distortion on leading edge of note

[Firestorm]

Indeed, the heater wiring to the 2nd output tube was crossed (Thank you Firestorm, good eye!). So, I re-wired it...and...darn, the problem is still there, I am dissapointed!

I did try, earlier this AM, doubling the capacitance of the first power supply filter cap. It is a 47uF 500V F+T. I wired-in an additional 47uF temporarily, to bring the total capacitance there to (nominally) 94uF. The two sum and difference tones got smaller by a factor of 2 or so. But, I can't believe this amp needs ~ 100uF there since all my other builds used ~50uF and were just fine (the other builds used two 100uF spragues in series, this amp uses just one 500V 47uF F+T). The unloaded B+ voltage in this amp is 475V.

Darn, indeed. But you did discover that you reduce the problem with more filtering. Gerald Weber always insisted you need two 220uF caps in series to keep amps from playing out of tune. Kevin O'Connor, on the other hand, says that sometimes increasing capacitance fixes it AND reducing it sometimes does (which I completely don't understand).

[Jana]

Have you tried changing the 2k2 resistor to a 22K (the one between the screen node and the PI node of the B+ rail)?

[pula58]

Have you tried changing the 2k2 resistor to a 22K (the one between the screen node and the PI node of the B+ rail)?

Hi Jana,
I changed the 2K2 resistor to 10k last week.
I really ought to post the current schematic. DONE

I tried doubling the screen filter capacitance - it had no effect.

Everything in the amp checks-out fine, except that it sounds bad with these intermodulation effects.

As I turn up the amps input signal (which I am applying directly to the phase inverter) everything looks fine until the grids of the output tubes are driven to conduction, that's when the sidebands start to appear.

Yet, If I measure the ripple on the first pwr supply filter cap, I do not see its ripple suddenly increase when the sidebands start to appear. It is as if something about the output tubes clipping that makes them more vulnerable to the normal ripple on the power supply. I do see that the pos peaks of the output tube plate waveforms start to show noticeable low freq (120Hz) behavior during the time that the output tube grids are clipped (conducting). But yet, I do not see any additional pwr supply ripple at the first filter cap (where the OT center tap is connected) once grid conduction starts occuring.

In a similar fashion, as I turn up the input signal (to the PI) the output tube screens start to show ripple (120Hz) during the time that the output of the amp clips (due to the grids of the output tubes being driven as far as they can go, and start to conduct). That is, during the time when the grid signal is flat (clipped) the negative peak of the output-tube screen voltage waveform ( measured at the tube socket) shows a little negative-going dimple that ripples at 120Hz (or maybe 60Hz, hard to eyeball it). If I back-off the signal generator a litttle, so that the grids no longer conduct, the ripple at the screens, and the ripple at the plates dissappears, and so do the side-bands in the FFT of the amps output.

[overtone]

I think Firestorm has a good point there.

Suggestion:
If you are adamant that you do not want to increase the first stage of filtering, then consider adding some series resistance to the choke - as a test. Calculate your wattage, or just test it with a 10W sand-block around 330-860 Ohms or so.

Side note:
The CBS schematic has 1000pf snubbers at the PA plates

Got to run,
best, tony

[Craig B]

AARRG--Had the same symptom a while back. After trying everything electrical I desperately tried mechanical troubleshooting. Wiggling the tubes while getting the distortion (needs three hands) helped me find a bad tube socket. Good luck--you have paid mucho dues and its time for success!
Craig

[jelle]

Is it possible that the two halves of the primary of the Output transformer are not matched well and that the common mode ripple rejection breaks down? This you would not see at the center tab.

[Tillydog]

Have you tried changing the 2k2 resistor to a 22K (the one between the screen node and the PI node of the B+ rail)?

...I changed the 2K2 resistor to 10k last week.
...I tried doubling the screen filter capacitance - it had no effect.

...If I measure the ripple on the first pwr supply filter cap, I do not see its ripple suddenly increase when the sidebands start to appear.

I think you need to look at your PI supply (your node 'C') to see if you have sufficient filtering (even with your 10k dropper). If you're seeing the 120Hz ripple on the PI output (especially with the NFB disconnected) it's not from your output stage. Check also that your PI ground (& bias supply ground FWIW).

0.02p

Andy

[pula58]

Is it possible that the two halves of the primary of the Output transformer are not matched well and that the common mode ripple rejection breaks down? This you would not see at the center tab.

