Grid Stoppers, Bypassed Grid Stoppers, & Blocking Distortion in pre-amp gain stages.

[wizard333]

Been looking at the whole grid stopper size issue in terms of both RF rejection and prevention of blocking distortion.

I'm not hearing any blocking distortion in my overdrive pre-amp designs and I'm using fairly small grid stoppers and want to go maybe even smaller. However, if I look at some designs, like older Marshall, I see big 470k grid stoppers.

How much is enough for this purpose? Assume 100k plate/1.5k bypassed CR, .0022uf or .0047uf coupling caps.

The other thing I don't quite get is, if part of the purpose of that grid stopper is to kill RF oscillation, but you put (again, example of old Marshall) a 470pf cap across it, haven't you just defeated your purpose and let all that RF through? I see some Dumble schematics doing the same thing with different values. What am I missing here?

[Steve Conner]

Blocking distortion is caused by the coupling capacitor taking a long time to discharge once grid current has stopped flowing.

So, as far as blocking distortion is concerned, the "R" in the RC time constant formula is the shunt resistance from grid to ground. Not the stopper resistor, which is in series with the grid, and therefore can only have an effect while grid current is flowing.

Yes, a capacitor across the stopper resistor would stop it from filtering out RF and snubbing oscillations. I guess Dumble was only interested in the tonal effect of it, which is to reduce the treble rolloff caused by an oversized stopper resistor in conjunction with the Miller capacitance of the tube.

12AX7 type tubes usually won't oscillate even without the grid stopper, and RF only needs to be filtered out of the first stage, so he would have got away with it.

[pdf64]

I like to use 10k-22k grid stoppers to mitigate for oscillation, on 12AX7 gain stages. Mounted on the tube base terminal. The 68k input mixers (when used) are best fitted there also. Bear in mind that the input socket switching terminals are usually arranged to parallel these up when the 'main' input is used, giving an effective value of 34k.
The above has allowed for even 4 gain stage pre-amps to be implemented without them being overly finickity regarding layout / screened wires.
Big 'grid stoppers' of say 100k - 470k can help to prevent a dc grid voltage building up when the stage is overdriven (due to grid - cathode diode chopping off the top half of the signal), see Blocking Distortion
because the 'grid stopper' acts to isolate the chopped off waveform at the grid from the coupling cap / grid leak node.
Used in this way, my understanding is that they aren't really acting as grid stoppers, this term being limited to resistors mounted on the tube base; rather they might be better described as series grid resistors.
So, as they're being used for blocking distortion (a dc/very low frequency issue) they can be bypassed for higher frequencies, assuming that doing so doesn't cause a oscillation problem. Pete.

[Merlinb]

How much is enough for this purpose? Assume 100k plate/1.5k bypassed CR, .0022uf or .0047uf coupling caps.

Well, if you were trying to be clever you would try and make the charging resistance equal to or greater than the discharging resistance. The discharging resistance is usually the grid leak plus the internal resistance of the previous valve.
The charging resistance will be the anode resistor of the previous valve, plus the following grid stopper (plus the input resistance of the overdriven valve, which I would estimate at 10k).

The other thing I don't quite get is, if part of the purpose of that grid stopper is to kill RF oscillation, but you put (again, example of old Marshall) a 470pf cap across it, haven't you just defeated your purpose and let all that RF through? I see some Dumble schematics doing the same thing with different values. What am I missing here?

Only the grid stopper on the input valve is there for primarily RF purposes. Those later valves have grid stoppers mainly for blocking or tone-shaping purposes.

[kg]

Only the grid stopper on the input valve is there for primarily RF purposes. Those later valves have grid stoppers for blocking or tone-shaping purposes.

that depends on how much gain you've got going on.

though they aren't "rf tubes" per se, common preamp tubes will happily oscillate far above AF range. this can be difficult to troubleshoot (or NOTICE even) without an o-scope.

for my preamps i always use some value of stopper, right on the socket. it's just good practice.

[wizard333]

Well, if you were trying to be clever you would try and make the charging resistance equal to or greater than the discharging resistance. The discharging resistance is usually the grid leak plus the internal resistance of the previous valve.
The charging resistance will be the anode resistor of the previous valve, plus the following grid stopper (plus the input resistance of the overdriven valve, which I would estimate at 10k).

