I have been studying various amplifier schematics over the past few weeks, and I have been reading a lot of threads, but I can't find a precise answer about what effect a cathode bypass resistor has on the sound quality of a hi-fi audio power amplifier. One old thread turned into a multi-page argument over whether to use one or not. It seems that opinions go both ways on that, and that's not my question. What I want to know, and can't really find answers for are:
1) What effect does using a cathode bypass capacitor have on the sound quality of an amplifier that has a circuit with one, and why?
2) If one is used in a circuit, how should the "ideal" size be calculated? I found this formula in an old closed thread: Choose a value where 1/(2πFC) = .5*R, where R=the cathode resistor value in ohms and F=low frequency cutoff (-3dB). That's fine if the formula is correct, but I have no idea why it should be that size or what effect more or less capacitance would have on sound quality. A quote from an old closed thread, "I remember a now fairly famous tube designer who ... always spec'd a 1000uf cathode bypass cap, people would ask why such a large cap? He would never answer." Add me to the list of those asking the question. Suppose hypothetically that the value is too low? Is bigger better? Is there a point where increasing size does not matter?
3) If it is correctly sized, is changing the type of the capacitor likely to make any difference in sound quality and if so, why? Hypothetically, what if it's the cheapest and worst quality electrolytic swapped out for a high quality electrolytic or some other more expensive type?
Edit: I am referring to push-pull class A amplifiers only, and only those with a single resitor and a single capacitor as in this test circuit from the 1955 article cited in posts below:
1) What effect does using a cathode bypass capacitor have on the sound quality of an amplifier that has a circuit with one, and why?
2) If one is used in a circuit, how should the "ideal" size be calculated? I found this formula in an old closed thread: Choose a value where 1/(2πFC) = .5*R, where R=the cathode resistor value in ohms and F=low frequency cutoff (-3dB). That's fine if the formula is correct, but I have no idea why it should be that size or what effect more or less capacitance would have on sound quality. A quote from an old closed thread, "I remember a now fairly famous tube designer who ... always spec'd a 1000uf cathode bypass cap, people would ask why such a large cap? He would never answer." Add me to the list of those asking the question. Suppose hypothetically that the value is too low? Is bigger better? Is there a point where increasing size does not matter?
3) If it is correctly sized, is changing the type of the capacitor likely to make any difference in sound quality and if so, why? Hypothetically, what if it's the cheapest and worst quality electrolytic swapped out for a high quality electrolytic or some other more expensive type?
Edit: I am referring to push-pull class A amplifiers only, and only those with a single resitor and a single capacitor as in this test circuit from the 1955 article cited in posts below:
Last edited:
"I remember a now fairly famous tube designer who ... always spec'd a 1000uf cathode bypass cap, people would ask why such a large cap? He would never answer."
I suppose that designer could hear still some sound difference between 500uF and 1000uF, and could not between 1000uF and 2000uF. That was a reason of his decision in favour of 1000uF.
And, in many tube amps design aspects, decisions are based on arguments like that. Why that kind of resistor or another? Less temperature coefficient ? Most frequently not, just preferable sound.
I suppose that designer could hear still some sound difference between 500uF and 1000uF, and could not between 1000uF and 2000uF. That was a reason of his decision in favour of 1000uF.
And, in many tube amps design aspects, decisions are based on arguments like that. Why that kind of resistor or another? Less temperature coefficient ? Most frequently not, just preferable sound.
Why would there be an "ideal" size, short of infinity? Why can't we have infinity?
Are you asking single-ended or push-pull? If P-P, class pure-A or tipping into AB? These are very different cases.
Are you asking for clean sound way less than clipping? Or a "nice" transition into and past clipping?
Can't you buy a 100-pack of capacitors and just try it?
Are you asking single-ended or push-pull? If P-P, class pure-A or tipping into AB? These are very different cases.
Are you asking for clean sound way less than clipping? Or a "nice" transition into and past clipping?
Can't you buy a 100-pack of capacitors and just try it?
