18W Plexi Crossover Distortion Woes

[tschwarz]

Hi everyone!

I have a friend's TAD 18W Plexi on the workbench and am trying to find the cause for some nasty hissing overtone "noise", especially when a two-string chord like a simple Amaj rings out. Human language isn't adequate to describe acoustic phenomena. Said friend used the term "nuclear fission" to describe it. Doesn't really help, does it? ;-) Well, maybe some oscilloscope plots do:



A closeup of the X-over - this is at 1kHz. The blue trace is the phase inverter output, the signal at the speaker is in yellow...



...and this is a snapshot of an Amaj ringing out:



I found the EL84s to be biased somewhat on the cold side - probably as a safety measure - keep grid bias in the safe zone when it's modulated - as well as to enhance the bias tremolo's effect. Also the OT impedance is 9.2k instead of the 8k most other circuits I know use at roughly 320V HT.

The circuit doesn't have an off-switch for the trem, the grids are just pulled down to ground fully when you drop the intensity. Same thing a switch would do - but:
I'm thinking with a switch (e.g a push/pull switch on the intensity pot) one could hook up a second cathode bias resistor in parallel to the 120R/5W, running the tubes slightly hotter - reducing X-over.

Anyone ever done something like this?



Cheers,
Tom

 

[Mongo Park]

I think you are on the right track. The trem has benn known to not work well when the preamp tubes are not middle of the road bias. My 6g2 has a trem switch which is like a tone cut, it lifts the ground on the pot, can’t remember which one maybe intensity. I find the trem for me sounds best with a clean bias preamp anyways. So I added preamp cathode bias options switches to give various degrees of clean bias, or not clean bias
Other than encouragement I don’t think this helps all that much.
Maybe try a different bias to see if it goes well with the trem before adding the switch.

 

[Wyatt]

How cold is the bias?

Cathode-biased, class AB amps should be biased to 100% dissipation at idle. The voltage shift when signal is applied (when we start playing the guitar) will drive the bias colder. So if you are biasing a cathode-biased amp like a fixed bias, it can drop into crossover distortion pretty easily once signal is applied.

 

[tschwarz]

Yes. I'm on board with that. The problem is the bias tremolo that modulates the operating point. So, in order to keep the EL84s healthy longer, the bias is moved quite a bit away from the 100%. I see the crossover problem disappear whenever the tremolo LFO pushes bias voltage lower. Hence the idea with a multi-purpose tremolo switch - trem off + hotter bias / trem on + cooler bias...

 

[2L man]

I think it is not necessary for the cathode bias AB1-class push pull to be 100%. Bias should make tubes somewhat close but still bit less hot than they come when tubes amplify the signal to max power. There OT impedance, B+1 and B+2 voltages and screen resistors should again somewhat match so that tubes do not overheat. Even the type of music and loudspeaker resonance can have significant effect.

For example if loadline does not approach double the tube max power there where operation turns to B (loadline turns more vertical) and bias is 100% tubes might run hotter when not playing.

Those oscilloscope pictures have so much distortion that I can't say what is the main mechanism?

 

[andrewRneumann]

I would also consider mismatched tubes as a possible culprit. Measure the OT primary resistance on both legs from the center tap. Then measure the voltage drop across the OT at idle for each power tube. Use Ohm’s Law to calculate the idle plate current in each power tube and report back!

Your idea of reducing the bias resistance by adding a resistor in parallel is also valid. Use alligator clips or tack solder in something to drop it a little. But please, report the voltages you are measuring so we can all see where the bias is right now.

I would think that bias tremolo doesn’t really put extra thermal stress on the tube because it spends just as much time under as it does over. The average heat is just the same. But I could be wrong about that—not really a tremolo guy myself.

Hope this helps.

 

[tschwarz]

I also did the appropriate amount of tube rolling. I tried 4 sets of EL84s (one of them came with the amp) plus one set of 6P14Ps. A (most likely matched) pair out of a Matchless DC30 sounded best - with a really noticeable difference in gain and output power. Sovteks from the 90s. I did measure a tube mismatch with the original TADs - roughly 11mA without an input signal. But the best set still showed crossover distortion...

 

[tschwarz]

I just discussed a possible mod with the owner: engaging the push/pull pot connects a 470R resistor in parallel to the cathode's 120R, which gives around 100R - the exact same value the AC15 a uses. It also reconnects the trem input to the EL84's grids to a trim pot, so the modulation can be limited respecting the new higher bias current / lower bias voltage. That way the footswitch would still work...

 

[tschwarz]

I think it is not necessary for the cathode bias AB1-class push pull to be 100%. Bias should make tubes somewhat close but still bit less hot than they come when tubes amplify the signal to max power. There OT impedance, B+1 and B+2 voltages and screen resistors should again somewhat match so that tubes do not overheat. Even the type of music and loudspeaker resonance can have significant effect.

For example if loadline does not approach double the tube max power there where operation turns to B (loadline turns more vertical) and bias is 100% tubes might run hotter when not playing.

Those oscilloscope pictures have so much distortion that I can't say what is the main mechanism?

I'm afraid I don't understand your question - "the main mechanism" for what?

