Regulated power supplies

I'm considering a regulated power supply for whatever my next amplifier ultimately turns out to be. My last amplifier (early JLH 10 watt SE) had a small buzz in the speakers which i reckoned could be removed with a regulated power supply - this became an "accredited" solution to the later revisions to the amplifier.

I have heard that regulated power supplies tend to reduce bass extension and generally strangle the soundstage a bit, but after reading Nelson Pass power supply article i thought about how this might be eradicated with some over-speccing of the supply.

I'm contemplating using multiple LT1084 regulators, each preceded by a 58,000uf capacitor (because i have lots available) and another 58,000uf capacitor connected to the output closeby the output transistor array.

I'm hoping that the capacitors downstream of the regulators will be safely charged via a current limited supply (max 5A) without the regulators shutting down (anyone have any experience of this?), not decided on how to limit the current surge to the pre-regulation capacitors, possibly a relay/timer/surge resistor combo.

Is this worth doing or pointless given that class A SE amplifiers are fed from a CCS anyway?


2001-07-30 1:35 am
Why dont oyu try a capacitance multiplier supply outlined at the ESP site ( I have one feeding a small class A, and I have no hum at all, it might also be due to the balanced interconnects that I have... but their is NO hum at all, i even tested it with a very sensitive oscilliscope. Also, do you have a toroidal trans? I hear they have lower noise/hum and stuff. I wouldn't know, i've never used a convemtional in an audio project. One last thing, how many of these 58KuF caps do you have? Have you considered a "battery" PS? I've never used one, but its an option if you have lots of caps... one downside, you have to pause music every 10-15 mins to "recharge" the caps... bummer
The cap multplier / zenor refference supply Yoda suggests is a very good idea. The only thing I would add is, design the supply for 50% to 100% more peak current than the amp requires. That way the power supply will always be able to deliver more that enough current to the amp. I would use a small value (220uf to 1000uf )at the power supply connection to the board but not at the output devices, and 4700uf at the output of the supply. Very large caps on the output of any regulator can cause problems for the regulator. Remember to use interconnect wire from P.S. to Amp. outputs that is as good as or better than the speaker cables you use. The signal current delivered to the speakers come from the power supply after all.

I'm in favor of using fairly large amounts of capacitance after the regulator circuit. That way the sound of the caps swamps the sound of the regulator circuit, laying to rest complaints that regulators "don't sound good." It also serves to buffer the current demands of the circuit, allowing more leeway in the design of the regulator.

I am also a big fan of large capacitors after the regulator stage. I also noticed a trend away from this as it may reduce the Speed of a high performance Wide-bandwidth regulator Circuit.It seems alot of folks are in favor of using just small film type capacitors and High Speed Shunt type regulators.I would be most interested in other Thoughts on this.
Prefer real caps to cap multiplier

I want to have a bottomless pit of energy available to the output devices - if they each have 58,000uf to draw from then i figure that's about as good as i can make it.

The cap multiplier is not going to be capable of delivering instantaneous current in this fashion, but then is that going to be a requirement given that the amplifier will draw a constant current?

Has anyone who has connected large post regulation caps had problems with the regulators shutting down, etc?
If the amp. has constant current draw then a large well of current is not required, just clean DC. Still, always over design the power supply. If you need 5 amps DC build it to deliver 7, 8 or even 10 amps. This means the transformer, rectifiers and a very large cap on the input to the regulator (your deep well of curent). Make it very fast, use hight speed series pass transistors. Use a stable reference voltage (a correctly done zenor diode reference is fine). Use a high current gain configuration like a darlington stack so that the base current in the zenor circuit stays within the zenor regulation range.

This is a brute force type of regulator NOT some feedback type which I agree would sound bad.



2000-12-31 9:31 am
you can always try to increase the time delay of the regulator to try and charge the capacitors more slowly. You can do that by building a simple regulator using extra power amp transisors, a zener and a few other components instead of LT1084 Usually power transistror are more sturdy than regulators. That's what I am doing for my 30W HIRAGA amp (which I am testing today!! eh eh eh! ;-)). You can build an adjustable version to compensate for the slightly different characteristics of the zeners.
The capacitance multiplier has to be one of the biggest scum in the history of electronics. If you can see past the misleading name what the circuit is doing is reducing the ripple by dropping the voltage. Since there is NO way to get something for nothing, the energy storage (ie physical "capacitance") is NOT really there.
Cap Multiplier

I knew that a cap multiplier circuit only gave "virtual" capacitance and had no stored energy and therefore already excluded it in favour of a regulator/big cap design.

