When large capacitors are used, they have an inherent component of inductance that can cause high frequency phase issues. A simple parallel capacitor like a .1 to 1u poly film can help overcome it. Not always needed, but nice trick to add if necessary. I chased a phantom issue in a SS amp 30yrs ago, solved by improving the power supply common ground and adding .1u caps across the large electrolytics.
If necessary? What diagnostics would show if it's necessary?
What is a "phantom" issue"?
Now if you had said you supplied a 15Khz or higher sine to the amp and had the output scoped as you switched your cap or ground changes in and out so that a direct result could be seen then your post would have some solid merit.
Whenever I see posts with "mights" and "cans" and "sometimes" and "issues" and "help" and no real data I just....
...but it is not in this case. I understand well RF problems, and I am 100% certain that it is not a plain inductance that smears soundstage. The only explanation I can accept is, combination of inductance and non-linearities of oxide dielectric.
I am (trying) thinking about a way to measure such non linearities.
hmm what if you put a resistor in series with the elecrolitic, put a -50 db soft white noise on the amp , and then mix the soft signal with much larger impulses of different frequencies.
If you measured an FFT of the white noise over the resistor, its bandwidth should/could change with the various impulses.
More or less like a HPF wich should be a constant but changes with the larger pulses.
or am I talkin ** here? ;-)
Paul
hmm what if you put a resistor in series with the elecrolitic, put a -50 db soft white noise on the amp , and then mix the soft signal with much larger impulses of different frequencies.
If you measured an FFT of the white noise over the resistor, its bandwidth should/could change with the various impulses.
More or less like a HPF wich should be a constant but changes with the larger pulses.
or am I talkin ** here? ;-)
Paul
When I was working in the CLIC design department at National Semiconductor, one of the tasks was to characterize the effective capacitance vs voltage of the caps formed by sputtering aluminum over oxide isolated P+epi tub which was dielectrically isolated from the rest of the circuit. The depletion depth was dependent on the voltage so the capacitance also varied with the voltage. The higher the voltage, the deeper the depletion depth, thus less capacitance. The study was in support of stability in linear op-amps where the cap is the dominant pole in the second stage. Small signal unity gain square wave response with the number of overshoot oscillations can indicate the phase stability of the op-amp. We has a stability issue in one op-amp, thus the study...
The phantom issue I was chasing in the SS power amp related low level hum and fatiguing when listening to it, a harshness. On the bench it was related to overshoot and ringing that was not related to the circuit design, but was power supply related. It was only detected with square wave response, overshoot and 4-5 oscillations that slowly decayed.
I was using computer surplus 10,000u caps 2 per voltage bank, thinking more is better with a 12 gage wire connecting the pair. We had lots of surplus stores in the area, and the caps were cheap. I thought that it was first circuit related, over damping the amp, the listening fatigue was gone but so was the hi end,then went after the power supply after a friend made the recommendation. I thought he was a little looped (it was the early 70's in California), but he was correct. I bypassed the large caps at the caps with .1mylar film and also added .1 mylar film at each of the power feeds to the pcb for each channel. I also went to an improved power bus ground between all 4 10,000u caps with a star ground point to remove ground loops which eliminate a little low level hum. I had several sins, in the original design, in the end I used it for over 10yrs before it died and I decided to replace it. I've made the change from the dark side, to the enlightened side, tubes..
The phantom issue I was chasing in the SS power amp related low level hum and fatiguing when listening to it, a harshness. On the bench it was related to overshoot and ringing that was not related to the circuit design, but was power supply related. It was only detected with square wave response, overshoot and 4-5 oscillations that slowly decayed.
