Lopsided Push-Pull 7591S Pair, 6U8A Splitter

[ktritty]

I have made a few mods to an Ampeg Gemini-2 G-15 Amplifier. I took out the vibrato circuitry, as well as the 'clean' preamp channel, and modded the reverb, all in an effort to simplify the amplifier and reduce noise by removing unused items.

(This amp now has 12AX7 preamp (follower-cascode), 12AX7 reverb (1/2 driver, 1/2 recovery), 6U8A booster/splitter (follower-cascode w/triode phase splitter), and a pair of JJ-7591S in push-pull)

All is going fairly well, and I have succeeded at making a few tweaks along the way to get the tone to be pretty nice in the clean range, and well-balanced reverb and much less idle noise than before.

For some reason, however, I am now having an odd problem when the amp goes up into the overdrive range. It goes from clean to overdrive much earlier than it used to at equivalent bias current, and one of the tubes gets hot (glows) along with the strumming while the other one remains constant or even cools off a bit if the volume is very loud. Also, the tone is shot, it is muddy overdrive and it is hollow (likely due to heavy push-pull phase imbalance).

I have done the following to troubleshoot:
1) swapped the push and pull tubes
2) swapped OT primaries
3) swapped cathode and anode following connections coming out of the 6U8A splitter tube.

The hot tube is always the one hooked up to the 6U8A anode output (push?), and the constant/colder tube is hooked up to the cathode (pull?).
So I don't think there is a problem with the OT (it ohms out okay too), nor with one of the 7591's. The problem occurs after trying several different 6U8's, and is more pronounced the higher the bias current is set.

I have done some testing with a DC volt meter at idle, and while strumming the guitar. The output from the splitter seems oddly asymmetric. I have attached a schematic. (Note: stock, the splitter triode has 55V cathode, 20V grid, 260V anode, with B4=320V. Also for 7591 stock B1=430V and bias is 40mA per tube but for the old style 7591's in the GT envelope). There is not a bad connection with my grid bias resistors, I have tested thoroughly, but oddly, the grid is always at 0V on the 6U8.

I appreciate any help in getting this circuit to operate in a more balanced fashion, and any pointers in making it more stable and less muddy sounding.

[Chuck H]

Try the same voltage testing with the power tubes pulled. It's hard to say what's responsible for the bias voltage imbalance otherwise.

Chuck

[gmoon]

Look familiar?

My Ampeg has the 7199 (vs the 6U8A), but check out this vid I posted in Feb, of the 7591's my Gemini II:
YouTube - 7591 tubes arcing and red-plating

Although I did a few mods (larger screen resistor, replaced the bias supply cap, replaced a few filter caps) that might have fixed this, the likeliest contributing factor was changing the tubes. 7591's seem to chug along fine after they're shot--volume-wise, but they're lifeless, and exhibit that glowing (the left tube is arcing, the right one is glowing.)

The behavior in the video is triggered by turning the tone pot only. And there's oscillation, attenuation and squeal the video camera doesn't pickup. It's stable at one setting, but too much volume or treble and it begins.

I pulled some old CMI 7591 tubes from an old Kalamazoo bass amp and the amp came alive--no "glowing" or arcing, with rich full tone. A different amp, really. So if you've only socket-swapped the same pair of power tubes, that might be the issue.

As noted, I did replace the screen resistor with a larger value one, since heat had obliterated the color code rings. But in-situ the old resistor still had the same value, despite the damage (tested cold, anyway.)

There was a tiny bit of glowing when the tubes were in the Kalamazoo amp, but replacing the greasy-looking bias supply caps fixed that...

[ktritty]

Although I did a few mods (larger screen resistor, replaced the bias supply cap, replaced a few filter caps) that might have fixed this, the likeliest contributing factor was changing the tubes.[/QUOTE]

Well, I hope the tubes aren't shot. They are quite new. And from day one, they have red-plated a bit anytime there is a hint of overdrive (even before my mods). Up until recently, they both did so evenly, and with great sound. Now, my situation looks pretty much exactly like yours from the video clip.

I am a bit skeptical of the tubes being the problem here since the right-hand one always goes red, even when they are swapped. Then, if you swap the leads from the 6U8 triode, the left one goes red (as I touched on above). My hunch is that there's something going goofy with the 6U8, that I need to change how it is setup somehow.

I, too, put in larger screen resistor (went from 470ohm to 1500ohm), with good results. One of the first mods I did to this unit. 2200ohm worked great too, but not much higher.

