Capacitor multiplier supply for ALEPH 3

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
hello, Im currently building an aleph 3 (the aleph 5 run to hot om my heatsinks) and im curios about building i supply with a capacitor multiplyer. but in not certain of how the circuit operates i only know that it uses the "miller" effect. It should be quite ideal when the aleph is a class A amp and draws a constant current.

I have found an article :
witch is quite informing. but i need a supply that can handle +-25V and about 2A. and im not sure that TIP3055 and TIP2955 can handle that?.... and i wonder can you "overkill" the circuit with a to high multiple ratio ? and will i need a soft start for this kind of a circuit as i would for a real 1F-2F cap?. it belive TIP3055/BD139 in an darlington config will result in an hfe of about 3500 that times 470µ is a quite nice capasitance....1,5F

and can i change the "real" output and input cap to 22 000µF variants (got a good stock of them) ?...the original uses a 4700µ and an 1000µ cap for that.


The Aleph does not draw a constant current from the power supply. The current in the supply rails, under load, will vary from near zero to twice the quiescent current (see my postings in other threads, Zen v Trimodal and A75 Mono).

The capacitor multiplier circuit will work well in your application, though use a complementary rather than darlington arrangement (ie a pnp pass device in the +ve supply and an npn in the -ve supply). This will reduce the voltage drop across the multiplier.

There should be no problems using your 22000uF capacitors on both the input and output of the circuit (though see the following re pass transistor). The initial diode peak current will be little worse than that due to the 22000uF input capacitor.

Regarding the pass transistors, the TIP2955/3055 should be adequate if you are using separate supplies for each channel since the maximum current draw will be 4A (assuming an Iq of 2A) and the voltage drop across the transistor will be about 3V. However, I would strongly recommend something a little more robust, preferably in a case with a better thermal characteristic. This will definitely be neccessary if you are using only a single supply for two amps.


If you are using the 'basic' capacitance multiplier in the ESP Audio Pages Project, increasing the base capacitance form 470uF to 1000uF will reduce the rms ripple voltage from about 2.5mV to about 1.25mV at a 4A current draw and with your 22000uF capacitors on the input and output. However, if you are using the 'improved' version with the additional filtering, my simulations show that increasing the base capacitance will actually cause the rms ripple to increase from about 350uV to 400uV.

The 'improved' version of the multiplier circuit needs more careful 'tuning' so as to get the full benefits of the additional filtering. Increasing the capacitance sizes without altering the associated resistances does not necessarily give any improvement in smoothing.

Hope this helps.


Your second query was posted whilst I was replying to your first one, hence the separate replies.

You are going to need both pnp and npn pass transistors whichever arrangement you use. The only difference is the supply rail in which they are used. Anyway, I'm not sure there is a problem as the data sheets I have for the TIP2955/TIP3055 show very similar hFEs for these two devices. If you are concerned about the total gain, use drivers and/or pass transistors with a higher hFE, eg MJE15029/MJE15031 drivers and MJL3281A/MJL1302A power transistors (these are only examples, there are many other altenatives).

You could also use MOSFETs as the pass transistors. These could be the same type as the ones in your Aleph output stage (particularly if you have some left over after matching). Some people prefer to keep to the same device in both the output stage and any regulator or similar feeding the amp.

Well i wouldnt know how to redesign the circuit for Mosfet:s but yes i got quite a good stock of them left.
Is it hard to remake the multiplier with mosfets? would it impeed operation?...

yes i know i will need both NPN and PNP devices but i have made the layout for the circuit using the darlington config but i havnt etched it so it is easily reconfigurable.


I have just one question here.
Why Nelson himself doesn't use regulated PSU in the Aleph series? It's cheeper, with lower noice... All looks fine at first glance. I have bitter experience with regulated supplies for OPamps. I tried to make some upgrades on my CD player. The result was that I put my previous configuration back. I think that sometimes "just caps" is the better solution.