I was using computer surplus 10,000u caps 2 per voltage bank, thinking more is better with a 12 gage wire connecting the pair. We had lots of surplus stores in the area, and the caps were cheap. I thought that it was first circuit related, over damping the amp, the listening fatigue was gone but so was the hi end,then went after the power supply after a friend made the recommendation. I thought he was a little looped (it was the early 70's in California), but he was correct. I bypassed the large caps at the caps with .1mylar film and also added .1 mylar film at each of the power feeds to the pcb for each channel. I also went to an improved power bus ground between all 4 10,000u caps with a star ground point to remove ground loops which eliminate a little low level hum. I had several sins, in the original design, in the end I used it for over 10yrs before it died and I decided to replace it. I've made the change from the dark side, to the enlightened side, tubes..
...Interesting. How much did the capacitance vary from the highest WV charge to the lowest (or 0 WV)? Is there a typical % deviation for that type of cap? I would assume that means a 100uf 450wv cap would have a higher capacitance the lower the voltage it is given. Subject it to 250v and it becomes a 102uf cap...
...I'm not seeing how this reference is related to power supply filtering, if you were testing OP amps with a clean square wave and looking for the output distortions. You seem to be trying to link this with your "phantom issue."...
...So are you saying the "hum" and "fatigue" was the result of "overshoot and ringing"? Was this noise from the SS rectifier that you then went on to try to filter with the giant caps? All of this visible with your scope as you measured the ringing...
...So the huge caps eliminated the "fatigue" but not the hum, but it did kill the upper frequency response. Then your friend suggested bypassing the main PS caps to bring back the high frequency response. What was the original PS capacity before you added the giant caps?
...So in the end, the improved ground fixed the hum, not a total of 40,000uf PS caps?...
So what did the bypass caps remove from the B+ that killed your high frquency response?
Last edited:
Sorry that I have not read EVERY page of this thread...so maybe this is already been ruled out, but...
Pretty much gut feel, but, it seems to me that you might be shunting a difference in the output transformer primaries (between the channels).
I think I recall that this power supply cap directly feeds both output primaries right?
It could be that there is a small difference in parasitics in the primary windings, between the two channels. The difference is unaffected (or not totally normalized) by the large electro with it's ESR but is being shunted by the small cap with low ESR. Since it is a difference that is being normalized between the channels, it might be noticed by a change in imaging.
You might be able to measure this by comparing phase distortion on each channel, with and without the bypass cap. It might be pretty subtle though.
Something to think about anyway.
Pretty much gut feel, but, it seems to me that you might be shunting a difference in the output transformer primaries (between the channels).
I think I recall that this power supply cap directly feeds both output primaries right?
It could be that there is a small difference in parasitics in the primary windings, between the two channels. The difference is unaffected (or not totally normalized) by the large electro with it's ESR but is being shunted by the small cap with low ESR. Since it is a difference that is being normalized between the channels, it might be noticed by a change in imaging.
You might be able to measure this by comparing phase distortion on each channel, with and without the bypass cap. It might be pretty subtle though.
Something to think about anyway.
John Curl promised, the man he knows well is going to apply some method used in video measurements... I did not hear about that since then.
When I observed FFT plot I tried to compare sidebands, but unfortunately intermodulation distortions were too low in order to see difference of lengths of sidebands, or my Tascam US-122 interface is so dumb, or I am so dumb to find a better method (that is most probable)... The amp is up and running, and I bought a flight case already to ship it to Europe (I hope it will land there safe this time). I do that with all my prototypes: let they play in someone's living room, or on concert stages, than die being disassembled for parts for other prototypes.
Well at work we would use a network analyzer or a digital scope and swept frequency source and matlab.
The NA is not useful at audio frequencies though, pretty much looks like DC to it. We are usually looking at high frequency (GHz) issues.
But basically you are looking for phase shift through the amp, that is not linear with frequency.
Never bothered to try to do this on an audio amp, other than just eyeballing input and output waveforms as the signal generator went through it's set sweep. Never cared enough, beyond what my ears were telling me. Maybe if a miracle happens and spare time falls through the clouds...
I do have a phase angle volt meter at home that I could use, but it is not a modern digital unit so it is a lot less convenient.
