I can see why more capacitance is needed for better bass under 120Hz per say, but I am not sure if the brand of the PSU caps will affect the sound all that much. Although I am a firm believer of better parts = better product/performance.
Excellent explanation. I never considered Low frequency music signals can drain the PSU caps! All I ever considered was lowering the ripple on the PSU.
Thank you. Yes, and it is even more important that just that. The PSU and decoupling caps' current IS THE SIGNAL that goes to the speakers.
That is why I wrote so many posts about trying to figure out how to get the correct minimum capacitances, AND about optimizing the actual signal path's layout and performance (at every frequency), which is the path of the current that goes directly from the PSU and decoupling caps to the speakers, through the power output devices. Click on the link in my sig:
Last edited:
And this is almost exactly what was seen. I had rigged up a very crude passive LP filter and was powering the amp off a couple of SMPS. Ripple wasn't an issue. The results were eye opening didn't expect to see all the high frequency content.
I remember your link from when you were helping with my first attempt at amp design. That's what prompted my experiment a little while later. 🙂
The difference in sound between Panasonic FC and Nichicon KG is so significant despite having better closer caps on amp, I will no longer design for or use FC's. Unless I want some mid-lofi sound. This is irrespective to size and amount of caps; and I recommend more smaller caps often.
And I'm saying that the composite signal, even with the usual LF content is usually faster than the charging cycle so it switches +/- enough during a charging cycle to discharge both rail caps roughly the same amount. A music signal is not a 40hz sine with a little wiggle on top from the rest of the music.
First, the problem, if it exists would be clipping. Bigger res caps do not change the sound of your bass! And a drop of rail voltage by a huge 20% will drop your clipping level by a paltry 1 or 2 db.
If there is a strong 40 Hz sine in the music, then yes, it will look like a 40 Hz sine with the rest of the wiggles summed with it, riding on top.
That won't usually be seen, though. Most real instruments have harmonic content. But if there is LF content, the effect is still much the same.
Also, you can decompose almost any signal into individual sine frequency components (Fourier analysis), and analyze them separately, and sum the results back together, and it is valid, for a linear time-invariant system.
So any low-enough frequency content can tax the caps the way it has been described, even when higher-frequency stuff is also present.
And remember, we are not saying "always". Some of us just want to be able to cover the worst-case possibilities.
If I did the math right, a "1 or 2 dB" voltage change means 12% to 26%. That is not "paltry", compared to 20%.
But where did you get those figures? It would depend on the reservoir capacitance value (among other things), which is what we are discussing, wouldn't it?
Anyway, it's been a couple of years since I looked at this stuff in detail, but I think I tend to agree that the main potential problem would just be earlier clipping, which just amounts to lowering the RATED max output power spec.
Last edited:
"But does this rule apply for Class A"
The cut-and-paste (and the link) I put up were from the Nelson Pass A40 class A design (read it).
Amps like the Crown DC300A, the GAS Ampzilla, the SAE 2400, etc, use the 90% values from the W C R equation (and were all designed before Nelson published the formula).
Amps like the Altec 9440A sound like a brick outhouse in the bass, despite having only a pair of 'puny' 10,000µF caps (with a pair of 10µF bypass caps per channel).
The cut-and-paste (and the link) I put up were from the Nelson Pass A40 class A design (read it).
Amps like the Crown DC300A, the GAS Ampzilla, the SAE 2400, etc, use the 90% values from the W C R equation (and were all designed before Nelson published the formula).
Amps like the Altec 9440A sound like a brick outhouse in the bass, despite having only a pair of 'puny' 10,000µF caps (with a pair of 10µF bypass caps per channel).
Howdy, Tom ... 😉
Capacitor quality is important - a simple program like PSU Designer can be an eye-opener in this regard: put a very low resistance at the output, and then vary only the internal resistance of the caps - and watch the level of ripple dramatically vary!! This sag could also be constantly altering the DC operating point in parts of the circuit, enough to cause other quality issues ... it is not a simple situation ...
Capacitor quality is important - a simple program like PSU Designer can be an eye-opener in this regard: put a very low resistance at the output, and then vary only the internal resistance of the caps - and watch the level of ripple dramatically vary!! This sag could also be constantly altering the DC operating point in parts of the circuit, enough to cause other quality issues ... it is not a simple situation ...
I want to point out that there's good enough and comprehensive here too. You'll have better performance without saturation potential of the transformer if it's not your only milk source. Also to get desired performance 600-750va with some capacitance would potentially take a much larger transformer when running almost no capacitance.
You might measure a difference in bass SPL, but only hear a stressed amp in the higher frequencies being less desirable. Perceived bass has a lot to do with the attack, and SPL can vary some. This I'd think would be very true given how important a low stress large transformer is considered to be in class A among Pass enthusiasts.
If your rails are very high obviously there's the potential to run less capacitance, given the overall charge is high. Unqualified to say, but I presume with class A this is tolerated better due to inherent topology qualities.
