Paul Ruby / Chuck H Zener Mod Question -- What Am I Doing Wrong?!?

[bob p]

I've got some questions about clamping cathode voltages using the zener mods. From what I've read the use of diodes to clamp cathode bias voltages was first described in R.E. Lang's 1960 patent, and recently we've been talking about the use of Zeners instead of regular diodes in the cathode clamping circuit in order to provide what I think of as "transitional bias" or "hybrid bias." Essentially we're talking about cathode bias at low drive levels which progresses to fixed bias at high drive levels. In these circles I think the idea was first brought up by Kg, but it's become popular due a lot of work that Chuck H has put into refining the mod.

My questions are about the Ruby / Chuck Zener mods for a cathode biased EL84 application. I'm hoping that someone who is familiar with Paul Ruby zener mods and the Chuck H zener mods can offer some insights. Today I 'scoped the amp to determine what value of zeners I need to use with my existing tube compliment, but I'm just not seeing the things on the scope that I think I should be looking for when 'scoping the amp to determine the Zener values. The specific test bed that I'm working on is an Orange Tiny Terror. A schematic is attached for reference.

The circuit is bone-stock. The B+ supply rail puts 317 VDC on the EL84 plates.
The cathode bias circuit is comprised of a 120R + 220uF passive elements.
The quiescent cathode voltage is 10.9 VDC.
Effective voltage across the tube is 317 - 10.9 = 306.1 VDC.
These voltages put us right in the sweet spot for chimey EL84 tone.

I followed Chuck's recommendations in this thread and other threads on the forum -- I tracked the rise in cathode voltage as I increased the amp's output, watching for the onset of crossover distortion in the output so that I could choose the cathode ZV and then determine the Ruby ZV which should be a couple of volts higher. Much to my surprise, the crossover distortion that I was looking for never happened.

What?!?

Here's what I mean: the following scope snapshot shows a top trace that corresponds to one of the EL84 input grids, and a bottom trace that corresponds to the 10R resistive load that's connected to the amp's 8R output jack.




When the snapshot was taken, the amp was being dimed. The cathode voltage had risen from 10.9VDC quiescent to 19.3VDC with the PPIMV wide open and a reading of 15.45 VAC on the 10R load. Where's the crossover distortion?

I think that part of the reason that I'm not seeing it is that I'm intentionally driving the EL84 grids with a fairly clean signal. I'm driving a low amplitude sine wave into the amp's input jack, at values anywhere from 200Hz to 1000Hz. When the photo was taken the amp was being driven at 500Hz. The MV control was set wide open, with the gain control set to minimum in order to provide as clean a signal as possible out of the very gainy preamp section.

This type of waveform, with a sine wave that was beginning to look a bit triangular, persisted as I continued to drive the amp with a larger and larger input signal. When driving with a fairly large sine wave, the amp still looked like these scope traces when running full-tilt. Of course, as soon as I turned the gain control open on the preamp all hell broke loose on the output; everything went square because I was feeding a highly clipped preamp signal into the EL84 grids, and they reproduced it well.

What surprised me about this is that although everyone has referenced rather ugly crossover distortion in EL84 cathode biased amps (especially Paul Ruby), I'm just not seeing it in this test specimen. I'm seeing plenty of harmonic distortion, where the sine waves are becoming peaky and a bit triangular instead of nice and round, but I'm not seeing any of the crossover that I was expecting.

Of course, in the absence of crossover it becomes difficult to determine exactly how to size the cathode clamping zeners or the Ruby mod zeners. What am I missing???

Personally, I'm having doubts about my test measurements and I'm thinking that I must be doing something wrong -- it's hard to imagine that I just got lucky, and that the amp just happens to be operating with sufficiently low plate voltage, and high enough bias current to have eliminated the crossover distortion from the output signal. For that to happen this specimen would have to be sitting right on the "ideal" point of push-pull class A operation, where the tubes are biased exactly in the middle portion of their transfer characteristics, so that neither tube cuts off during any portion of the cycle.

Did I just get lucky or am I completely missing something? Either way, with these results I'm having a hard time deciding what sort of ZV might be useful. As it stands now, I can't see how the Zener mod would really be helpful.

Anyone?

[bob p]

Here's another try at the schematic upload.

