Re: Need EL34 grounded-grid curves

Patrick Turner <info@xxxxxxxxxxxxxxxxxx> wrote in
news:4609D915.7CF9766D@xxxxxxxxxxxxxxxxxx:

There are 2 x EL34 and 28 SS devices in that circuit of yours.
Worked out how each and everyone of them work?

A bunch of these solid state devices are constant current sources/sinks,
and the Vbe multiplier in the second stage. There are only four SS per
side that do gain on the signal: the differential input pair, the
common-base voltage shifters (to allow DC coupling), the main voltage
gain common source stage, and the cathode drivers.

There is also a lot of NFB in action, and slight changes of bias may
not
make a visible
difference to X-Y lines on a CRO.

I'm sure it has an effect. But the NFB is not that much, unless you're
counting degenerative feedback such as the source resistors on the input
pair. The global feedback cuts gain only in half. Notice that the
feedback goes to the sources, not gates, of the input LTP. The designer
told me he did this because the otherwise needed megaohm resistors have
too much parasitic reactance. Actually I'm not sure how the resistors
are calculated. I know I have more gain than needed even now, as my DAC
puts out enough voltage swing to clip this amp. So given the spare
gain, I'm not sure if it's better to increase the global NFB or any of
the three local degenerative feedbacks instead.

All differential amps are PP amps, because one side turns on while the
other side turns off,

_Neither side turns off in this amp_. Each side swings about 750 V
peak-peak here, for a total differential of 1500 V, and both sides are
always conducting.

and the even number harmonic currents cancel at the cathodes,
emitters,

Yes, I understand that. I guess I was under the impression that PP
referred to when the load is _not_ between the two sides as is here. I
got that impression from the usual SS PP output stages where the two
sides are connected and drive one side of the load, with the other
grounded, whereas in this case the harmonic cancellation happens not at
the source/emitter connection as in the SS amp, but in the load itself.

Actually some of the cancellation is through the feedback going to the
other side as well (throught the LTP's source connection). It makes
distortion from one side appear common mode on the other side and
cancels out in the load. Even if I lower supply voltage on one side so
it clips, the other side stretches up the peak so the differential
output remains a sine. I only tried this in SPICE as I don't have a
good way to try it in the amp because I use one supply for each side.

The load is indeed a capacitor and dc drift won't worry the signal,
but you may find dc drift occurs during low level op and high level
op, and you have to analyse why this occurs if it does.

I assume by op you mean operating point?

ESL usually only have a significantly pure C load which dominates the
loading at HF above 7kHz
because a treble panel is typically only 200pF measured stator to
stator.

Well I'm driving headphones, and the Stax Omega 2 (to which I'll upgrade
in a few months) response is supposed to be from 6 Hz to 41 kHz. Stax
only provides a single number, not a graph w.r.t. frequency. I'm not
sure how I can measure it, as my LCR meter uses a fixed measuring
frequency. The data*** says 170 Kohm at 10 kHz, so assuming no
resistance, that's 93.6 pF.

If you were to understand the working, you could work out
at what dc anode current and at what ac signal level where there was
distortion
caused by cap charge and discharge.
200pF at 20kHz is 39k ohms and would seem to me to be an easy load
for a pair of class A1 EL34 with even a fairly low bias current.
As the C load is increased, the C load current increases.

So assuming the headphone worst case is not much below the data***'s
single given number, I shouldn't have any problem with cutting from the
original 25 to 20 mA per side. The headphones are drawing a tiny amount
of that current; it really was chosen for a specific nonlinearity and
the guy obviously had no concern with efficiency, or he woudln't have
used speaker tubes for headphones.

If I were to try to scale this up for ESL, then speaker current would be
significant w.r.t. bias, so then it would make more sense to measure the
specific impedance vs frequency for the speaker.
.