Tremolux vs. Vibrolux tremolo?

[Wombaticus]

OK, I think it's time to admit that I've gotten myself in over my head, and have reached the limits of my knowledge.

I've been banging my head against this for a few days now... shielded the runs, kept a good physical distance between LFO and audio circuits, tried a variety of different bias points, etc. and no matter what I do, I can always still induce the helicopter effect and its attendant huge bias shift and B+ sag. And my scope is down pending replacement of PSU electrolytics, so I'm unfortunately a bit in the dark here about what is really going on.

I'm still wondering if there is something fundamentally flawed with the circuit itself, as opposed to a layout issue. Certainly the LND150 behaves a bit differently in the model than a tube, especially in terms of how the waveform looks at the LND150 gate, although at the end of the day the model still comes up with a reasonable enough looking waveform on the neg bias.

I'm also worried in the back of my mind about the maximum grid circuit resistance to ground on the 6V6s... although this part of the circuit is very similar indeed to the venerable 5G9, it does seem like it wildly busts the datasheet values. Could this be contributing to a bias runaway in the powertubes themselves?

[loudthud]

There are a couple of things I would keep in mind when trying to diagnose this problem.

1) If there is feedback around the power amp, try disconnecting or just decreasing it. You are injecting a signal into an amplifier with feedback and the feedback is going to fight that signal. You are ideally changing the gain of the feedback amplifier, that can cause stability issues.

2) The power supply filtering is a suspect. It can act as a feedback path between stages and the LFO. Try a regulated power supply for the LFO or PI as an experiment. I've got one of those old 0-400V Heathkit power supplies that comes in handy from time to time. The screen supply for the power tubes could also cause problems.

3) The LFO signal doesn't stop at the power tube grids. It trys to go all the way to the PI plates through the coupling caps. In fact, the power tube grid leak resistors form a voltage divider with the guitar signal path back to the PI. Try some lower value coupling caps or grid leak resistors. If the impedance at the PI outputs is imbalanced as with a concertina PI, the LFO signal won't balance at the power tube grids. Try to balance the LFO signal at the power tubes so it will cancel at the OT.

[Wombaticus]

There are a couple of things I would keep in mind when trying to diagnose this problem.

1) If there is feedback around the power amp, try disconnecting or just decreasing it. You are injecting a signal into an amplifier with feedback and the feedback is going to fight that signal. You are ideally changing the gain of the feedback amplifier, that can cause stability issues.

2) The power supply filtering is a suspect. It can act as a feedback path between stages and the LFO. Try a regulated power supply for the LFO or PI as an experiment. I've got one of those old 0-400V Heathkit power supplies that comes in handy from time to time. The screen supply for the power tubes could also cause problems.

3) The LFO signal doesn't stop at the power tube grids. It trys to go all the way to the PI plates through the coupling caps. In fact, the power tube grid leak resistors form a voltage divider with the guitar signal path back to the PI. Try some lower value coupling caps or grid leak resistors. If the impedance at the PI outputs is imbalanced as with a concertina PI, the LFO signal won't balance at the power tube grids. Try to balance the LFO signal at the power tubes so it will cancel at the OT.

Some great suggestions there, thanks! The amp doesn't have global NFB, but right now, the trem circuit is powered from the same supply cap as the power tubes screens (as was the case with the original 5G9 circuit). I don't have a regulated power supply handy, but maybe I can try test clipping in a bunch more parallel capacitance at the screen PSU node, and see if that helps.

I hadn't thought about the interaction between the LTP and the LFO. First thing I check when I get home will be to confirm that I got the LTP right, and didn't accidentally swap the 82K and 100K load resistors, which would of course unbalance it horribly. The coupling cap sizes and grid ref. resistors are borrowed directly from the 5G9 --- so theoretically, if I can make my FET circuit behave close enough to the tube circuit, these SHOULD be OK size, but of course, it obviously isn't behaving like the tube equivalent is it? So it'll bear looking at for sure.

