Strange pulse of square wave into cap load

[kward]

Given the changes in coupling but still seeing the issue really leads me to believe it’s power supply related. Another indicator of a PS issue is both channels oscillate in sync--but were both channels loaded identically on their outputs, or did you excite only the loaded channel while leaving the other channel unloaded?

If it’s still happening when feeding a signal directly into the power amp gain stage then there’s not a whole lot more to check to isolate it. Some additional things to try:

Remove voltage amp tube from power amp section and retest (driving the amp from the grid of the inverter tube).
Remove voltage amp and inverter tubes from power amp section and retest (driving the amp from the grid of one of the output tubes)
Reinstall gain and inverter tubes but remove feedback (temporarily) and retest (driving the amp from the grid of the voltage gain stage--careful! amp will be super sensitive w/o feedback).
Try clipping in a smallish (say 22 uF/450V) cap in parallel with one cap in the PS at a time, but ensure the clipped in cap has a solid ground connection on the "noisy" end of the ground bus and retest by driving grid of voltage amp directly. Careful with paralleling that cap with the upper cap in the PS doubler. It's negative terminal is not at ground potential, so the clipped in cap should go from the pos terminal of that cap to the other side of the secondary. This test is simply verifying each PS cap is working correctly and has a solid ground connection. If one of the caps has a bad ground it's effectively out of the circuit.

There might be a few more tests you could conger up to try to isolate which stage is causing the issue....

Also stupid question but I gotta ask. Is the positive feedback between first and second power amp stages connected correctly with the correct resistor values? Here:

 

[grindfix]

Given the changes in coupling but still seeing the issue really leads me to believe it’s power supply related. Another indicator of a PS issue is both channels oscillate in sync--but were both channels loaded identically on their outputs, or did you excite only the loaded channel while leaving the other channel unloaded?

the clip with both channels was actually completely identically unloaded and signal fed into Aux jacks into both channels. I thought it was easier to see on the screen without a load but then I discovered that it does pretty much the same into cap only load as well.
Regarding cathode side of voltage amp / inverter. When I was initially checking receiver out, I went through all resistors except tuner section and didn’t find any out of their tolerance. I will check tomorrow to see if there’s a connection error in those circled but I doubt it.
I’ll run those tests with additional cap and will follow up.
Thanks guys!
Be safe

 

[GordonW]

I get that part. What I'm saying is you have too low of a rolloff point. If you want to keep low frequency oscillation from occuring it needs to be rolled off more if it's there for some reason, so you need to start at a higher frequency for the -3dB point. More like 20hz, not 3hz or 6hz or 12hz. That's a start. Then do you even have an option to seperate the PS into another filter and tap point for decoupling the stages? Adding a resistor and filter cap to a stage will change the plate load and voltages. Probably the best solution though if that is the source of the LF oscillation.

Good luck, stay well.

Well- the point is, that every other example of these amps worked for DECADES without issues, with these cap values and rolloff points. It was made with those values from the factory. And this is Harman-Kardon- who did not generally do random design screw-ups, too.

Something else has to be wrong- so I'd be very dis-inclined to go too far from stock cap values.

Regards,
Gordon.

 

[grindfix]

Answers to Kevin's suggestions.
I tried bypassing every PS cap with 22/500v cap without improvement.

Removed voltage amp tube and fed input into grid of inverter, still doing it.

Removed both driver tubes and fed signal into grid of one of the output tubes. Result was so garbled that I really can't tell if that pulsing is still there or not. I may have to redo that again.

Removed feedback and reinstalled driver tubes. Waveform looks like garbage but pulse is there. Something I noticed, with feedback disconnected, this pulse stops completely if dialed at following frequencies: 24, 40, 60, 80, 120, 160, 200, 240, 280Hz. I did not go further as I definitely see the pattern.

Another thing noticed today that pulse does not start immediately. It takes a minute or two and sometimes one channel starts pulsing and another one catches up within a minute. Sometimes opposite channel seems to start first. Both stop pulsing as soon as power switch is turned off while signal is still present and wave is going down.

Driver tubes cathode resistors are connected correctly with one exception that I forgot about from my previous testings. I even wrote this discrepancy on schematic but I have printed several pages during several months working on this HK that I did not see it until re-tested this evening. Voltage amp 330ohm cathode resistor is in fact 270 ohm as installed from the factory. Red paint is visible on the connection of these resistors to the ground bus. Both channels same. Red, violet, brown and gold. Could that affect the stage biasing to produce this oscillation?

 

[kward]

Removed voltage amp tube and fed input into grid of inverter, still doing it.

Removed feedback and reinstalled driver tubes. Waveform looks like garbage but pulse is there

Both [channels] stop pulsing as soon as power switch is turned off while signal is still present and wave is going down.

These above statements lead me to believe that the problem is in the output stage. It's hard to imagine how that could actually be, but the evidence points to it. Where did you get your output tubes? I wonder if the output tubes themselves are the cause of this? I'm not sure if the 7355 has beam forming plates or if it is a true pentode. I think they are beam forming plates for the "screen" function. But if the screens aren't stable, it could cause what you are seeing.

One test for sure to validate is to see if you can prove the signal is stable (not pusling) as it comes out of the inverter. By disconnecting feedback and then pulsing the amp as you have been, what does your scope show if probes are put on the outputs of the inverter vs on the speaker outputs? If the scope shows stability on the outputs of the inverter, but pulses on the speaker outputs (with FB disconnected) I think you know the problem is in the output stage or in the power supply section feeding the output stage (either screens or plates).

Something to think about anyway...

