Need help understanding source of distortion

[acheld]

I’ve been asked to provide a “very clean”  amp to serve as the wet side of a wet/dry rig; the player prefers Princetons as a rule, and is asking for this build after hearing a Princeton clone, with Reverb and Trem deleted, and with quad 6L6GC on stage.

I’m in the prototyping stage now, and after going through the effort of deleting the Reverb and Trem sections
on my own, have built something very similar to Robrob’s version. (Schematic attached with voltages).

While testing this build, I found something unexpected happening to signal passing through the portion of the circuit which used to enclose the reverb and trem.    On the schematic, this is the section between TP4 and TP5.

What I see on the scope at TP4 is clean signal, with very little distortion (all scope images are at Volume=max).   

However,  the images from TP5 show asymmetrical distortion at 80Hz, which gradually disappears by 500Hz, and then reappears above 5kHz or so. 

I fully expected to see signal strength diminished, due to the 3M resistor and the small caps — but did NOT expect the appearance of the distortion.   What would be the cause of this?    One theory might be that the caps strip out the very high frequencies from the signal, which is otherwise attenuated by the 3M resistor — but I don’t think this is right. 

So why did I even check the signal in these positions?  I would like to simplify this network if I can, and tune it to get max clean driving signal at the output stage.    But what I’m seeing — I don’t understand.

Note:  Due to the limit for attachments, scope images from TP5 are in the following post. 

all scope images taken with Vol=max.   The input signal is in yellow at the top.

[acheld]

Follow on to add the TP5 scope images.

[shooter]

you're measuring across a filter network and directly to the grid of a tube, guessing you're seeing the result of loading/impedance mis-matching.
try comparing TP 4 with TP6.  use TP 6 to set the "clean level" with the volume pot so the added gain does trip you up.

[PRR]

> at Volume=max...... asymmetrical distortion

If you were playing, you would turn it down a notch. Your 'guitar' is too strong to stay clean for that volume setting. Just-barely, since even the small response un-flatness puts it out and in distortion. (And what is the response shape of that tone stack?)

[shooter]

> at Volume=max...... asymmetrical distortion

> missed that part

[acheld]

Vol is at max, but you can see that at TP4 the signal is pretty clean.  It's only at TP5 that it looks (mildly) ugly.

Maybe there is differential phase delay causing this distortion ?? 

Over the weekend I'll record some images of the signal at both TP4 and 5, and TP5 and 6.   I can only do two channels at a time.

Maybe also try some different probes at different impedances at TP5 (eg, x1, x10 and x100) across TP5 to test the loading theory.   

[pdf64]

1. My gues is that you’re loading the amp with a speaker or equivalent. I suggest to use a resistive load for such tests.
2. At those settings, loaded with a speaker, an input of 200mV seems to be overdriving the output valves? Reduce the input such that they aren’t overdriven.

[acheld]

you’re loading the amp with a speaker or equivalent

Oh my spouse would have a fit (!) if I did testing with a speaker . . . so no, just a big 8 ohm resistive load.

The output valves aren't being overdriven much (if any) at this point.   They certainly were when I had 2 12AX7's in the preamp, as expected.

I'll post more scope images over the weekend.   My main question for now was why the distortion after the remnant of the "deleted reverb."



 

[PRR]

Vol is at max, but you can see that at TP4 the signal is pretty clean.  It's only at TP5 that it looks (mildly) ugly.

At TP5, V2-A grid is going positive. At TP4 the 3Meg resistor masks the positive grid distortion.

[HotBluePlates]

... the images from TP5 show asymmetrical distortion at 80Hz, which gradually disappears by 500Hz, and then reappears above 5kHz or so.  ...

The following is meant to "connect all the dots"; I apologize if it covers anything you already knew.


  -  V2-A cathode is at 2.36vdc according to your schematic notes.
  -  Every tube distorts when the peak drive signal equals or exceeds its bias voltage (and often before reaching that level).


  -  TP5 at 80Hz has 7.36v peak-to-peak according to your scope, or 7.36v / 2 = 3.68v peak.  That exceeds the 2.36v of bias.  Yes, it is distorting because you're overdriving the grid, which then draws grid current and squashes the top of the waveform.

  -  TP5 at 500Hz has 3.84v peak-to-peak according to your scope, or 3.84v / 2 = 1.92v peak.  This is just-below a level that would cause the tube biased with 2.36v to start pulling grid-current.  The tube is clean (just barely).

  -  TP5 at 5kHz has 6.48v peak-to-peak according to your scope, or 6.48v / 2 = 3.24v peak.  That exceeds the 2.36v of bias.  Again, it is distorting because you're overdriving the grid, which then draws grid current and squashes the top of the waveform.

  -  The 500Hz signal is remaining clean due to the lower signal level, which is a byproduct of the tone stack which is notched at/near 500Hz for most control settings.


So it appears everything you show is "normal behavior" because you're overdriving a tube stage.  Keep an eye on peak signal levels vs tube bias voltages to understand when you're accidentally creating an unexpected condition.

[PRR]

All that numeric typing...... when we could have just turned it down a notch and observed clean-up.

"Usually" 20mV (5mV-50mV) into a dimed amp will distort somewhere. 200mV and full-up is abuse.

[pdf64]

…
The output valves aren't being overdriven much (if any) at this point.   …

But the waveform at tp5 was replicated at the amp output?
If so, perhaps the stages after tp5 don’t have sufficient gain / available signal swing to at least fully drive the output valves to their AB1 limit?