[tubeswell]

Have you tried upping R3 to 1M? That will decrease the output voltage of the LFO.

Another idea - You could build the LFO circuit as per the 5G9 but just substitute the 12AX7 for 2 x IRF820. The only changes that you would make are that with the source follower part you may need a 12V-15V (zener) diode between the source and the gate to prevent a voltage spike destroying the MOSFET, and you may also need a 100-200R resistor in series with the gate (from the LFO circuit)

(@ loudthud - there is no GFB circuit in the 5G9.)

[Wombaticus]

Have you tried upping R3 to 1M? That will decrease the output voltage of the LFO.

Another idea - You could build the LFO circuit as per the 5G9 but just substitute the 12AX7 for 2 x IRF820. The only changes that you would make are that with the source follower part you may need a 12V-15V (zener) diode between the source and the gate to prevent a voltage spike destroying the MOSFET, and you may also need a 100-200R resistor in series with the gate (from the LFO circuit)

(@ loudthud - there is no GFB circuit in the 5G9.)

TW - I'm slightly confused by your comment as the circuit is already completely SS, which is probably why I'm having these problems. It has a LND150 for the LFO, and a MOSFET (not too dissimilar to the IRF820, but with built in gate protection and an insulated case) as the source follower: http://www.wombatamps.com/Tremolo_FET.pdf.

I had thought about switching R3 to 1Meg, but until the helicoptering starts, the apparent depth of the trem isn't (audibly) particularly heavy. But perhaps I'll make the change anyway and see what's what.....

[tubeswell]

I was meaning something like this

[Wombaticus]

2) The power supply filtering is a suspect. It can act as a feedback path between stages and the LFO. Try a regulated power supply for the LFO or PI as an experiment. I've got one of those old 0-400V Heathkit power supplies that comes in handy from time to time. The screen supply for the power tubes could also cause problems.

Looks like power supply filtering is a major contributor! When I clip in a 47uF cap in parallel with the 20uF cap, it seems to pretty much completely stabilize things at reasonable bias settings. Now, it may be, of course, that this is just putting a band-aid on things, as there is still the question of why 20uF is evidently sufficient for the 5G9 but not for my version with the FET based trem.
I'll investigate more in the morning if perhaps my choke may has significantly lower reactance than the original? I'm using a Hammond 158Q, which is 5H inductance - I thought that would be plenty, but I don't recall properly checking the original's spec.
Of course, there's a lot of other stuff different too, from layout to ground scheme, etc., etc...

[tubeswell]

Hiya Paul - is that you also posting about this over at AX84?

[Wombaticus]

Hiya Paul - is that you also posting about this over at AX84?

Hiya Paul - is that you also posting about this over at AX84?

TW - yes, AX84 is my more usual haunt. I had noticed your postings popping up in the AC-coupled cathode follower thread. There are some bright folks posting over there (I believe you've already encountered Iain Hartney, for instance - the guy is cornucopia of ideas), and I hope you stick around.

BTW, I think I have this helicoptering problem close to licked. Celeste Hall over at AX84 pointed out that the amount of current through the source follower could be contributing to the problem, and sure enough, changing the source resistor from 100K to 220K seems to have dramatically reduced the tendency to helicopter with the original 20uF cap. I'm pretty sure the combo of the 220K resistor and a 47uF cap will be a complete cure.

The trem, sure enough, has a nice hypnotic quality now. The remaining points to address with the amp are getting the speed control tweaked slightly so that there are oscillations throughout the complete range of motion of the 2Meg pot... but this I think is mostly a matter of tweaking the series resistance, and perhaps slightly adjusting the gain of the LND150.


Aside from that, there are a couple of other bits of weirdness yet to be resolved. The bass in general seems to be quite flabby. The low E is a bit like a rubber band, and distorts quite unpleasantly -- I'm not sure yet why. I used the default coupling cap sizes, which seem pretty reasonable (22nF for the PI input and outputs)...... Might also be lead dress or the OT, I suppose.