[Edited by Asen on 04-17-2001 at 04:02 PM]
Geoff: I am afraid I can not agree with opinion that Aleph 3 behaves as ordinary AB amp (current varies from almost zero to ..). As I am going to build it so I made some simulations. Supply current oscillates around the basic current value of the bias current (ie cca 2A). Without signal requires cca 2A. When driven consuption rises (peak cca 4A). Of course the wave falls close to zero but for very short time. Average value is far from ordinary AB (B)amp.
Capacitance multiplier really looks like good solution. You can not achieve low ripple at reasonable price. 100mF at 25V gives cca 2V ripple, using multiplier I can get ( 2x10m, 2x2m2 in RC) less than 5mV ripple. How fast this supply is I don´t know. Feeding 2 channel from one supply can cause problem. Supply should handle more than 4A continous a 8A peak current - more than single circuit with one darlington is able to do. So I prefere separate supply for each channel.
Does anybody know where I can get more models for CircuitMaker 2000 ?

We are not in disagreement. Yes, the average current on the supply rails is different for different types of output stage, but what matters when designing a power supply, capacitance multiplier, regulator etc is predominantly the peak current. Many people seem to think that a Class-A amp has a constant current draw from the supply rail(s) equal to the quiescent current which is not the case.


The low ripple voltage figures I was qoting (under 1mV rms) were obtained by simulation of the capacitor multiplier circuit - see Project No 15 at the ESP Audio Pages (

The transformer secondary voltage was set at 23Vrms to give a dc output from the capacitor multiplier of 24V (a source resistance of 0.27ohms was assumed). Standard high current diodes were used and the load resistance was set at 6ohm to give a 4A current draw. The input and output capacitors were 22000uF (since this was the value 'hifi' wished to use). The remainder of the circuit was exactly as per Project 15 (using spice models for the BD139 and TIP3055).

Hope this answers your query, if not email me any further questions.

Capacitance multiplier – regulated PSU


I don’t agree with you completely. The word is how to multiply the capacitance using a transistor but I do not see basic differences between an ordinary transistor regulator and a cap multiplier. If you take a regulator consisting of a transistor and a zener diod, and put a big cap in parallel with the zener you will get a cap multiplier with output voltage, determined by the zener voltage. You have a voltage drop but you’ll have such drop in any cap multiplier. Otherwise you will not reduce the ripple.
So the transistor characteristics are critical here, especially when we are talking about dynamic current sources (Aleph case). These are just theories but if you take a look at the Return of Zen article, Nelson Pass reports better results with LC PSU. Notice that the Zen amp is with constant current source.
Timo Crist also thinks that supplies with inductors sound better than regulated PCUs or (if you like) capacitance multipliers. (I’ve exchanged several mails with him)

I'm sorry, but I'm not sure what it is that you don't agree with. However, as I've started a reply, I may as well add a few comments.

Yes, I agree that a power supply with a choke input filter is better than one with a capacitor input filter. I don't know whether the difference is audible as I haven't had the opportunity to judge this, but the choke input filter reduces the electrical stresses on the transformer and the diodes by removing the high current pulses. From an engineering standpoint this must be a good thing. However, choke input filters are, I believe, somewhat difficult to get working correctly. They also require the use of a large, purpose designed inductor that many DIYers will find hard to obtain.

One of the main advantages of the capacitance multipler over a voltage regulator is that it 'floats' with variations in the supply voltage. With a voltage regulator you have to design for worst case conditions so the output voltage must be set relative to the lowest expected input voltage (allowing for the dropout voltage of the regulator) but the regulator must be capable of dissipating a much higher voltage when the supply is at its maximum (because the output voltage is fixed). The volt drop across a regulator will vary from about 3V to in excess of 10V and heatsinking and the pass transistor will have to be sized on the highest figure. The capacitance multiplier doesn't suffer from this problem because it maintains a relatively constant volt drop of about 3V so smaller transistors and heatsinks can be used.

If you add the zener diode to the capacitance multiplier as you suggest, it will lose the ability to 'float' and will need be designed for worst case conditions as with any other regulator.