The NA is not useful at audio frequencies though, pretty much looks like DC to it. We are usually looking at high frequency (GHz) issues.
But basically you are looking for phase shift through the amp, that is not linear with frequency.
Never bothered to try to do this on an audio amp, other than just eyeballing input and output waveforms as the signal generator went through it's set sweep. Never cared enough, beyond what my ears were telling me. Maybe if a miracle happens and spare time falls through the clouds...
I do have a phase angle volt meter at home that I could use, but it is not a modern digital unit so it is a lot less convenient.
So are we talking about phase intermodulation distortion (PIM) or someting else? If PIM, I believe it's caused by phase shift that is not linear with amplitude, and it shows up as sidebands as does "regular" IM distortion.
I'm refering to papers by Matti Otala, JC, Ed Cherry, and others.
A simple 2 tone test signal will produce the distortion. The trick seems to be in the method for isolating PIM from regular amplitude IM distortion.
I'm refering to papers by Matti Otala, JC, Ed Cherry, and others.
A simple 2 tone test signal will produce the distortion. The trick seems to be in the method for isolating PIM from regular amplitude IM distortion.
Well, yes and no.
I would certainly lump PIM into the general category of phase distortion.
But what I was thinking of is frequency based non-linearity and shows up with a single, swept tone.
Now, recall that I was speculating about what the root cause of the imaging difference Welbourn was hearing when adding the bypass caps. I was speculating that it might be differing parasitics in the output transformers primary and then speculating that it might be manifesting itself as a difference (between channels) in phase distortion. Is that speculation squared?
But it could be an amplitude-based affect...
I would certainly lump PIM into the general category of phase distortion.
But what I was thinking of is frequency based non-linearity and shows up with a single, swept tone.
Now, recall that I was speculating about what the root cause of the imaging difference Welbourn was hearing when adding the bypass caps. I was speculating that it might be differing parasitics in the output transformers primary and then speculating that it might be manifesting itself as a difference (between channels) in phase distortion. Is that speculation squared?
But it could be an amplitude-based affect...
Hello Michael,
ELI the ICE man says that voltage leads current in an inductive circuit and vice versa in a capacitive circuit.
Let a ME type take a swing at this.
Seems like decreasing power output at the frequency extremes due to power factor could decrease the illusion of realism. You think?
Could be, depending on the mechanism of the filter poles (reduced bandwidth) dampening may be greater on the positive swing than the negative that would create frequency dependent nonlinearity. Seems like that may be near the definition of distortion.
Just a couple off the top for fun!
DT
All just for fun!
Very close to how I understand PIM distortion!
If the power bandwidth is limited vs. small signal bandwidth, then indeed we have nonlinear phase wrt amplitude. Nonlinearity wrt amplitude is what causes intermodulation distortion. Phase intermodulation distortion is caused by phase nonlinearity wrt amplitude. (BTW, in my earlier post I shoud have said "phase that is not linear with amplitude", not "phase shift...")
To come to the point, I'm not convinced it's necessary to invoke PIM distortion to explain a smearing of the stereo image produced by an amplifier with a common B+ connection (particularly a switching amplifier). There may be other explanations.
Cheers,
Michael
Socratically, I would ask if the gain bandwidth product of the amplifier was improved by adding the power supply bypass cap.
Last edited:
How about this for an idea:
Measure the frequency and phase response of the amp as accurately as you can. From this estimate where the poles and zeros are (e.g. by curve fitting).
Repeat at different power levels.
Plot the movement of poles and zeros with power level.
See how this correlates with (a) listening and (b) bypassing big caps.
Try to determine which pole/zero comes from which part of the circuit so you know what to try tuning.
Measure the frequency and phase response of the amp as accurately as you can. From this estimate where the poles and zeros are (e.g. by curve fitting).
Repeat at different power levels.
Plot the movement of poles and zeros with power level.
See how this correlates with (a) listening and (b) bypassing big caps.
Try to determine which pole/zero comes from which part of the circuit so you know what to try tuning.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.