Hello, I'm just simulating a SE tube amplifier with quite heavy GNFB, and, under simulation, if I decrease the interstage coupling cap (from 470n to 100n with R=150K), I get a slight increase (+0.5dB) in the low band bass response (< 50 Hz).
If I decrease to 1/10, 47nF, I get even a slight peak around 20-30Hz.
How this can be possible?
Are these still good basses?
This is only applying the GNFB. Without GNFB, decreasing the cap value, I decrease the low band frequency response, as I was expecting.
If I decrease to 1/10, 47nF, I get even a slight peak around 20-30Hz.
How this can be possible?
Are these still good basses?
This is only applying the GNFB. Without GNFB, decreasing the cap value, I decrease the low band frequency response, as I was expecting.
In my opinion, sticking to around 47n (0.047uf) for coupling capacitors from plate to the next (output tube) grid in tube amplifiers is the best value in terms of speed and frequency response.
I don't understand why a large value (470n/0.47uf) is ever needed for such designs, yet I see some commercial designs using that value.
To me, it's counter-active and a waste, because who really needs to amplify sub-sonic noises which can hinder good listening, over-work tubes to respond to such low frequencies, and take up more under-chassis space.
What people don't seem to realize, is that the changing charge resulting from a large capacitor causes lags in response, making the sound appear soft, giving that fake "tube sound".
And I prefer accurate sound in my amps.
I've rebuilt amps that have used 220n/0.22uf coupling caps with typical 220K-470K grid leak resistors, and installed 0.047uf caps, and the amp performs much better, and still have great bass response.
Because in each case, the "more is better" rule isn't justified.
I don't understand why a large value (470n/0.47uf) is ever needed for such designs, yet I see some commercial designs using that value.
To me, it's counter-active and a waste, because who really needs to amplify sub-sonic noises which can hinder good listening, over-work tubes to respond to such low frequencies, and take up more under-chassis space.
What people don't seem to realize, is that the changing charge resulting from a large capacitor causes lags in response, making the sound appear soft, giving that fake "tube sound".
And I prefer accurate sound in my amps.
I've rebuilt amps that have used 220n/0.22uf coupling caps with typical 220K-470K grid leak resistors, and installed 0.047uf caps, and the amp performs much better, and still have great bass response.
Because in each case, the "more is better" rule isn't justified.
Everyone is entitled to his opinion but is this statement a fact in LF? It is the same voltage passing and no increase in current happens with double size coupling capacitor.
There is another pole in the loop that is higher in frequency. When you raise the frequency
of the coupling capacitor's pole (by decreasing its value), the phase shift in the loop
is increased in part of the LF range, causing peaking and potential instability if raised too high.
Last edited:
What I'm saying is.... this doesn't have anything to do with current or voltage levels.
It has to do with the speed of the capacitor's ability to fluctuate/change potential with the complex audio frequencies that "pass through" it.
A large capacitor naturally takes more time to charge/discharge than a smaller one, with a given resistive load, right?
Like swinging a heavy door back and forth - the strength needed to do it is more than for a small door.
Electrons in such a situation react the same way - you've got more to move to change polarity.
Besides, who needs to move electrons at sub-sonic frequencies?
There is no musical benefit to that nonsense. - rumble, woofer pumping, etc.
It only wastes precious amplifier power that's really needed for what a person can hear.
Case in point - a lot of amplifiers are stated as being "DC to 100kHz" - but WHY?
When music only requires 30Hz to 15kHz.
This is intended in the spirit of utmost politeness, but I think you have some misconceptions about how a coupling capacitor functions. Using a bigger coupling capacitor minimizes voltage changes between the electrodes at low frequencies.
A section in a book that I read by Douglas Self comes to mind where he demonstrates that electrolytic capacitors generate significant distortion when used as coupling capacitors unless they are of sufficiently large value that essentially no signal voltage appears across them at frequencies of interest. This involves making them large enough to couple signals well below the audio band. This also applies to capacitors of other types but distortion is generally proportionally smaller with other types.
Not to mention that when a feedback loop is involved, phase shifts are also a concern in addition to flatness of response.
That would (probably) explain why a small bypass on a coupling capacitor leaves an impression. I never liked it and now I can see why.
ygg-it, just moving coupling cap corner frequencies around to tweak roll-off response in the audio range is not the full story, especially in an amp with gnfb.
The high pass filters from coupling caps and the output transformer need to work in with the stability requirement for your target level of gnfb. In some amps the OPT high pass is designed to be very low, and well below the coupling cap filter frequencies, with the aim of the coupling cap drop off managing the stability requirement. But the flip side is that the OPT has to have exceptional primary inductance over a very wide signal level range, so as not to encroach on any gain or phase margin. That can require the coupling caps to have a fairly high corner frequency and so the distortion of practical low frequencies can increase (as closed loop gain is falling) and so any low note or sound could noticeably splatter harmonics in to a much more discernible frequency range.
Many amps end up with OPT's that cause gain and phase changes in the 10-50Hz region, and so some then push the coupling cap corner frequencies well below the OPT roll-off region.
One vintage technique to manage theses interactions is to insert a low frequency step/shelf filter to become the dominant gain reduction mechanism to ensure adequate gain and phase margins. In that case, the 'normal' coupling caps can be increased to make their corner frequencies very low, and the step/shelf response is high enough for the OPT not to cause instability. But again that causes a net reduction in feedback level down in the practical bass region, and hence higher distortion levels for low bass notes, and a change in speaker damping (which may or may not align with what the bass speaker wants).
