I just converted my Keithley 417 to a solid-state front end, with good
results so far. This is my last electrometer that had a 5886 tube front
end, so I'll be getting rid of my maintenance stock of these tubes soon.
If anyone needs 5886s for maintenance of old Keithley and other gear
while keeping them in original stock condition, let me know.

So far it looks like I'm done with tubes for these applications. As the
new design is refined, I'll put up info if anyone is interested in
possible upgrades to the old gear. The new setup is looking good, with
only about 1-2 fA bias current at room temp, about ten times better than
the original spec.

Some background is here:

https://funwithtubes.groups.io/g/main/message/65994

Ed

Hi Lou.

I'll be putting up more info as the project progresses. This is kind of
an experiment with how some of the modern CMOS opamps are specified, how
they actually behave, and how to get the best performance from them.
I've been investigating these over the years, and it's quite a
complicated story. The bottom line so far is that some fairly mundane
parts are capable of phenomenal performance, once you know what's
inside. The key issues are the package and pinout versus application,
the input protection design, and of course temperature.

The 417 upgrade is based on an LMC6001B, which is about a twenty or more
year old design, that was promoted as an "electrometer-grade" opamp with
25 fA maximum bias current (for the premium "A" version). I got a bunch
of free sample, lesser performance "B" (100 fA) ones many years ago, and
decided to start there. The pinout is amenable to use in the inverting
mode, which is right for the 417, to replace the entire amplifier
circuit. From the specs, I expected it to be on par with the original
5886 tube, but found that it had only a few fA input bias at room temp,
and was even better (1-2 fA) after washing the input section. It's now
settles to somewhere below 1 fA as far as I can tell, after burning in
for a few days. This is with symmetric supplies +/-6.2V from 1N821 TC
Zeners, and a voltage follower buffer amp to minimize load-dependent
(feedback divider, meter movement, and possible external load up to 1
mA) self-heating of the LMC6001. The non-inerting input is only a few
k-ohms to common, which protects against leakage from the adjacent
negative supply pin. The pin 1 on this part is a no-connect, so has very
low leakage issues against the high-Z inverting input. So, the inverting
input is shielded by a floating pin (which could be grounded, except
that would increase the capacitance) on one side, and on the other side
by the non-nverting input that can only be up to a few mV away.

The opamp approach is also much more stable than the 5886 type, in terms
of offset voltage and drift - it's rock solid in comparison.

One problem with replacing the original circuits with opamps, is the
huge open-loop voltage gain. The original circuits had maybe 10,000
tops, while almost any modern opamps have ten to a hundred times as
much, and some even more. This makes the compensation and stabilization
trickier, especially considering that the original amplifiers often have
various compensation loops within their discrete stages, and also range
dependent. Using a single opamp gain block needs a different setup, and
that's the biggest problem right now.

Ed

I just did a quickie experiment to get higher resolution. I rigged up an
external meter movement about the same size as the internal one, hooked
to the monitor out, which is 3V FS. The external meter is set up for
about ten times the sensitivity, so 300 mV FS. I wanted to see how the
noise looks, with similar meter ballistics, and expand the zone around
zero to assess the bias current. The net result is the external meter
reads about 10 fA FS, and the noise and drift are still low enough to be
quite usable. The latest bias current estimate is about 50-100 aA,
slowly dithering around less than plus one percent of FS, with
occasional jumps to about 500 aA or less. I can't tell yet if these
jumps are part of the 1/f noise, or line noise and transients getting
through the power supplies.

Anyway, presuming I didn't make any mistakes in my measuring and
figuring, this is quite impressive. I hope I just didn't get lucky with
one particular part, but I think the others will be similar. I will
eventually be checking them all, and other part types, in test setups.

For reference, I confirmed the scaling. With the 417 set up for maximum
sensitivity, it's 100 fA FS. Putting the suppression supply at 100 mV
through the top 1E12 resistor, it reads near full scale. Setting at 10
mV, the internal meter reads ten percent of FS, and the external meter
reads near its full scale. BTW I can set the suppression directly with
mV resolution. Long ago, I changed the last digit pot to a ten-turn
precision type with a kilodial indicator - it's nice.

