Learning a thing or two! WARNING! Forecast: WINDY

[Chuck H]

I've been reworking a design of mine and making a few already known, but new to me discoveries. This amp uses a cold biased triode as the first clipping stage and last preamp stage behind the LTP. Well, I flipped the phase as part of an FX loop upgrade and the amp lost a little mojo. I figured it would be easy to get back with a little tweaking but nothing seemed to get that certain "something" back. Fortunately I have another amp with the same ORIGINAL schematic (different layout) on hand. So I took it apart and put it on the bench for some testing. After much exploration I finally discovered what had been missing tonally in the reworked amp. The original design drove the PI with the hot side of the cold biased stage positive. The peak voltage was driving the PI all the way into asymmetry on the duty cycle. Flipping the phase on that stage has the clipped waveform on the positive PI swing. The PI is still clipping well ahead of that last triode and just as wide, but it's not being driven into asymmetry. The reworked amp has a nice symmetrical square wave and the original is a little wonky. And I like the wonky one better! It sounds a little more saturated and "hairy" at the same clipping level. I'm not willing to rework the rework. I'm too happy with some of the functional and noise floor improvements. So now I need to find a way to put the wonky back in At first I considered simply altering the drive from either side of the LTP, but that would do funny things to the clean tone too. I also considered separating the bias for the LTP triodes. That would also mess with the clean tone, BUT!.. If I do BOTH I can balance the output for most of the useful clean range and still have an offset duty cycle when clipping Done for the day. I'll get to experimenting with this soon enough.

More to follow...

[uneumann]

I've been reworking a design of mine and making a few already known, but new to me discoveries. This amp uses a cold biased triode as the first clipping stage and last preamp stage behind the LTP. Well, I flipped the phase as part of an FX loop upgrade and the amp lost a little mojo. ......

More to follow...

Schematic?

Happy New Year!

[Chuck H]

Good call. Ok, other changes have been made, but the only one significant to this thread is the rewiring of the effects loop shown in the red boxes below. The first incarnation was plate driven and was a little noisy and lost too much HF if cables longer than a couple of feet were used. The advantage to changing the loop drive to the cathode were three fold (sort of). It's notably quieter because there's less series resistance, it'll drive about any amount of cable you want now with no change in fidelity and flipping the clipped waveform mitigates voltage peaking on later stages while still achieving the clipped level I wanted. That last one is where the symmetrical/asymmetrical issue arose. So that's the double edged sword.

[uneumann]

Good call. Ok, other changes have been made, but the only one significant to this thread is the rewiring of the effects loop shown in the red boxes below. The first incarnation was plate driven and was a little noisy and lost too much HF if cables longer than a couple of feet were used. The advantage to changing the loop drive to the cathode were three fold (sort of). It's notably quieter because there's less series resistance, it'll drive about any amount of cable you want now with no change in fidelity and flipping the clipped waveform mitigates voltage peaking on later stages while still achieving the clipped level I wanted. That last one is where the symmetrical/asymmetrical issue arose. So that's the double edged sword.

I get the asymmetric waveform phase problem into the PI - it's easy to overlook - but as you've shown, it can/does make a difference. If you favor the plate drive, it cries out for a simple buffer. The 100k/22k divider is the high source impedance that's killing your ability to drive a cable. A small and inexpensive LDN150 or KSP44 (sorry if this causes pain) as a follower could drive just the source jack (after the divider). That would give you plenty of buffer and since the signal levels are low - there's no chance of signal corruption (clipping) by the SS device. You could even drive the return shorting link with the original signal so there is no change at all when the FX loop is unused. Alternately, drive the send jack with the follower and drive the shorting return link from the plate. Most FX devices completely screwup the signal phase anyhow, so there's not much argument for sending one phase or another.

[Chuck H]

I knew where the high source impedance was. I only included the effects loop because I have an OCD aversion to a hanging triode in an amp Some years later now I just thought I'd make it work correctly. And the new circuit works great. I don't necessarily want to flip the phase back the way it was. Driving an LTP into a bent up duty cycle probably isn't the best thing to do anyhow. I'll experiment with alternatives for asymmetry. It would be nice to control the actual amount anyway. Then I could "tune" it for the exact amount of effect that sounds best to me. The OP was really just about discovering what went missing, what caused it and the subjective difference between symmetry and asymmetry as it relates to distortion tone!

[Chuck H]

Update...

Well, trying to artificially bend the duty cycle, that is, other than enormous voltage peaks causing grid conduction, has proved a lot more tricky than I thought it would be. Because the LTP operates as a cathode coupled circuit with a resulting high degree of local NFB I'm finding it difficult to imbalance the operation of either side without unwanted anomalous behavior.

What I AM discovering is many of the different mechanisms that affect the duty cycle of the PI other than extreme voltage peaks. Like the amount of resistance and the size of the filter cap in the HV rail. The input coupling cap (because the LTP input impedance changes radically when clipping) and most surprising was the size of the grid decoupling capacitor to the non inverted side. A small value here has faster recovery, but doesn't decouple to a low enough frequency for "balanced" performance. But a large-ish value slows the time constant of the circuit on and affects recovery accuracy when the signal is modulated. What I'm finding is that you really have to see this in action. Which is probably why all the "scope shot" pictures haven't really told the story for me. I'm still closing in on any sort of formula for predictable and desirable manipulation of a LTP that's clipping. This is a fun ride so far and I'm learning quite a lot. More to follow.

