Wescor Psychrometric Microvoltmeter( ~Keithley 155)

[TheSteve]

I recently purchased a pair of Wescor Psychrometric Microvoltmeters off a popular online auction site. TiN found them and pointed them out, and they looked interesting.
They are clearly a re-badged Keithley 155 Null Detector-Microvoltmeter with a few modifications made to them.
As you can see from the pictures there is an extra battery added(original 1.35 volt Mercury cell was still in one of them). That cell feeds a small daughter board mounted to the front panel which then leads to the two extra switches mounted to the front. The daughter board outputs go to the + front panel post as well as what they have labeled as "const". The "const" post is where the original guard connection would have been. Looking at the wiring it appears the guard connection is now tied to the - front panel post.

My plan is to restore both of them to proper operation as a Keithley 155 Null Detector-Microvoltmeter.
I have the service docs and schematics of the K155 but if anyone has any internal pictures of an original 155 showing the wiring it would be very helpful and would speed up the process.

[TheSteve]

A few more pics:

[KE5FX]

Interesting piece of hardware.  From the title, I expected something with a Church of Scientology property tag.    Apparently this sort of thing is used for HVAC measurements?

[TheSteve]

Interesting piece of hardware.  From the title, I expected something with a Church of Scientology property tag.    Apparently this sort of thing is used for HVAC measurements?

hah, that would probably be "Psychometric" instead of "Psychrometric". Just rolls off the tongue doesn't it?

There isn't much info out there on these units but I assume the mod applies a bias current/voltage. I see one application was for measuring dew point.

[TheSteve]

Spent some time today cleaning up one of the units. The binding posts will need some proper treatment. The outer case could use a new paint job and some new screws will be installed.
The original batteries have all been removed, the foam cleaned out, and 4 new 9 volts installed.

Tracing through the wiring it looks as though the meter works exactly as normal with the toggle switches in the default positions.
However The mod still needs to be removed to get rid of the extra solder/switch connections and to restore the guard post.

[maginnovision]

Pretty sad I didn't see or know about this. Been looking for one(but not 500$...) at least you're taking care of them though. They look pretty good. 

[branadic]

Pretty sad I didn't see or know about this. Been looking for one(but not 500$...) at least you're taking care of them though. They look pretty good. 

Agree, he is a lucky one. I'm also searching for a unit for quite some time now, but I'm also not willing to pay $500 for one. Congrats.
Maybe we should design a modern version of such a unit?

-branadic-

[Kleinstein]

A modern micro-voltmeter could be an interesting project, possibly not even that difficult. One has to find a balance between input bias and related current noise and on the other side voltage noise.  So there is likely not one good solution, but more like different ones for low impedance source and high impedance sources.

For the batteries I would consider a holder for normal AA cells. Not sure they really need 4x9 V. 2 x 12 V might be enough.

[TimFox]

"Psychrometric" refers to measurements such as wet-bulb thermometers for relative humidity.

[TheSteve]

Both meters are now operating 100%. The ancient batteries in the second one are working fine. I may still respray the outer casing and replace a few more screws. The original 155 also appears to have a pair of 0.01 uF 1100 VDC caps in series between - and case. These units only have a single capacitor so I will need to order suitable replacements.

Removing the Wescor customization is very easy. Removed are two toggle switches, a battery holder and a small PCB. The two soldered black wires are clipped off of the - post. The wire coming out of the "const" post is clipped and then the copper spade connected to the blue colored teflon wire for guard is removed from the - post and put on the const(now guard post). Then desolder the red wire from the power switch that goes to the toggle switches. Lastly desolder the reddish/brown wire that connects to the toggle switches(and + post) and solder it to the power switch where the red wire was removed. The two front panel holes are easily filled with screws etc.

[z01z]

Here are some pics to show the wiring, I hope they are of some use. Also, sorry about the quality...

[TheSteve]

Thank you for the pictures. They are indeed very helpful. The 0.01 uF caps between guard / - and case / - makes more sense. The schematic does not show the capacitors are connected guard and makes them appear they are in series.

[Echo88]

Tried to simulate the K155 according to schematic.
Sadly it doesnt work and doesnt care about input voltage or offset-adjustment. It settles always at about -7V with a bit oscillation.
Maybe i missed something or the transistors and P-Channel-FETs arent suitably chosen to work in this application?