The high pass filters from coupling caps and the output transformer need to work in with the stability requirement for your target level of gnfb. In some amps the OPT high pass is designed to be very low, and well below the coupling cap filter frequencies, with the aim of the coupling cap drop off managing the stability requirement. But the flip side is that the OPT has to have exceptional primary inductance over a very wide signal level range, so as not to encroach on any gain or phase margin. That can require the coupling caps to have a fairly high corner frequency and so the distortion of practical low frequencies can increase (as closed loop gain is falling) and so any low note or sound could noticeably splatter harmonics in to a much more discernible frequency range.
Many amps end up with OPT's that cause gain and phase changes in the 10-50Hz region, and so some then push the coupling cap corner frequencies well below the OPT roll-off region.
One vintage technique to manage theses interactions is to insert a low frequency step/shelf filter to become the dominant gain reduction mechanism to ensure adequate gain and phase margins. In that case, the 'normal' coupling caps can be increased to make their corner frequencies very low, and the step/shelf response is high enough for the OPT not to cause instability. But again that causes a net reduction in feedback level down in the practical bass region, and hence higher distortion levels for low bass notes, and a change in speaker damping (which may or may not align with what the bass speaker wants).
Last edited:
Ah, but yes there is.
The AC potential fluctuates, certainly, with the audio, and the DC naturally fluctuates too from the driving tube plate.
And those are "normal operating conditions".
Understand electronics, and you'll understand this.
A properly sized coupling capacitor does not charge or discharge in a properly designed circuit. It should have a constant DC voltage across it, but the AC potential should be the same on either side of the cap. Of course, this is in an ideal circuit with an ideal cap, neither of which exist in real life.
In a real amplifier there are several non-ideal factors that could affect the way a coupling cap interacts with the circuit, the usual ESR, ESL and dissipation factor issues in real caps, and other reactive loading effects can create some unusual situations in a real amp.
The OP saw some unusual behavior in a simulation, which usually assumes ideal components unless a "real" component model is used. These "real" parts are only as good as their models, which vary from "close" to totally worthless when it comes to some of the minor effects seen in tube circuits.
There are two places where simply installing a larger coupling cap in a real amp can really mess up the amp leading to a change in frequency response, or even total low frequency instability.
Swapping out the coupling caps for bigger parts will lower the low frequency response, often below the audio range. If the power supply is not properly decoupled well below this new LF limit the amp can exhibit a low frequency oscillation, or instability. The amp may now have gain at 1 to 10 Hz, but some of the low frequency energy can also pass from stage to stage backwards through the power supply. Ever see some audiophile's woofer cones move slowly back and forth while playing a CD. It can be somewhat normal when playing a warped record, but not on a CD or even nothing at all. Often this low frequency can't pass through the OPT, but the output tube's idle current won't hold steady and changes at a periodic rate. Add some GNFB to this situation and you could have an unstable mess.
Another situation where a fat coupling cap will hurt you is when the amp is over driven, even momentarily. Normally a class A or class AB1 amp does not see grid current in any stage. All of a sudden a fat bass drum transient hits the amp and it attempts to drive the grid of a capacitor coupled output tube positive. At this point the grid becomes a low impedance circuit to the cathode and demands current. It gets this current by bleeding charge off the coupling cap thus reducing the instantaneous voltage on it's grid. The bass drum hit passes and the amp attempts to return to normal, but the coupling cap has lost a few volts of charge. This will cause a temporary bias shift pushing the output tube closer to cutoff until the charge can be restored through the grid circuit resistance of the output tube. This resistance can be anywhere from 47K to 1 meg, which with a fat coupling cap can lead to a long recovery time. The impedance of a conducting grid can be under 1 K ohm, so the "upset" time can be short. So a short overload can lead to a long recovery. Add GNFB and the situation gets worse. This is called blocking distortion.
I agree with most of tubelab's explanation on larger-than-needed coupling caps.
It verifies my learned education of tubes, and even of solid state builds.
Perhaps my use of the word "charge" was mistakenly used, and the fact of "recovery time" is more in line with what we're discussing here.
But it remains a fact that using such over-rated caps causes several issues which can easily be resolved by using sensible values in builds, focusing on the quality of hearable audio within a sensible bandwidth.
0.01, 0.02, 0.047 uF output coupling caps have been "the norm" with the majority of tube amplifier design for decades, all you need do is go and look at those time-tested designs in service manuals.
The occasional "bloated" capacitors used in other designs I suspect were either a marketing tool to inflate frequency response specs, or what the manufacturer had on hand from the vender.
I see no reason that the OP cannot use the 0.047 caps in his design and be quite happy with it.
It verifies my learned education of tubes, and even of solid state builds.
Perhaps my use of the word "charge" was mistakenly used, and the fact of "recovery time" is more in line with what we're discussing here.
But it remains a fact that using such over-rated caps causes several issues which can easily be resolved by using sensible values in builds, focusing on the quality of hearable audio within a sensible bandwidth.
0.01, 0.02, 0.047 uF output coupling caps have been "the norm" with the majority of tube amplifier design for decades, all you need do is go and look at those time-tested designs in service manuals.
The occasional "bloated" capacitors used in other designs I suspect were either a marketing tool to inflate frequency response specs, or what the manufacturer had on hand from the vender.
I see no reason that the OP cannot use the 0.047 caps in his design and be quite happy with it.
The cap value is just half the story with respect to blocking distortion, the other half is related to the resistances that charge and discharge any dc level shift on the cap. It's not valid to just refer to the cap value as 'too high', if the circuit has similar time-constants to another circuit with lower cap values. So a 'fact' is only half a fact really.