Cheers,
Tom

 

[printer2]

Going too hot on a fixed bias amp can mess with a trem circuit, on a cathode biased amp I am not sure, something to try. A switch could be used if the oscillator has a problem with a colder bias.

The main mechanism of the distortion I would assume. The scope trace looks like an 18 Watt being overdriven. And the correct term is fizz if I remember properly. A few different ways to skin that cat.

18 Watt filter fizz - Google Search

 

[BenTobith]

1974 circuit is cathode biased and designed to be biased at or above 100%. The main areas for failure and improvement in this circuit in the modern context (modern wall voltages, crappy new production tubes) is to beef up the screen grid resistors (at least 470R/3W or 1K/3W). Shielded cable from both inputs, as well. A large value cathode bypass cap for the output can also help with crossover distortion, apparently (like 2000uF).

 

[froggie]

Hello, did you mesure the voltages on your amp or are they the same as the schematic shown ? On the schematic Ik would be 11/120=92 mA, 46 each side and assuming a screen current of 5 ma, anode dissipation is some 0.041x306=12.5W ! You are hot biased (on the schematic...). What is your real cathode voltage ?
As the output signal is symetric, the impedance of the transformer seems right for me.

 

[breubo]

May want to try the first stage supply cap. I had one a number of years ago that audio bleed through the power supply. Weird sound. Worth a try.

 

[Volcanicash01]

Yes. I'm on board with that. The problem is the bias tremolo that modulates the operating point. So, in order to keep the EL84s healthy longer, the bias is moved quite a bit away from the 100%. I see the crossover problem disappear whenever the tremolo LFO pushes bias voltage lower. Hence the idea with a multi-purpose tremolo switch - trem off + hotter bias / trem on + cooler bias...

Very interesting tech talk. On a different subject, non tube related, has anyone out there got a Cube Street Ex? If so, what EQ settings do you use on the lead channel to get a nice creamy distortion, instead of a harsh boxy sound? (There is no master volume to play around with) Or is this the nature of the beast? I just do solo gigs these days. I would like to ditch the pedals.

 

[andrewRneumann]

Must read if you haven’t: https://www.aikenamps.com/index.php/what-is-crossover-distortion

A class AB amplifier usually uses some amount of negative voltage feedback in order to reduce the amount of crossover distortion in the output waveform. This negative voltage feedback will take out some of the "notch", at least until the point where the output stage is driven to clipping, which opens up the feedback loop because there is no more "excess gain", allowing the full crossover distortion to reappear in the output waveform.

 

[SanoTele]

Hi everyone!

I have a friend's TAD 18W Plexi on the workbench and am trying to find the cause for some nasty hissing overtone "noise", especially when a two-string chord like a simple Amaj rings out. Human language isn't adequate to describe acoustic phenomena. Said friend used the term "nuclear fission" to describe it. Doesn't really help, does it? ;-) Well, maybe some oscilloscope plots do:

View attachment 983596

A closeup of the X-over - this is at 1kHz. The blue trace is the phase inverter output, the signal at the speaker is in yellow...

View attachment 983597

...and this is a snapshot of an Amaj ringing out:

View attachment 983598

I found the EL84s to be biased somewhat on the cold side - probably as a safety measure - keep grid bias in the safe zone when it's modulated - as well as to enhance the bias tremolo's effect. Also the OT impedance is 9.2k instead of the 8k most other circuits I know use at roughly 320V HT.

The circuit doesn't have an off-switch for the trem, the grids are just pulled down to ground fully when you drop the intensity. Same thing a switch would do - but:
I'm thinking with a switch (e.g a push/pull switch on the intensity pot) one could hook up a second cathode bias resistor in parallel to the 120R/5W, running the tubes slightly hotter - reducing X-over.

Anyone ever done something like this?



Cheers,
Tom

Regarding the noise, not the tremolo switch mod, I would try different tubes in each position first, then check the coupling caps. My Princeton Reverb had an ungodly overtone, It turned out to be a coupling cap.

 

[2L man]

I'm afraid I don't understand your question - "the main mechanism" for what?

Cheers,
Tom

I keep everything simple to find where the distortion comes. I am not saying you don’t but mixing two frequencies make it more difficult. On third picture those consecutive positive waveforms are so different that for me it is impossible to say anything of the cross over distortion.

Phase inverter output is already way too distorted. Power amp meet the max current and max voltage what the operative voltage allows and there comes output what make tube amps so interesting

Adjust level so that power tube drive signal is pure sine wave!

 

[TequilaCaster]

Going too hot on a fixed bias amp can mess with a trem circuit, on a cathode biased amp I am not sure, something to try. A switch could be used if the oscillator has a problem with a colder bias.

The main mechanism of the distortion I would assume. The scope trace looks like an 18 Watt being overdriven. And the correct term is fizz if I remember properly. A few different ways to skin that cat.

18 Watt filter fizz - Google Search

Yes indeed, the 'Paul Ruby 18-Watter Buzz Fix' is the solution.
@tschwartz I sent you a tdpri message yesterday with a .pdf that explains it. You may not have seen it yet. Click on the little envelope icon on the tdpri menu bar. I see you are a fairly new member.

 

[Mark1406]

Maybe this is what you are looking for?

Paul Ruby mod

 

[andrewRneumann]

I'd love to hear some audio of nuclear fission. Please post.