18 watt fizz: please help me diagnose this?

[JdJ]

I would do that first and see how much it sorts your signal out.. might not, but at least you then know for sure its the amp and not the Scope/probe causing all those spikes.

I hear ya man. I was sort of hoping for the instant gratification of the scope... which made for a hasty introduction for sure.

The probe is now compensated for- there was a slight under compensation before. Unfortunately (as my ears were telling me) the fizz is alive and well.

Thanks for the suggestions.

Tom - now that I've completely hijacked your post for my own purposes - any luck scoping your amp out over the weekend?

Cheers,

Josh

[JdJ]

I'll be interested to see how your next tests go and how things look with a resistive load vs the inductive one. Also just humour me and move things around in the power stage & OT with a chopstick with the spike up on the CRO and see if it changes. I don't think it will but just in case.

I chopsticked as you recommended with the spike on the CRO - no change. I did notice some oscillation on the EL84 plates - I'm not sure if it was my time base on the scope, or something weird, but the waveforms seemed to be oscillating vertically 1/2 a DIV - same with the probe on the speaker jack.

Here they are - without the Zeners in this time & back to the 820ohm blocking resistors on the output grids.

PI Plates




Non Inductive 8ohm Load


[/img]

[TomOlsen1]

Hi guys

Well, there is good news. We hooked up the amp to the scope and followed the method outlined in the Paul Ruby doc. The trick was the speaker impedance load test. And our scope gave us the same crossover distortion read out with the spikes. So with this knowledge I headed out to the electronics store to buy the diodes.

And it is sounding much better!! I still haven't found the best values for all of the points on the VVR, but the fizz is mostly gone. Is there a rule of thumb for the values of the zeners compared to the cathode voltage on the el84's?

I have yet to try the huge cap trick, but so far the Paul Ruby mod is spot on. Thank you all!

And Josh, it's all good, I hope all our amps are in top condition soon.

Tom

(Also, I swear the amp has got even louder after the fix????)

[katopan]

Tom - Good news!

I did notice some oscillation on the EL84 plates - I'm not sure if it was my time base on the scope, or something weird, but the waveforms seemed to be oscillating vertically 1/2 a DIV - same with the probe on the speaker jack.

The vertical oscillation is likely to be the 50/60Hz hum that the test frequency signal is riding on. Nothing to worry about - just makes accurate measurements a challenge unless you've got a dig CRO with waveform hold button.

Here they are - without the Zeners in this time & back to the 820ohm blocking resistors on the output grids.

Hope you mean 8k2! 820ohm wouldn't block much.

Josh I hate to say it but your PI plates look really good, exactly as they should. Output has more than normal crossover distortion which spikes with a speaker load. Looks a lot like what Paul Ruby's doc shows. Although he is a champ for sharing the fix with everyone, it's a shame at the end that it's not clear why the "PAM is magic". Anyway, you tried the zeners with no change right?

[stevesuk]

Tom,
The same as the EL84's Cathode voltage or perhaps 1 volt above.

Steve UK
www.valvepower.co.uk

[TomOlsen1]

Thanks Steve!

The amp is sounding HUGE. I am so happy, and will be even more so once I've tweaked the zener values. I'm going to install a dual wafer rotary switch on the back of the amp to select the appropriate zener value according to where the VVR is set. But so far so good. And input two is really smoking now, whereas before the fix it was unusable.

The only real problem is that the amp is making many of my dirtboxes redundant.

Tom
(p.s. I can't wait to use it in a session.)

[JdJ]

Great news Tom! So the waveform resembled my spikey speaker loaded examples?

Hmmm... does 1/2 watt vs 1 watt make any difference? I closely followed the schematic & layout - bands on the diodes facing each other - 4007 to ground zener to +. I'd say they are working judging by my PI traces from earlier, right?

Hope you mean 8k2! 820ohm wouldn't block much.

Guilty as charged 8k2 - I'll edit for clarity...

Josh I hate to say it but your PI plates look really good, exactly as they should.

That's what I thought too. They closely resemble everything else I've seen so far.

Anyway, you tried the zeners with no change right?

Unfortunately, yes. 9.1v, 15v & 16v (have 3 sets made up for quick installation).

Could the screen some how be affecting crossover distortion? I am still using a 100ohm shared resistor. Maybe I'll try a 1k.

I also want to go back to a lower value cathode resistor for testing. If it is grid conduction, or just purely biased too cold the lower value (higher cathode voltage) should noticeably reduce the distortion in either case...

