I received a Private Message from a diyAudio member, containing this request. Since I believe that ten thousand heads are better than one, I'm posting the request here in the Forum, to access the collective wisdom and insight of everyone here on diyAudio. Please assist this member if you're able. Thanks very much --MJ
I wish to make a cap multiplier for class A amplifiers.
I found various attempts in this forum and decided to focus on Smooth Like Butter cap mux, based on Complementary Feedback Pair configuration [aka Sziklai]. (LINK) In that thread I saw posts about the possibility of CFP oscillations, and I am concerned that it might be serious.
Can you please suggest a good yet oscillation free Cap mux? Or help me to refine the SLB with components for making it more stable? I noted the following:
1. Ballast resistor in the emitter of main pass transistor -- 0.33R is enough?
2. The expression "gm degeneration" was mentioned, but I didn't understand.
For a Class A amp with reasonable PSRR, there is no need for a low Zout cap multiplier.
Certainly not worth the oscillation risk of a Sziklai pair, which can happen with one amp and not another.
I personally would just use a simple one like that on the JLH Class A Amplifier Site.
There are better choices of active devices, like Darlingtons from Sanken or Onsemi.
Although I do not think they will make a huge difference either.
https://sound-au.com/tcaas/jlhcapmult.htm
Cheers,
Patrick
Certainly not worth the oscillation risk of a Sziklai pair, which can happen with one amp and not another.
I personally would just use a simple one like that on the JLH Class A Amplifier Site.
There are better choices of active devices, like Darlingtons from Sanken or Onsemi.
Although I do not think they will make a huge difference either.
https://sound-au.com/tcaas/jlhcapmult.htm
Cheers,
Patrick
The objective is to reduce he amount of capacitance in the PSU without causing hum in the amplifier.
Rod Elliott in his Project 15 recommends Sziklai pair for lower dissipation and improved performance.
This has been implemented successfully by the designers of SLB and it works well; with a minor possibility of oscillation.
Thanks and regards,
Sumesh
Then why ask for alternatives ?
The famous Salas Shunt love by many here, for example, oscillates wildly in the F5-HA, as found out by Stixx :
https://www.diyaudio.com/community/threads/f5-headamp.271926/post-4848225
No such problems with a normal Darlington cap multiplier.
Cheers,
Patrick
This is a very elaborate power supply. How much current does it need to provide?
30,000uF per rail is enough (USD $48 to $208-worth quan-4, 63V / 80V at DigiKey) that some reasonable people would not see the need or benefit to the capacitance multiplier, even if it did not pose an instability risk -- especially serving a Class A amp.
In addition to the considerable component cost, your time spent building and testing it, while probably enjoyable, still represents a cost.
If you persist in building this, please consider lower values for C15, C17, C16, and C18. Am also puzzled by the trimpots -- what do they do?
Cheers
30,000uF per rail is enough (USD $48 to $208-worth quan-4, 63V / 80V at DigiKey) that some reasonable people would not see the need or benefit to the capacitance multiplier, even if it did not pose an instability risk -- especially serving a Class A amp.
In addition to the considerable component cost, your time spent building and testing it, while probably enjoyable, still represents a cost.
If you persist in building this, please consider lower values for C15, C17, C16, and C18. Am also puzzled by the trimpots -- what do they do?
Cheers
It seems that you never built a high power cap multiplier using a darlington...it needs either huge heatsinks with a large input to output voltage difference or huge capacitors...
For ringing a ferrite bead on base pins or some 10 windings of ennameled copper in the air in the transistor's base can solve the problem.Actually even normal darlingtons need ferrite beads as they can oscillate from time to time too.A darlington behavior in a power class ab amplifier isn't very far from one in a cap multiplier.Harman kardon pm665 vxi is an example and actually most famous harman kardon amps would have those inductors.If in pm645..665 you have a sziklay pair driving the final transistors, in hk990 you have normal 3ef stages using the same inductors paralleled with 10 ohms resistors...pm665 also have a 15 ohms resistor inside the sziklay pair...Besides most amps have zobel networks that prevent ringing due to the load...a cap multiplier might very well need a zobel too.It all depends on the nature of the load.
A conventional Darlington cap multiplier will reduce LF ripple by 30-40 dB, and bit more in the mid band. It does little at frequencies above 100-200 kHz.
You lose 1-1.5 volts (load dependent). It works by effectively shifting ripple on the rails to the -3dB RC time constant of the cap multiplier. Usually you set this to << 1 Hz and it works on the premise that noise at this frequency is inaudible and the amp PSRR is very good.
it won’t make up for shoddy wiring or solve earth loop problems, so if you have a noisy amp, you should make sure it’s absolutely silent before moving on to add a cap multiplier.