[g1]

Can you try a bigger input signal so you clip the output more? I believe the x-over is supposed to increase with power amp clipping.

[bob p]

Thanks for posting.

Yesterday I tried overdriving the EL84 input but the results didn't come out as I had expected --

Driving the input to the amp with a larger signal from the signal generator doesn't really help much, because I've already got the output stage dimed, and increasing the "gain" control on the amp to allow more signal through introduces such a nasty preamp signal to the EL84 grid that the waveforms don't seem to be all that useful for diagnostics. All that I get is unrecognizable hash (see description below) that eventually causes the output wave forms to square up when driven hard enough. Even with ridiculous amounts of overdrive that cause square wave clipping on the outputs, I'm just not seeing the zero notching that I expected.

What's odd is that I never saw this type of signal on the output:



^ The above picture is not one of mine -- is a captured snippet from Paul Ruby's article on his diode mod. His photo clearly demonstrates crossover distortion in what he describes as a "good" 18 watter on the left and a "bad" 18 watter on the right. My UUT didn't look like that. I was never able to get the obvious notching at the zero crossover no matter how hard I drove the amp. Even when I drove the signal hard enough to fully square up the output, there just weren't the crossover notches I was expecting. The rises and falls were always smooth.

The last time that I looked at this topic was several years ago (I've been sidelined for a while). To get a better handle on how much EL84 grid drive was needed to do the Chuck H mod, I asked Chuck about drive levels in this thread. I was trying to get a handle on the amplitude of the EL84 drive signal that's necessary for doing the ZV measurement, relative to the reference specs for the tube, to determine a working test point. I never got a definite answer about the necessary EL84 input drive level, so I'm sort of flying blind here.

What's odd is that I'm not seeing the crossover appear no matter how hard I drive the amp.

I was never able to replicate the notching that I was expecting to see based upon Paul Ruby's article, or based on Chuck's comments. I never got that sharp jagged crossover at all, even when flooding the EL84 grids with highly distorted preamp signal. With a dirty, highly distorted preamp signal what I did find was severe asymmetric distortion, where the rising face of the sine wave developed a very steep slope to a narrow peak, and the falling face of the sine wave smoothly developed a progressively shallower slope leading up to the crossover point and going negative to form a much wider negative peak. The jagged steps at the zero crossover distortion never appeared.

When I drove the signal even harder, to the point that I clipped off the asymmetric peaks, I was able to get square waves that looked... well... SQUARE. Rapid rise, clamped peak, rapid fall, no crossover notching. I'm stumped.

Unfortunately I wasn't able to complete the job of taking photos because I was driving the amp so hard that I sacrificed the miniature T250mA HT fuse during my drive tests. That put an end to my testing. I haven't been able to find the 5x20mm replacements locally -- my part of the world hasn't changed over to mini-fuses. So now I'm in a wait state pending a mail order parts shipment. I thought this down time would be a good opportunity to ask why I'm not able to replicate the crossover notches, even though I'm driving the amp hard enough to pop the HT fuse.

This amp has never been buzzy sounding to me, so I guess my inability to find notching under hard drive reinforces what my ears have told me about the amp -- it's character is one of cascaded preamp gain and the output just doesn't have a severe crossover problem. If that's the case then doing the PR zener mod might not help all that much (in terms of de-buzzing a non-buzzy amp), though I guess it should help to tighten up damping at high drive levels. I can see where the Chuck H mod might still be helpful in order to change the feel of the amp to fixed bias at higher drive levels, but in driving the amp I never found a clear indication of what the ZV should be. I'd really like to get a better handle on understanding of the data, rather than proceeding through blind trial and error.

thanks.

TACH=CONFIG]39712[/ATTACH]

[mhuss]

I suspect the preamp is clipping/limiting the signal before it gets large enough to really slam the PI (and then the outputs). If that's true, then the crossover 'nasty' described in the mod is a moot point for the TT. If you want to experiment more, perhaps try injecting the sine wave signal at the point between the 68k and .0001, bypassing the first stage altogether.

[loudthud]

Cold bias of the phase inverter would limit the positive excursions of it's output. Try tweeking the 1.2K resistor downward. The 47K tail resistor might also be limiting headroom of the PI.