Any other ideas?

[katopan]

From your first CRO images, the symmetrical PI plates showed that the zeners (when you had them installed) were discharging the PI to PA coupling caps in the other direction as they should. But your output still had crossover distortion, and that is where your buzz is coming from. Next step would be to really explore where the various PA voltages move under heavy signal vs. their bias value and work out how the crossover distortion is being generated.

[JdJ]

Next step would be to really explore where the various PA voltages move under heavy signal vs. their bias value and work out how the crossover distortion is being generated.

I'm not sure how to go about doing this - roughly the same idea as the power tubes? Should I be comparing Grid vs Cathode on the PA tubes with a signal? I was assuming that the PA & PI were okay based on my traces...

Thanks for all your help!

[katopan]

I knew you'd ask that. I've never had to to trouble shoot a buzz on my own amps, so if anyone else wants to jump in they are welcome. But I'm happy to keep trying to help based on theory.

If I were in your shoes I'd be trying to work out why you've got crossover distortion buzz that the zener fix doesn't fix. I'm trusting that you put the zeners in right (just double check), and as SteveS said you want diodes with a zener voltage of about 1V above the cathode bias voltage.

To see the mechanism of crossover distortion in action set up your CRO with both channels probing each of the EL84 grids. With no signal adjust the trace vertically to set them to the same zero point (traces on top of each other, lined up with one of the display grid lines). Now as you turn up the signal the two waveforms should be opposite phase to each other and the point that they cross over each other should be around your zero line. As you turn up the signal and the grids start clipping, if the coupling caps are charging up you'll get a shift downwards of the crossover point. Make a note of how many volts it goes down at full signal. Then move one of the probes to the cathode (leave the vertical trace adj where it is) and see what it does with no signal vs. full signal. It's not clear from the published EL84 data what the grid voltage for cutoff is, but for 18W voltages it'll be something around -18V to -20V grid to cathode. So if the difference between the grid 'zero' (crossover point of the signals) as it moves and the cathode voltage as it moves gets more than the cutoff voltage, it'll result in crossover distortion. Earlier you said you had 12.9V on the cathode - so it'll start distorting if the combined movement of grids and cathode is any more than cutoff minus 12.9V, or 5-7V depending on your EL84s. The Ruby doc said he was getting grid movement of 17V. Anyway you'll see the point when that happens. The thing to check is it mostly from movement of the grids or of the cathode voltage. The Ruby zener fix is all about grid voltage movement, and won't do anything for cathode voltage movement.

Hope I haven't lost you with all that.

[JdJ]

Katopan - I truly appreciate the amount of time you've spent posting and thinking about this problem. I owe you several drinks at the very least.

I will check again that my zeners are in properly, but I am 99.9% sure they are.

I'm pretty sure I get what you are saying. I am basically looking for the grid conduction point and the resulting bias shift as well as what the cathode voltage is relative to the grid at various points. I'll give it a shot although I only have 1 probe right now. I think I should still be able to do it by switching the probe back & forth while noting where the zero crossing is.

I'm not sure when I'll have time to experiment - but I'll post back when I do.

Many many thanks!

Josh

[JdJ]

I managed to find some time tonight to check out what's going on per Katopan's suggestion. I measured the grid dc bias and cathode voltage over a range of volume settings with a 333hz 400mV test signal.




My estimation is that the power tubes are going into cutoff in between 2 & 4 on the volume knob with this test signal. Weird! I wonder why I am having such a severe case of Rubyitis?! Still, I would expect the zeners to take care of this problem... (they are not in as of this test)

[zaphod_phil]

But your output still had crossover distortion, and that is where your buzz is coming from.

Yes, but 18W amps are *meant* to have a good bit of crossover distortion, as that what contributes so much harmonic richness to their tone. Let's be wary of focusing too much on crossover distortion, when the 18W buzz problem actually has more to do with blocking distortion.

[katopan]

Phil - Agree, in fact I was strongly pro 'crossover distortion adds to the tone' in the thread about all that a while ago. Have had thoughts on another mechanism for crossover distortion from the OT inductance kickback when one half of the primary stops conducting. Also been wondering if that has waveform and therefore tonal differences (different positioning, smoother transition) to the crossover distortion from cold bias / PI signal blocking distortion / rising cathode voltage. Had no time to explore all that on the bench. Anyway, I'm just trying to give Josh a way forward.

