Cathode bypass capacitors: Electrolytic vs. film?

[Chuck H]

I see some Hifi designs (particularly those described in 'Audio Valve Amplifiers by Rainer zer Linde) where every electrolytic is bypassed with a low-value film cap, even in the PSU - usually 220nf or 470nf. I can't detect any A-B difference with/without no matter how hard I listen. I'm not convinced of the value in doing this in a guitar amp. If there's anything it's outside of my hearing range. Either that, or I'm not a True Believer.

Same here. And "I want to believe" (thanks Mulder ) Still, I don't hear it, BUT...

I've often thought that on some amps that dare to decouple like phase stages at the same node bypassing that node filter with a film cap might be a good idea. Some such amps, like the Fender HR series amps are prone to oscillations due to this design. These amps might benefit greatly by reducing HF impedance at that filter node. I've suggested it here, but no one has tried it yet. I don't do repairs and I don't design amps with filter nodes sharing like phase stages or I already would have tested this.

[Leo_Gnardo]

I see some Hifi designs (particularly those described in 'Audio Valve Amplifiers by Rainer zer Linde) where every electrolytic is bypassed with a low-value film cap, even in the PSU - usually 220nf or 470nf. I can't detect any A-B difference with/without no matter how hard I listen. I'm not convinced of the value in doing this in a guitar amp. If there's anything it's outside of my hearing range. Either that, or I'm not a True Believer.

This practice of paralleling el caps with films, usually in the ratio of 100:1 or thereabouts, I first noticed late 70's in Audio Amateur magazine. At that time quite a hubbub had started over the sonic qualities of capacitors in audio circuits. The reason given for parallel power supply caps was, they not only smooth the rippled DC from the rectifier, but they are also tasked with decoupling amp stages from each other. Electrolytic caps may be the bees knees for filtering lower frequencies up to a couple kilohertz, not so much for higher frequencies. Parallel film caps allow full frequency decoupling, well beyond the audio band, presumably preventing preamp stages from "talking to each other" via their power supply connections, suppressing noise and preventing ultrasonic oscillations. At least, that was the theory at the time.

Contributors to AA included audio luminaries such as John Curl, Erno Borbely, and Nelson Pass. It was taken pretty seriously, not a place for "nuts" to bloviate.

[bob p]

Electrolyics are great at grounding LF line noise (AC) but they are not so great at grounding HF content because their HF response is so poor, and it gets worse with age. If they were available in the 70s, large value film caps (0:1 bypass) would have been a better option for the PSU than using an electrolytic (1:0 bypass), but even today they're large and expensive and impractical. The happy medium is comprised of using inexpensive, small, high-density electrolytic caps for AC-DC conversion and bypassing them with economical film caps that have a complimentary frequency response to eliminate HF line noise. The hybrid design using the 100:1 ratio only originated because it wasn't economical to obtain film caps in the range of 1:0, 1:1, to 10:1 ratios. Even today 100:1 still represents a sweet spot in terms of cost effectiveness.

This method of bypass is very common in PSU design. As Leo mentioned, it's been standard for decades. I see it present in many modern tube amp schematics and I see it missing in many others -- and totally absent from builds that are made by blindly parroting designs from the 50s and 60s. Today you'd have to be a pretty backwards tube amp designer not to consider bypassing the first stage of a Pi filter with a film cap. Usually that alone is sufficient for the PSU, as AC line noise is what we're worried about. Once the line noise is gone, it's gone -- and bypassing sequential nodes isn't necessary.

If another stage should inject it's own noise (oscillation) into the circuit then bypassing the PSU node caps isn't an effective way to solve the problem. There are more efficient methods of improving the margin of stability with lower value caps that are less expensive, most effectively accomplished via slugging from grid to ground at the dominant poles.

To answer Chuck's question, I think that the reason that bypassing is not done at the PSU nodes is because it's more effective to do the HF slugging at a stage's input (grid-ground) than at a stage's output (anode-anode in a PI like a Marshall) or at across the anode resistors (anode-ground) or the PSU node caps (B+ to ground).

Mesa uses a combined approach in some of their flagship high gain heads that use a large amount of NFB; they'll bypass the PI anode-to-anode (like an old Tweed Fender), in addition to bypassing the anode resistors (equivalent to anode-ground). In the SF era Fender chose a much more simple, efficient, cost effective option of placing the shunt capacitance directly at the 6L6 grids. Both designs lower the dominant poles to about the same frequency, which is well above the range of our bandwidth limited signals. In spite of the ongoing lamentations of BF gurus, these capacitors shunts are inaudible as they only have effects in the ultrasonic range.

