DSL40CR blocking distortion, maybe? (warning, nerdy circuit stuff ahead)

[TheKman76]

Hello Marshall techie brains trust!

I have some nerding to work through and I'd like to know if anyone sees any flaws in my thinking.

There is some concern from some MarkII users that the Ultra channel can crap out a little when the gain is above about 2 o'clock. Personally I don't push the gain that hard, but it got me thinking about the apparent complexity of the circuit after V1A (or V1:2 on the schematic) and why all these extra components are there. For context, I've already removed C68 from mine for tonal reasons, among other changes.

1) R94 - This looks to be merely a ground reference to reduce pop when switching channels. Does really need to be there if there's a muting circuit? The answer may well be yes.
2) R93 - if C68 is already removed, does this need to be there at all. I think not.
3) C60 - WTF? I'll circle back to this.
4) R90 - Also WTF?

So I'm thinking that the "crap out" some users refer to is in fact blocking distortion caused by C60. This is rather a huge cap for simply coupling and a 1Meg grid leak is large in this context, especially without any other stopper.

If the divider R90 & R91 are left in place and C60 and R90 are removed, does this not constitute a better design? I'm all about keeping the signal path as simple as is possible.

Anyone?

 

[Pete Farrington]

Some valves have more grid current than is ideal, and across grid leak resistance that puts a small V DC on the grid.
That can cause a pop when stuff gets switched in the grid circuit, which people find very annoying.
Hence C60.
Bias shift due to signal rectification when overdriven will tend to be reduced by the large unbypassed cathode resistor, hence bias shift / blocking distortion may not be a problem.

But yes, I guess that pushed to extremes, it could still tend towards that.

 

[TheKman76]

Bias shift due to signal rectification when overdriven will tend to be reduced by the large unbypassed cathode resistor

Well made, sir. I didn't spot that.

Regarding C60, if there's always a ground referenced AC coupled connection to the grid of V2a, surely this can't be a problem? I don't see how this would prevent popping.

Hypothetically, if R94 is left in place and C60 & R90 are removed (shorting C60 of course), how is the circuit function any different? The milliseconds of switching time with a 470K grid leak and microamps of grid current would be irrelevant in the context of 10's of volts of signal at this stage. Surely? And, of course, there's a muting circuit too.

I'm more than happy to have my misunderstanding pointed out to me, but a large cap directly connected to a grid where clipping is very likely is a recipe for bias shift. Even with large cathode swing there's just nowhere for the cap to discharge quickly.

 

[FleshOnGear]

Well made, sir. I didn't spot that.

Regarding C60, if there's always a ground referenced AC coupled connection to the grid of V2a, surely this can't be a problem? I don't see how this would prevent popping.

Hypothetically, if R94 is left in place and C60 & R90 are removed (shorting C60 of course), how is the circuit function any different? The milliseconds of switching time with a 470K grid leak and microamps of grid current would be irrelevant in the context of 10's of volts of signal at this stage. Surely? And, of course, there's a muting circuit too.

I'm more than happy to have my misunderstanding pointed out to me, but a large cap directly connected to a grid where clipping is very likely is a recipe for bias shift. Even with large cathode swing there's just nowhere for the cap to discharge quickly.

One thing you could try, to see if it remedies the crapping out sound, is hooking up diodes from V2:1’s grid to ground - with cathodes pointing to the grid and anodes to ground. With the large unbypassed cathode resistor you’ll probably need like 3 LEDs plus a regular silicon diode in series.

I know what you’re thinking, but hear me out. The idea is to use diodes oriented to conduct when the grid is negative, with a large enough voltage drop that they don’t turn on until V2:1 is in cutoff. That way, if there’s a negative charge on C60 that would cause bias shift, that negative voltage is shunted to ground on the downswing, but this happens after V2:1 is cutoff so you don’t hear any diode clipping. The bias shift is then eliminated.

Try it temporarily, and if it doesn’t help, you know you’re barking up the wrong tree. Just a thought. Good luck!

 

[TheKman76]

I see what you're saying, limit grid voltage and provide a low current path to leak the grid. A perfectly reasonable solution.
What I don't really understand is why C60 and R90 are there at all.

I did a quick simulation of removing them and found it also increased gain by 2dB. So reducing R91 to 330k brought it back in line, also providing a lower impedance grid leak for V2a.

Would this not be a better solution?

 

[FleshOnGear]

I see what you're saying, limit grid voltage and provide a low current path to leak the grid. A perfectly reasonable solution.
What I don't really understand is why C60 and R90 are there at all.

I did a quick simulation of removing them and found it also increased gain by 2dB. So reducing R91 to 330k brought it back in line, also providing a lower impedance grid leak for V2a.

Would this not be a better solution?

As usual I was overthinking it. I was thinking of the diodes as an easy test to see if blocking distortion was indeed occurring, but it would probably be very simple to bypass C60 with a wire without even pulling the board.

I think @Pete Farrington is right that C60 is there to keep DC from the grid off of the switch. I just bypassed a cap in a similar position in solid state amp, and it didn’t make the switch pop, though.

 

[Pete Farrington]

Regarding C60, if there's always a ground referenced AC coupled connection to the grid of V2a, surely this can't be a problem? I don't see how this would prevent popping.

Grid current in the order of pico or nano amps won’t cause much grid voltage.
However, there isn’t a type limiting value for it, and with some 12AX7, it can be up near the microamp range.
That puts enough V DC onto the grid to cause noisy pots and, pertinent to this, pops when stuff in the grid circuit gets switched.
So my guess is that C60 is there to stop that.

 

[TheKman76]

As usual I was overthinking it.

I may be stiff competition for the over-thinking-it crown.

... and with some 12AX7, it can be up near the microamp range.

You've got me thinking about the dynamic behavior of the switching. I suppose if the channel was switched during an instant of heavy clipping on V2a, this would be problematic. At the same time, would you notice this if the stage was clipping?
Pot noise from DC is an interesting idea. There are three gain stages after this relay, so every little bit of noise induced will be audible.

Just to satisfy my own curiosity I'm going to try bypassing C60 and see what happens. If there's meaningful DC on the grid I should be able to measure it before and after bypassing. Will also test for popping.

Thanks folks, for helping work this through.

 

[TheKman76]

Also trying to find an example of a Marshall amp with switching that doesn't have a C60 type cap reveals this:



This JVM410H uses RL3 to switch the second gain stage V6a out of circuit. No cap. With the 10k cold clipper there is also the potential to have higher DC on the grid too.

The Defence rests, your honour.