Re: Hybrid Circuit

Sounds like you've invented the cascode in hybrid form...or indeed a tetrode,
with a follower for impedance conversion.

Triodes of course have no "internal feedback" in any sense other than that in
which it can be said that resistors do, or a dog with a bone. If you stick to
the meaning of the word in books like RDH4 and in engineering convention, you
could avoid much confusion.

Feedback is not sensitivity to load. It is what you use to defeat sensitivity to
load. Alternatively you can remove the load. Removing the load is not removing
feedback, it is just removing the load.

"Degeneration" is quite a good word, but really not necessary because
sensitivity to load is generally assumed, in the absence of feedback, which is
generally used to defeat it.

I drop a stone in air and its acceleration declines in proportion to its
velocity, due to drag. Stones don't have "internal feedback", but they are
sensitive to load.

I pass a current through a resistor and it develops a voltage which counters the
voltage I apply. Resistors don't have internal feedback.

When the men rushed out to clear the falling snow from the track between runs,
they weren't removing "internal feedback" from the bobsleighs.

Setting the cruise control on your car does not remove the load, but defeats its
effect on your speed. This is the application of feedback, not its removal.
Removing the load has the same result as applying feedback, not removing it.

Removing the load from your triode does not remove "internal feedback". It
removes the load. You can't do that by removing feedback, although you could
remove its effects by *adding* feedback.

Yes there is a common-language meaning, blah-de-blah...don't want to begrudge
that to a drowning man in search of a straw. But if you mix it up with
engineering design you'll stop making sense to yourself.

cheers, Ian


"flipper" <flipper@xxxxxxxx> wrote in message
news:puh812diqta36bjjasjkqmivj36fh66g72@xxxxxxxxxx
On Sun, 12 Mar 2006 12:55:16 GMT, "Ian Iveson"
<IanIveson.home@xxxxxxxxxxxxxxxx> wrote:

Flipper,

Try a centre-tapped choke in your simulation instead of the mirror. It is able
to generate extra voltage and can't lock up or switch off. Hope I'm not
infringing Patrick's copyright...

AFAICS we are saying the same thing about the mirror...just I was talking in
impedance and you in current. I don't think the two valves are coupled if no
current change occurs...how does one "see" the other in that case?

I am afraid I have lost interest in the circuit. It doesn't work, and I am
happy
to take your word for it that it can't.

Well, I said it can't work 'balanced'. It does work if you offset it,
like you mentioned.

One part of my project I am looking forward to even less now...converting the
output of a DAC into a bias supply capable of up to -100V. Unavoidably SS
analogue. Things seem to blow up so quickly. Valves illustrate their problems
in
technicolor, but SS just flashes and goes smokey black. How am I supposed to
know
what's gone wrong?

Hehe. I hadn't thought of it that way but, now that you mention it,
they start off as mysterious bits of sand inside little black boxes
and, when things go wrong, turn into just as mysterious broken bits of
sand inside charred black boxes. Or, as the saying goes, they don't
work if you let the smoke out.

Sounds like you're either exceeding the breakdown voltage specs, in
one direction or the other, or running too much current through one
pin or the other. And if you're using bipolar you have to watch out
for reverse biasing things. Remember our discussion about one of the
transistors in the current mirror turning into a diode under certain
conditions? That can be catastrophic if something else doesn't limit
the current. In that case the tube was only pulling a little current
so the transistor would survive but if there was a low impedance load
there it would be curly smoke time. Just an example of how things can
go 'poof' if the circuit wanders into an unexpected state and why
you'll sometimes see protection diodes and zeners sprinkled around
things.

Also, at 100 volts you're entering the region where one needs to take
into account the "safe operating area." The current carrying capacity
drops off at high voltages, more than just what one would calculate
from the thermal power dissipation.

Sim it up in spice.

(btw, MY current mirror PP design uses only one triode for gain AND
the phase splitter. Simm'ed open loop gain is 306 from a single 13EM7
unit 1 whose amplification factor is 68 with no solid state gain
stages. Solid state parts are one CCS, one current source cascode, and
one current mirror. Figure that out ;))

Hmm...you would need a load line with a positive slope. You must be
cheating:-)

Hehe. Well, the load line would, indeed, need to be positive if those
plate curves still applied but I majically altered them ;) For one,
the 'load' isn't on the plate, or on the tube at all, yet it's still
the tube driving the load =O:)

The first part of the secret is simple. Plate current goes into a
current mirror with the load on the other side of the mirror. No pesky
triode 'internal feedback' since plate voltage remains constant and
the load can then be referenced to any voltage one likes; meaning the
load R can be arbitrarily large without affecting either plate voltage
or current. The tube is 'isolated' from the load it's driving so, you
see, the plate doesn't see the load voltage like it does in the normal
plate curves.

The second secret is bleeding off 'bias' current from the mirror so
the voltages on the second side don't have to be quite so dramatic for
a large R. I.E. One might have a 1mArms signal riding on an average of
7mA through the triode and those 'extra' mA beyond what's in the
signal just cause a voltage drop across the R (5 times the signal of
interest) that serves no purpose, from the perspective of the signal,
so bleed off, say, 5mA before it get's to the load R. R is then
operating almost entirely on just the signal and can be even bigger
(for any given V). Of course, the triode is oblivious to all those
shenanigans because it's still operating into that one mirror
transistor and couldn't care less what R is.

Note that none of those solid state devices provide gain. A current
mirror just reflects 1 to 1 and a constant current source just sources
constant current. All the gain is from the triode's (signal only) gm
going into the now isolated R. They do, of course, introduce some of
their own distortion, as nothing is perfect (current mirrors are darn
good, though), but I then close the loop back to the triode (cathode)
so it's the controlling device.

The actual circuit is constructed a bit different than what the
description implies but the gist is correct and my first cut at it was
done that way. I later rearranged things to simplify.

I think you should post the design in a new thread. Perhaps you could get a
wider discussion going before everyone disappears?

Might as well leave it here since it's a hybrid too.


cheers, Ian






.