Of course, even as true believers in bypass, we know that we're not supposed to hear any of this in a guitar amp; these changes address closed loop stability by taking away ultrasonic HF content that isn't supposed to be there in the first place. In an amp with properly designed NFB the CL frequency response is going to be far beyond the range of our bandwidth limited input signals.

[nsubulysses]

Reading this thread is awesome. I posted my thought on it and it was abandoned for two months. I hope the OP is stil around to benefit from the good conversation

[Chuck H]

Reading this thread is awesome. I posted my thought on it and it was abandoned for two months. I hope the OP is stil around to benefit from the good conversation

He's around. A regular. Last activity one week ago so I'll guess this will all be seen if he checks his posts at all.

[bob p]

Reading this thread is awesome. I posted my thought on it and it was abandoned for two months. I hope the OP is stil around to benefit from the good conversation

sometimes it takes a while before somebody kicks the can hard enough to get it some hang time, like tubeswell did in this case.

[bob p]

If we are talking about a BF or SF preamp I've always liked to use 2.2uF film caps instead of the 25uF electrolytics for the first gain stage. Or use a center-off SPDT switch to toggle between 2.2uF, 0uF and 0.68uF (Dan Torres sold such a switch as his Triple Killer kit. Only I think he included a 22uF e-cap...)

using the Rk value of 1k5 and the simplified formula that doesn't include gm, those 3 caps would give corner frequencies of 4.2, 48 and 1500 Hz, respectively.

I'd have to guess that the Torres mod options would give you the illusion of higher HF "gain" by selectively throwing away LF content. That's OK, but he does this at the expense of impairing hum rejection in the first stage, where it's most important in defining the limits of the amp's signal to noise ratio. I wonder if the Triple Killer Kit could result in a noisier amp. If so that mod could amount to shooting yourself in the foot when it comes to SNR.

I take a different tack. I prefer to bias using big fat caps in the first stage for two reasons: 1) to eliminate LF distortion artifacts, and 2) to provide optimal heater noise rejection in the first stage, which is a make it or break it stage when it comes to defining the amp's SNR. Diode biasing is better yet.

All things considered I think one can get better results by focusing on optimizing SNR in the first stage, and then waiting to do the frequency response shaping in later stages once you have a strong noise-free signal. IMO establishing a low noise floor is a key element in designing a good sounding amp.

I think that this Dan Torres mod is a good example of the different concerns that prevail when modding amps vs. designing amps.

[nsubulysses]

sometimes it takes a while before somebody kicks the can hard enough to get it some hang time, like tubeswell did in this case.

everyone elevating the explanation to a much higher level than I could ever offer

[nsubulysses]

I'd have to guess that the Torres mod options would give you the illusion of higher HF "gain" by selectively throwing away LF content. That's OK, but he does this at the expense of impairing hum rejection in the first stage, where it's most important in defining the limits of the amp's signal to noise ratio. I wonder if the Triple Killer Kit could result in a noisier amp. If so that mod could amount to shooting yourself in the foot when it comes to SNR.

I hear this mentioned on MEF pretty often about fully bypassed first stage for less hum. I'm sure it's really a good thing but I tried to adopt this and feel like I went on a 3 week journey of making my amp sound worse. I'm sure it can be done with good results but maybe my amp is not a good candidate, or I didnt' hone it well. Or maybe I was too far along designing the later stages of the preamp to make such a big change to the first gain stage.

In theory or reality, is it ACTUALLY possible to obtain the exact same sound by toggling coupling and bypass cap values? What I thought was "best sound" for my first gain stage was 1.5uF cathode bypass cap and .033uF coupling cap. Tube is in parallel with 1.5K cathode resistor and 75K anode resistor. switching to 47uF cathode cap and .0047uF coupling cap i could get a similar sound. In ways it was better. A little more clear and articulate. But I ended up deciding it was more worse than better. It felt too compressed and thick. The little cathode cap and big coupling cap sounded more dynamic and airy. A little less defined but a little more open and dynamic sounding. also the tone stack seemed to have less range and usability, like the sound was too defined by the first stage or something. I ended up going back to 1.5uF cathode and .033uF coupling cap. it sounded a little more loose and open and I think with big .033uF coupling caps for first two stages it gave the tone stack a lot to work with.

I don't know the math of this. I just did it by ear.

edit: I almost started a thread a few weeks ago to ask other people experience with this cap toggle between cathode and coupling caps and I was gonna call it a Chuck H mod. Maybe it works better for a cleaner amp instead of high gain. I would like to try it agian in the future even though it didn't work out this time around

[Chuck H]

To answer Chuck's question, I think that the reason that bypassing is not done at the PSU nodes is because it's more effective to do the HF slugging at a stage's input (grid-ground) than at a stage's output (anode-anode in a PI like a Marshall) or at across the anode resistors (anode-ground) or the PSU node caps (B+ to ground).

