Treble shunt - before or after grid-stoppers?

[R.G.]

Interesting discussion. For clarification, a series R followed by a shunt C is usually called a low-pass filter, so I would think that the response would be called a high frequency roll-off, no? Just want to make sure I'm not missing something here...

You are correct. An R followed by a C shunt to AC ground is a low pass filter. It is indeed a high frequency rolloff, as it's the highs that get cut. I just can't type and think at the same time. Doh!

[Gaz]

It is in general a bad idea to use extra grid capacitance for stopping oscillation in the audio band. It is better to make the output tubes go well above audio, control the gain/phase response with a single compensation cap either on the phase inverter [my italics], in the feedback network, or in the transformer implicitly. Grid stoppers are useful for killing oscillations at RF, where the "black magic" stuff takes hold. Adding more low frequency rolloffs down in the audio band can make oscillations worse.

R.G., are you implying that strapping a cap across the phase inverter anode resistors (in a long-tail pair) doesn't cause a phase shift in the feedback loop? If so, that's great because one cap is cheaper than 4!

[Chuck H]

Well. from my (admittedly low tech) point of view, since the FB loop is typically inserted at the cathode of the PI there will be some corse interaction. Since the shunt cap is past the plates (post PI) there must be some phase shift due to the altered freq response within the loop itself, no??? But it's probably mice nuts and can be neglected in this application.

Chuck

[Gaz]

Thanks for the reply, Chuck. Yes, it must be within the feedback loop, of course, and cause a phase shift.

I was just looking over Merlin's book, where he discusses phase shifts in the FB loop at length. I clearly skimmed that section He insists that adding capacitance at the inputs of the power tubes is the best place to stabilize a FB amplifier because it's the most predictable. He notes that adding capacitance across the phase inverter anodes relies too much on its balance (I don't understand this completely). He doesn't mention raising the grid-stoppers, however, but perhaps because it's a noisier solution??

I think I might go ahead and add the caps to ground. An aside: Can I simply wire the shunt caps from the power tube grid pin to the cathode pin if the cathode is grounded through a 1-ohm resistor for bias measurement? In short, is ground still ground through the 1-ohm resistor? Stupid question perhaps, but I needed to ask

[martin manning]

I think I might go ahead and add the caps to ground. An aside: Can I simply wire the shunt caps from the power tube grid pin to the cathode pin if the cathode is grounded through a 1-ohm resistor for bias measurement? In short, is ground still ground through the 1-ohm resistor?

Isn't AC ground in this context the bias supply voltage, i.e. the junction of the bias feed resistors?

[R.G.]

Thanks for the reply, Chuck. Yes, it must be within the feedback loop, of course, and cause a phase shift.

Strictly speaking, it can be anywhere in the forward amplifier gain path that causes the gain to fall below unity before the phase shift accumulates to cause positive feedback through the feedback path. If that happens, it's stable.

There is a second way to handle gain and phase shift stability. If you can't reduce the gain below unity with a single pole rolloff, you can reduce the *entire* open loop gain of the forward path. This lowers the entire gain at all frequencies. If you could not get below unity gain with a simple compensation cap, you can lower the gain by perhaps UN-bypassing a cathode resistor bypass cap or lowering the value of a plate resistor to make the whole thing simply have less gain. This lowers the high, middle and low frequency gains together, but it can duck the high frequency gain below the point where it oscillates. The only bad thing about this approach is that with less gain to work with, you get less gain to correct distortion, so less of the good effects of negative feedback you were going to all the trouble for.

Tube amps are forced to do this to some degree anyway. The output transformer by itself has two out of the three phase shifts needed to cause oscillation, so unless you can have a really, really, really good OT with the phase shifts way high in frequency, you must either (1) give up on using feedback compensation entirely or (2) lower the gain until it's stable with the OT only (or both 1 and 2 together) or (3) be really crafty and perhaps lucky. This single issue is why tube amps with OTs can never get to the high gains and low THD of a non-transformer amplifier, and the reason behind the many-decade push to get output-transformerless amps.

I was just looking over Merlin's book, ... He insists that adding capacitance at the inputs of the power tubes is the best place to stabilize a FB amplifier because it's the most predictable.

That's a judgement call, as are all statements that say "best", but probably a good one. Predictability of compensation in the face of component variability (...uh, I think I'll put in some EL34s tonight, and didn't I read on the internet that 12AY7s are better than 12AX7s because they sound juicier?) and aging is an almost priceless commodity in amplifier design.

He notes that adding capacitance across the phase inverter anodes relies too much on its balance

The gain of each half of a PI can be and is quite different, especially in the differential-amplifier version that's commonly used in Fenders, for one example. The cathode biasing does not make this a good, balanced diff-amp.

He doesn't mention raising the grid-stoppers, however, but perhaps because it's a noisier solution??

He may not have thought about it. It's an unusual choice, and has its own set of issues. I like it because it offers more predictability in many cases. It does add noise, but the voltage gain from the grids of the output tubes to the speaker is much smaller than the gain from any prior part of the amp; noise impact gets less the further toward the output you go.

Can I simply wire the shunt caps from the power tube grid pin to the cathode pin if the cathode is grounded through a 1-ohm resistor for bias measurement?

Yes. This directly reduces the grid-to-cathode voltage by bypassing current around it.

In short, is ground still ground through the 1-ohm resistor?

That is a reasonable question, but it is easier to answer if it's stated another way. I'd say "Does the 1 ohm resistor between the cathode and ground change what a capacitor connected to the grid sees very much over it being connected directly to ground?"

The answer is that it depends on the other parts. The cathode resistor is "multiplied" from the point of view of the grid by the gain of the tube, so it appears to be a higher resistance than it is because the tube's plate current is raising and lowering its voltage in synch with the grid voltage. A rule of thumb is that it appears to be voltage-gain-times bigger, like a Miller capacitor. That's not strictly correct, but it is a handy approximation.

The cathode resistor appears to be tens-of times bigger than it is, so it could "look" like maybe 10 to 50 ohms to a capacitor connected to the grid. However, the grid stopper is hundreds to thousands of times bigger than the cathode resistor, so the cathode resistor being in series with the cap too can be neglected as not changing what the current flowing through the grid stopper by any significant amount. The amplified cathode resistor *does* add a highpass step at the frequency of the capacitor and the amplified cathode resistor, but this is so high in frequency that it doesn't need to be considered for audio purposes. The grid stopper and cap to cathode have long since dropped the gain into the mud.

Isn't AC ground in this context the bias supply voltage, i.e. the junction of the bias feed resistors?

Maybe. Ideally, you want to keep all current into and out of the bias supply, because you do not want to feed an error voltage to everything connected to the bias. And bias supply bypassing is almost criminally neglected in many amplifier designs. You're right that the bias supply is telling the grid where "AC ground" is, but it's a better design to leave the bias supply alone, and bypass anything you really want to be "ground" to the star point in the amp. Using the bias supply as a ground reference means that the bypass cap on the bias supply is inserted in series with the "ground" for everything connected there. The quality of the bypass cap then determines the quality of that grounding.