I'm interested in any comments you may have on the design of this power supply. They seem to have thrown everything but the kitchen sink at it, especially in the way of capacitors. Your thoughts on which capacitors are of value (aside from the obvious big filter caps), and which may be fluff?
Audio Amplifier Power Supply Design - Troy Huebner/John DeCelles
My interest is in upgrading an older amp with the hope of reducing noise some, and improving channel separation. Will replace the main caps, but after that wondering what else is of value...
Audio Amplifier Power Supply Design - Troy Huebner/John DeCelles
My interest is in upgrading an older amp with the hope of reducing noise some, and improving channel separation. Will replace the main caps, but after that wondering what else is of value...
A quick look suggests to me that they don't really understand power supplies, beyond elementary textbook level. On the other hand, they seem to have followed every possible audio fad. They don't talk about the need for mains-related capacitors for C1-C4, and I'm not sure all these are needed anyway. They don't appear to have given much thought to proper grounding. The PCB design appears to assume that a thick trace can be regarded as zero ohms - not true when charging pulses are around! Other capacitors seem to appear on the basis that caps must be added between every pair of circuit nodes (well, almost!).
I am surprised. NS application notes used to be much better than this. Maybe this is the new generation of engineers?
I am surprised. NS application notes used to be much better than this. Maybe this is the new generation of engineers?
hi Ron,
Found your post while searching the same thing,I have found a cap .1uf across the primary,a inrush current limiter inline 10ohm, fuse,secondary with a .1 cap,to a bridge with .1 film across each diode ,10000uf cap .1 cap 100uf to a reg in,out the reg to a .1 cap and found 3 different caps here 220uf,or100uf,or 47uf, also it is suggested to bypass at each opamp and major digital chips 10 to 47 uf at the opamp and .1 and 10 uf at the digitals,I also read a .33 across the + and _ on the opamps help the sonics,Now this is just what I have found PLEASE add to it or any Ideas are welcome,I haven't tried this ALL out yet. I have put the .33uf on some opamps and noticed better sound on the op amps(bass and mids improved), seems to be some talk about the kinds of caps too Solid polymers (Nicholson web site) show more ripple rejection,so should this be our .1 cap to use?
I have a bunch of elna's and cerafines and there ez to get so I use them for the middle values and rubycons for the big electrolytic s, I have 10 tantalum in bypass places, can the polymers replace those?I have had that type of cap short(tant) and burn in just a few years old, from what I have found on transformers the E type seems to be most use and recommended,the torrid are smaller and maybe more efficient but also more nosier than the E types I am not a GURU I just read DIY audio alot,Happy hunting!
Cheers!
Found your post while searching the same thing,I have found a cap .1uf across the primary,a inrush current limiter inline 10ohm, fuse,secondary with a .1 cap,to a bridge with .1 film across each diode ,10000uf cap .1 cap 100uf to a reg in,out the reg to a .1 cap and found 3 different caps here 220uf,or100uf,or 47uf, also it is suggested to bypass at each opamp and major digital chips 10 to 47 uf at the opamp and .1 and 10 uf at the digitals,I also read a .33 across the + and _ on the opamps help the sonics,Now this is just what I have found PLEASE add to it or any Ideas are welcome,I haven't tried this ALL out yet. I have put the .33uf on some opamps and noticed better sound on the op amps(bass and mids improved), seems to be some talk about the kinds of caps too Solid polymers (Nicholson web site) show more ripple rejection,so should this be our .1 cap to use?
I have a bunch of elna's and cerafines and there ez to get so I use them for the middle values and rubycons for the big electrolytic s, I have 10 tantalum in bypass places, can the polymers replace those?I have had that type of cap short(tant) and burn in just a few years old, from what I have found on transformers the E type seems to be most use and recommended,the torrid are smaller and maybe more efficient but also more nosier than the E types I am not a GURU I just read DIY audio alot,Happy hunting!
Cheers!