Disagree (peacefully),
The HR series amps are quite specifically known to oscillate in the HF due to decoupling failure. Some of this was due to lower quality electrolytic caps, but most of the problem is in the design. An improvement in HF decoupling performance at the like phase node in the preamp would benefit the design. I'm sure of it. Further, I doubt it would be as audible a difference as any sort of "bleeder" caps placed from signal path to ground or across the plate resistor. To preserve the original tone of these amps and still achieve acceptable performance without oscillation I believe that a film cap of lower HF impedance across the like phase shared filter node would absolutely be the cheaper and least audible solution.

EDIT: I'm totally in your camp about SNR in the first gain stage (or even the second in cascaded amps). Something I've incidentally gravitated towards and only recently realized I was doing. I've never used DC filaments and, as such, have often sought to deal with any noise on the early stages. Since no significant clipping is achieved until after the second triode in most designs it only makes good sense to consider the priority for those stages to be amplification without any extra noise. Tone shaping can come after that (BF type circuits not withstanding ).

[octal]

The one thing no one is discussing here is that an "undersized" cathode bias cap has a shelving response -i.e. the stage will have the "fully bypassed" max gain above a certain frequency and the gain will taper off below that frequency at a -6dB/octave slope but hit a plateau at some point (as the stage will always have a minimum unbypassed gain determined by the value of the cathode resistor and the type of tube.) On the other hand, using a coupling cap to cut bass, will just have a steady -6dB/octave slope starting at the knee frequency and continuing down to near DC.

[Chuck H]

The one thing no one is discussing here is that an "undersized" cathode bias cap has a shelving response -i.e. the stage will have the "fully bypassed" max gain above a certain frequency and the gain will taper off below that frequency at a -6dB/octave slope but hit a plateau at some point (as the stage will always have a minimum unbypassed gain determined by the value of the cathode resistor and the type of tube.) On the other hand, using a coupling cap to cut bass, will just have a steady -6dB/octave slope starting at the knee frequency and continuing down to near DC.

You beat me to it! I was going to mention this while reading the posts this morning and then there you were

Creating that "shelf" isn't hard either. It's just a coupling cap followed by a resistor with a parallel smaller value cap. The trick is mating the values to get the response you want relative to the following load impedance. And, if it's an early stage in a high gainer it's a good idea to keep series resistance as low as practical. So in that case you could be trading hum for hiss.

On line calculators or spice programs make the design chore for simple circuits like this a snap.

[Mick Bailey]

A guy could go insane through knob-twiddling induced anxiety just by building an amp with (say) a 12-position rotary switch with a bunch of caps, plus a pot instead of a cathode resistor. Make it one for each stage. I nearly went off the rails just by installing trimpots without even bringing in cap values. Every time I played that amp I got my new favourite sound. Then the next day I'm sure pixies got in and altered everything to make it sound less-than-optimal.

[Old Tele man]

Can't speak for the HiFi tweekers, but paralleling a BIG electrolytic cap with a small tantalum or ceramic cap (100:1) was PRIMARILY the byproduct of computer circuits, where the BIG cap served as a voltage "reservoir" and the small cap was the noise decoupler. Later, three (3) caps were found to work even better, with a BIG reservoir cap (100:1), a medium "ripple/switching" (10:1), and a small ceramic "transient/spike" decoupler...especially when CPU clocks moved above 1Mhz. Today, it's ALL about "tuned" traces and wavelengths between components.

All this "sharp-edge" technology is not needed in our 82Hz-7KHz guitar amps...compared to computers, our amps are literally "down" in the 'mud' spectrum by comparison.

[Leo_Gnardo]

Can't speak for the HiFi tweekers, but paralleling a BIG electrolytic cap with a small tantalum or ceramic cap (100:1) was PRIMARILY the byproduct of computer circuits, where the BIG cap served as a voltage "reservoir" and the small cap was the noise decoupler. Later, three (3) caps were found to work even better, with a BIG reservoir cap (100:1), a medium "ripple/switching" (10:1), and a small ceramic "transient & spike" decoupler...especially when CPU clocks moved above 1Mhz. Today, it's ALL about "tuned" traces and wavelengths between components.

All this "sharp-edge" technology is not needed in our 82Hz-7KHz guitar amps...compared to computers, our amps are literally in the 'mud' spectrum by comparison.

Let's also mention "local" decoupling caps, typically I'd find at least one 0.1 uF disc across the power rail on a small computer board, up to a couple dozen peppered across a large board.