K155-Measurement range in this simulation is 1V.

http://w.ko4bb.com/getsimple/index.php?id=download&file=06_Misc_Test_Equipment/Keithley/Keithley_155_Schematics.pdf

Edit: Thanks to Kleinstein the simulation now works. 
Edit2: Added less cluttered and labeled version.

[Kleinstein]

The choice of p-MOSFETs for the chopper is totally off.  These should be a rather small type.  Even the BSS84 as the smallest JFET I have in my standard list if on the large side, but the simulation than works. I would expect a FET more like 1/10 the size.

One could reduce the supply filtering to speed up the simulation.

P.S. From the Schematics the FETs used in the chopper look more like n-Channel with separate substrate.
P.S. According to the description the FETs are p-channel, and p-channel also works the right way around.

[Echo88]

My hero! 
Indeed BSS84 work.
DC-Amp-Feedbackresistor R56 (R150 in schematic) is correctly noted as 1GR in the schematic, but wrong in the BOM with 109R. Also C19 (C113 in schematic) is 2.2nF in BOM, but the 22nF seems to be the correct value as stated in the schematic and according to simulation.
Changed the .asc-file to work.

What would you suggest to change if one wanted to build a more modern K155-equivalent with at least the same specs?
Matched p-MOSFETs like the ALD1102 at the input and a few OP-AMPs like OP07 to substitute the factory selected transistors?
Due to the low used frequency and the small p-MOSFETs one gets very low input charge injection/bias-current.
If wed use big FETs like here on page 8: http://www.janascard.cz/PDF/Design%20of%20ultra%20low%20noise%20amplifiers.pdf
and increased chopper-frequency then wed get way lower noise, but more charge injection/bias-current/EMI i suppose?

https://www.mouser.de/datasheet/2/8/ALD1102-387.pdf

Never seen a thoroughly tested/specced DIY-nullmeter.
Conrads DIY-nullmeter seems to miss the input-bias-spec? http://conradhoffman.com/mini_metro_lab.html

[texaspyro]

i thought this thing can measure water density in the air?

I thought the thing could measure somebody's tendency to become a mass murderer...   

[Kleinstein]

For a DIY null-meter I see several possible solutions:
1) a low bias AZ OP, e.g. AD8628 (not very low noise, but relatively low current noise). It may need some bias current compensation.

2) a circuit similar to the K155, but with modern parts. I would prefer n-channel FETs over p-channel. One could also try JFETs instead of the MOSFETs. For the AC amplifier I would prefer JFETs. The circuit from the link, using a MAX4693 for the chopper is also not that bad. With adjusting the supply to the switch chip only could have a way to trim charge injection and thus bias current.

3) Using a polarity reversing chopper with 4 FETs, e.g. like the input amplifier of the Datron1281 (chopper with 4x 2N4117 JFETs and JFET differential amplifier, 4066 for demodulation)

4) Chopper with Photo-fets (e.g. H11F1)

Integrated CMOS choppers use a rather high frequency, because the CMOS amplifier has quite some 1/f noises and because they want to avoid large filter / coupling caps. A discrete solution has no problem with caps in the µF range. For low bias current, I prefer a rather slow chopping (e.g. 30-500 Hz) and an AC amplifier that also works well for low frequency (e.g. audio grade JFETs).

[Echo88]

You raise some good points.
I think theres no AZ-OP-Amp with the necessary low current noise (since its specced wrong by the manufacturers) and id like to keep the design as free as possible from switching artifacts, since id love to avoid that headache.

I think the following basic designs are possible:
Slow discrete chopper (like K155)
JFET-Op-Amp-Frontend based on a very good fitting type (found none with needed specs)
JFET-Frontend with Autozero (like suggested D1281-design)

ADA4625 came to mind, but its current noise is too high and produces much noise at the simulated 752A-balance-bridge. Bridge consisting of 2x100k and 2x100K in my case (Fluke uses 2x40k and 2x120k) led to 131nVrms.

