Page 93 - Analog Circuit Design Art, Science, and Personalities

The Story of the P2

miles of that Rustrak paper. When the P2A came along, it was able to do as good as
1 or 2 pV p-p for an hour. But Bruce Seddon, one of the senior engineers, was al-
ways skeptical about the possibility ofa P2 having VOs stability that good. He com-
puted that a single microvolt was worth about 600 electrons on each varactor. Since
a varactor diode had really a rather shallow slope, you could compute that a 1 mV
DC input would cause a 0.03 pF imbalance in a 200 pF bridge. And a 1 pV DC
imbalance would cause a 0.00003 pF imbalance. Needless to say, that was a pre-
posterous situation. You could compute that even a couple of atoms of shift on the
components nearest to the varactors would cause worse imbalance than that. But we
measured a lot of P~s, and a lot ofP2As, and some of them would hold better than 1
pV p-p for an hour or two. Bruce always was incredulous about that.
Now, if you trimmed the offset voltage to zero, the input current was pretty small,
about 5 or 10 pA typical, and 100 pA guaranteed max. Some people would pay a
surcharge for selected units with extra-low input current. But many people would
just crank the input offset pot over to the side-perhaps a few millivolts-and get
the input current down to less than 1 PA. It wasn't perfectly stable if somebody sud-
denly turned on the airconditioner, but under constant ambient conditions, it was
better than all but the best electrometer-tube amplifiers.
In addition to having low DC errors, the P2 had fairly decent low noise. The P2A
was guaranteed to be better than 1 ,uV rms in the bandwidth 1-100 Hz, and many
P2s were almost that good. Now, how can an operational amplifier have noise as
good as that, right where most solid-state amplifiers have many microvolts of noise
p-p? The fact is that the varactors transform, or down-convert, the noise of the first
RF transistor at (5 Mcps to 5.0001 Mcps), down to input noise at the inputs of the
P2: in the frequency range (0-100 cps). Those varactors really did provide the
advantages of a parametric amplifier. And those germaniums weren't bad at S Mcps,
so the P2 did a respectably good job for low noise. It Look many years before its
performance was matched by FET amplifiers.
The P2 was assembled with its two little PC boards rivetted securely togethcr and
then installed in a cast aluminum case. Then the whole cavity was tilled with room-
temperature-vulcanizing (RTV) silastic material. It did seem to keep things at a
constant temperature, and if there was much moisture, the RTV did seem to help
keep it off the boards. Still, on some moist days, they could not get the P2s to pass a
100 pA final test, so they would just set them aside and wait for drier weather.
When we built the P2A, we did not use RTV, because at +8S "C, the RTV would
expand and pop the P2 right out of its case. We just used several heavy coats of
Humiseal, and that gave very good results. I don't think moisture gave us much
problem on the P2A.
According to some of Bob's friends, Bob said that he could tell when the women
assembling the P2s were menstruating. He thought it was the amount of sweat that
would cause corrosion or leaky printed-circuit boards. He could check out failurc
rates versus serial numbers versus the initials of the assembler, and the yield would
go up and down every 28 days. I know I was impressed that there M'ere always two
inspectors, inspecting the PC boards after they were hand soldcrcd. They could spot
badly solderedjoints and cold-soldered joints. and mark them with a red pen, to go
back and get resoldered and touched up, because a P2 would sometimes run really
badly, noisy and flaky, if there wcrc cold-soldered joints on the board.
To this day, 1 still have the dismantled carcasses ofa few P2s. That is because
Bob Malter decreed that if you could not get a P2 10 meet spec, after you had tried
evcrything. the technicians would pull off thc valuable components-the trim-pot
and thc piston capacitor-for re-use. Then the transistors and transformers would
be removcd, so that even a competitor who wantcd to raid our trash cans would not
74