Excellent, thats exactly what I was looking for. What confuses me there is, since the stopper is between the grid and the grid leak, why wouldn't it be considered part of the discharging resistance? What is the typical internal resistance for a 12AX7 or 12 AY7?

So with a grid leak of 470k, ignoring the internal resistance of the previous stage, and a Rp on the previous stage of 100k, add the in put resistance of 10k, the Rs would have to be 360k or more to prevent blocking? That seems quite high in terms of what its going to do for the high end attenuation; for example I calculated the high end attenuation in my setup for a 220k RS to have a -3db point right around 6khz.

With a 1M pot serving as a grid leak, that gets even worse.

I think using that method as a fail safe is going to end up with too much high end attenuation; is there something else to go for here that can avoid that? How does the size of the blocking cap come into play in the equation?

Only the grid stopper on the input valve is there for primarily RF purposes. Those later valves have grid stoppers mainly for blocking or tone-shaping purposes.

So one never need worry about RF inside the chasis, even with 6-8" of unshielded wire between stages? Ok.

12AX7 type tubes usually won't oscillate even without the grid stopper

Are there certain situations where you've seen this to be more prone to happening?


Thanks!

[Merlinb]

What confuses me there is, since the stopper is between the grid and the grid leak, why wouldn't it be considered part of the discharging resistance?

Because the grid acts like a diode. Current can go into it, but not the other way.

What is the typical internal resistance for a 12AX7 or 12 AY7?

Check the data sheet. ~65k for a 12AX7

So with a grid leak of 470k, ignoring the internal resistance of the previous stage, and a Rp on the previous stage of 100k, add the in put resistance of 10k, the Rs would have to be 360k or more to prevent blocking?

Yep, but you can bypass it with a small cap to retain the highs. It is mainly at low freq's that blocking occurs, so the added cap doesn't hinder the system much.

How does the size of the blocking cap come into play in the equation?

It determines the highest freq at which blocking becomes possible. The smaller the coupling cap, the higher in freq the blocking can occur, BUT the quicker it can recover, so it;s a trade off.

So one never need worry about RF inside the chasis, even with 6-8" of unshielded wire between stages?

Depends. If the layout is good, and spacious, then it shouldn't be major a problem except in super high gain designs.

[wizard333]

Yep, but you can bypass it with a small cap to retain the highs. It is mainly at low freq's that blocking occurs, so the added cap doesn't hinder the system much.

How would I calculate the resulting frequency response and knee points? I can calculate the high end roll-off due to the grid stopper and capacitance in the tube, but not sure how to add in the affect of the bypass cap across the grid stopper.

[Merlinb]

How would I calculate the resulting frequency response and knee points? I can calculate the high end roll-off due to the grid stopper and capacitance in the tube, but not sure how to add in the affect of the bypass cap across the grid stopper.

The maths is a bit long-winded. However, all you really need to see is that by adding a cap in parallel with the grid stopper you are creating a capacitor divider with a resistor in parallel with the upper arm (like the opposite of a vol pot with bright cap), creating a shelf filter.

For example, if you have a 470k grid stopper and 100pF tube capacitance, the cut-off freq' is 3.4kHz and the attenuation increases indefinitely.

If you then add 100pF in parallel with the 470k then you have created a 100pF-100pF capacitor divider, so the maximum possible attenuation is now 0.5 or -6dB. The loss at 3.4kHz increases a little, but by less than 1dB so you can basically ignore it.

If you used 1nF then the loss would be only 1-(100pF/(100pF+1000pF)) = 0.9 or -0.8dB

I suggest you play with a circuit SIM program.

[wizard333]

Long winded math doesn't bother me if you want to post it I'd use it! The calc I got for the cut-off frequency is from Aiken's site.

Which SIM program would you recommend?

Thanks!

[Merlinb]

Which SIM program would you recommend?

I prefer Tina: Tina - TI Software Informer: Latest version download, news and info about this DesignSoft program. A circuit simulation program based on a SPICE engine.

There is a free student edition which supposedly doesn't have valve models, but if someone sends you a TINA file with valves already in, it seems to work...
Tina - TI Software Informer: Latest version download, news and info about this DesignSoft program. A circuit simulation program based on a SPICE engine.