Fun new zener tweak!

[Chuck H]

I have an amp on the bench right now that was doing something "I" found most annoying. The LTP PI would skew hard upon clipping. I know they do this, but the imbalance just happened to trigger at the same time as the onset of power tube grid conduction. The power tubes are clipping well ahead of the PI, with nice symmetrical flat lines top and bottom, but before getting to a crunchy square wave the PI impedance would tank, the PI balance would skew and the onset of grid conduction would only occur on one power tube. So the waveform looked like this:



I know you can look at a bunch of old Fenders and see a similar wave form, but I just didn't like the look of it. With the one tube biased all chilly and the other tanking bottom end. So I tried a bunch of stuff to reduce the affects. Or at least get the amp to NOT do all the bad things at the same damn time. Changing PI bias, tail resistor, plate load balance/imbalance, PI voltage, different PI tubes, and, of course I swapped the power tubes just to see if that would change the anomaly, etc. Even changing the NFB circuit had little affect. Adding a large grid stop to the PI grid had very limited affect. So I thought on it this week while I was at my day job and came up with some ideas. Most didn't work or at least not well enough. But one did! I measured the AC volts on the PI input at the onset of PI clipping at just over 3V. The signal going into the PI could be turned up to about 8.5V. I want the amp to grind so just reducing the PI input wasn't going to work. Instead I installed face to face 5.1V zeners across the PI grids. Bam! Nice square, symmetrical waveform that doesn't introduce any diode clipping because the power tubes are well into clipping before that, so, also no affect on clean tones. No more impedance tank or imbalance on the PI either with it's cursory LF roll off and the tiny 2V of grid conduction is somewhat even on each power tube so the push/pull is much more even. I connected a pot in series with the diodes and found that adding about 8k of resistance really dialed it in.



In fact the wave form may be TOO square and symmetrical now. I haven't heard it yet. I may need to increase the ZV and add more resistance to lessen the affect if the clipping sounds too hollow. BUT! This could be just the thing for those guys that like to play metal through their 18W Marshall clones and Tiny Terrors or if you just want a more hard rock/less "bluesy" overdrive from a BF type amp!

That's it on this one for now. I'll report on how it sounds when I can.

[Chuck H]

I've done some more testing and the results are rather surprising "to me". The circuit below is how I have it wired:



Interesting to me is that the actual scope reading of the PI don't show any obvious clipping.?. The grid voltages dropped about a volt and the input AC voltage at max still reads at 7.8V.?. I would have guessed the input voltage to drop by the zener voltage plus the diode drop. That would be about 5.8V at the input. Ok... So there's something happening here I don't understand. That happens. But I'd like to understand! Obviously the two grids are shunted at 5.1V to 5.8V differential (5.1 plus the typical .7v drop of the other diode if the difference is one direction). Both grids seem to be behaving as normal for a LTP but there is no impedance tanking or imbalance at the top of the swing! What have I stumbled onto? I need some circuit interpretation from the higher minds. This looks too good to be true. I'm still waiting for Allen Funt to step in.?.

[J M Fahey]

1) THANKS for testing and sharing

2) haven't digged very much into it , but just thinking aloud, typically all grids in guitar amps get routinely overdriven , specially meaning they get positive and rectify.
All those receiving signal through a coupling cap will charge it to peak voltage and said voltage will remain during a time determined by the cap value and grid reference resistance.
Said voltage will wreak havoc on bias, of course; generally tending to cut off the involved tube, shifting the "idle" current and working point and making it clip unsymmetrically, producing tons of even harmonics.

In an LTPI an extra phenomenon happens: since total current through both triodes is constant (or nearly) , if one tube nears cutoff, the other tube nears saturation, its working point also shifts and then, (now we are entering Hypothesis area):

maybe that explains the second triode wonky waveform.

To boot, both triodes have the same "working space" at idle or while clean, from idle plate voltage to almost +V to almost cathode voltage (minus tube saturation voltage, say 20 volts, should check the datasheets) , and the same heavy job: driving power tube grids which clip way before the PI tubes would, so in principle they should clip roughly symmetrical.

BUT you keep rising PI drive signal, it starts to unbalance, both PI triodes move opposite ways (I mean DC, not only AC) , and I guess one finds it easier to keep driving those clipping power tube grids, while the other fails miserably and clips in a funky way.

All this is idle theory of course, one way I see to try to explain how a symmetrical load (identical power tubes in both sides) fed from a basically symmetrical driver (both PI tubes) behaves symmetrically while clean, stays reasonably so at the start of (power tube) clipping but then becomes so unsymmetrical when heavily overdriven.

