> How can be properly done a transformerless varimu.
It can be done. But the transformers add a LOT of advantages, not easy to do without.
The output transformer is often omitted in budget limiters. The vari-gain plates swing a very large voltage (100V) as it goes into limiting, so the next stage needs a BIG common-mode range. Except: you can "hide" a slow attack by letting the next stage clip transients that the vari-gain control path doesn't catch.
The input transformer is great. It gives a very balanced and reasonably low-Z audio drive, yet the impedance to the control voltage is "infinite". This greatly simplifies the control path, which can terminate in a high impedance.
If you just capacitor coupled from a phase-splitter, and injected control voltage after the caps, the control voltage would have to charge those caps. That forces very slow response.
Doing it in opamps is possible. But if we assume we want a differental line in and a gain control, I can't find a solution using less than 5 opamps. True, 5 opamps cost nothing today, and one of them does not carry audio so does not have to be very good.
There is probably a discrete equivalent, but it would not be a trivial design.
Holding the grids at ground and using a variable current source up the cathodes simply does not work. Before you get much gain-change or level, distortion pops out all over.
Using a second pair of tubes to control bias and cancel thump gets very complicated before it gets any better. It would take great dedication to make it work, and wealthy customers to make it pay.
> it's all in polish but the diagrams are pretty clear.
Ah, good old single-ended vari-gain. It works. It thumps, and violently at short attack times. It has large 2nd harmonic; or rather, THD becomes "high" about 20-30dB before a push-pull vari-gain stage's THD would get high. It may actually be "musical", but for wide dynamic range material it won't be clean.
> EF89 is listed as a variable transconductance tube.
ALL tubes and transistors are variable transconductance. Reductio ad absurdum: at zero current you always get zero transconductance. And there is no current where Gm suddenly changes from zero to book-value. Instead Gm increases as current increases. You can't avoid it.
> creative use of the suppressor grid!
Varying other grids has been done, and can work. Radiotron 4th has some notes on this. One interesting trick is: one of the penta-grid (AM converter) tubes has sharp cutoff on one grid, remote cutoff on another. Drive audio into the sharp cutoff (linear) grids, and cut the Gm with the remote cutoff grids. Now the signal grids can be driven any convenient way, and the control voltage sees an infinite impedance, not even the stray leakage of a transformer. Yet I don't know any actual boxes which do this.
The circuit Rob points to (his is from Pop Electronics but it is also shown in RDH4 in a different drawing) is a bit more clever. The cross-coupled 6BA6 don't cancel audio distortion (which will be non-trivial) but do semi-cancel thump. Both V2 and V3 are driven off, but V3 increases V2's Screen voltage which increases V2's current, and vice versa, so V2's current varies very little (apparently you can trim it to zero as a first-order approximation) and the output thump may be very small.
Rob, do you have the rest of the diagram and the part-list? If you feed a line-level, omit the magic-eye, use crystal diodes instead of 6AL5, it is just three dirt-cheap tubes. It will need a buffered output in modern life, and possible balancing, but the core is very simple. It might make a dandy guitar or vocal limiter and flavor-adder.