 

[grindfix]

These above statements lead me to believe that the problem is in the output stage. It's hard to imagine how that could actually be, but the evidence points to it. Where did you get your output tubes? I wonder if the output tubes themselves are the cause of this? I'm not sure if the 7355 has beam forming plates or if it is a true pentode. I think they are beam forming plates for the "screen" function. But if the screens aren't stable, it could cause what you are seeing.

One test for sure to validate is to see if you can prove the signal is stable (not pusling) as it comes out of the inverter. By disconnecting feedback and then pulsing the amp as you have been, what does your scope show if probes are put on the outputs of the inverter vs on the speaker outputs? If the scope shows stability on the outputs of the inverter, but pulses on the speaker outputs (with FB disconnected) I think you know the problem is in the output stage or in the power supply section feeding the output stage (either screens or plates).

Something to think about anyway...

This morning, as I was re-running my tests in my head, I was thinking that what if screens are causing this and get close to arcing point. I don't have stability resistors installed as I thought they are not necessary in cathode biased output stage. But this one is not a fully cathode biased unit. Even less of that after zener installation I think.

7355 tubes are original to this unit. They are Harman Kardon branded. IBAM biased to 52mA with 3 x 11v zener diodes across cathode filament string.

Kevin, just out of curiosity, have you ever measured feedback level in your A500?

 

[kward]

I have not. And I have modified mine from stock so I will never know what the original feedback level was on mine.

 

[grindfix]

I think it's scope sync-lock problem. Either the level or line/TV or scope timebase... The signal is shifting horizontally. Like a heartbeat.

... Is the scope and the sig gen both using the same signal ground point?

Scope test leads are grounded at the amp chassis. Signal generator is connected to Aux input via RCA Y-adapter. Do you mean scope needs another ground? I believe there is a ground lug on the scope but I've never used it and this is the only amp that behaves this way.

 

[GordonW]

Rethinking this- I wonder if anyone else has one of these amps, and could check if their amp does the same thing.

I know that's a long shot- but it would tell us if it was something endemic to this model, or a problem just with this unit. That tells us a lot of what needs to be done... it'll be a different solution, depending on it's just the one, or all of them that do it...

Regards,
Gordon.

 

[grindfix]

A500 would probably qualify for this test as well. Same circuit minus tuner stuff. Schematic for A500 is drawn a little differently but I think all voltages and component values are same.

 

[GordonW]

The A50K and A500 have different output transformers (FT3824638) than the TA5000X (FT4725698). It's unlikely that they would behave in an identical fashion.

I've never observed instability of any kind on A50Ks or A500s.

Regards,
Gordon.

 

[whoaru99]

it does it with cap only load
It does it with output open
It does it with cap + 8ohm in series.
Cap + 8ohm in parallel doesn’t bother it.

Is an amp expected to produce a proper square wave into a singularly or largely capacitive load? I'm a rank amateur compared to those otherwise replying but your test sure seems out of the norm to me.

What value cap?

 

[grindfix]

Is an amp expected to produce a proper square wave into a singularly or largely capacitive load? I'm a rank amateur compared to those otherwise replying but your test sure seems out of the norm to me.

What value cap?

At such a low frequency (20Hz) proper square wave is not possible even into resistive load. The goal is that square wave signal puts amp onto its toes and any possible instability will show-up quicker where it could possibly be missed using sine wave as a stability test input.

 

[grindfix]

It’s doing it out of inverter with feedback disconnected.

here is the clip of what screen is doing regardless of the signal presence. All scope settings can be seen too.

 

[GordonW]

Is there any way to try a different scope? Just to rule that out...

Regards,
Gordon.

 

[grindfix]

I can borrow identical scope as mine from a friend but I think scope is not an issue though. My scope stays "on" while amp is getting turned "on" and "off" while I'm trying different methods to tame this condition. Every single time amp is turned "on" and signal is present - waveform starts to "dance" after about a minute or two. If I remove signal and let flat line stabilize, which takes just a few low bounces, then turn signal back on - "dance" resumes right away. So it seems something within the amp causes it when amp reaches normal warmed-up condition. It never does it immediately as signal starts to appear on the output during the warm-up.

I feel that this below is what hiding the answer:

Something I noticed, with feedback disconnected, this pulse stops completely if dialed at following frequencies: 24, 40, 60, 80, 120, 160, 200, 240, 280Hz. I did not go further as I definitely see the pattern.

 

[gadget73]

and it doesn't do this direct out of the sig gen? What about at the output of the voltage amp?

 

[grindfix]

and it doesn't do this direct out of the sig gen? What about at the output of the voltage amp?

It doesn't do it out of generator. I think I scoped voltage amp output and it was doing it but that was before I removed feedback. I'll try again tonight.

 

[grindfix]

I borrowed scope from a friend. It is identical to mine but it has all of it's switches and pots oxidized and it's extremely difficult to get waveform to stop moving across the screen.
But the same dancing pattern on the output is there.

After going between scopes, I think it IS signal generator after all! If I scope signal generator output directly, without it being connected to the amp - there is no pulse. But if I scope at the back of AUX RCA jacks, where signal generator is connected, I see same pattern. It's on much smaller scale and harder to see at such a low level without feedback but it responds to the same frequency pattern I mentioned earlier. My guess is that scope input impedance is minimal and this "dance" is not showing itself when scoped directly but when amplifier input impedance is loading the generator signal it's easier to see. Then each amplification stage amplifies it and it becomes much easier to see?

I will continue tomorrow and will try with feedback re-connected. I think that will allow me to scope input and output simultaneously so I can take a short video.

 

[grindfix]

Actually, I think I have an idea. I’ll borrow signal generator from my friend tomorrow and will run each amp channel from a separate signal generator and see if both channels are pulsing on the output in sync to each other.