Fun With Low Leakage/Bias Current: Femtompere, Electrometer, Keithley 617

[Hydron]

I did a sweep on mine and there are big jumps in absolute error at certain points corresponding to most significant few bits, so a few spot checks won't tell the full story. That said though the range around 0V is the most critical so if you do a few more measurements close in that's probably enough.

I did end up changing to the LTC part and it's certainly better but probably not enough better to be worth it, though the part I removed went into another instrument which had a much worse AD7541 clone. Might try and dig up my measurements and post them here.

Tweaking the 0 offset as Alex suggests above was a bigger and much cheaper improvement - mine is padded by a parallel resistor now but I've got a trimmer sitting in my Digikey cart that will go in place of the resistor. You could also change the range trimmer for a finer adjustment one - there's a 4-turn option available that should be drop in compatible though I haven't tried that.

Edit: added DAC sweep plots for K617 Vsource and K181 "analogue" out (this is the "other instrument" - also uses a AD7541 type DAC for the output, note it has a different output circuit/adjustments to the K617 hence the different plot characteristics). All have error in LSB as vertical axis (note the total range changes between DACs), horizontal is output voltage. I've tweaked the K617 "original DAC" plots for better offset/range correction to make it fair (hadn't fixed the offset before getting the data). The K617 started with a 1992 datecode MP7623JN which I swapped to a LTC7541AKN, and the K181 started with a 1980 datecode MP7621JN which was changed to the MP7623JN I pulled from the K617.

[Finderbinder]

Interesting finding - electrometer op amp ADA4530-1
https://www.analog.com/media/en/technical-documentation/data-sheets/ada4530-1.pdf
Someone thinking to upgrade JFET / LMC662AIN to this? 

As this one was found only by chance maybe there exist even more interesting electrometer "specialized" amplifiers 

[Hydron]

OPA928 is what I'm going to have a go with - similar specs but lower noise and slightly cheaper. The ADA part has been used already to replace the input FETs and opamp - search this thread.

[Finderbinder]

I haven't read the whole thread yet. But I try 

[Finderbinder]

Some confusion here.
Are you and others choosing Coto 9002-05-00 sure it is shielded? As I see in the datasheet regarding part marking, 9002-05-00 is unshielded, without diode, 9002-05-10 is shielded, without diode.

Based on the great summary from MiDi regarding the relays, there are currently three options on stock:
Meder Electronic/ Standex HI05-1A66 (Mouser) ~12.75$ each
Coto 9002-05-00 (Mouser) ~6.92$ each
Coto 1240-0197 Ebay (NOS) ~31$ each + shipping+handling, total >~300$ for 6relays

With 1240-0197 from eBay the risk and price is too high. The Meder has a higher specified insulation resistance but I think the missing shield is a draw back. Therefore, I'll get some Coto 9002 as r6502 did and will report back.

[Hydron]

Yes the 9002 has an *electostatic* shield, which is different to the magnetic shield in the -10 part. Magnetic shield is not required (or probably desirable) here.

Key is to count the pins - you want the 6-pin version (2 pins are for the electrostatic shield).

On a similar note - if anyone has a spare or broken one of these COTOs it would be great to see what the inside looks like. (Or even better, can take an x-ray).

[Alex Nikitin]

The COTO relay mentioned is not really an electrometer grade, the MEDER barely is. I have some custom Crydom relays which are truly outstanding but with 9V coils, unfortunately.

Cheers

Alex

[Finderbinder]

We are waiting eagerly to see results 

OPA928 is what I'm going to have a go with - similar specs but lower noise and slightly cheaper.

[Hydron]

The COTO relay mentioned is not really an electrometer grade, the MEDER barely is. I have some custom Crydom relays which are truly outstanding but with 9V coils, unfortunately.

Cheers

Alex

Sadly the COTO 9002 seems to be the best that's reasonably available however - we're just lucky it seems to exceed it's specs by enough to be usable in this application. Thankfully my 617's original relays seem to be in good condition so no need to swap them out.

I did find some stock of Keithley custom COTOs here, but their prices are painful and the website is worse: https://products.testco-inc.com/#/app/layout/itemlisting?item_search=COTO
They were good to deal with over the phone though - I picked up a couple of COTO 1240-0229s for my K236-K237 conversion in August (now apparently $30 ea, annoyingly it was $45 when I got them ). That particular part could easily have been replaced by a lesser relay but I was in the area for work and it's nice to put the correct part in rather than bodging around fly-wiring stuff. Next time I'm in Sunnyvale (pretty likely I will be at some point in the next few months) maybe I can do a group buy for UK folks if there's any interest.

[Finderbinder]

Can I take plots through GPIB, or should I use 2V output and some DAQ?

[Kleinstein]

Taking the digital data via GPIB is the better way. AFAIK the analog output is from a DAC and the digital data and thus less accurate.

[Hydron]

Analogue output is from the input to the ADC, so is true analogue but uncalibrated. It's also referenced to the input low rather than earth or floating, so whatever you use to log it must be floating or using the same reference potential. If 5.5 digits is enough then the internal ADC doesn't seem to be too noisy, so easiest is just using gpib.

[David Hess]

OPA928 is what I'm going to have a go with - similar specs but lower noise and slightly cheaper. The ADA part has been used already to replace the input FETs and opamp - search this thread.

Transimpedance amplifiers are not usually my thing, but I am designing a quick and dirty picoammeter and noticed something about operational amplifier selection including the OPA928.  It also seems increasingly likely that I may end up rebuilding a Keithley 617 for my own use.  I am thinking now that a 2 operational amplifier differential amplifier would be a better configuration to replace the dual MOSFET with the gain adjusted for stability of the feedback loop.

CMOS parts usually lack detailed information about their input current noise, (1) and I wondered why the OPA928 only gives input current noise at 0.1 Hz.  Most parts give it at 1 kHz, but lack a graph.  Something I never noticed before is that CMOS input current noise *falls* at lower frequency, where it rises for JFET and Bipolar input parts.  Reverse flicker noise?  So the 0.07 fA/SqrtHz specification at 0.1 Hz of the OPA928 should be comparable to the claimed 4 fA/SqrtHz specification of the LMC6081 at 1 kHz, and maybe inferior to the claimed 0.2 and 0.13 fA/SqrtHz at 1 kHz of the LMC6061 and LMC6001.

I say claimed because I think the LMC6081, LMC6061, and LMC6001 should all have the same input current noise, especially the LMC6081 and LMC6001 because as far as I know, the LMC6001 is just a graded LMC6081.  Maybe someone here has an idea of what is going on with these parts.

The input bias current versus temperature for all of these parts displays the typical doubling of current every 10C at higher temperatures, indicating leakage of a PN junction which I assume is the input protection diodes.  The difference with these parts is that the input protection networks are bootstrapped, so this only appears at higher temperatures.  If this is the case, then shouldn't the low frequency input current noise increase into the typical 1/f curve at higher temperatures where diode leakage becomes significant?

Where input current noise dominates, it may be best to make a spot noise measurement of current noise between 0.1 and 10 Hz.  I am not longer sanguine that the datasheets can be trusted on this subject.

(1) Linear Technology comes to the rescue, as usual.  Their LTC6244 datasheet gives input current noise versus frequency for a CMOS input part with a relative high 1 picoamp input bias current and shows what I described.  The OPA928 datasheet shows it, but I thought it might be unique when I first saw it.

[Finderbinder]

At last my 10h log every 0.3s. Temperature drift clearly visible. Suspicious electrolytic capacitors were replaced, no other mods.