Josh - Now you've tested that, put in the zeners (ideally around 14V ones if you're cathode bias is still 13V) and see if anything changes. We're trying to work out why you're still getting visible inductive spikes and resulting fizz even with the Ruby zener fix in place.

[stevesuk]

JdJ,
Have you got a photograph of how your zener/diodes are installed ?

Steve UK
www.valvepower.co.uk

[zaphod_phil]

Also are you using 1/2W or 1W zeners? I suspect that 1/2W zeners will be more effective, as they will tend to operate more readily past their knee point. The lower wattage the better.

[JdJ]

JdJ,
Have you got a photograph of how your zener/diodes are installed ?

I don't have a photo currently, but I have now verified that they are in correctly both against other pictures, the schematic and through testing.

Also are you using 1/2W or 1W zeners?

They are 1/2 watt.

I don't have the chart that I made this morning with me right now, but the grid bias shift completely disappears with the zeners in!

I also went back to the original 130 ohm cathode resistor which is resulting in too much dissipation (13.5watts) but seems to be keeping the tube from cutoff longer (I guess that's obvious to everybody...) since the grid takes longer to enter conduction.

Strangely the problem is still there - both visually and on the scope. I will say that it sounds better, but still I hear the phasey fizz with the guitar wide open on anything above 6 or so on the amp's volume - although it isn't as bad. You can still see a spike on the scope both with the guitar plugged into the amp and with the test signal.

I'm starting to wonder if I should look into trying to get my B+/plate voltage down so I can drop the cathode current/voltage keeping it closer to class a without over dissipating. Still I feel like there's something I'm missing....

Thanks for chiming in Steve & Phil.

Josh

[katopan]

Now you've got this testing down, try out the big cathode cap and see what effect it has on the cathode voltage rise. Your chart had a rise from 14V to 28V.

Another thought just to try and rule it out - have you played it through a different speaker to the Vintage 10s you mentioned in the other thread?

[JdJ]

Now you've got this testing down, try out the big cathode cap and see what effect it has on the cathode voltage rise. Your chart had a rise from 14V to 28V.

Now with the 130ohm cathode resistor the rise is from 12 to 17.5. With the 15v zeners in the grid dc varies slightly + to just under 2 volts. I do notice that if I measure the grid dc on the other side of the PI, it measures 4 VDC. Enough to cause a fizz?

The 2200uF cap didn't seem to change the fizz much at all. But I do know I definitely prefer the sound and feel of the 500uF so I went back to it afterwards. I also tried a shared 750ohm screen resistor just for kicks instead of the 100ohm that was in there. I ended up changing that back too after it didn't do anything to alter the fizz.

Another thought just to try and rule it out - have you played it through a different speaker to the Vintage 10s you mentioned in the other thread?

I tried an 8 ohm speaker that's in an old music man. It attenuated the fizz a bit, but just because it is so muddy. The only thing else I had around is a SVT classic 410 which soundd horrible (surprise!) and still had the fizz. I also tried the Vintage 10s individually at 8 ohms which made no difference. I did notice the spike changing a bit on the scope when I ran the speaker selector at 16 with one of the 8 ohm V10s. It was attenuated a bit. A clue perhaps?


Anyone have time to hook up a scope to your amp with a 400mV 333hz signal and an inductive load to see if you get any spikes?

I'm fresh out of ideas (besides the lower B+ one which I have yet to try).

[JdJ]

Next update -

I tried strapping some 470k resistors across the existing pt grid ref resistors to get them down to around 235k - no luck. Next I put some 10k resistors in series with the existing 8k2 grid blockers - still nothing so I reverted to the stock grid ground refs and grid blockers.

Then I swapped the pt plate to transformer primary leads - now the nastier overshoot spike shows up on the negative swing at the OT secondary. There's still one 180degrees out too - it's just smaller. It's almost like I need a super fast limiter or something...

Then I replaced the PI tube with a lower gain 12ax7-wa (it had a new production Mullard in there prior). The spike is less pronounced but still quite present - same with a JAN 5751. This seems to make a bigger difference that the grid blockers/refs. That's kind of surprising to me, but I haven't tried to work out any math.

I'm open to suggestions if anyone has any ideas. I wish I had another OT transformer to try, but I'm not about to spring for another one. It has the Vintage GDS in there right now.

Could it be coupling caps between the PI and PT grids?

Thanks,

Josh