As you’ve built a new power amp and are concerned with getting the most out of it, you may be better starting by fully analysing and making sure of that.

So I suggest to remove V1, inject the input signal at tp5.
Move the output valve grid stoppers to the bias supply output, so that they idle at a suitable point and the HT supply is  loaded as normal.
Then set the input for the max unclipped signal at tps 8&9, measure the p-p voltage and gain (from tp5).
Then check that the p-p voltage is more than twice the bias voltage.
The gain will show the minimum clean signal level needed from the V1 circuit.

[acheld]

You guys are making great points, and it's much appreciated.  It never occurred to me that by overdriving the grid, it would distort the signal at the grid.

"Usually" 20mV (5mV-50mV) into a dimed amp will distort somewhere. 200mV and full-up is abuse.

Yeah, you're right.   Need to bring that back to reality.  I should look at typical output for Lollar humbucker/neck + vintage tele single coil/bridge. 

So, I'm going to:

a)  Test different scope probes (x1, x10, x100) to see if that makes a difference at TP5.  Probably not, but will look anyway.
b)  Dial in (more realistic) input signal.
c)  Re biasing for each of these pre-amp stages aiming to to get that ideal signal at the output stages. 

Move the output valve grid stoppers to the bias supply output, so that they idle at a suitable point and the HT supply is  loaded as normal.

Not sure what you mean by this  . . .  could you explain?  I understood the rest (I think), and the procedure makes sense.

I have limited time right now, but will provide results as I can get them.

[acheld]

I measured the output of one of my teles with setup similar to my buddy, with humbucker neck and single coil bridge.

Scope input is across a 1M resistor alone (not plugged into an amp).   Volume and tone knobs dimed.

I'm kind of amazed -- I'm seeing about 600mV P-P.   The three chords you're seeing are typical shell voicings he plays. 

Single string amplitude is lower, though the B string easily gets into the 500mV P-P.

Am I measuring this incorrectly?    I don't think so . . . but these voltages are higher than expected. 

[HotBluePlates]

I'm kind of amazed -- I'm seeing about 600mV P-P.  ...

Single string amplitude is lower, though the B string easily gets into the 500mV P-P. ...

Now do the same thing with a True RMS multimeter, or with your scope's numerical reading set to display RMS volts.  It will be much lower (like 100-200mV).

When I owned a 1967 Princeton Reverb, a Strat pickup would make it distort when the amp Volume reached 4.  Brief distortion on peaks would have happened earlier, but wouldn't have been obvious enough to notice.  Which is why turning the am-volume full-up is not realistic.  And as PRR points out, once you start analyzing gain & headroom of each stage of a circuit, you find 1960s amps could distort the output tubes with as little as 20mV at the input jack.

It never occurred to me that by overdriving the grid, it would distort the signal at the grid.

RDH4

Pages 18-21 talk about grid current, but the key point is on Page 19: at something between -1 and 0v grid-to-cathode, the tube will begin passing grid current.

Once grid current starts, it ceases being ∞Ω as we're used to thinking, and drops to 1k-10kΩ.  Regardless, that new very-low impedance squashes the output of the previous stage.

RDH4 notes this happens when the grid is driven positive of the cathode, and even begins somewhat before the grid gets to that point.  This is why only the positive peaks of your grid signal where squashed/clipped.

[tubeswell]

You’ve got 3 preamp gain stages and a cathodyne, so the signal is going to distort in the preamp. And with 3 preamp stages in front, the cathodyne will be prone to gain spike distortion and frequency doubling distortion. If you want less distortion, use the 3rd gain stage and the cathodyne as a LTP. This would be equivalent to ‘2 and a 1/2’ gain stages driving the 6L6s - the amp will be loud and clean

[shooter]

Now do the same thing with a True RMS

when I set up an amp I typically use 80-100mVac_RMS.  Set all pots to 5, scope the input jack and the speaker.
Increase the input signal til ch2 (speaker) just starts to distort, like the TP5 you originally posted.  record the input signal value, then reset to 80-100mV
next turn the gain pot til ch 2 just distorts, record pot #, then set back to 5
repeat with each pot, including TS.  that will give you a pretty good starting point for "middle of the road" values.
Then I move on to tweaking the stages/pots to give me what i'm looking for from quiet to full roar.

[PRR]

...I'm seeing about 600mV P-P.

Which is just about 200mV RMS.

And if he plays that strong all night long, amp full up, he is wanting a bit of distortion.

[pdf64]

…

Move the output valve grid stoppers to the bias supply output, so that they idle at a suitable point and the HT supply is  loaded as normal.

Not sure what you mean by this  . . .  could you explain?  …

The point is to allow the output valves to idle whilst the preceding stages are tested to assess their capabilities. If the output valves were left in the signal path, it would be harder / extra work, as we would need to check whether clipping was due to the cathodyne reaching its limit, or the output valves were being pushed into grid current clipping. Both conditions might look the same on the scope.

A further point came to mind; scoping inside a feedback loop can result in confusing waveforms, especially at the onset of clipping, as the waveforms are a composite of the audio signal and the error correction signal.
I find it’s much better / easier to open any global feedback loop and get the circuit optimised to get the best performance without negative feedback providing a crutch to help the output waveform whilst messing up the waveforms inside the loop.