[Chuck H]

Bob,
I think you have a few things skewing your results. One is that the TT stays closer to class A operation than many 2xel84 amps. It can do this because of the lower plate voltage. That helps a lot WRT crossover distortion. Also, the TT has a 500k pot load parallel with a 220k resistor load for each grid, for a grid to ground resistance of only 150k as compared to the average 2xel84 amp, usually between 220k and 470k. That makes for a lower time constant discharging the PI coupling caps. That reduces crossover distortion too. BUT...

You report seeing a cathode voltage rise of 9V with no flattening of the wave form. That's clear evidence that the amp is cooling the bias and moving into class AB. Then there's the PI coupling cap value, indicated as .1uf on the posted schematic. That value would bump the discharge time constant back up quite a lot. How much? Dunno. Math is involved and I don't have the formula off the top of my head . My point is that the two niggles I listed so far may be moot.

So why don't you see crossover distortion? I think it's because you're driving the amp differently than a guitar would. If the flattening of the wave form has a tilt to it (most amps do) the signal will look to be forming a point that leans to one side before the cutoff and saturation clips make themselves apparent. This is very much like the wave form in your post.

Looking at the TT circuit I wouldn't expect the preamp to clip prior to the power amp, but perhaps just barely after. So why is your preamp showing a nasty wave form before the power amp clips? I think you may be driving it with too much signal and then attenuating with the preamp gain control. It's also possible that the funky dual 500k gain control allows the preamp to clip prior to the PI. In either case the power tubes aren't clipping before the preamp is clipping and that makes a proper measurement impossible. So do this:

Plug into the amp with the guitar full up. Five all the knobs on the amp and play. If the amp isn't distorting, turn up the preamp until it starts to. Note the level of distortion. Now plug the signal generator into the amp and adjust the output level of the signal generator so it produces about the same level of distortion. Use that signal generator level for all future testing.

Now, with all the amp settings and signal generator settings as described above, scope the power amp. Is it clipping? Scope the preamp. Is it clipping? With the test arranged as described is it possible to clip the power amp prior to the preamp? This is paramount to the issue at hand. Report your findings and we can continue from there.

I want to close by saying that you mentioned the TT has never sounded buzzy. So I suppose you're intention is to tighten up the response by reducing cathode bias sag when the amp is clipping, right? Ok, first, if you don't hear offensive buzzy tone you don't need the PR mod (even if crossover distortion were evident). Not all amps present offensive tone with crossover distortion. Ergo Paul Ruby's "good amp/ bad amp" scenario where both amps exhibit crossover distortion. I think the difference may be relative to symmetry of the crossover distortion, but that's another topic. So, with the information presented in this thread I think it's already possible to determine a cathode resistor zener value. Do this:

With the amp and signal generator set up as previously discussed, scope the output and adjust the amps preamp gain control until you see the onset of clipping. Measure the cathode voltage under these circumstances. Use a zener value one volt higher than your measurement. It's as simple as that and it doesn't really matter what the gain structure of the amp does. This will give you fixed bias for your clipped tones. For better or worse All things tone are subjective.

[Malcolm Irving]

... . The cathode voltage had risen from 10.9VDC quiescent to 19.3VDC with the PPIMV wide open and a reading of 15.45 VAC on the 10R load.
[ATTACH=CONFIG]39708[/ATTACH]

The bias shift from 10.9V to 19.3V is enough to shift the EL84 into class C, but that is not showing up on the output trace on your scope shot. The output wave on the scope is pretty close to a sine wave, but you are getting 15.45x15.45/10 = 24 watts output - doesn't seem to add up (it's only a 15 watt amp).

Is there any way there is some mistake in the way the scope is set up, i.e. you are not really seeing the proper voltage across your 10 ohm dummy load?

[Chuck H]

Even if the scope isn't set up exactly correct, which I assumed and actually do all the time when I just want to look at the wave form, the trace should still be representative of what's happening at the dummy load where the measurement was taken. So I also suspect the shown trace isn't representative of the same instance in testing where the bias voltage was raised. IME that usually doesn't happen significantly until the onset of clipping. Which the trace doesn't really show. So there's definitely some dirt in the test results. I tried interpret what I could from it, but clean test results would be better. No offense Bob! I've confused issues here with my own dirty test results on occasion. Members here have always been kind about it.