The other issue is a weird interaction between the volumes and the tone (aside from the usual tweed interactivity, which I'm well familiar with). Right now, with either volume turned above about 90%, the tone knob acts normally. But if neither volume is above 90%, it acts normally throughout most of its range, but there is then a very abrupt point in the tone knob travel when it goes below about 10% when something weird happens. You go through a seemingly very small region of travel, there is suddenly scratching like DC on the pot (there is not), then it *very* abruptly starts to hum like crazy. It's weird...it's like a switch.... there is absolutely no swell in the hum with the knob... turn the knob just a micro-hair, and it clicks and scratches and starts humming -- but turn it just a micro-hair the other way, and it stops. The weird thing is that if I set the tone knob in the region where the hum happens, and then transition one of the volume knobs back and forth across about the 90% range of the their travel, I get a similar switch-like scratch-and-hum phenomenon (so it's not just a bad tone pot with a glitch in its taper). The only thing I can think of is that is somehow related to a log-to-linear transition in the taper of the log pots I'm using (CTS 1Meg log), but I've used these in other tweed builds without a similar issue. You ever encounter this?

[Wombaticus]

All right, subbing out the apparently crappola CTS pots with some nice PEC 2W milspec pots has cured the weird pot hum issue. Remaining issue is to figure out why the bass notes don't sound right..... reflowed any suspect solder joints, subbed in different tubes, chopsticked extensively, all to no avail. Will need to investigate possible coupling cap issues in the morning I think....

[tubeswell]

FWIW I went to .047uF coupling cap after V1 on the bright channel IRRC. I also went to 2.2uF Rk on V1. You might have to play around with those and see how it goes.

[Wombaticus]

FWIW I went to .047uF coupling cap after V1 on the bright channel IRRC. I also went to 2.2uF Rk on V1. You might have to play around with those and see how it goes.

Sorry, did you mean a 2.2 KOhm Rk, or 2.2uF Ck?

First thing I'm gonna try, I think, is to put a 100K grid stop on the LTP. I already added 1.5K grid stops on the 6V6s when I built it. Since it seems OK until overdriven, I'm thinking this is a blocking distortion issue. If that doesn't do it, I'll try a 47nF or 22nF coupling cap on one of the channels after V1 and see where that gets me....

[tubeswell]

Sorry, did you mean a 2.2 KOhm Rk, or 2.2uF Ck?

Ooops yeah I did

a 100K grid stop on the LTP. I already added 1.5K grid stops on the 6V6s when I built it. Since it seems OK until overdriven, I'm thinking this is a blocking distortion issue.

Are you using a 12AY7 in V1?

[Wombaticus]

Ooops yeah I did



Are you using a 12AY7 in V1?

er... you still didn't answer the question whether you meant capacitance or resistance!

I've tried both the 12AY7 and the 12AX7 in this amp. In my 5E3, I much prefer the 12AY7, but I think the 12AX7 sounds better in V1 for this one. My comments though were based on using a 12AY7.

Today I added the grid stop to the LTP, and tried a 22nF coupling cap on V1, which helped a bit but without solving the problem.
Poking around in the amp, I noticed that the ground leads for the shielded runs to the power tubes were a bit microphonic... which I thought was odd. So I disconnected the shield grounds (which were to the chassis). Lo and behold, a fairly significant improvement! The weird thing is, the ground leads test with infinite resistance to the grids themselves (as they should) so it's not some sort of partial short to ground. Also the capacitance of the ground lead / shield to the signal cable is too small to reliably measure with my multimeter, but based on measurements of much longer lengths, each run is unlikely to have more than about 50pF of capacitance between shield and signal... which would, in any case, just shave off some high end in the context.

Sure would be nice to have a reasonable explanation. Anybody have any theories? Either way, I'm getting closer.

[tubeswell]

er... you still didn't answer the question whether you meant capacitance or resistance! .

2.2uF Ck

Re: the wiring/layout - can you post any gutshots?