If someone read only one part of this whole thread, then I wish it was this:
We have a winner- you have to get up pretty early in the morning to fool Daniel.
I recently learned a bit of trivia, admittedly with little practical application, that made me question much of what I supposedly "know" about bypassing caps with other caps.
Digressing for a minute, I suspect old time engineers often better understood the underlying mechanisms or relationships of components better than we do. They had to slog through the equations by hand, on paper, maybe with the help of a slide rule. The interaction of the variables was much clearer than we get with simplfications, Spice sims and on-line tools. They also spent more time confirming their results with actual measurements. Today we have so much faith in what's on the monitor that we rarely confirm it, especially if it's something basic.
OK, what made me question the depth of my knowledge was the simple matter of paralleling two capacitors. We've all been taught that when you parallel two caps you simply add the values. Turns out that's a handy simplification that works in 99.9% of cases. What it doesn't take into account are losses.
Without writing a book on the matter, caps can be modeled with the losses in series or parallel, a perfect negative reactance and a resistance. We usually use the series model, but actually the parallel model is better suited to high losses and can be handier for calculations. Dissipation factor (loss tangent), ESR, EPR, power factor and several other parameters all define losses, and you can convert between any of them at a given frequency.
When losses are low, they can usually be ignored, but not when they're high. Where do you find high losses? Electrolytic caps at moderate and higher frequencies.
Doing the full parallel calculation I noticed that the combination of a big electrolytic and small film cap (at 1 kHz) resulted in less capacitance than the big one alone. This seemed an odd and probably erroneous result, but putting the caps on a good enough LCR meter to resolve the difference confirmed it. As the small film bypass was attached, the total capacitance went down, not up. Weird but true. (caveat- for the series model)
Sometime I'll put up a spreadsheet to show why, but my point is that the gut is sometimes wrong about these things, and particularly with bypass caps. What actually happens needs to be proven by experiment, and you have to know to what degree formulas are approximations.
It seems so logical to think a small high performance cap will shunt high frequencies across the big slow high-inductance electrolytic, but you have to do the numbers or make the measurements. Gut feelings and what seems logical on the surface fail badly on this one.
I recently learned a bit of trivia, admittedly with little practical application, that made me question much of what I supposedly "know" about bypassing caps with other caps.
Digressing for a minute, I suspect old time engineers often better understood the underlying mechanisms or relationships of components better than we do. They had to slog through the equations by hand, on paper, maybe with the help of a slide rule. The interaction of the variables was much clearer than we get with simplfications, Spice sims and on-line tools. They also spent more time confirming their results with actual measurements. Today we have so much faith in what's on the monitor that we rarely confirm it, especially if it's something basic.
OK, what made me question the depth of my knowledge was the simple matter of paralleling two capacitors. We've all been taught that when you parallel two caps you simply add the values. Turns out that's a handy simplification that works in 99.9% of cases. What it doesn't take into account are losses.
Without writing a book on the matter, caps can be modeled with the losses in series or parallel, a perfect negative reactance and a resistance. We usually use the series model, but actually the parallel model is better suited to high losses and can be handier for calculations. Dissipation factor (loss tangent), ESR, EPR, power factor and several other parameters all define losses, and you can convert between any of them at a given frequency.
When losses are low, they can usually be ignored, but not when they're high. Where do you find high losses? Electrolytic caps at moderate and higher frequencies.
Doing the full parallel calculation I noticed that the combination of a big electrolytic and small film cap (at 1 kHz) resulted in less capacitance than the big one alone. This seemed an odd and probably erroneous result, but putting the caps on a good enough LCR meter to resolve the difference confirmed it. As the small film bypass was attached, the total capacitance went down, not up. Weird but true. (caveat- for the series model)
Sometime I'll put up a spreadsheet to show why, but my point is that the gut is sometimes wrong about these things, and particularly with bypass caps. What actually happens needs to be proven by experiment, and you have to know to what degree formulas are approximations.
It seems so logical to think a small high performance cap will shunt high frequencies across the big slow high-inductance electrolytic, but you have to do the numbers or make the measurements. Gut feelings and what seems logical on the surface fail badly on this one.
Capacitors still have relatively low impedance above their self-resonance, but they don't behave like capacitors. So measuring capacitance at too high a frequency can have strange results. If they are resonating then they may also confuse an LCR meter depending on how it works. It would be interesting to look at the exact test setup. A plain LF sine impedance measurement would show increased capacitance.
Yes, if two caps show apparently less capacitance than the sum of them both this is a good sign that inductance in one of them is approaching a parallel resonance wth the capacitance of the other one. Or the cap meter is inadequate or faulty.