[g1]

Any chance there is an issue with the load or it's connections?
The high undistorted output Malcolm Irving mentioned plus the lack of x-over makes me think of the load.

[bob p]

Thanks for everyone's input.

Regarding the scope traces -- I cant re-do any tests to verify them because I'm waiting for the HT fuses to come in. It looks like I'll be stopped for a week or so.

Regarding the load -- it was nothing fancy. Just a 10R 100W ohmite vitreous enamel resistor connected to the amp's 8R output via 18 ga lamp cord and a phone plug. Voltage measurement on the load was taken with a handheld DVM clamped across the 8R output jack, not with the scope. Another handheld DVM was strapped across the cathode resistor so that I could get concurrent Vk and Vload measurements. No voltages were measured off of the scope, so I wasn't just doing something silly like measuring peak-peak voltages. I only used the scope to trace waveforms. While I have reason to doubt the scope traces I don't have reason to doubt the metered voltages.

My preliminary test used a 100mV / 1KHz signal at the amp's input jack. Master volume was dimed and preamp was cracked open. Quiescent Vk was 10.9 VDC. At the onset of flattening of the top of the 1KHz wave form Vk rose from 10.9 to 12.5 VDC. The output DVM concurrently read 11.8 VAC on the 10R load when Vk read 12.5 VDC. Doing the math, P = I^2 / R = (11.8)^2/10 = 13.9W. Seems reasonable.

I did try overdriving the amp to get nasty square waves, by overdriving the input with the generator and by unleashing the preamp "gain" control. I took it all the way, overdriving the input and diming all of the amp's controls. The result was square waves with a rapid rise, sloped peaks, and a rapid fall. This test blew the 250mA HT fuse and cut short my playtime. When the HT fuse blew the tranny was singing, the Vk meter read 19.3VDC and the output meter read 15.45 VDC.

I never saw crossover. I can't explain why. I can only guess that there was an error in my scope setup, and I'll have to look into that after the fuses come in.

Thanks again.

[teemuk]

All in all, IMHO, a cathode-biased amp with a post phase inverter master volume control isn't probably the best example of an amp that produces distinct DC bias shifting during overdrive. First, you need to turn that master volume control up to levels where power tube grids truly will get overdriven, draw current, and asymetrically clip the signal to enforce DC offset shifting.

...But then there's the cathode biasing feature, which both tries to "autobias" tubes to their proper operating range (fighting against any "dynamic shifts") and enforces DC currents to share the same ground return path through common cathode resistor, which further reduces effects of individual bias offset tilt of power tubes, which then would result to crossover distortion. Not to mention, it is probably biased very close to class-A so it needs moderately more overdrive to introduce similar magnitude of crossover distortion than an amp biased closer to class-B.

So yes, you need to ovedrive that kind of amp MUCH MORE to achieve your goal, and even then amps with "hot" cathode bias aren't best examples in displaying DC bias shifting during excessive overdrive. The fixed bias -scheme is much more sensitive to effects of dynamic grid-to-cathode (bias) voltage shifts and in it (at least not very often) the power tubes do not share a common high-ohmic resistor in their DC return path either so tube bias at each half of the push-pull can "float" more freely during excessive overdrive that disrupts intented bias.

[Mike Sulzer]

If you really want to understand this, I think you need to make some extensive measurements, and I am not sure it is worth it.

One thing to consider is that when the dc cathode voltage rises, the effect is a result of increases in both the plate and screen currents. If the increase in screen current is significant (and I suspect that it is) then this turns the tube on more. That is, you move to a different set of plate characteristics with more current at the same control grid voltages. This could counteract the tendency for the tube to turn off, and thus prevent crossover distortion. One check on this would be to double the size of the screen resistors and see what happens. I suggest increasing them a bit in any case: your dissipation is getting up there.

[Malcolm Irving]

... the effect is a result of increases in both the plate and screen currents. If the increase in screen current is significant (and I suspect that it is) then this turns the tube on more. ...

There would need to be an increase in screen voltage to turn the tube on more. But when screen current increases, the screen voltage drops.

[Mike Sulzer]

There would need to be an increase in screen voltage to turn the tube on more. But when screen current increases, the screen voltage drops.

I agree, an do not know what to make of this.