[loudthud]

Try changing the tail resistor, 47K in both your schematics.

[Chuck H]

I did try that. The difference wasn't really noteworthy. I could change the plate loads as well and Bob's your uncle. Un-balancing isn't hard. What I'm after is balance when the amp isn't clipping and a dynamic imbalance when it is. By dynamic I mean the duty cycle would (ideally) shift slowly enough with signal modulation that it can be heard as "swirl". I'm not even sure I need (or want) much of this, but the "phenomenon" fascinates me a little.

The trouble for me is that with THIS amp I'm driving a pair of EL84's. They don't need much drive voltage and the LTP can easily pound them into G2 current if not checked (the rest of the circuit such as it is). So I'm exploring for clever ways to recreate the phenomenon without the unwanted side effects.

One thing I really like about the Trainwreck amps is that they swirl in a useful and musical way. Ken (RIP) was onto something with many of his circuit specifics. As simple a circuit as it is, it has been tuned to swirl and most of the effect is coming from the LTP. Of course I could just build THAT amp, but there are some things about it I DON'T like too. So I'm experimenting to find either a happy medium or, ideally, a way to force the effect in a controlled way without the unwanted side effects.

It doesn't help that this amp uses (and really does need) the Paul Ruby circuit because it does mitigate grid conduction at the power tubes that would otherwise create another imbalance of it's own. The trouble with this for me is that with el84's this causes G2 over current unless you use big screen grid R's. I much prefer the bigger and more dynamic tone of 470R G2 resistors. 1k G2 resistors made the amp sound too small on clean and only slightly overdriven tones.

Still experimenting and reporting. Not having a problem now. Just having fun. The amp is the bees knees. If I can make it "swirl" just a little without losing anything else it has it might be a good thing. Can't know if I don't try.

Thanks

[g1]

I much prefer the bigger and more dynamic tone of 470R G2 resistors. 1k G2 resistors made the amp sound too small on clean and only slightly overdriven tones.

Lot's of values to try between those two, you don't have to stick with standards.

[pdf64]

...with el84's this causes G2 over current unless you use big screen grid R's...

Even with that big 2k2 dropper between HT nodes A and B?

[Chuck H]

Even with that big 2k2 dropper between HT nodes A and B?

OT is 8k. Not to mention the rather higher reflected impedance from a real guitar speaker. At some high level of grid drive dissipation through the tube does try to find a lower impedance load. I've seen it a lot. Screen circuits are nearly always a lower impedance load than the plate circuit. Even Marshall amps with their low-ish primary impedance OT, but high grid drive levels are somewhat known for this. This can be another source of duty cycle imbalance because of more extreme, and unbalanced grid conduction. If I don't want to abuse the screens my usual fix is to reduce drive levels and increase resistance in the screen circuit.

2.2k gives me a decent screen voltage for the high-ish plate voltage and I like the attack envelope the amp has now. That and el84's don't like to be abused TOO much (though I do abuse them) or they go all jangly in only a couple of months.

[uneumann]

If I can make it "swirl" just a little without losing anything else it has it might be a good thing. Can't know if I don't try.

Just wondering how your experiments with different PI input ground cap values went. Do you have any conclusions for what values work well and what to avoid?

[Chuck H]

Ok... This whole thing just became a mute exercise. Sort of...

I'm still interested in implementing "swirl" for the right application. And I'll continue to experiment with ways to do it predictably, BUT,

I just go to hear the amp I've been working on and compare it more extensively side by side with the "control sample" which would be my personal amp of many years (this had been impossible thus far because of my current living arrangements that includes caring for the in-laws). This new incarnation I've been working on does so many things well that I've wanted from the original. Simply increasing gain a bit at the top end of the adjustment got me into the saturation I was missing and while there isn't much "swirl" I really like the tight, focused tone. My proto swirls a little more, but lacks the bigger sound and articulation of my recent work and I think the trade off is well worth it. At least for me. And especially since I've also been chasing the biggest possible tone that still grinds well from a pair of el84's. I think there's only so much one can expect from a single, simple amp design.

[Chuck H]

Just wondering how your experiments with different PI input ground cap values went. Do you have any conclusions for what values work well and what to avoid?

I still have the amp open on the bench. Let me re examine the tests to be sure of the results before posting. They're interesting so I want to be accurate, if not technical about it.

[Chuck H]

At some high level of grid drive dissipation through the tube does try to find a lower impedance load. I've seen it a lot. Screen circuits are nearly always a lower impedance load than the plate circuit. Even Marshall amps with their low-ish primary impedance OT, but high grid drive levels are somewhat known for this.

I want to correct this. It has been a notional consideration for me, but not proven. In Fact, when I consider it, it's probable that the classic Marshall's don't push the power tube grids such as to cause G2 conduction. It's more probable that because they are typically head style amps there is a failure to plug them into a proper load. That would make the shortest current path the G2 circuit and that's more likely why they blow screen grid resistors. Though I've heard it cited before the way I reported it above I'm not at all certain it's accurate and I wanted to be clear about that.

Everything else in the post I'm still ok with