You might measure a difference in bass SPL, but only hear a stressed amp in the higher frequencies being less desirable. Perceived bass has a lot to do with the attack, and SPL can vary some. This I'd think would be very true given how important a low stress large transformer is considered to be in class A among Pass enthusiasts.
If your rails are very high obviously there's the potential to run less capacitance, given the overall charge is high. Unqualified to say, but I presume with class A this is tolerated better due to inherent topology qualities.
You missed the point. Look at a music waveform for 1/120 second. Does it stay all positive ( or negative)? I don't think so. If it dosnt then the problem you describe does not exist.
Are my eyes deceiving me or are there large steel fixing bolts running through the core of those inductors?
Back in the old days I used to just copy what a saw other amps using. If the stereo amp was 100 watts rms per channel, I'd go with 2 caps (bipolar supply) that were about 10000uF each and call it good. When I got a job in video engineering at Tektronix they taught me the formula (during the job interview):
I = C (dv/dt)
I had seen that formula back in school but never got active with it.
You can rearrange this formula to be
C = I / (dv/dt)
I is current (at full power), dv is change in voltage. dt is change in time (1/120 of a second).
There are a few issues going on here: the above formula is the professional way to calculate the cap value, but you have to choose what you want dv to be. "dv" is how much voltage droop you are willing to have during the 1/120 of a second.
A high feedback power amp circuit will have a significant amount of "active" power supply rejection, so you could have 10% ripple on the rails when the amp is at clipping on the bench, and the amp circuit may well virtually eliminate all of the ripple from the signal path. It depends on the amp circuit design. There's also the issue of phase shift at the lowest frequencies, and the possibility of "motorboating" oscillations at a sub-audio frequency. Again it depends on the amp circuit.
Caps that aren't bypassed with 0.1uF caps right at the amp circuit could contribute to high frequency instability too. Large electrolytics and the wires going to them can be significantly inductive at frequencies above the audio range. The phase margin is more predictable if the amp circuit sees zero ohms from DC to infra-red when it looks at the power supply. How things are grounded can make a big difference with that. So it's hard to give a clean simple answer to your question. Take a shot and then try to verify the phase margin at both frequency extremes, both before and slightly into clipping on the bench.
I like to use a 10kHZ squarewave to look at the high freqs for stability. There shouldn't be very much ringing at all. Going in and out of clipping slowly with a sinewave should be completely free of any spurious oscillations.
I = C (dv/dt)
I had seen that formula back in school but never got active with it.
You can rearrange this formula to be
C = I / (dv/dt)
I is current (at full power), dv is change in voltage. dt is change in time (1/120 of a second).
There are a few issues going on here: the above formula is the professional way to calculate the cap value, but you have to choose what you want dv to be. "dv" is how much voltage droop you are willing to have during the 1/120 of a second.
A high feedback power amp circuit will have a significant amount of "active" power supply rejection, so you could have 10% ripple on the rails when the amp is at clipping on the bench, and the amp circuit may well virtually eliminate all of the ripple from the signal path. It depends on the amp circuit design. There's also the issue of phase shift at the lowest frequencies, and the possibility of "motorboating" oscillations at a sub-audio frequency. Again it depends on the amp circuit.
Caps that aren't bypassed with 0.1uF caps right at the amp circuit could contribute to high frequency instability too. Large electrolytics and the wires going to them can be significantly inductive at frequencies above the audio range. The phase margin is more predictable if the amp circuit sees zero ohms from DC to infra-red when it looks at the power supply. How things are grounded can make a big difference with that. So it's hard to give a clean simple answer to your question. Take a shot and then try to verify the phase margin at both frequency extremes, both before and slightly into clipping on the bench.
I like to use a 10kHZ squarewave to look at the high freqs for stability. There shouldn't be very much ringing at all. Going in and out of clipping slowly with a sinewave should be completely free of any spurious oscillations.
Incorrect. Sorry. I have been through all of this, in depth. I don't feel like explaining it all, yet again. Read the related posts. Or just try it, which is the antithesis of "engineering". Try no caps at all, and see when clipping starts. You will see that some capacitance is necessary, in order to raise the rated max output power RATING. So then you will have to find out what the MINIMUM capacitance is, for any given max output power RATING. And there is a MINIMUM, for the worst-case signal. I have already derived (and verified) the equation. You should take the time to read the links I already posted.
Did you even read what I said ? Which part of it is incorrect ? I don't care what you have derived, my statement is correct. Not my fault if you don't understand.
You also missed the point in post #46. My numbers don't depend on anything. A drop in rail voltage decreases the clipping level. Why, dosnt matter. And what I was try to point out with the numbers I did in my head ( and from your math it looks like I was close) is that a rather large rail voltAge drop of 20% will only decrease your max power output by around 2db.
- Status
- Not open for further replies.