BTW I'd recommend not rushing out to obtain LMC6001s just yet. I see
they are quite expensive new, and there are alternatives that are much
cheaper and may work even better. I have a a number of candidate parts
and data collected many years ago, that are not organized. As I sort
through my old notes and parts, I'm getting things in order, and will
report more info later.

The short list is here, from what I've recalled so far. Always get PDIP
package if possible - surface mount won't work well. Neither will TO-99
or ceramic (if such exist anymore), without silicone treatment.

LMC6041 single, similar to LMC6001, possibly much lower bias if used
properly.
LMC6042 dual version of '41, pinout of duals can be used advantageously
for voltage followers.
LMC662 dual, similar to '6042.

I vaguely recall that there were a few others in the LMC family with low
bias, tending to be LMC6-something-something -1 or -2, singles and
duals. Don't bother with quad opamps in any of these - the pinouts are
useless for ultra-high Z applications. There were also some from
Burr-Brown and AD, back in the old days. There may be some much newer
types, but the LMCs are still in production after all these years. The
line is owned by TI, since they acquired National years ago.

The input bias current is paramount, of course, but depends on not just
the specs, but on how they actually work inside, how they are packaged,
and how they are applied. It's not easy to predict or discern real
performance to expect, without experimenting.

The input protection devices and structures of the CMOS opamps are key
to getting low bias current. The package and pinout determine the
external limits of performance. You can't do anything about these, but
you can choose the best trade-offs for an application. The more you can
figure out about the input behavior, the better you can use it.

Once you have some parts to try out, first study the protection devices.
For instance, I know that the LMC6001 has single junction diode clamps
to the rails, protecting the MOS input gates, and some series resistance
from there to the outside world. I studied this long ago, planning to
use one as a logging cell for another project. Chances are that similar
ones are the same, but sometimes they're optimized for other features,
and use different protection schemes.

Once you know what the input looks like, the main thing is to see how
the bias current responds to the power supply voltages, and with it, the
input common-mode voltage. Here's an interesting thing that I plan to
study: I have a couple of electrometer boards from junked equipment,
that use the LMC6041. I studied the circuits to see about re-using them
whole, and found they ran the opamp from locally regulated +3V, and -10V

• very asymmetrically. Maybe they found this best for minimizing bias
  current, or maybe it's just coincidence, and they needed it for the
  input or output voltage ranges in the rest of the system. Indeed,
  looking at the datasheet, it appears the bias goes way up as common-mode
  goes toward the negative supply, although the listed conditions are not
  well defined, and kind of ambiguous. I assume that for some reason, the
  input protection in this part is asymmetric, unlike the LMC6001, or
  there are some other things going on in there - like maybe because the
  '6041 has rail to rail output, while the '6001 does not.

That's all for now. Have fun.

Ed

Joel and Lou,

I think I successfully sent you both the schematic pdfs - should be an
individual pdf for each of 4 pages. Let me know if they were not received.

Ed

I have posted some stuff at the geigercounters group, related to the
noise hits on the 417, so there is now some parallel info there too,
beginning here:

https://groups.io/g/GeigerCounters/message/6843

I'll have more experiment and design info to report in a couple days.

Ed

The schematics are up now, posted here:

https://groups.io/g/GeigerCounters/files/Keithley%20417%20Electrometer%20Upgrade

I'll be reporting a bunch of new info soon too.

Ed