See pages 5 and 6 of this thread for David Berning's take. I think he's the "famous designer" who proposed a 1000uF bypass cap. ;-)
http://www.dcaudiodiy.com/phpBB3/viewtopic.php?f=4&t=595&hilit=berning+bypass+capacitor&start=40
He made the same recommendation to me in regard to my Williamson amplifiers, to the effect of the larger ther bypass cap the better. Even in a Class A amp a bypass cap can reduce intermodulation distortion. See page 21:
https://worldradiohistory.com/Archive-All-Audio/Archive-Audio/50s/Audio-1955-Nov.pdf
http://www.dcaudiodiy.com/phpBB3/viewtopic.php?f=4&t=595&hilit=berning+bypass+capacitor&start=40
He made the same recommendation to me in regard to my Williamson amplifiers, to the effect of the larger ther bypass cap the better. Even in a Class A amp a bypass cap can reduce intermodulation distortion. See page 21:
https://worldradiohistory.com/Archive-All-Audio/Archive-Audio/50s/Audio-1955-Nov.pdf
The equation is wrong. You get first-order roll-off below a corner frequency fc = (gm + 1/R)/(2πC), where gm is the transconductance of the valve - sum of the transconductances to the anode and to the screen grid if it is a pentode. So it's not just the cathode resistor that matters, but also (and usually mainly) the transconductance. This follows from network analysis, where I neglected the valve's internal resistance for simplicity. Turning the equation inside out, C = (gm + 1/R)/(2πfc)
You usually want fc to be far (two octaves or more) below the audio band, because you already get some roll-off and phase shift at frequencies above fc.
When C approaches infinity, the roll-off and phase shift at audio frequencies approach zero, but the time the voltage needs to settle at power-up approaches infinity. Depending on the rest of the circuit, you may also get a nasty current peak through the valve. I wouldn't make the time constant RC more than a second or so, short compared to the time that the valve needs to heat up.
I'm assuming a stage working in class A with small signals. For an output stage working in class-AB or class-B, you get rectification effects that change the bias point during a loud passage. The smaller the capacitor, the quicker it recovers to its normal bias point after the loud passage.
Last edited:
That must be him!
That article is fantastic. It is quite a detailed analysis, requiring an understanding of inter-modulation distortion, which I always have found to be both critical and often overlooked - for example any specification sheet for an inexpensive solid state amplifier focusing only on vanishingly low THD. There is a whole lot more to the story, like IMD and harmonics, which the article covers in great detail.
Turning the equation inside out, C = (gm + 1/R)/(2πfc). Could you quickly give one example calculation, perhaps for something common like an EL34 or KT88? Just to make sure I have a full understanding of the units and so I can save it for future reference when I can't remember all of this.
Thanks to both of you. Those two links, and the proper formula, comprehensively and definitively answer the question as far as I am concerned. I can't believe two people came up with a totally comprehensive answer so quickly. Thank you!
Now when I look at a schematic, which I am doing more and more these days, I know exactly what that little capacitor is doing. I won't say which way I would lean if designing or modifying a circuit, but I already have an informed opinion, which is a good starting point. 😉 Stumbling along with a box full of components with different values, testing one thing and another endlessly without first understanding what you are doing and why, is counterproductive for both the brain and the wallet!
Attachments
Last edited:
A comparatively minor effect is the distortion added by a capacitor's linearity with applied voltage. Although a properly biased electrolytic capacitor has reasonably small distortion with signal voltages much less than bias, less signal voltage is still more better. Folk love to rhapsodize about coupling capacitors (BTW, a cathode bypass is included as a series element) but if no voltage appears across an element (say, a capacitor) it can't cause any distortion. From solely this perspective, the bigger the less badder.
In long loop feedback amplifiers, the time constant at the output valve's cathode(s) is just another nuisance to be brute-force removed as far away as practical from a dominant low frequency pole.
All good fortune,
Chris
In long loop feedback amplifiers, the time constant at the output valve's cathode(s) is just another nuisance to be brute-force removed as far away as practical from a dominant low frequency pole.
All good fortune,
Chris
As my depth of understanding has increased, I now understand why it is that way rather than taking someone's word for it.
Yes, but I'm not to the point of addressing that issue in my mind yet. I'm still working on relatively "simple" topics like tube bias, push-pull phase splitting (split load and long tail), grid blocking resistors, tube power supply basics (B+ and grids), reading tube specification sheets, and now the bypass capacitor and resistor combination. The answer on that one surprised me. It's a whole lot to digest all at once, but I am managing.