I am also a little more enlightened (thanks Geoff) regarding the current requirements of the output devices for an SE amplifier running from dual voltage rails which has strengthened my desire to use a really big regulated supply.

I was considering soft starting the post regulator caps by ramping up the adjust pin on the regulator over a few seconds, the pre-regulator caps could be slow fed via a thermistor or two in the supply line to the toroid.

Still don't know if the regs will holds out with that amount of capacitance on the end of them though...
Don't all linear regulators reduce hum by dropping the output voltage below the ripple, or does magic come into play here?

Isn't the goal of any good power supply;

1. Very low output impedance at the frequency specturm present in the load.
2. More than enough current reserve to supply the load's needs.
3. A stable voltage output.

I have used Zenor referenced, cap filtered, non-feedback, series pass type power supplies with 100 to 500 watt RF power amps for years. They work better than any other type I've tried. Unless you spend a lot of money on a HP or Lamda.

Yes, you do need a big cap but it goes in front of the regulator to provide the extra current required when the transformer reaches it's limit.
It seems that most people hear are in favor of Big caps Prior to the regulator and Small of No caps after the Regulator. I gess if one is using RF transistors then No Cap upon the output of the Regulator.

Regarding cap Multipliers These work verry well for me on Low Current Voltage gain stages I do not use a Zener as the Voltage ref. since doing so Changes the Circuit from a Cap. Multi. To a Series Pass Non Feedback type of Regulator and as sutch i dont think will then qualify as a cap. multi.Morover Regarding the Stored energy in a cap Multi. I removed the Current source suppling Base Deive to the Pass Transistor and the Circuit Kept supply ing Voltage for a long Time Considering the small capacitence used on the Base. A true cap Multi. just amplifies the capacitence of the Capacitor upon the base of the series pass transistor by the Hfe of the Transistor, The Effect is to Create a Very low Impedance whitch it dose very well. Again This is with only the capacitor upon the base and the Current source Suppling Base drive. No resistors or zeners.I have not rried this type of circuit on a High Current Output stage, But for low current Voltage gain stages thay work Great.
An interesting discussion is on. In my experience, a zener referenced darlington series pass regulator works very well. The darlington is made up of TIP31C/32C and MJE3055/ MJE2955 for non-critical applications or faster transistors if required. When used to power a 60watt class A/B amplifier I noticed that the amplifier continued to sing for a relatively long time when power was switched off as compared to a real capacitor bank of 20,000uf per line; the regulator only had a 2,200uf cap at the input, a 100uf cap at the base of the TIP transistor and a 470uf cap at the output. My rough estimation was that the virtual capacitance should be over 30,000uf and it sure seemed to be.

The sound of the amplifier also becomes more detailed and when a larger transformer and higher capacitance is used at the input of the requlator, bass does not seem to suffer audibly. The slam and the dynamics from such a power supply are assuredly better than from a capacitor bank, this could be due to the fact that the sag in voltage is less than from capacitors alone, provided there is atleast 5volt (more prefered, but watch the dissipation) drop across the regulator/capacitance multiplier.

Sometimes, inferior transistors or zeners make such a supply noisy and changing these components for better brands have solved the problem.

Many years ago, when I was tinkering around with a Philips TDA1520A IC audio amplifier, I used such a power source as described above. The amplifier worked well. Then by a brilliant idea, (atleast I thought it was), I changed the base capacitor from 100uf to 2200uf. When I switched on the amp, there was a flame from the power supply board, I lost all the components on the power supply board (the base caps had split up physically) and the ICs too. I put back new components only to witness the disaster again. I had an identical board supplying power to the preamp, but all was well there. Then I realised that the virtual capacitor was indeed working as a true capacitor and when I changed the value back to 100uf, all was well. The power on surge was too high I suppose. A brake resistor/relay or thermistor arrangement might have helped. But I did not try further.

Try it out for yourself. Any comments?