An extra detail which might explain this, is that the first PI triode gets as strong a signal as the earlier tube can give it, even over a hundred Vpp , while the second one gets none at the grid, (NFB does not count much, it's a 1:10 to 1:20 attenuated sample of the speaker voltage , a couple V at most, as mild as can be), so NOW we have a very unsymmetrical event happening.

Now the back to back Zeners avoid this horror story from happening.

[Chuck H]

Thanks for stepping in Juan. Sometimes I get windy and people run, screaming.

A lot of what you suppose is certainly true. I've seen it on the scope a few times. LTP's get wonky in proportion to how much the input triode is overdriven. Once it starts clipping the input impedance tanks and the input cap knee frequency rises. Some amps seem to handle this better than others. Or at least they LOOK like they do on the scope. I think what is actually happening is that earlier stages are clipping and this prevents the PI from actually clipping even though it may appear that way if you weren't looking elsewhere.

On this amp there is another preamp configuration that DOES clip ahead of the PI and pretty much right at the point of power tube clipping. In this mode the final wave form always looked dandy. And it hasn't changed one bit with the zener addition. All the zener circuit seems to be doing is stabilizing the PI in the range where it would usually get ugly.

I still need to see how it sounds. My bench is in my home and I have in laws that live downstairs now (Damn kids and all their crazy music!) They know it gets loud sometimes when I have a project but I try to keep it to a minimum. It helps to warn them ahead of time so I don't need to pick them up off the floor anymore

[trobbins]

As the zener approaches DC zener bias voltage turn-on it may exhibit a noticeable soft turn-on resistance as current starts being shunted - depending on the zener voltage rating. At 5V6, I measured a very rapid turn on characteristic, compared to 3V9 and below where it is quite soft. 5V1 may still be quite abrupt for shunt current above 10uA. I had to go to 3V9 to get a soft knee which I could then use for cloning the old SiC varistors used in Maggies.

A such, if you wanted to soften the shunting effect a bit then you could use say 2x 3V3 instead of 5V1, or put a 3V3 in series with a 5V1 or 4V7.

[Chuck H]

Ok... I just tone tested this thing. Picked up the in laws and got them back into their chairs and am reporting back as promised

It's what I expected, with the exception that the difference with the circuit in or out is smaller than I thought it would be. I did install the circuit because I didn't like the way the wave form LOOKED and not because of how the amp sounded so I guess it's alright.

With the circuit in, the clipping is a little smoother, tighter and more focused with slightly more even sustain across the spectrum. A little more controlled sounding. With the circuit out the amp is just a little woolier. Less even sustain across the spectrum but with some notes accentuated more. A little more "hair" on the notes.

These differences are small.

I don't even know which tone I prefer From a functional standpoint I know the output tubes will wear more evenly. The even sustain should be a nice feature for getting consistent performance in different venues. So for these reasons I'll install the circuit permanent.

Thanks for hanging in if you red all this And if anyone can tell me more about how this circuit does what it does that would be cool too.

[Chuck H]

As the zener approaches DC zener bias voltage turn-on it may exhibit a noticeable soft turn-on resistance as current starts being shunted - depending on the zener voltage rating. At 5V6, I measured a very rapid turn on characteristic, compared to 3V9 and below where it is quite soft. 5V1 may still be quite abrupt for shunt current above 10uA. I had to go to 3V9 to get a soft knee which I could then use for cloning the old SiC varistors used in Maggies.

A such, if you wanted to soften the shunting effect a bit then you could use say 2x 3V3 instead of 5V1, or put a 3V3 in series with a 5V1 or 4V7.

Thanks trobbins. There is NO visible diode switching with the 5.1zV for some ding dang reason. With a 6.2 there is a soft jump of the crossover notches on the rise lines. Not too abrupt. At 6.8zV the jump in position is greater. In fact the notches move across center. First bellow 0 and then jump above 0. I didn't want this because it means there is a moment when the wave is perfectly symmetrical and that might cause some weird phasey thing. You could actually hear the sing of the OT into the dummy load get really quiet at the zero crossover point. With the 5.1zV the notches move a just a tad asymmetrical and only to one side of zero. I think this should have a more consistent harmonic and distortion order content. I didn't have smaller zV diodes to try. Reading AC into the PI grid I get a little over 3V as the power tubes begin to compress. I figure the further I can stay above that with the zV the less likelihood of any audible switching artifacts. That is, these were my considerations when I implemented the circuit. What it's actually doing seems to be another matter.

[J M Fahey]

You could actually hear the sing of the OT into the dummy load get really quiet at the zero crossover point.

Interesting.

I didn't have smaller zV diodes to try.

