Choosing loop gain and feedback level for NFB

[Tooboob]

I switch between 6L6s and EL34s by re-biasing, and I hear a clear difference, but the gm of the EL34s is much higher.

Is the best optimization of the level of feedback done by ear, or do you try to optimize for best input signal level at max power (however you'd define that), or what?

Is there an adjustment internal to the standard LTP for voltage gain, or do you mess with the bias resistor values or ... ?

[NateS]

The resistor to ground at the base of the tail is the bottom leg of the feedback divider. It may be the presence control on some amps and in parallel with it on others. You could parallel that resistor to reduce the amount of nfb for the el34s. So long as its not a huge drop the taper shouldn't get too weird. Since one end goes to ground, it gives you several options on how to switch it.

I'm messing with the LTP right now, still trying to understand adjusting the gain, output swing and biasing. I cant seem to get the 140V swing I want from a 390v B+ no matter what i do with the load, quiescent current, and biasing. The parameters are very interactive, and even with massive amounts of feedback, I still had to pad the signal way down to keep it from clipping in the PI. Since I actually don't need any more gain, just inversion, I'm thinking about a split load mosfet PI. The output from the tone stack on the 2204, even clean, according to my simulation is stout enough to drive the grids as hard as I want, provided it's buffered. (I had to go back as far as the screen tap to get enough voltage to drive it though)

[Chuck H]

Toob, it's best done by ear. If you're trying to get the same tone and response from both tube types that negates the point of switching EL34's with grind like crazy with the same drive typically used for 6L6's. Just do what sounds good and gives you the difference you expect. Typical NFB ratios used in guitar amps is 20:1 at the divider from an 8 ohm tap. All things in between (OT specs, PI gain, power tube GM, etc.) affect the reality of the actual feedback voltage. Just experiment and settle on what seems to work best for both tube types or idealize for one type and take what you get with the other. A "proper" circuit for this doesn't exist because guitar amps are signal generators, not reference amps.

NateS, have you tried reducing PI plate voltage or the plate resistor values? Other than a blunt cross line feedback resistor these are the only two ways to significantly reduce PI output that I know of. Messing with the bias will only result in earlier clipping of one side of the wave form or the other. This can work if you don't overdrive the amp. If you do overdive the amp with a greatly skewed PI bias you may get audible PI cutoff clipping before the power tubes are driven into cutoff. It usually doesn't sound good, but that's subjective. Of course you could also use a lower gain tube. A 12ay7 makes a dandy LTP.

[jazbo8]

Also bear in mind that with the NFB engaged, the waveforms out of the PI may look funky, so like Chuck said, tweak by ear is the way to go here (assuming the incoming signal to the PI is relatively clean).

[Dave H]

I'm messing with the LTP right now, still trying to understand adjusting the gain, output swing and biasing. I cant seem to get the 140V swing I want from a 390v B+ no matter what i do with the load, quiescent current, and biasing...

140V p-p may be possible if it’s carefully set up. What value are the power tube grid leak resistors and how are you measuring the PI output swing? If the power tubes are operating it will probably clip at their grids first so it won’t be measuring the max available p-p PI output. If the power tubes are removed the B+ will rise giving a high reading for the PI output. To get the true operating value I’d disconnect the power tube grids from the grid leak resistors and connect them to the bias voltage (fixed bias) or ground (cathode biased)

Edit - I've just read Chuck's post. Nate if you actually want less output from the PI forget what I just said

[NateS]

I'm still simulating - I can do that with a little one asleep on my shoulder. Soldering and beating up EL34's not so much. In my 2204 sim, I see about 140vpp as max clean output from the tone stack, and it happen to match my drive requirements. I think the problem I was encountering was the max signal input to the LTP clips around 10vpp, I wasn't able to pad it down enought to avoid clipping in the PI and still get the output swing.

I tried a mosfet SLPI last night and appears to work beautifully - I can drive the grid buffers to +20v from -44v bias (or even cutoff). I just need to figure out if I want to ditch the NFB loop or figure out how to incorporate it. Being a 100% NFB amp with symmetric outputs - it seems like the SLPI has all the makings of a summing amp.

[NateS]

I tried a couple methods for injecting NFB into the SLPI, I really thought buffering NFB signal as a virtual ground to the SLPI was going to work. It almost did, but it had poor balance. Next I went "Ok: discrete because it has really high voltage swings. Feedback. Summing amp.... and before long I realized I was once again looking at the familiar looking differential amp just for a mixing amp, in front of the phase splitter. Inelegant, so...

I'm thinking that NFB around the LTPI is probably my best bet to getting something approaching a unity gain phase splitter capable of 140Vpp swings without clipping, that provides convenient mixing capabilty for global NFB.