[sdiy] Tubes in Synths [ was : Harry's Nightmare! ]

[jh.]

>Typical problem of this simple phase splitter - unequal output resistances.
>This
>is also of concern in push-pull tube amps, where at high frequencies the
amp
>isn't truely push-pull anymore, because of the different phase shift.
>
>However there are phase splitters that manage to have a low output
>resistance at
>both outputs. The circuit I have in mind here is the phase splitter of the
>wiliamson amp. Basically a simple phase splitter that is preceded by a
>cathode
>follower. Ok, you'd need two triodes per stage then. Many stages would be
>expensive.
>Yet, I think I'll try that one day.

But I think a kathode follower has less than unity gain, hasn't it ?
In an amp, you can compensate this (if you need), but for a phaser
this is problematic.
There are other tube phase splitters. Differential amp configuration
for instance. But then you have two high Z outputs.
OTOH three tubes are enough to build an opamp, so you can use the
opamp approach for a tube phaser of course.


>>I'm not a tube expert at all. This is just where I stopped to
>>follow the tube path when I built a univibe-style
>>(transistor phase splitter) phaser some years ago.
>>(That was the Neptune Phaser, and it's all transistor in the
>>end.)
>
>I'm a bit surprised here, because this circuit has also unequal output
>impedances. The emitter output is having current feedback, while the
>collector has not. (Similar situation as for a triode, although the
>impedances are lower.)

I think in a transistor circuit this doesn't matter too much:
Both, the emitter and the collector see the same RC network
parallel to the Rc and Re resistors. So the same RC combination
that attenuates the collector voltage also decreases the local feedback
from the emitter resistor. For an ideal transistor, the two effects
might even totally cancel. In real life, the cancellation is not
perfect, but much better for transistors than for tubes. The trick is
to make Rc and Re as low as possible and to add an emitter follower
or source follower in front of each splitter. This works remarkably
well with typical RC networks where the R is a LDR.
I can't say it wouldn't work with tubes (meanwhile I saw Don
mentioned existing tube phasers), but it must be much more difficult
to get deep notches over the whole frequency range.

JH.