Of course, when switching between the parallel and series models for an impedance you have to think about which is appropriate to your situation. Otherwise confusion can occur.
Of course, when switching between the parallel and series models for an impedance you have to think about which is appropriate to your situation. Otherwise confusion can occur.
All this talk of parallel film and electrolytic caps may confuse OP old'n'cranky, who was, if I remember correctly, repairing an old receiver. Old e-caps need to be cut out and thrown away, not paralleled with new ones, whether electrolytic or film. The model of a 35 year old e-cap that I most respect is that of a bottle of salt water with two electrodes dipped in it. That is, a low ohm resistor, a short that comes and goes as the water flows back and forth establishing paths between the electrodes or not. Shorts do not reliable tuned circuits or amplifiers make.
True, especially about low impedance above self-resonance. I don't know the fundamental measurement that the meter I used makes, but it happily converts it to any form needed. If the cap goes inductive, it will just read the negative capacitance. It's all math! IMO, sometimes it's better to think in terms of reactance, positive and negative, and loss.
A LF measurement will of course give the expected results, because the losses are low there. At higher frequencies (heck, only 1 kHz) the decrease is real, both by measurement and by calculation. As I said though, it's more of a curiosity of the model and math, than anything useful, since plus or minus a dozen uF on a 16,000 uF cap doesn't amount to a hill of beans.
Yes, we wandered off to the subject of bypassing, but obviously bad parts need to go before thinking about anything else.
"NEW- with patented NaCl short circuit technology!"
It would be unusual to find an electrolytic forming part of a tuned circuit.indianajo said:
If repair is the aim then simply replace the relevant old electrolytic with a new electrolytic.
feedback shunt capacitors
Can the feedback shunt capacitors C304 and C354 be replaced with 50VDC 30VAC polyester caps (3x WIMA 10uf)? Is the ac rating too low? Thanks!
Can the feedback shunt capacitors C304 and C354 be replaced with 50VDC 30VAC polyester caps (3x WIMA 10uf)? Is the ac rating too low? Thanks!
An externally hosted image should be here but it was not working when we last tested it.
You can change from 100uF to 30uF if you want to modify the LF behaviour of the amplifier by moving the rolloff from 2.6Hz to 8.5hz. Why do you want to do this?
Why are there so many OEM schematics with the emitter and collector swapped on some transistors? 😕
The DC blocking capacitor in the lower leg of the NFB should see nearly zero DC voltage and nearly zero AC voltage if the amplifier is working correctly and you choose the right value.
Put in too low a value and it will see a bigger AC voltage.
If the amplifier output fails to one supply rail, then the DC voltage can rise to supply rail.
But, if you bypass the capacitor with a pair of inverse parallel diodes, that DC voltage is divided down by the NFB ratio 100k:620
If you consider that 400mVpk is too low for the bypass voltage, use 4 diodes in inverse parallel to give you 800mVpk.
Put in too low a value and it will see a bigger AC voltage.
If the amplifier output fails to one supply rail, then the DC voltage can rise to supply rail.
But, if you bypass the capacitor with a pair of inverse parallel diodes, that DC voltage is divided down by the NFB ratio 100k:620
If you consider that 400mVpk is too low for the bypass voltage, use 4 diodes in inverse parallel to give you 800mVpk.
Datasheet for the STK shows 22uf+1k in this position so 30uf should be ok as value, I wonder about the voltage ratings. I replaced the FB cap with 220uf Nichicon FGold cap and the resistor with 1k to lower the gain.
Sounds better but something's not right with the treble eg cymbals, soundstage and so on. Then I bypassed the Nichicon with 0.1uf polyester and things improved eg I have a nice sounstage but the amplifier is harsh and sibilant (got a peak in the 7k region).
So is 30vac enough rating for these caps?
Sounds better but something's not right with the treble eg cymbals, soundstage and so on. Then I bypassed the Nichicon with 0.1uf polyester and things improved eg I have a nice sounstage but the amplifier is harsh and sibilant (got a peak in the 7k region).
So is 30vac enough rating for these caps?
'Fixed' the problem by using some OSCON SP I had on hand. Sounds smooth now. OSCON seems to sound ok in analog for signal coupling but not in power supply.
On topic - replacing low value elytics in amp input is highly recommended. WIMA MKS2 in 5mm pitch up to 10uf is available at TME for example.
On topic - replacing low value elytics in amp input is highly recommended. WIMA MKS2 in 5mm pitch up to 10uf is available at TME for example.
Last edited:
- Status
- Not open for further replies.