I still haven't read it, Bruce, but at least I worked on the instrument and made some progress.
Since they're nearby and have them in stock for cheap, I bought a couple of 151-0083-00 from Vernonia Northern. Thanks, VN! I also bought a couple of generic 2N964's just for fun.
(Hakan is also sending some parts. Thanks again!)
The replacement -0083's performed worse than the original,
and the generic 2N964's worse still. Eventually I discovered the secret sauce.
1. LEAD LENGTH. As you insert Q80 deeper in the socket, loop gain rises.
Cut the leads to absolute minimum. Also grind the socket down.
I'm thinking of removing the socket and sinking individual contacts into
the board, or even omitting it altogether and soldering the transistor.
2. LEAD DRESS. Move the T88 red output wire (positive bridge strobe) as far
away from the bridge output as possible.
Adjust in the following order.
(a) BRIDGE STANDOFF for 2V positive or negative input without baseline lift;
(b) BRIDGE BALANCE for no baseline shift across mVOLTS/CM;
(c) SNAPOFF CURRENT for 350ps risetime;
(d) MEMORY GATE WIDTH for maximum loop gain;
(e) MEMORY BALANCE for no baseline shift across SMOOTHING;
(f) LOOP GAIN to maximum.
Until this point, comparing loop gain is meaningless.
After performing the above, my 1S1 maximum loop gain is 1.18 .
(g) LOOP GAIN back to 1 after measuring, recording, comparing, etc.
I was using three original sampling bridge diodes and one borrowed from the
VintageTek Museum's 1S1, which I'm also working on. Both quads are
152-0433-00, which is silicon. I don't know if they're PN or Schottky.
VT's 1S1 also topped out at 1.18, in spite of the drastic difference in snapoff
current - 13mA vs 40mA. That difference only occurs when VT's Q80 and
VT's D87 are installed together - substitute either part and you have to
crank it back up to 40mA to get the specified 350ps risetime.
Same thing happens when VT's Q80 and D87 are in my 1S1. Some kind of magic there,
those two particular parts are just made for each other.
I've been thinking about this circuit. That is, the bridge sample strobe circuit,
a snapoff diode driven by an avalanche transistor. Except for a 1nF coupling cap,
there's nothing between the collector and the cathode except circuit board trace.
You can follow along by downloading a manual from W140 et. al. Go to
schematic 1, "SAMPLER".
Between pulses, the transistor is off, collector at -19V and the cap charged.
The diode is forward-biased, so the junction is full of charge carriers.
When the transistor avalanches, it wants to go from high-voltage no current
to low-voltage high current, RIGHT NOW, but the diode stolidly continues to
conduct until the last carrier is out, holding the collector at voltage. Only then
can the collector move, and that edge becomes the bridge strobe. We want
that step to be as big and fast as possible, to get maximum sampling efficiency
and therefore maximum loop gain.
The transistor and the diode are the classic case of irresistible force vs immovable object.
I suppose it's the transistor that gives, via collector region resistance.
I suppose trace inductance does a little bit to isolate the two, but in that case,
wouldn't you think that less inductance would mean less stored energy and a smaller
pulse? But the opposite seems to be true, since shorter transistor leads equals
higher loop gain. I don't know what to make of it.
But my instrument is well again.
Do you think the cap resonates with the stray inductance? Makes my head
hurt to think of it.