buildafriend said:
I think I understood what Ian was saying mostly but now I'm trying to figure out why the thump happens instead of just what the work around is.
It's largely the mismatch of the tubes. If you look at the plate curves, the datasheet gives you 'typical' characteristics, but each tube is slightly different from any other, and also varies over time due to wear and how nicely the cathode is activated at that moment. So when the grid-cathode voltage is changed, even by the exact amount for both tubes (as will happen with the center tapped grid input transformer), the plate voltage for each tube will not change by the same amount. So, the center tapped plate transformer will see a net difference in current from each side, which looks like an output signal, so it will make it to the secondary.
So, the thump is less from any imperfect application of the control voltage, but rather how the tubes act when they're driven out of the normal bias point - two tubes might match up well at zero gain reduction, but have different curves at the reduced bias levels.
Part of me wonders if there is anything that can be implemented to come from the output to do the job better than the rectifier tube with enough precision to not upset the input valve's sensitive grid/cathode ratio and biasing.
One possibility would be to use a more precise way to trim the symmetry of the two tubes, to try to get their curves over a wide range of control voltages to match up. The commonly used 'one knob' balance trim pot controls the static grid voltage offset, but not so much the gain of the CV to the grids. I could see using a more complex circuit that not only allows the static grid bias offset to be trimmed, but also the gain of the control voltage to be adjusted. In other words, you could try to trim the static and linear control voltage to reduce 'thump'. This essentially adds a gain trim to the control voltage, to the existing static DC trim.
Stepping back to reality, this would be fairly annoying to trim out, but it could be done. However, as the tubes age, these trims, imperfect as they are, will slowly become wrong, since the errors can't be fully trimmed by a zeroth and first order trim anyway, and will need to be readjusted. It would also make the tube selection/matching process extremely annoying, requiring a pile of tubes from which a suitable pair could then be trimmed from.
I could imagine that this complex and painful cal procedure would reduce the thump by 6-10dB, but is that such a great prize? It's not going to go away, so the 'thump' is pretty much tolerated.
Overall, most vari-mu feedback limiters can be helped out by much more clever release circuits rather than worrying about the thump that's buried inside of 'full scale' output signals. The hole that gets punched from excess LF or too slow of a release time is IMHO a better thing to worry about. I'm restoring a few RCA BA-6A limiters now, and fortunately, I have a bunch of new 6SK7 gain reduction tubes to select among. Given some quality tubes, matched for static characteristics and trimmed in the limiter itself, you can get the BA-6A to work pretty well, enough so that the thump is not a big deal. Still, all bets are off after 500 hours of use, but they're NOS tubes, so hopefully, they'll drift together and they'll get re-trimmed every few months. Most folks who want the BA-6A find that these quirks are part of their character, so as long as you try to make the unit basically clean and well behaved in its stock configuration, the inevitable errors are desirable.