Try Red Leds : 1.9V each and rounder clipping than a Zener (although sharper shoulder than with regular Si diodes).

I bet your Mod will become popular with fast finger shredders

[guitician]

There doesn't seem to be much DC conduction path for those diodes. So I would think they're not going to perform like Zeners having an avalanche of current.

[Chuck H]

There doesn't seem to be much DC conduction path for those diodes. So I would think they're not going to perform like Zeners having an avalanche of current.

No there isn't. I thought there might be a DC thing happening and I hadn't measured voltage differentials between the two circuit points. But your right. Not enough DC shift to allow conduction at any operating point.

I just did some more testing and it's clear that there IS clipping of the input grid . But there's also something else happening I haven't been able to pin down. Perhaps some interaction between the grids restores some local NFB. This would seem odd since the non inverted triode grid is AC grounded.?. But simply clipping the PI input signal to ground doesn't yield the same positive result I get with this circuit. I'm probably not the right guy to analyze this, but I'm still looking into it anyway

[J M Fahey]

Zener diodes avalanche always when a certain ("Zener") voltage is applied across legs, be it DC or AC instantaneous value.

It's not a "DC only" thing.

Otherwise they wouldn't clip higher than Zener AC peak voltages applied to them.

[loudthud]

Traynor had an interesting tweek to this circut. They dropped the grid leaks as low as 47K and added a grid stopper to the driven grid. Said it smoothed out the distortion. Did you try any grid stoppers?

The Trainwreck Express is another case. The PI is driven directly from a cold clipper that limits positive drive.

[Dave H]

I've done some more testing and the results are rather surprising "to me". The circuit below is how I have it wired:

[ATTACH=CONFIG]34977[/ATTACH]

Interesting to me is that the actual scope reading of the PI don't show any obvious clipping.?. The grid voltages dropped about a volt and the input AC voltage at max still reads at 7.8V.?. I would have guessed the input voltage to drop by the zener voltage plus the diode drop. That would be about 5.8V at the input. Ok... So there's something happening here I don't understand.

The zeners must clip if the voltage across them is greater than the zener voltage but you are not measuring across them. You are measuring from input grid to ground. To see the clipping you'd need to use a differential scope connected across the diodes. The right hand grid of the PI isn't AC coupled to ground, it's AC coupled to to the feedback resistor input. When the diodes clip both grids of the PI will be driven common mode by the input signal which is why you don't see the input voltage clip at the zener voltage (plus a diode drop).

[Chuck H]

The zeners must clip if the voltage across them is greater than the zener voltage but you are not measuring across them. You are measuring from input grid to ground. To see the clipping you'd need to use a differential scope connected across the diodes. The right hand grid of the PI isn't AC coupled to ground, it's AC coupled to to the feedback resistor input. When the diodes clip both grids of the PI will be driven common mode by the input signal which is why you don't see the input voltage clip at the zener voltage (plus a diode drop).

I just saw clouds part and I heard a sound... AAAAAAAHHHHHHHHHhhh...

Capitol. I have a better idea where to look in the circuit for info now. For instance... The voltage at the bottom of the decoupling cap for the non inverted triodes grid is roughly 2V... 5.1(ZV)+.7(FVD)+2(circuit elevation)= 7.8V BING BING BING BING BING! Exactly the peak input voltage with this circuit in place.

Ok... I need to button this thing up pretty soon but I'll poke it a little more before I do and report anything interesting.

[Chuck H]

A little more testing...

While the input to the PI IS being clipped, it's being clipped above the threshold where the PI itself begins to clip. So there is no change performance up to that point. But even so this reduced input voltage peaking seems to allow the PI to clip symmetrically up to full output and beyond that into square wave operation. This might explain a few things. One is the difference between the two most classic designs, Marshall 1959(and slight variants) and the BF Fender. Or, if you're a purist, the Fender 5f6a and the Fender black face models. This amp has two "modes". A Vox TB preamp and a Fender BF preamp. The TB preamp had always exhibited more symmetrical clipping. It's worth noting that the TB preamp topology is the same as the Bassman/Marshall. That is, gain stage, gain stage, cathode follower, tonestack, PI. This preamp is clipping near the same point as the PI and that keeps the PI from going all wonky. Where BF topology actually puts a larger clean signal into the PI and does cause more asymmetrical PI behavior. Blues players like BF amps and rock players like Marshalls (No whining please. This is a generalization.) It's also worth noting that most players prefer the vibrato channel on BF amps even if they don't use the reverb or tremolo. Hmmm... I don't know how much clipping happens at that mixer stage in the BF design, but if it's enough to give the PI some relief then that could be one reason for the vibrato channel preference. Anyhoo... Just thinking out loud now.