[Edited by Samuel Jayaraj on 08-18-2001 at 07:10 AM]
Samuel: Great report BTW. Normaly i would have thought that the Zener would have Bleed off the Stored Charge upon the Capacitor, But i gess this did not happen in your Case. Wow using a darlington as the pass Device The huge hfe of darlingtons seem to me to be Great in this type of Circuit.Depending upon the Voltages involved an alternative to the Soft start relay is to use a Constant Current source to provide Base drive and to Bias the Zener. A Nice Soft Start of the cap. multi Circuit can be had just by the Time constant of a limited amount of Current charging the Capacitor on the Base. I used this method once and worked Very well. Did you try replacing the cap upon the Output of the cap. Multi with a small value film cap? If so what did this Do to the Slam and did Microdynamics get better. I suspect thay would, But you never Know.Thanks for your input Samuel, I found it most interesting.

Thanks for the comments. Appreciate your alternate ideas of soft-starting the capacitor-multiplier. I did not try any of the methods in the particular case I had described.

I have not tried shunting the output capacitor with a film type assuming that the capacitor on the amplifier board, usually a film type is in anycase playing the same function although it is physically connected through a longer wire than if on the PSU board.

Whenever, I build amplifiers requiring two supply voltages, a higher voltage for the input/driver stage and a lower one for the output stage, I use this type of capacitance multiplier supply for the input/driver stage (and I get to regulate the supply as well) and a conventional supply for the output stage. In my opinion, this works very well for overall sound quality.
My First DIY use of a Cap Multi was for the Driver Only stage Of my Leach Dirivitive 250 Watt Amp. I designed my Own + & - 85 Volts Via a discreet Component Regulator of Similar topology to A series pass Transistor andOpamp but the Opamp is replaced with a single End Transistor gain stage. The output and driver stages operate off = & - 80 Wolts Lower like you Do. I decided to attempt to isolate the Driver stage From the output transistor Rails well i Installed a Cap multi with a 3,800uF ultra low ESR Electrolytic by passed with a 0.5uF Polypropylene To do This The sound Quality with it removed and connected was Drmaticly Noticeable Merry mutch so at the verge of Clipping 350 Plus actual wattes into 8 Ohms.
Beware of stability when using regulators

When using regulators, you may run into stability problems if you mess around with too much capacitance etc. at the wrong point because they have feedback loops which you don't have control over.

My preference for power amps these days is for un-regulated, overspecced, inductor fed passive units. If you are very serious, consider multiple stages of same. Try Duncan's PSUD2 for designing such units. You can find it at and I highly recommend it. In it's current inception it really delivers! I know because I was beta testing since version 12 (which was also pretty good)

Another Application i am working on is for A Opamp/ Buffer Combo Battery Powered Headphone Amp using Virtual Ground Driver the same output current as both output stages I first tried 15,000uF x 4 then replaced them with 4.7uFSolens. Big Difference in speed and Articulation the Bass had proper Sustain where as prior it was delayed. I gess a virtual ground driver can ce considered like a regulator.
Has anyone checked out Jeff McCaulay's capacitance multiplier which appeared in the June 2001 issue of Electronics World (Circuit Ideas pages). Seems very good.

I made this circuit and briefly tested it this morning. Since my application was to only power a discrete buffer stage which works of + - 32volts, I had not used much capacitance, post rectifier/output of capacitance multiplier. I found the voltage setting to be quite stable - off load; when I connected a 10k resistor to the output to put a static load on the supply, I found the voltage drop from 32 to 19 volts. This may have to do with the actual capacitors not being sufficient, particularly, post rectifier. I need to verify this.

The author says that this circuit was used to bias a Class-A amplifier at 2 amps and using it reduced ripple from 2 volts pk-pk to a few microvolts.

My idea is to use this capacitance multiplier followed by a 317/337 regulator to power some preamplifiers, like the Pass DIY Opamps, which I have at hand.

Anyone else wants to join the fray?
I don't believe this anecdotal evidence. The operation of a circuit is pure physics. If the mains power is switched off, then where does the current to run your amp come from? There must be an actual reservoir of charge somewhere to provide that current. But, transistors cannot store charge, nor can transformers or zeners, or magical pixie dust. That charge must come from an actual, real capacitor, which will store Q=CV Coulombs of charge.

The cap. multiplier circuit unquestionably provides only virtual capacitance.