Last edited:
All caps used are legit for noise filtering, the purpose of some if not all is listed in the app note. You could say that with each added the return becomes less and less but so it goes with capacitors. They generally design for components easy to source (substitutions will be ok, no exotic or especially high performance parts per their kind/type are used).
If your particular transformer, bridge rectifier, incoming mains AC power, wiring, etc does not have or introduce as much noise as someone else's does, you may achieve similar or better results without all the capacitors. Which you omit depends on which portions of your power supply have less noise introduced.
It's an arbitrary amount of components, you could add more and different types of filtering or remove some of what is there, making it increase or decrease in complexity, size and cost. Then consider the amp, this PSU is proposed for powering some of their amp chips but your amp may not meet or might exceed noise rejection of those chips.
If your particular transformer, bridge rectifier, incoming mains AC power, wiring, etc does not have or introduce as much noise as someone else's does, you may achieve similar or better results without all the capacitors. Which you omit depends on which portions of your power supply have less noise introduced.
It's an arbitrary amount of components, you could add more and different types of filtering or remove some of what is there, making it increase or decrease in complexity, size and cost. Then consider the amp, this PSU is proposed for powering some of their amp chips but your amp may not meet or might exceed noise rejection of those chips.
Last edited:
If they were really worried about noise they would have used a proper mains filter (which would include the correct capacitor type) rather than a single cap across the mains supply. Some of their bypass caps might actually inject noise into the ground connection. They admit that their component values were not carefully chosen but they simply copied everyone else. No, IMHO this PSU was intended to push audiophile buttons rather than be an example of good design.
^ It's a better design than most people on DIYaudio are using (or perhaps I should write, "more complete" since we can view the PSU as modular so some will think you only have two modules being PSU or Amp while others will just consider these the only two PCBs present along with other external circuitry) with the exception of higher impedance from use of only a single 20,000uF bulk cap per rail, so I don't see your point.
It isn't valid in the discussion that caring about noise means using a line rated capacitor - less safe perhaps, but no less effective filtering noise. I do not see this design as dependent on specific component part #s, line rated caps can be chosen as well as other parts' substituted as designed. While a full blown mains filter would do better, I also consider the possibility that since these can be purchased as modules and often include a handy IEC socket with chassis mounting ears for a chassis power cord connection, that it isn't necessarily considered part of the power supply itself.
Pushing audiophile buttons would almost certainly include use of audio grade capacitors, certain models of soft recovery diodes.
We might say it's a "generic" PSU example, in that it leverages component types common world wide so in countries where access to parts isn't so good it might provide good performance : price ratio.
Of course if you feel another design is better for that parameter, please do share a schematic...
It isn't valid in the discussion that caring about noise means using a line rated capacitor - less safe perhaps, but no less effective filtering noise. I do not see this design as dependent on specific component part #s, line rated caps can be chosen as well as other parts' substituted as designed. While a full blown mains filter would do better, I also consider the possibility that since these can be purchased as modules and often include a handy IEC socket with chassis mounting ears for a chassis power cord connection, that it isn't necessarily considered part of the power supply itself.
Pushing audiophile buttons would almost certainly include use of audio grade capacitors, certain models of soft recovery diodes.
We might say it's a "generic" PSU example, in that it leverages component types common world wide so in countries where access to parts isn't so good it might provide good performance : price ratio.
Of course if you feel another design is better for that parameter, please do share a schematic...
Last edited:
They shipped out Bob Pease. Its not only the app notes which have taken a dive, the datasheets equally have been going downhill. Be interesting to see if acquisition by TI continues the slide.
<edit>
I have a schematic on my blog
Last edited:
I'm thinking what I will do is add the small ceramics Cs9 and Cs10 to filter out the higher frequency noise from the secondary AC circuit. The amp currently has a small ceramic across the AC feed, which I will remove as it seems two caps to ground would be more effective. Also will add the snubber caps around the bridge Cs11-14. Probably will use metalized polycarbonate instead of polyester, as there is little difference in cost. The main caps will be upgraded from two 10,000 uF units to two 30,000 uF.
