Paralleling PI output coupling caps with a switch

[Chuck H]

The difference between the two cap values is about 1dB at the onset of audible frequency. That might be audible to a tiny percentage of the population... EXCEPT!.. Guitar speakers also have very poor efficiency at 20Hz. So the actual difference in dB at any PRODUCED audio frequency becomes a small fraction of a dB. And absolutely NOT audible unless you're the sort of person that hears things because they want to rather than because they are real. That said...

I don't really know if there may be an audible difference under clipping conditions for the PI where the input impedance drops and these cap values might have more of an effect in the audible range. I know others here that have more tech chops could evaluate this circumstance better so I hope to hear from them about this.

[Chuck H]

Regarding the switching circuit I mentioned above. I didn't take any charging or discharging of the switched cap into account when I proposed it (fig 1). If the cap must remain in a charged state it seems easiest to employ a low value resistor in parallel rather than a high value resistor in series (which would require more complicated switching to keep the cap charged). This would be fig 2. Corrections and opinions accepted as usual





EDIT: Don't use these. Differ to the circuit shown below.

[Helmholtz]

When applying a DCV (i.e. at turn-on) to a series wiring of 2 caps, a common transient current will flow, charging each of the caps to a voltage inversely proportional to its capacitance. The sum of the cap voltages must equal the applied voltage.
E.g. if the applied voltage is 220V, the 150n cap will charge to around 70V and the 68n cap to around 150V.
When one of the caps is shorted via the switch, the other cap must charge to the full voltage.
The voltage jump will result in a pop.
The 1k resistor in fig 2 will not prevent a voltage across the switch when it's open.

[Chuck H]

When applying a DCV (i.e. at turn-on) to a series wiring of 2 caps, a common transient current will flow, charging each of the caps to a voltage inversely proportional to its capacitance. The sum of the cap voltages must equal the applied voltage.
E.g. if the applied voltage is 220V, the 150n cap will charge to around 70V and the 68n cap to around 150V.
When one of the caps is shorted via the switch, the other cap must charge to the full voltage.
The voltage jump will result in a pop.
The 1k resistor in fig 2 will not prevent a voltage across the switch when it's open.

Well then for plactical and uncomplicated reasons I say go with figure 1 and accept the pop. Otherwise I'm open to suggestions.

[Helmholtz]

I'd go with a parallel combination (2x 68n) where the second cap is connected to the first one via a large value (100k or greater) series resistor.
The resistor makes sure the second cap is always charged.
The switch is then used to short the resistor for paralleling the caps.

[Chuck H]

So like this:



I have no idea why I didn't mentally see the parallel arrangement could still isolate the switch from HV. Total brain fart.

NOTE: seven wants values switched between .047u and .15u

For the resistor 100k looked just a little soft in the simulation so I bumped to 220k and it looked fine.

[Chuck H]

Follow up...

This has been a good exercise in how to switch cap values under a given circumstance, BUT...

There is literally only about 1dB difference at 20Hz for these values with this switching for the PI output. And if you can hear that with a guitar played through a gutar amp and guitar amp speaker your ears are better than mine. And that said...

Still no address to my inquiry above regarding the PI impedance vs power tube input under clipping conditions and possibly making these cap values more audible. My simulations aren't able to articulate things like that and haven't been able to find finite information about these impedance ratios when clipped to make a determination. So this will require an actual engineer who can evaluate a circuit on it's own in both circumstances. So let's just say it's the standard LTP for a guitar amp with 1M grid leaks and 10k to 47k tail. Any global feedback considerations welcome as well. This is just too deep and complicated for my level in the craft.

But my guess is that the difference between these two values, .15uf and .047uf, still won't be significant if audible at all.

[Helmholtz]

[QUOTE=Chuck H;n992507]
Still no address to my inquiry above regarding the PI impedance vs power tube input under clipping conditions and possibly making these cap values more audible.
QUOTE]

When driven into grid conduction the power tube input impedance might get as low as 10k/pair (very rough estimation).

Another consideration might be the influence of coupling cap value on blocking distortion.

I think differences are best evaluated by listening tests.
All of my favorite amps use 0.1µ power tube coupling caps.
Seems I like them better than 0.022µ caps, though I didn't do extensive cap testing.

[Chuck H]

Ok. Well now it makes a little more sense then. I just did a very blunt test by changing the grid leaks on an amp model from 220k to 10k. With the 10k there's just over 1dB of difference between .047 and .15 coupling caps at 82Hz. That by itself probably doesn't make enough difference, BUT we have to also recognize that with the lager value the amp is working hard to make extra power at frequencies below 82Hz as well (almost 6dB at 20Hz). And I think this must affect it's behavior. If not actually the tone. I mean, if the amp is making all the power it can but more of that is in the LF that leaves less power for all higher frequencies doesn't it? I might expect the effect on audible frequencies to be a sort of compression.

[Helmholtz]

Where would frequencies considerable below 82Hz (or 73Hz with drop D tuning) come from?
Are you thinking of ghost notes?

[Chuck H]

Not just ghost notes. I always expected there must be undertones below fundamentals due beating from difference tones. You can really hear it in some players that use heavy distortion But this would be the only circumstance. If the amp isn't clipping hard I don't think it matters.

EDIT: When I first started to notice this Mesa was the only "high gain" amp. There are now dozens of manufacturers. But back then I heard the effect almost exclusively from Marshall amps turned up to eleven.

[Helmholtz]

So generally undesirable (to me at least) IMD artefacts.
Not very likely with a "low gain" amp like an AC30.

[Chuck H]

Probably not with an AC30. But those notes do happen at all gain levels to some degree. I just thought of a good analog. I think we've all tuned our guitar with harmonics at one time or another. These would be notes that are above the fundamental and yet we can clearly hear the beating oscillation at a frequency much lower than the fundamental. Even with a purely clean signal. And this happens all through the subharmonic spectrum. I've read a lot from very scholarly dudes poo pooing subharmonics but I still believe.

Not that you can hear it much with guitar speakers. But there has to be some reason to prefer .1 caps to .022 caps from the PI. And it's probably not that 1dB at 82Hz only under clipping conditions because this difference is tiny and dependent. I don't have the physics chops to back this theory but it might explain this where nothing else does as yet.

[tedmich]

All this micro tweaking does usually end up just being left one way, as has been mentioned!
To whit, the JSX presence/depth dual knob I added to my Ultra stays at one setting, and a famous DIP tweaker amp (the PV VTM, described nicely here) see's very little switching once "dialed in"

[Chuck H]

I just read a wiki on subharmonic undertones and thought to share this:



Now this speaks mostly to acoustic resonance but there may be analogs with signal processed through a guitar amp. And...

I'm remembering an experiment I did when I was some younger and had time on my hands. The details are lost to time because I wasn't doing serious research or keeping records. Ok, the experiment:

I plugged a signal generator into a dirt box and observed a square wave on the oscilloscope. Then I put an EQ in the signal chain after the clipped signal and tweaked it to expose an almost perfect sine wave at a frequency lower than the fundamental. So this wouldn't be a sum/difference note unless it was strictly the sum/difference within the harrmonic order of the single clipped note.