Don't forget tha fact that a cathode bypass capacitor is actually in series to the audio signal. The PSU last bypass capacitor is also in series. Their quality has direct influence on the audio signal at all frequencies.
You can use any consistent set of units. I would use the SI base units so I don't have to think about consistency, but you can decide otherwise if you want to.
To give an absurd example: if for whatever reason you would prefer to express frequency in cycles per fortnight, you can do so, but then the unit of capacitance becomes ampere fortnight per volt if the units of resistance and transconductance are ohm and siemens (siemens is also known as mho or A/V).
Less absurd example: if you would like to express frequency in hertz and transconductance in millisiemens, then the unit of resistance becomes kilo-ohm and the unit of capacitance millifarad. The units of resistance and transconductance are then each other's reciprocal, and the unit of frequency is one cycle per the product of the units of resistance and capacitance.
Example:
ECC83 with a transconductance of 1.6 mS and a 2.2 kΩ cathode resistor, lower limit of the audio range 16 Hz, cut-off frequency two octaves lower:
SI base units:
fc = 16 Hz/2/2 = 4 Hz
C = (0.0016 S + 1/(2200 Ω))/(2π 4 Hz) ≈ 0.00008174776623 F = 81.74776623 μF
Cycles per fortnight:
fc = 16 Hz/2/2 = 4 Hz = 4 838 400 cycles per fortnight
C = (0.0016 S + 1/(2200 ohm))/(2π 4838400 cycles per fortnight) ≈ 0.000 000 000 067 582 478 69 ampere fortnight per volt ≈ 0.000 081 747 766 23 ampere second per volt = 81.74776623 μF
kilo-ohm, millisiemens, millifarad:
fc = 16 Hz/2/2 = 4 Hz
C = (1.6 mS + 1/(2.2 kΩ))/(2π 4 Hz) ≈ 0.08174776623 mF = 81.74776623 μF
A cathode bypass cap doesn't just effect the "sound" though, in some designs it will affect stability. I've tried 1000u cathode bypass caps in some of my amps just because someone gave me a bag of them cheap, I ended up pulling them out and using something more sensible like 47u or 100u.
There's a difference also between calculations, tests and sitting down and listening to music through an amp, your ears being the final arbiter. On several of my amps that I've built I've ended up changing cathode caps and coupling cap values after listening to the amp over a period of several weeks because bass response was wrong or whatever.
Lastly, you see a pattern emerging after looking at amp schematics, EG 47k anode resistors on 6SN7 type valves, and 10n/1M for interstage coupling between V1 and V2, so often I don't bother drawing numerous load lines and doing several calculations and just chuck in a 47k, 47u 10n or whatever.
There's a difference also between calculations, tests and sitting down and listening to music through an amp, your ears being the final arbiter. On several of my amps that I've built I've ended up changing cathode caps and coupling cap values after listening to the amp over a period of several weeks because bass response was wrong or whatever.
Lastly, you see a pattern emerging after looking at amp schematics, EG 47k anode resistors on 6SN7 type valves, and 10n/1M for interstage coupling between V1 and V2, so often I don't bother drawing numerous load lines and doing several calculations and just chuck in a 47k, 47u 10n or whatever.
The colorization of the loudspeaker will block anything we would hear from a cathode cap,
provided the cap is not defective, Spend your time & $$$ on the speaker, the ROI is far greater,
provided the cap is not defective, Spend your time & $$$ on the speaker, the ROI is far greater,
Bypass caps matter a lot. Ideally you avoid them. You can do this with DHTs by using filament bias. Not the output stages like 2a3 or 300b, but the input stages like 26, 4P1L, 2P29L, 10Y etc etc. Since it's good to use DHTs where you can, this is a double benefit.
Where you have to use a bypass cap, try something like a 40uF DC Link cap like Kemet make. I've always found this gave me better sound, e.g. with 2a3 outputs and the like.
Where you have to use a bypass cap, try something like a 40uF DC Link cap like Kemet make. I've always found this gave me better sound, e.g. with 2a3 outputs and the like.
For the curious & others here is a link to an interesting paper published in Audio Magazine, Nov 1955.
They cover a cap cathode bypass PP application but some of the measurements & conclusions are interesting NTL.
https://worldradiohistory.com/Archive-All-Audio/Archive-Audio/50s/Audio-1955-Nov.pdf
BTW, I did get olde, so at 90 I've seen a lot of both good & crap.
