distortion on leading edge of note

[Firestorm]

I just reviewed this whole thread and I think Jana solved it several pages ago. The complaint is that you get distortion at the point where the grids are driven positive wrt to the cathode (0v). At that point the grids start to conduct. That is the definition of Class AB2. Successful Class AB2 amps pretty much always use direct-coupled drivers to provide the current needed. So I think you have to reduce PI gain, or add a driver, or maybe just kick up the negative bias voltage (make it more negative) so the signal can't take you past the 0 point.

Still interesting that it manifests itself as 120Hz.

[pula58]

I just reviewed this whole thread and I think Jana solved it several pages ago. The complaint is that you get distortion at the point where the grids are driven positive wrt to the cathode (0v). At that point the grids start to conduct. That is the definition of Class AB2. Successful Class AB2 amps pretty much always use direct-coupled drivers to provide the current needed. So I think you have to reduce PI gain, or add a driver, or maybe just kick up the negative bias voltage (make it more negative) so the signal can't take you past the 0 point.

Still interesting that it manifests itself as 120Hz.

Hey Firestorm,
Indeed, I backed off the bias a little bit on the output tubes (i.e., I made the grid bias voltage more negative) and I could reduce the sidebands a little bit. The amp is slightly better but I can still hear the sidebands when I really dig into the guitar strings. Obviously, driving the output tubes into grid conduction is not something I plan to do, but there are transients (like when hitting the strings real hard) that appear to drive the grids into conduction, and there are these nasty sidebands.

Any idea what could be causing them (the sidebands)? I have never noticed this effect in any of the Fender amps (blackface and Silverface)I have played, and, if anything, this amp I'm working on (trying to perfect) has larger power supply filter caps:
47uF first cap ---- 4H choke------47uF----10k----22uF.....7.5K.....22uF

"A" voltage is 433V
"B" voltage is 431V (the plates voltages at the sockets are a couple hundred mV higher than the screens at the sockets)
"C" voltage is 355V
"D" voltage is 322

[Firestorm]

Where is it biased now (% dissipation at idle)? Were your previous similar amps also single-channel?

Bottom line, I think, is that by hitting the strings hard you create a transient that takes grids past where they want to be. If you could get it to compress a little, it might help. A resistor in series with the rectifier output or maybe remove the bypass cap on the mixer stage. In two-channel AB763s, the channel mixing results in a loss of about 6dB. You're running unattenuated.

[JazzGuitarGimp]

From what I have read, the only way to get the power amp stage into overdrive is by attempting to push the output grids more positive than the cathodes. The main purpose of grid stoppers in this situation is to limit the current into the grids. But, it is the attempt to push the grids past the cathodes that causes the plates to flat-top. I think you're very close to the answer - I'm just not convinced the problem is too much gain in the PI.

[Structo]

Do you still have 100K resistors on both PI plates?

[pula58]

Where is it biased now (% dissipation at idle)? Were your previous similar amps also single-channel?

Bottom line, I think, is that by hitting the strings hard you create a transient that takes grids past where they want to be. If you could get it to compress a little, it might help. A resistor in series with the rectifier output or maybe remove the bypass cap on the mixer stage. In two-channel AB763s, the channel mixing results in a loss of about 6dB. You're running unattenuated.

It was biased at about 68% previously, I backed it down to 59%.

There is a 50ohm 10W resistor in series with each diode string i nthe rectifier.

I do have more gain than a Fender due to the fact that
1) I only have one channel (roughly 2X the gain right there)
2) No tremelo circuit and its associated load on the signal.

But, because there is more gain, I simpy turn the vol knob down for the same output sound pressure.

[pula58]

Do you still have 100K resistors on both PI plates?

Yes.

I understand how driving the grids into conduction is a distortion source...but i don't understand how I get these sum and difference terms of 120Hz and the test signal freq when the grids conduct.

with 300Hz test freq I get (when amp is driven just to point where grids start clipping)

180Hz (i.e. 300Hz - 120Hz)
420 Hz (i.e., 300Hz + 120Hz)

[ChrisM]

Where is it biased now (% dissipation at idle)? Were your previous similar amps also single-channel?

Bottom line, I think, is that by hitting the strings hard you create a transient that takes grids past where they want to be. If you could get it to compress a little, it might help. A resistor in series with the rectifier output or maybe remove the bypass cap on the mixer stage. In two-channel AB763s, the channel mixing results in a loss of about 6dB. You're running unattenuated.

It was biased at about 68% previously, I backed it down to 59%.

There is a 50ohm 10W resistor in series with each diode string i nthe rectifier.

I do have more gain than a Fender dure to the fact that
1) I only have one channel (roughly 2X the gain right there)
2) No tremelo circuit and its associated load on the signal.

But, because there is more gain, I simpy turn the vol knob down for the same output sound pressure.

What's the resistor doing?
Using it to drop voltage, introduce some sag?

[pula58]

Where is it biased now (% dissipation at idle)? Were your previous similar amps also single-channel?

Bottom line, I think, is that by hitting the strings hard you create a transient that takes grids past where they want to be. If you could get it to compress a little, it might help. A resistor in series with the rectifier output or maybe remove the bypass cap on the mixer stage. In two-channel AB763s, the channel mixing results in a loss of about 6dB. You're running unattenuated.

It was biased at about 68% previously, I backed it down to 59%.

There is a 50ohm 10W resistor in series with each diode string in the rectifier.

I do have more gain than a Fender dure to the fact that
1) I only have one channel (roughly 2X the gain right there)
2) No tremelo circuit and its associated load on the signal.

But, because there is more gain, I simpy turn the vol knob down for the same output sound pressure.

What's the resistor doing?
Using it to drop voltage, introduce some sag?

I wanted to drop some voltage, and get some sag, like a rect tube, but, without the hassles of the tube failing, rect tube vibration noises, etc.

[martin manning]

There is a 50ohm 10W resistor in series with each diode string i nthe rectifier.

Have you installed these in an amp that did not display this +/-120Hz phenomenon? What happens if you short them out?

[ChrisM]

Short them out for now.

[Firestorm]

Short them out for now.

Or at least try a single resistor after the junction of the rectifier legs (the way I've usually seen it done).

[pula58]

I'll try two experiments

1) short out the 50ohm "sag' R's
2) use just one"sag" R

And see what happens, but, I seem to recall that I had this problem before I added those R's...but we shall just make sure....will report-back tomorrow...

Thanks!

P.

[pula58]

I'll try two experiments

1) short out the 50ohm "sag' R's
2) use just one"sag" R

And see what happens, but, I seem to recall that I had this problem before I added those R's...but we shall just make sure....will report-back tomorrow...

Thanks!

P.

Both experiments (shorting-out (bypassing) the 50 ohm R's, or trying to use just one shared (by both diode strings) 50 Ohm res) yielded no improvements in the problem. Arg.

see lab video here:
http://www.youtube.com/watch?v=F_WfEl3Q ... e=youtu.be

[cxx]

Seems like the normally rejected 120 Hz ripple is modulating your output. There aren't that many ways for it to get in. I might suspect that it is getting in via the bias. That would be the most sensitive place. I would look at the coupling caps as a possible way that could happen. Seems like the only component that you haven't tested. I would also look at the bias supply/power tube grids with the scope during the distortion.

You could put a pi filter before the power tube plate tap to significantly reduce the ripple like on a single ended amp. I don't see that as a solution but more as a test. You might also temporarily convert to cathode bias to see if that eliminates the issue.

There is a ground missing on the schematic for the pt cathodes?