Anode follower as +/- trim stage

[atavacron]

I must be missing something. It’s not this simple, is it? Probably looking at the JJ E88CC.




Let’s call that ^ the back end of a mic pre or line stage. It would feed an output section, WCF + transformer for example. The point of the second stage is only to trim after the gain of the first stage, not add another order of magnitude. I set up both halves of the valve the same in order to match plate resistance (~2k4).

• Do the two plate resistances factor in to the divider values, thus changing the end stop values around the pot?

• Each plate is loaded by 75% of the ohmic value of the divider string, right?

• Does it matter what the gain of the second stage is, as it’s 100% fed back at center?

• Does second stage noise mostly cancel at unity?

• Accurate enough to use a detented pot?

• Anybody got an example that worked IRL?

Here’s what I assume to be the op amp equivalent of the second stage.



Here’s a Broskie high current example, cathode-driven on both sides, probably 10-20% of my source resistances (though the 12AU7 & 6DJ8 would present a mismatch). He found VERY low THD using two SRPPs. Ignore the 47u + 665R PSU feedback.



• Is it safe to assume similar or better distortion characteristics from two regular common cathode stages doing the same internal job?

And finally, here’s a line stage example, same topology as my sketch in the area of interest. Ignore the I/O.

 

[ruffrecords]

I do not think it is necessary to run the 6922 that hot and you should probably multiply the trim network resistor values by 10. Might be worth placing a couple of high valye resistors (1 Meg) from the wiper to each side to ensure it does not go open loop if the wiper becomes disconnected.


Cheers

Ian

 

[atavacron]

I do not think it is necessary to run the 6922 that hot and you should probably multiply the trim network resistor values by 10. Might be worth placing a couple of high valye resistors (1 Meg) from the wiper to each side to ensure it does not go open loop if the wiper becomes disconnected.

Noted!

20K pot with 25K5 end stops nets a ~200kHz bandwidth. 25K pot with 31K6 end stops ~175kHz.

Main thing is, I can’t tell if I’m supposed to subtract the plate resistances from the end stops to obtain the desired +/- spread

 

[ruffrecords]

Noted!

20K pot with 25K5 end stops nets a ~200kHz bandwidth. 25K pot with 31K6 end stops ~175kHz.

Main thing is, I can’t tell if I’m supposed to subtract the plate resistances from the end stops to obtain the desired +/- spread

The output plate resistance does not matter because the NFB is taken after it. But the source plate resistance is in series with the left hand fixed resistor so that one may need to be reduced.

Your biggest problem is going to be distortion. The pot wiper is a virtual earth - not a very good one but not far from the feedback arm divided by about 30 (estimated open loop gain). The the pot turned to the left the virtual earth is 22K/30 so about 680 ohm. The total load seem by the first stage is therefore 12K7 + 680 so a little over 13K which is rather heavy. The same applies to the second stage with the pot fully to the right - hence my recommendation to make the network values 10 times bigger There may be some serendipitous equal and opposite cancellation in the next stage but I would not bank on it for real world tubes,

You don't need more than 75KHz of bandwidth so 50K and 63K2 will probably do the trick.

Cheers

Ian

 

[atavacron]

Your biggest problem is going to be distortion.

Points taken. Getting warmer?



Current is still on the high side at 5mA each, but roughly in the low distortion zone cited at this link. I fear I may have doubled the normal bias voltage though, judging from data sheets.

Also, for bandwidth calc, wouldn’t the source impedance seen by the second grid be the parallel result of the two sides of the trim network? 96kHz ish worst case (center) here.

 

[ruffrecords]

Points taken. Getting warmer?


Current is still on the high side at 5mA each, but roughly in the low distortion zone cited at this link. I fear I may have doubled the normal bias voltage though, judging from data sheets.

5mA quiescent at 90V plate voltage is fine.

Also, for bandwidth calc, wouldn’t the source impedance seen by the second grid be the parallel result of the two sides of the trim network? 96kHz ish worst case (center) here.

It would be if it weren't for the 100% negative feedback at that setting. Easiest way to find out is to sim it in LTspice.

Cheers

IAn

 

[atavacron]

Thanks Ian!

And yeah, I did blow the cathode values and voltages. Had to ditch the calculator I was using and draw some load lines. Supply is actually 176.6V. 18K Rp & 390R Rk nets 84V and 2V respectively at 5mA.



Would using a 25K pot and 31K6 end stops (well, 28K and 31K6) promote distortion? The first triode’s plate would be looking into 11k25 altogether, between its Rp and the OLG math of the divider string. I see 33.7dBu of swing available on the load line, and 5mA into 11k25 is good to 34.2dBu. So in a 28dBu “maximum” system, there’s over 5dB headroom. There’s ~1.1dB lost vs the same 18K / 390R triode looking into a 1M load, I think.

Am I overestimating the drive capabilities of the humble E88CC plate?

 

[atavacron]

Hello just reporting back to the thread for posterity that:

A) this is a job for CCSs instead of plate loads, so I will be going in that direction, and

B) the actual plate resistance is more like 4k2, thus making for a string that’s 1K trim - 27K - 25K pot - 31K6 (left to right)