28 cm of snow here yesterday AM, so busy again on the tractor.
It seems there can be a lot more than audio in one's life. 😱
They cover a cap cathode bypass PP application but some of the measurements & conclusions are interesting NTL.
https://worldradiohistory.com/Archive-All-Audio/Archive-Audio/50s/Audio-1955-Nov.pdf
BTW, I did get olde, so at 90 I've seen a lot of both good & crap.
28 cm of snow here yesterday AM, so busy again on the tractor.
It seems there can be a lot more than audio in one's life. 😱
Perfect. Thanks! I find that if I come back to formulas a year (or more) later without a textbook example, I sometimes have a hard time remembering how they work.
Yes, we were referring to that excellent paper, which was linked in post #5 above. I recommend it to everyone.
I have the same speakers that I designed and built in 1994. I went very, very, very deep into loudspeaker design back then. Papers published by Paul Klipsch helped a lot, regardless of whether people like his horn speakers or not. Also back issues from AES, which I used to read at the public library. They had them all bound in the reference section for some reason, which was fortunate for me, along with other professional audio publications covering theory. I have repowered the speakers twice over the years and will be doing so again soon. I have absolutely no desire to change them. I have always wanted to delve into different amplifier designs, but until recently, I didn't have the time or the cumulative knowledge required to do so.
Last edited:
The by pass capacitor has increased gain compared to the gain lost across just a resistor.
I wouldnt expect to hear any quality difference unless extra gain took signal into distortion.
I wouldnt expect to hear any quality difference unless extra gain took signal into distortion.
I've been sloppy with terminology, I meant the SI base and coherent derived units.
Without a bypass capacitor there is degenerative feedback as the cathode voltage tries to follow the grid voltage. This decreases distortion. It also increases the drive requirement on the driver tube.
With a bypass capacitor there is no degenerative feedback. The cathode is essentially locked at the bias voltage and the grid is more sensitive to the input drive so there is less requirement on the driver tube.
I don't have a PP amp on the bench, however for the SE amp I currently have on the bench at one watt out shows:
Configuration V-G rms THD at 8 ohm resistor output
No Bypass 48.8 0.730%
With Bypass 19.2 1.849%
In order to achieve 8W out I will need 8Vrms into the speaker, so I will need 8/2.828 = 2.828 times more drive at the grid of the output tube.
With no bypass capacitor, I will need 48.8 *2.828 = 138Vrms! That is a 390V swing from peak to peak!
With a bypass capacitor, I will need 19.2 * 2.828 = 54.3Vrms, much easier to achieve. 154V p-p, much easier to achieve, AND, the driver will contribute much less distortion.
My current driver can deliver 0.13%THD at 50Vrms. It can not provide the 138Vrms drive under any circumstances.
There is not free lunch.
It is all tradeoffs.
With a bypass capacitor there is no degenerative feedback. The cathode is essentially locked at the bias voltage and the grid is more sensitive to the input drive so there is less requirement on the driver tube.
I don't have a PP amp on the bench, however for the SE amp I currently have on the bench at one watt out shows:
Configuration V-G rms THD at 8 ohm resistor output
No Bypass 48.8 0.730%
With Bypass 19.2 1.849%
In order to achieve 8W out I will need 8Vrms into the speaker, so I will need 8/2.828 = 2.828 times more drive at the grid of the output tube.
With no bypass capacitor, I will need 48.8 *2.828 = 138Vrms! That is a 390V swing from peak to peak!
With a bypass capacitor, I will need 19.2 * 2.828 = 54.3Vrms, much easier to achieve. 154V p-p, much easier to achieve, AND, the driver will contribute much less distortion.
My current driver can deliver 0.13%THD at 50Vrms. It can not provide the 138Vrms drive under any circumstances.
There is not free lunch.
It is all tradeoffs.
Thanks. I hadn't really considered that, but I will add it to my notes. I do understand the technical operation of the capacitor, but nobody mentioned that detail.
I think the articles and discussions linked in previous posts definitively address the possible consequences to sound quality, specifically IMD vs THD with and without cap. Discussion of the type and quality of cap could go on endlessly I suppose.
The proper value for the cap if the decision is made to use one also has been addressed in detail.
Thanks everyone!