This is now over a hundred times better than the original spec for the
417's grid current. When it's all said and done, I'll likely add a 10X
meter switch arrangement in the 417, to get a 10 fA FS range capability
too. It may be good even at much higher sensitivity, but 10X is pretty
good, and I likely won't need any more, practically speaking. Besides,
putting a DVM (and some filtering) on the output can expand it even
more, if ever necessary.

One problem I did find, is there's some line ripple appearing in the
output signal, which I need to investigate. It's pretty small - about
one percent FS - and the meters don't notice it, but it still bugs me,
since it's way bigger than the apparent random noise.

Ed

The 417 info is easy to find. Try "keithley 417 manual" in your search.
The Tektronix one pops up up first, out of many:

https://www.tek.com/manual/keithley-models-416-and-417-instruction-manual

Ed

Update.
The 417 has been running out in the garage for some time, and I've been
occasionally monitoring operation. It's uncontrolled temperature and
humidity, so I get to see a decent range of conditions. In the mornings,
after cooler ambient, the bias current appears about 1 fA or less, then
as things warm up, it gets down to the 0-+/-200 aA range at some ideal
condition, then increase to about 1 fA max in the other direction later
in the day.

On top of this are occasional jumps up to an extra fA or two in either
direction. Also some 120 Hz line ripple that was about 1% FS p-p, which
I have reduced greatly with some temporary changes in the wiring. Again,
the drift and jumps may be from many factors that I can't pinpoint yet.
Besides temperature, there is the random noise and 1/f noise at the
front, and line regulation and transient susceptibility to consider, and
possibly noise at the back end gain ranging switch contact noise, and
connection pin noise. The 417 is a plug-in type deal, and the business
end connects to the main box through a DB-15 connector, and has
provision to run on an extension cable too. Everything adds to the
situation, and the signals are at very low levels, so difficult to
assess cause and effect.

One thing that I think I've determined, is that some of the jumps may be
due to occasional ionizing radiation hits. Last night it was running
around 600 aA, and I placed a bag of thorium mantles on top of the
plug-in's back end, where all the action is. It promptly ran up to about
3 fA and stayed, then went back to the original range when it was removed.

Also, days ago, it seemed to get into a high leakage mode sometimes
especially after being cool over night. The bias would stick around 2-3
fA then much later in the day finally settle down. I believe I have
fixed this problem by changing the first stage compensation cap. It's
hard to say for sure the effect of changes on things like this, because
sometimes the act of simply taking something apart and putting it back
together (necessary each time plug-in is worked on) is actually what
makes the difference.

The opamp front end is temporarily way over-damped with a small cap
(actually two stages of C)  from the output to inverting input, so I
could get stable DC operation. It's mechanically not possible to work on
the plug-in in place, so each group of changes I want to try means
removing it and opening the case, doing the stuff, then putting it all
back. Ideally, this could be done while on a proper extension cable,
which I need to build. I tried a standard DB-15 straight-through cable,
but it all oscillated horribly. The cable needs to be built the right
way to eliminate crosstalk  Until then, I do only a little at a time,
with no live testing available.

The compensation cap is a 1 pF glass type (because I don't have any
polystyrene ones that low). With cleaning, but no silicone treatment,
its body resistance is probably around E11-E13 ohms tops, comparable to
the highest feedback R E12 ohms. To isolate the opamp's DC output from
the glass resistance, I used an extra stage 1000 pF, then a 100 meg R to
common/guard, then the 1 pF cap. This way, the glass sees nearly zero DC
at the output end, minimizing leakage current, while the capacitive
current goes right through. I suspect that the mica cap I used initially
may have been way too leaky (virtually all my parts of all types are old
used pulls, so may be of questionable condition). Replacing it with a
well washed polystyrene one seems to have fixed it, and I haven't seen
the behavior since - although it may be just coincidence, as outlined above.

I'll have more to say later. I pulled out some of my other electrometers
too, and have some interesting observations.

Ed