I remain skeptical about the value of small film capacitors in parallel with the big electrolytics. However, to cover all bases I think I will add a 10 uF metalized polycarbonate to both the positive and negative rails of each board.
This approach would seem to address 4 areas:
1. Noise in the AC
2. Noise generated by the diodes
3. Marginal capacity of the main filter caps
4. The mysterious one of film caps in parallel with huge electrolytics
I remain skeptical about the value of small film capacitors in parallel with the big electrolytics. However, to cover all bases I think I will add a 10 uF metalized polycarbonate to both the positive and negative rails of each board.
This approach would seem to address 4 areas:
1. Noise in the AC
2. Noise generated by the diodes
3. Marginal capacity of the main filter caps
4. The mysterious one of film caps in parallel with huge electrolytics
^ There is no mystery about film caps in parallel, the app note clearly mentions they are for high frequency filtering. Capacitor ESR and ESL rises with frequency. Chipamps have good PSRR but it drops as frequency rises. However, it is somewhat of a coincidence that they are in parallel, you'd put them where noise occurs, is picked up, etc, it is irrelevant in that context where the bulk electrolytics are located. In other words the less useful thing is to truly place them in parallel by soldering them across the bulk electrolytic's leads rather than where noise is introduced.
Last edited:
The issue I have with small film caps in parallel is that at high frequency (say 20 kHz), they still have significantly higher impedance than a large electrolytic. A couple I have been looking at have the following characteristics:
CDE 30,000 uF
Pure capacitance impedance at 20,000 Hz - 0.000265 ohms
ESR at 20,000 - 0. 0156 ohms
Total impedance - 0.015865 ohms
EPCOS 10 uF metalized polypropylene
Capacitance impedance at 20 k - 0.795775
ESR at 20 k - 0.0029
Total impedance - 0.798675
So yes, putting that film in parallel will reduce impedance of the electrolytic alone to 0.015556 ohms or a change of 2%. Not much, and that is with a 10 uF film cap which is very large compared to the more commonly used 0.1 uF. That is why I wonder about the real effect of adding them??
On location of the cap, I would agree it should be close to the source. Do you see the rectifier bridge as the only source, or could the power output stage of the amp be introducing noise or ripple on the PCB board power rails?
Last edited:
^ Why would we say only 20KHz, wouldn't you want to filter as high as possible? Is this coming from the idea that only noise within an audible frequency range would matter?
The source is anywhere you pick up noise. That's near active components like the amp chip, diodes, board traces without a proper ground plane, non-shielded or twisted wiring. Some might say "I can't hear the difference", but by the same token many would say they can't hear a difference between 10,000 uF and 30,000uF bulk capacitance or two similarly enough spec'd diodes in a PSU rectifier.
I would not suggest use of 10uF film caps, too large and costly. You can get 0.1uF to 1.0uF dirt cheap in bulk or at surplus electronics sites, no need for some particular brand and type with premium price at an electronic component supplier unless you're doing something else with them like signal coupling... or just leave them out of the design, a lot of things you could put in an amp are optional and subject to debate whether it ultimately matters.
The source is anywhere you pick up noise. That's near active components like the amp chip, diodes, board traces without a proper ground plane, non-shielded or twisted wiring. Some might say "I can't hear the difference", but by the same token many would say they can't hear a difference between 10,000 uF and 30,000uF bulk capacitance or two similarly enough spec'd diodes in a PSU rectifier.
I would not suggest use of 10uF film caps, too large and costly. You can get 0.1uF to 1.0uF dirt cheap in bulk or at surplus electronics sites, no need for some particular brand and type with premium price at an electronic component supplier unless you're doing something else with them like signal coupling... or just leave them out of the design, a lot of things you could put in an amp are optional and subject to debate whether it ultimately matters.
Last edited:
You can't just add ESR and capacitive reactance, as they are in quadrature. However, your general point is good - that bypasses don't do very much in or near the audio band. If you want to add a bypass the best place is on the amp board, as there it will keep RF under control. If there isn't room there, then don't use a bypass at all.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.