 600V for your KT88's, even a few kV for an oscilloscope or electrometer.

Update: 17/11/08. Now I note from the excellent Hack-A-Day site that somebody has used ten flash modules in parallel to charge a large bank of high-voltage electrolytic caps.
But even if you know exactly what you are doing this is still frought with truly deadly dangers.

Apart from the obvious one of jolting yourself, or your cat, instantly into the hereafter, this rig is being used for high discharge experiments such as vapourising aluminium. This literally explosive release of energy produces intense heat, light - including rich ultraviolet, and generally a spray of white-hot metal globs.

Far beit from me to discourage experimentation - like Hunter S. “insanity works for me”. People do mad things and I've done my share, but the difference between a game and an experiment is that nobody gets hurt with the latter. Ear, eye, face, hand protection, insulating barriers, bleeders, neon tell-tails and so forth are not optional.

There are good ideas, then there are just ideas. Unless you are fooling with photo-multipliers, electron microscopes, or amateur radiography, this is not a “good” idea.

There are all sorts of things, sensible and dumb you can do, just remember that this isn't a Van de Graff generator, a Whimshurst machine, or a car ignition coil we are talking about, but a very serious mains-powered supply, potentially deadly at any voltage.

Valid HTML 4.01 Transitional <<<OzValveAmps