The stock schematic has B1 at 440V, but before I changed anything, it was at 510V (now 515V after mods). The other voltages were also higher than the stock schematic, and after my mods, they are all only slightly higher yet (within 20% at worst, for the preamp, which was one of my goals to get a bit more gain). I think this is because the PT was sized for 117VAC back in the day, and I typically get 125-128VAC from my house (close to a power plant). That accounts for most (9%) of the 12% hike across the board that I am measuring.

I will post again after doing the voltage measurements without the 7591's, and also I will dig out my oscilloscope and see what else I can figure.

[ktritty]

Try the same voltage testing with the power tubes pulled. It's hard to say what's responsible for the bias voltage imbalance otherwise.

Chuck

I will do that. One question first: When I pull the 7591's, shouldn't I hook a dummy resistor onto the lead that went into each control grid? If so, what value, and should it go to ground or some other connection? it seems to me this would help with accuracy to simulate everything without the 7591's.

[Chuck H]

There is no need for a dummy resistor anywhere when the power tubes are pulled.

But... I read on your last post that swapping the leads from the PI DOES make the OTHER tube red plate. Perhaps you mentioned it before but it got past me. This is suspect. But a drive imbalance wouldn't cause the red plating. So I suspect that something about the circuit on that end of the PI is altering the BIAS of whatever tube it's connected to. I'll look at the schem again. You may want to check the decoupling cap on that end of the PI for leakage.

FWIW I originally thought it may be related to the power tube socket. But if you can get the problem to change sides by swapping the PI leads then it's probably the PI circuit affecting the bias of the tube it feeds. So you really don't need to do the voltage test with the power tubes pulled. That was to determine a power tube socket or bias supply issue.

Chuck

[Chuck H]

Ok... I just looked at the schem again and something has me confused. If the bias supply is -25V (the B7 node says -23V but since you have -25V on the plate driven tube I'm guessing that it's acually -25V at B7 now) then how can you have -41 volts on the grid of the cathode driven tube?!? Are you sure that read is correct and it isn't +41 volts there? That I could understand due to a component failure. But additional -voltage?!? Where did that come from? Very confusing.

Check the voltage at the B7 node so we know if it's -25 volts.

Replace both the decoupling caps from the PI. I think one is leaking.

Chuck

Edit: Also, because you've always had red plating before, you should make the bias voltage more negative. If you can post the area of YOUR schem with the bias supply on it I'll tell you how to do it. The schem at 'Schematic Heaven' doesn't seem to match yours.

[gmoon]

I, too, put in larger screen resistor (went from 470ohm to 1500ohm), with good results. One of the first mods I did to this unit. 2200ohm worked great too, but not much higher.

The stock schematic has B1 at 440V, but before I changed anything, it was at 510V (now 515V after mods). The other voltages were also higher than the stock schematic, and after my mods, they are all only slightly higher yet (within 20% at worst, for the preamp, which was one of my goals to get a bit more gain). I think this is because the PT was sized for 117VAC back in the day, and I typically get 125-128VAC from my house (close to a power plant). That accounts for most (9%) of the 12% hike across the board that I am measuring.

That's exactly the screen resistor value I'm using now, too.

B1 in my amp is ~490V. I do have some zeners for the CT to put that right; still on the to-do list.

I think the socket behavior was identical, despite swapping the old power tubes (a-b, b-a.) I suppose I could reinsert the old tubes and check if the arcing/redplating stays on the same socket.

In my Gemini, the 7591 cathodes are connected to ground via a 100 ohm bias "balancing" POT, so I never thought too much about which tube was arcing and which was redplating. The POT wouldn't correct the problem, and asymmetry was assumed... Since the amp still functioned reasonably (to a point--it was loud, so the PI, OT, etc., were all working, AFAIK), new tubes seemed like the simplest thing to try first.

I'm sure there's some asymmetry in the phase inverter stage outputs, even when it was brand-new. How that "plays" on fried 7591's is beyond my ken...
----------------
edit:

I should also note that the amp itself had been modded (one preamp channel was routed into the other) and the 7591's hadn't been changed since the 1970s. Also, during the un-modding the arcing disappeared, and the old tubes were just replating. ( discussed in this thread. )

So I can't really retest for arcing anymore. But both tubes still replated above 3 o'clock when a loud note was played.

[Chuck H]

The important thing to notice on ktritty's schem is that the bias voltages are lopsided. Drive from the PI can't do this. Since this is the most likely culprit for the problem balancing the PI drive won't help much. You may also benefit from replacing the PI decoupling caps.

Though I think it's very suspect that you both have a similar problem and film caps going leaky isn't that common. It would be a hell of a coincidence. The schem at 'Schematic Heaven' doesn't seem to match the 7591 amp. Can one of you post a schem so I can look at the balance and bias circuits???