51.3 ohms and 56.7 ohms blue to CT and Red to CT. These values seem reasonable.

[pula58]

I think Firestorm has a good point there.

Suggestion:
If you are adamant that you do not want to increase the first stage of filtering, then consider adding some series resistance to the choke - as a test. Calculate your wattage, or just test it with a 10W sand-block around 330-860 Ohms or so.

Side note:
The CBS schematic has 1000pf snubbers at the PA plates

Got to run,
best, tony

Thanks for your post Tony
I don't know what you mean by "PA plates?

I tried increasing the filter cap for the output tube screens (I doubled it) and it had no effect. Wouldn't putting resistance in-series with the choke be equivalent to upping the screen filter capacitance? Or, are you thinking that the screen V's are too high and they need to be dropped a bit?

[pula58]

Have you tried changing the 2k2 resistor to a 22K (the one between the screen node and the PI node of the B+ rail)?

...I changed the 2K2 resistor to 10k last week.
...I tried doubling the screen filter capacitance - it had no effect.

...If I measure the ripple on the first pwr supply filter cap, I do not see its ripple suddenly increase when the sidebands start to appear.

I think you need to look at your PI supply (your node 'C') to see if you have sufficient filtering (even with your 10k dropper). If you're seeing the 120Hz ripple on the PI output (especially with the NFB disconnected) it's not from your output stage. Check also that your PI ground (& bias supply ground FWIW).

0.02p

Andy

Thanks for your post Andy.
The outputs of the PI don't show the sidebands (test tone is 300Hz, the sidebands are at 300-120 and 300+12) when the feedback is disconnected.

The weird thing about this is that the supply ripple at "A" and "B" (and "C") don't get suddenly larger as the output tube grids clip. So, there is no excee ripple or anything like that.

The only thing that makes sense to me is this cooked-up theory. When the grids clip, during that time there is no feedback loop in terms of the PI and the output tubes. That is, once the grid clips, if the PI tried to drive it higher it can't, so, the feedback loop is effectively broken. With no feedback, perhaps the output tube 120Hz (on the plates) is no longer suppresed by the feedback.

But if this were true, then all my amps (with the same circuit) would do this behavior. But, they do not.

[Tillydog]

The outputs of the PI don't show the sidebands

Sorry, misread a previous post as the PI output showing a 120Hz ripple, but it was your output screens. My mistake.

Good luck!

Andy

[Firestorm]

But if this were true, then all my amps (with the same circuit) would do this behavior. But, they do not.

Have you compared the level of drive signal relative to plate voltage in your previous amps to this one? There is a point where the positive going signal cannot drive the plate any higher- it's limited by the supply voltage - and the negative going signal cannot pull the plate any lower - it's limited by the saturation voltage of the tube type.

I haven't thought through the implications of trying to exeed this (except that clearly you shift into class AB2).

[jelle]

Is it possible that the two halves of the primary of the Output transformer are not matched well and that the common mode ripple rejection breaks down? This you would not see at the center tab.

51.3 ohms and 56.7 ohms blue to CT and Red to CT. These values seem reasonable.

The resistance is not relevant here, it is the impedance that counts.

[Structo]

PA=Power Amp

I know this is a shot in the dark and may not be anything but I was wondering why you have your filter caps with the negative side of the caps towards the tubes?

Then you have your B+ wires draped along the control side of the amp.
I may be confused by the color coding but I would change that.

I'll bet if you clean up the lead dress and keep power wires from signal wires you may be able to stop the harmonics.

I had a similar problem with one of my D type amps.
I had done a not so good job on the wiring.
I ripped it all out and redid it like Howard and the noise went away.

Like I said, shot in the dark but lead dress can really screw up phasing and cross talk.

[overtone]

Yes Structo is correct, by PA I mean the output section, power amp.

I noticed that you had no improvement by upping the filtering of the screens, but I am not suggesting upping the uF of the second filter.
It is the first reservoir filter value and the choke that are your friends here.
You noticed a half-improvement by doubling the value of the first reservoir cap of the CLC.
You can also improve the ripple rejection by adding a resistance in series with the choke.
Either before or after, it does not matter and is easy to do.
Maybe you can get the 120 Hz modulation out without stiffening the amp too much or even adding any more uF than you want.

You can see it work in PSU designer too if you fudge the resistance of the choke.
Merlin Blencowe has also written about this in his book on amp power supplies.
Worth a quick shot, I think.

But it would be great if it was only lead dress, try that suggestion too.

All the best, tony