You mention transistor characteristics. The capacitance multiplier can be constructed using the same BJT output devices as the amp that it is feeding. There is even a way that a dual supply rail can be constructed so that only one polarity of transistor is required. In this way, if you were using say two 2N3055 devices in a JLH Class-A amp, you could use a further two 2N3055s in the capacitance multiplier and so the supply would have the same characteristics as the amplifier itself.

I view the capacitance multiplier as a salvation for those of us who cannot afford the large capacitors that would be needed to achieve the same ripple factor. Of course, the larger capacitors are probably a better solution because they remove the active devices, but they also cause problems with the power supply design due to the high inrush current unless suitable precautions are taken.

I hope we have not been talking at cross purposes.

Couldn't resist tossing in another thought experiment...
Suppose one were to set up a relatively small bulk power supply as a capacitance multiplier, then follow it with a standard regulator (followed by a cap bank--I, for one, don't like the sound of regulators dropped straight onto gain devices). The capacitance multiplier takes a lot of the yukkies off the top and the regulator ties the voltage to a set value. Yes, it's more complicated than a straight cap bank, but semi's are much cheaper than large caps. Might be worth playing with.


There is some kind of misunderstanding, due maybe to my poor English. I'll try to be clearer this time.
The reason I've shared some thoughts in this discussion is my conviction that transistor type of PSUs are not suitable for amps like the Aleph (never mind if they are capacitance multipliers or regulators). Transistors have poorer dynamic behaviour than caps. I agree that one can use the same devices like in the amp circuit itself, but (I'll cite Nelson Pass again) the active devices are the main source of distortion (99%). Putting them in the PSU you put one more source of distortion in the whole circuit.
I myself would use such PSU for the Andrea's Follower99 and intended to use something similar for the JLH 10W. Since I've received the simulations you were kind to make for me, warning me that the circuit will draw up to 3 A when the idling current is 2 A, I've given up the idea of using a transistor PSU. I still believe though that a capacitance multiplier will be a good solution for the Andrea's Follower due to the constant current source.
As for the inrush current - an elegant solution is the use of thermistors, or resistors with relay delay circuit (as mentioned in the ESP site).

I admit that money are of great importance here and we all are pushed to do compromises sometimes.

[Edited by Asen on 04-22-2001 at 06:27 AM]
Multipliers, Regulators, Passive

Hi All,

as Asen has already mentioned, i don't like the sound of series regulators. I admit that i haven't played with cap multipliers. I intended to at some point but have lost interest.
The series regulators that i tried were high performance ICs by Linear Technology (LT), no discrete designs.

At the moment my amps use passive supplies, Aleph (3) (clone, some mods) was upgraded to 0.28 F cap bank with choke input filter and three-phase power. IMHO this is close to the best possible passive PSU.
AKSA (very good AB amp from ) uses 10000uF per rail and channel, no inductors, but three phase.
Three phase power increases the effectivity of supply filter caps about ten fold because the DC content in the rectified AC is very high.

In the regulator department i will try shunt regulators to power a 5W single-ended (similar to Andrea Ciuffoli's PF) constant-current amp. This will waste alot of power... because the shunt regulator has to have the same quiescent current as the load. Therefore, efficiency for the amp circuit is cut in half again.
The shunt regulator i intend to use is similar to the SuperReg as designed by Allen Wright ( ). It is fed by a constant current source (supplies 2x load current).
To the best of my knowledge this is the only full class A regulator. It can source and sink current. The CCS at its input isolates the shunt and load from the transformer and filter caps, diode spikes etc due to its high impedance.
Allen Wright sells a book, the Preamp Cook Book, which describes the SuperReg and other important bits of power supply design in detail. The book is very nice and more than worth the price.

Geoff, the difference with a choke input filter is easily audible. Even relatively small chokes, for example 2 mH air core, make a worthwhile difference.
However, i think it is not adviseable to use choke input for non-class a circuits with unsteady current draw because of the varying voltage drop. Although, i would have no problems with a small coil to kill some of the diode noise in any amp. Maybe you can try with some old crossover coils or something.

Have a nice day,
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.