The D1281-idea seems good: Low noise N-JFETs, Autozero all x seconds instead of chopping.
Assuming that one uses a normal DMM to measure the Nullmeter-output-voltage and doesnt need a continous signal like from a chopper-design.
How is the measurement cycle in the D1281? Measuring, then shorting the input path with the N4117A or does it also switch input polarity to average thermal EMF?
The H11F1 are indeed interesting to use as input switches, maybe one can obtain much reduced charge injection compared to JFET-switches.
Never seen actual charge injection measurement data of them anywhere. 
Attached is the nullmeter bridge-simulation with the used ADA4625.

https://www.analog.com/media/en/technical-documentation/data-sheets/ADA4625-1-4625-2.pdf
https://www.analog.com/en/license/spice-models?mediaPath=media/en/simulation-models/spice-models/ADA4625-1.cir&modelType=spice-models
https://eu.flukecal.com/de/literature/product-manuals/752a-instruction-manual

2SK170,LSK389A

[Kleinstein]

The DA1281 input uses relatively slow chopping. I don't know the frequency, but it should be something around 1 KHz or so. The chopper (4 x 2N4117 FETs as switches) is reversing the polarity. So it produces a kind of alternating offset and the relatively slow ADC is used to average over this.
The 4 switch solution is usually lower voltage noise, but can be slightly higher current noise. However there is quite some compensation in the charge injection, so that the switching spikes may be even lower, especially if well tuned / adjusted.


A possibly suitable JFET OP for slow (possibly 1- 10 Hz range) chopping could be the OPA140 or OPA145. There are also discrete JFETs that could be used. This also applies to the 2 switch chopper like in the K155. Anyway, with a resistance bridge like 40 K - there is already quite some noise. So no need to have a super low noise amplifier. 40 K source resistance already have some 25 nV per square root hertz of noise.
This is also a challenge in designing a filter for the spikes.

The H11F1 and similar have essentially no charge injection. However they are slow and the heat from the LED lead to a thermal gradient and thus thermal EMF. I don't know how much this would be.

[MiDi]

@Echo88: you have already posted some research for H11F-Series.

Seems you have to decide which death you want to die: charge injection vs. leakage/ (low frequency) input impedance.

For fast switching AZ-OP AMPS it could be sufficient providing low enough impedance at the op input at and above the switching frequency to bring charge injection to tolerable limits.
For low frequency choppers, filtering could become quite bulky.

[Echo88]

I was under the wrong impression that the D1281 is no chopper design, but uses the 2N4117 as slow autozero-switches (as in 100NPLC -> 2s measurement of input, 2s measuring offset).
Cant find the Clock-L frequency in the manual, maybe i should run an OCR-tool across the pdf to make it searchable...meh.
Attached a D1281 simulation, which doesnt yet work.
Youre right, the high bridge-resistance completely completely destroys the low-current-noise-approach with its thermal-noise of 128nVrms -> ~850nVpp for a bridge consisting of 4x 100k-resistors alone.
So either a slow-chopped low noise JFET-OP like OPA140 + an additional gain OP or a discrete JFET-frontend like D1281...which looks nice since its already designed and working.
Dont know if id need the bootstrap-supply though for increased CMRR in this situation?

During playing with my K155 i found that it doesnt have 0.1-10Hz-spec, but instead a specced rise-time of 5s at the 1µV-range which roughly explains the measured noise of ~900nVpp (measured with 34465A at K155 analog output over half an hour) measured at an 900kR-resistor.

Is an amplifier-concept which uses a slow automatic offset compensation possible, which would short the input every 10s and then after measuring offset it would be automatically compensated?

Edit: Attached working simulation of a chopper taken from http://www.janascard.cz/PDF/Design%20of%20ultra%20low%20noise%20amplifiers.pdf page 8. Need to attach D1281_simulation...
Edit2: Attached non working simulation chopper taken from https://www.eevblog.com/forum/metrology/low-frequency-noise-of-zero-drift-amplifiers/?action=dlattach;attach=663045

[Echo88]

Found a paper about the H11F1 when it was invented: inner details, more graphs and a few example schematics:
https://ieeexplore.ieee.org/document/4043078 available via scihub
Interestingly "no charge-injection" is never mentioned.
Figure 8 "Offset voltage" is relevant (missing from the contemporary datasheet), my 220µV measured with activated emitter fits perfectly.

Not the perfect place for this note in this thread, found no better fitting thread so far.