Chuck

[gmoon]

The important thing to notice on ktritty's schem is that the bias voltages are lopsided. Drive from the PI can't do this. Since this is the most likely culprit for the problem balancing the PI drive won't help much. You may also benefit from replacing the PI decoupling caps.

Though I think it's very suspect that you both have a similar problem and film caps going leaky isn't that common. It would be a hell of a coincidence. The schem at 'Schematic Heaven' doesn't seem to match the 7591 amp. Can one of you post a schem so I can look at the balance and bias circuits???
Chuck

There are several variations on the G-15--this is the 7199/7591 version and matches the one in the cab. Ktritty's will be different.

I checked the bias voltages right on the grids like Ktritty (at different volumes and with guitar input), and the values are rock-steady, ~-21V with only 0.1V variation between the tubes.

My tests were relative to the chassis ground, not the cathodes themselves, which have a small amouny of cathode bias "lift" from the bias "balancing" POT.

These tests were with the good tubes, of course.

[ktritty]

Ok... I just looked at the schem again and something has me confused. If the bias supply is -25V (the B7 node says -23V but since you have -25V on the plate driven tube I'm guessing that it's acually -25V at B7 now) then how can you have -41 volts on the grid of the cathode driven tube?!? Are you sure that read is correct and it isn't +41 volts there? That I could understand due to a component failure. But additional -voltage?!? Where did that come from? Very confusing.

Check the voltage at the B7 node so we know if it's -25 volts.

I think this is a result of my cheap DC meter getting fooled. I had the same thoughts... yes, it read -41 which should not be possible, but is very repeatable. I had a very busy day and was not able to try everything we discussed yet, but some of it. I got my old oscilloscope working (apologies, I had previously forgotten I had it), and did some poking around. Of most interest is with the scope probe on AC/DC mode, connected to the control grid of each 7591, comparing results when picking the same simple pattern on my testing guitar.

I think I figured out why my meter came up with -41Volts. In overdrive, the waveform of the 7591 hooked up to the cathode is very bottom-heavy, and when you set the scope to super fast trace (to create blur effect) it appears as if the center of the wave (blur area) drops, which probably fools my DC meter. The other 7591 input does not exhibit this behavior at all, its center stays put.

I will have more time within a couple days to try replacing the caps and do some more testing without the 7591's just to be concrete on our thinking. I will post a power supply schematic later after I can dig one up. All of my testing from here forward will be run using a variac so that my input is 120 on the nose. (That gave B1 of 490V, for reference). I will also update the schematic I previously posted. Work is hectic, it may take a day or two.

Thank you all so much so far for your willingness to help me!

[ktritty]

Bad cap, fixed most of it

Okay, I had some time to work on it. I have attached updated schematic of the whole amp (in sections). [By the way, can somebody tell me how to remove my previous attachment from 5 days ago?] I still need to redraw them and renumerate, as a number of mods have left them a touch sloppy.

In any case, the plate-driven decoupling capacitor (C-19) was indeed shot. It was original and only rated at 400V. Replaced with a 600V orange drop. It tested fine at idle, but leaked signal badly. I replaced both C-19 and C-20, and the amp now sounds better than it ever has, but still doesn't quite have the feel I know I can get, but I'm much closer.

I never knew that decoupling caps could be bad without giving popping or hissing. The 7591's still redplate a bit, but evenly now; I plan to turn down the bias a little bit more (spec was 40mA per tube, which puts the tube Wa at about 22W, above the 19W spec max. Currently at 32mA, will try 28mA next).

I am still troubled by a few of the voltage readings, as they are out of the spec ranges for the 6U8 triode, and the pilot lamp (trivial to fix), and the 7591 screens. I think I can handle everything from here, with the exception of the screens. How should I lower that voltage? The value of R-47 (power supply) would have to be increased to about 12Kohm to do the job, but isn't that way too high for a screen resistor? My other thought is to add a bleeder, connecting the screens to ground, but wouldn't that also affect the tone? The maximum screen voltage spec'd by JJ is 440V. Perhaps I should just leave it alone at 485V? Or would that shorten the life of the 7591's?

Thanks

[Enzo]

C19 is a coupling cap, C27 is a decoupling cap. A coupling cap is there to couple two stages together for the signal to pass. A decoupling cap is used to prevent the signal from leaking through the power supply into other stages.

Thought you'd want to know.

[Chuck H]

Hmmm... I've seen "coupling" caps called "decoupling" caps here many times then. I always assumed it wan an indication of their use to keep DC from the signal chain. But I'm happy for the correction and will try to observe the distinction.

Chuck