On a class A amp with good PSRR with high constant rail draw, you will get some ripple ‘leakage’ easily observable with a sound card - a cap multiplier will suppress that significantly - so in a practical case it will go from -90dBr to -115 dBr (measurements quoted are for the kx2 class A amp).
The standard Darlington configuration is reliable and with a base stopper not readily prone to oscillation. Be aware, a cap multiplier raises the power supply output impedance above that of a straight repetitive+ cap bank.
Whether you want to go to that effort or you can hear it is a separate discussion. YMMV
😊
You lose 1-1.5 volts (load dependent). It works by effectively shifting ripple on the rails to the -3dB RC time constant of the cap multiplier. Usually you set this to << 1 Hz and it works on the premise that noise at this frequency is inaudible and the amp PSRR is very good.
it won’t make up for shoddy wiring or solve earth loop problems, so if you have a noisy amp, you should make sure it’s absolutely silent before moving on to add a cap multiplier.
On a class A amp with good PSRR with high constant rail draw, you will get some ripple ‘leakage’ easily observable with a sound card - a cap multiplier will suppress that significantly - so in a practical case it will go from -90dBr to -115 dBr (measurements quoted are for the kx2 class A amp).
The standard Darlington configuration is reliable and with a base stopper not readily prone to oscillation. Be aware, a cap multiplier raises the power supply output impedance above that of a straight repetitive+ cap bank.
Whether you want to go to that effort or you can hear it is a separate discussion. YMMV
😊
Last edited:
Never measured that as i don't have 200khz ripples anywhere in my systems to fight with, but I'd like to know where is the 200khz ripple found to be reduced: in front of the cap multiplier or as a load?
I sim'd it as a load and it seems that the darlingtons cap mult are quite effective at reducing 200khz load ripple .I placed a 1 amp 200khz current sink and the output voltage varied by about 16mv at 200khz with 10v dc input and 3 v ac 100hz ripple. Audionote dac 4 used a very high hfe multiple darlington after a lifted ground 7805 to supply all the digital work and i think they wouldn't try that if it wasn't effective at high frequencies.Yet they used base resistors to ground for each transistors to speed up the base discharge which in turn asked for more input voltage and more transistors to get effective...
I'd say that sziklay pairs need to be treated accordingly as in any power amp using them and they should be fine.
Besides i never recommend the use of audio circuits with very poor psrr right from the start .If your audio circuit has good performance it never needs fast or overly complicated regulators.Really good circuits don't need regulation at all other than for keeping a safe voltage supply for the circuit.
I strive to keep regulators for voltage regulation not to improve audio circuits and that allows me to improve my amplifier design to the point that i don't need expensive or finicky regulators. Most of my low voltage circuits are now supplied with a single 78xx ic , sometimes with zenner ground lifting and virtual grounds and i can't complain about anything.I think this is the way to go even for the most advanced circuits.If you don't like the brutality of 78xx ic's you can use Technics slp999 78xx with op-amp aid and i bet you'll be fine IF your audio circuit is properly designed.I think that using those very high quality ldo modern regulators for rf use in the analog audio domain is trying to hide poor analog design skills instead of enhacing some esoterical properties of audio products.
Last edited:
You can quite easily do a sim that will give you an indication of the ripple rejection afforded by a cap multiplier. Then you can build one and measure it. If you have a decent amp, you can drive it to near full power with a 50 kHz square wave which will have plenty of HF harmonics and monitor your amplifier supply rails. The 200 kHz reference point is simply to reflect on the fact that you cannot expect a cap multiplier to work as well at 200 kHz as it does at LF.
Maybe it would be a good idea to start another thread with a different title, perhaps like
How can I simulate instability/oscillation of a Sziklai capacitance multiplier in LTSPICE?
There are lots of people on diyAudio who are very experienced using "Tian Probe" simulations, and they may be able to assist you in creating a sim that provides insight.
How can I simulate instability/oscillation of a Sziklai capacitance multiplier in LTSPICE?
There are lots of people on diyAudio who are very experienced using "Tian Probe" simulations, and they may be able to assist you in creating a sim that provides insight.
Thank you @mark!
I think you're right, in the end a sziklay pair is a circuit with global feedback that can be checked the same as an audio amplifier for stability.Never used Tian probes either.Maybe now's the time for a first!I remember failing miserably with a very simple headphones amp having a tone eq placed in the feedback path that sim'd absolutely fine for me and ringed like hell once it was built...
I think you're right, in the end a sziklay pair is a circuit with global feedback that can be checked the same as an audio amplifier for stability.Never used Tian probes either.Maybe now's the time for a first!I remember failing miserably with a very simple headphones amp having a tone eq placed in the feedback path that sim'd absolutely fine for me and ringed like hell once it was built...
Last edited: