Electronics > Metrology

Strange asymmetric nV noise of Keithley 148

(1/5) > >>

yus:
I recently bought a Keithley 418 with tube (not FET) input. I bought it out of pure curiosity, I do not need it for living.
I am a physicist by education (programmer by occupation) and I was curious if it can really measure voltages below 10nV.

On arrival I had to:
1) fix the power cable connection, fortunately I noticed an issue before a major AC 220V short arrived,
2) hack the chopper+demodulator oscillator : to be able to set the required 94Hz, I adjusted R320 to a slightly lower value*. The oscillator runs stably within specified +-0.1Hz instability,
3) hack the 94Hz amplifier-demodulator : to be able to set the required response time to about 0.25sec for 100nV step signal, I adjusted R169 to a slightly lower value*,
(*) adjusted to a slightly lower value = to be about 25% lower. This was a quick hack to verify if the instrument is doing something reasonable at all settings.
What good I have after the hacks above:
1) The calibration appears to be very close to the specs even on the lowest 10 nV scale.
2) 24 hour drift I estimate to be of about +-6nV max non cumulative (=around initial value at 0 time). It may probably be up to 2 times higher than a better new instrument or up to the specs, please see the picture of the expected long term drift from Keithley document The-measurement-of-nanovolts, attached pic 5. I do not want to actually run it for 24 hours due to the ugly annoying 2KHz noise of its power supply.

What I do not understand and consider an issue is the "short-time" noise I see with the input shorted with a copper wire thermally isolated at the lowest 10nV scale. I would expect the "short time" noise to be up-down symmetric. Obviously it is not. Almost always the voltage slowly rises and then sharply drops 1-2 nV on average. Please see the attached pictures 1-2. This is not what the 1/f "flicker-noise" is supposed to look like. This is also not what I see on Fig 13.b , attached pic 5.
Chopper+demodulator oscillator as well as amplifier-demodulator at first sight are not supposed to cause any up-down asymmetry even if they may  function to some degree incorrectly. Since the max long term drift is below 6nV, no asymmetry is noticeable at 100nV scale or above.
If the mechanics in the chopper is functioning somewhat incorrectly, say it is worn above specs, it would cause increased noise/drift, but also symmetric.
All the transistors are probably old germanium ones, I cannot even find their specs anywhere, probably noisy and leaky, still no asymmetry is expected.

On the pictures and supporting doc-s:
5) Expected short-term and long-term drift of Keithley 148 from Keithley document The-measurement-of-nanovolts, by Julius Praglin. In the same document there is the actual amplifier+demodulator+power_supply schematics for the tube variant. Vertical scales indicated a=6nV, b=2nV, c=2nV.
1-2) The actual "short-term" noise/drift on 10nV scale. Where after cleaning I can see occasional rare up-jumps and/or down-drift, encircled with green ellipses, otherwise it is always slow rises and sharp drops. Do not pay attention to the "needles" on the pictures, they are picked up from a pc-power-supply and a lamp power supply nearby. It will change nothing if I turn them off, already verified. 200mV per division vertical scale of the linear output corresponds to 2nV input. Horizontal scale is 100s per division.
3) What was cleaned up physically in the nanovolt circuitry,
4) What was cleaned up shown on schematics,
6) Keithley 148 manual with FET input schematics, see: https://www.tek.com/en/manual/model-148-nanovoltmeter-instruction-manual-29029-rev
7) Keithley document The-measurement-of-nanovolts, by Julius Praglin with Keithley 148 tube input schematics, will upload if requested.

My suspects for the asymmetry were:
* Oxides at the nV circuitry. So I cleaned what I suspected could be the most critical contacts: input contacts, chopper wire contacts outside + 1Ohm resistor in the feedback divider, see the green ellipses in pictures 3-4. This may have changed the "short-term" noise/drift a slight bit in that on rare occasions some minor up-jumps appeared as well as some minor slow movements down, green eclipses on pics 1-2.
* The chopper innner contacts may be oxidised,
* Maybe some strange "out of phase" asymmetry of the demodulator, though I cannot imagine how it may cause any up-down asymmetry.
* Even if the oscillator were not stable, I would had not being able to imagine how it would have caused any up-down asymmetry.
====
My final guess is that to get a better idea I would have to completely disassemble and clean all the nV circuitry including the chopper inner contacts.
The complete careful disassembly is necessary because bending some of the copper wires can easily break them, especially the short ones of the red
feedback 1Ohm divider resistor. This sounds like an 8 hour project. May be I find some day for it in the future.

Why large jumps is an issue and slow drift is no issue for the accuracy:
If no jumps then after zeroing, I would expect some drift of the order of the square root of time. Then the actual resolution would indeed be better than 1 nV as Keythly 148 manual says, if zeroing and measurement time is limited by say 1 minute. That is zero and measure something quickly enough before the instrument drifted too much. With jumps it is just possible to say that the measurement accuracy is +- the jump value, with the average jump of about less than 3nV. But on the 10nV scale the 3nV random jump may be several times above say 1nv signal.

Questions:
==========
* If you have Keithly 148, could you please upload a picture of its "short-term noise" with input shorted by thermo-isolated copper wire for at least 20-30 min at the corresponding scale ?
* What is your idea of what can cause this type of up-down noise asymmetry ?

doktor pyta:
You have probably seen my old post: https://www.eevblog.com/forum/repair/keithley-148-nanovoltmeter-repair/msg1393448/#msg1393448
From my experience nanovoltmeters are very troublesome to repair.
Intermittent, oxidized copper to copper crimped contacts are common issue.
The electromechanical chopper have ~2000h of service life so you may be facing wear out of the contacts.

yus:

--- Quote from: doktor pyta on May 16, 2023, 08:21:08 am ---You have probably seen my old post ...

--- End quote ---
First of all thanks a lot for answering. Yes, I have read your post with lots of attention before deciding to buy this thing.
If you still have data of that post, could you upload it please ?
Later you posted the noise after replacing the electromechanical chopper. Unfortunately the vertical scale is 10nV so it is
not really possible to see much details. From that little that I could see on picture record ca. 13h sampled every 5 sec
the long term drift  is +-6nv and I do not see any jumps.

--- Quote from: doktor pyta on May 16, 2023, 08:21:08 am ---The electromechanical chopper have ~2000h of service life so you may be facing wear out of the contacts.
--- End quote ---
Keithley 148 manual says that a worn electromechanical chopper will be seen as drift in tens of nanovolts.
I have the long-time drift of about +-6nV same as what you posted after the chopper replacement.
Before replacement with a bad chopper it looks like you had about +-35nV long-time drift, that agrees to what Keithley manual says.
from your noise picture record ca 22h sampled with 90s interval it is difficult to guess if there are any jumps or not.
Response time (10%-90% of signal) for Keithley 148 at the 10nV scale is about 5 seconds as can be seen from pic 5.c .
It would be nice to have the noise data at 10nV scale with at least 5 sec sampling, 1sec would be even better.
I also have thoughts on the claimed 2000h of service time for the electromechanical chopper. The cars from those years 1970, many still run well despite 50 years of service. So the milage can vary, I guess.

Questions: If you still have Keithley 148, do you see any nV jumps up or down at 10nV scale? What is the "short-time" noise of your device ? Is it less than 2nV peak-to-peak over say 5 min time at 10nV scale as Keithley manual says (picture 5.b of my initial post) ?

The noise I have, it looks to me as if there is some "nanovolt zener diode" there. And a random nanovolt thermal noise source is charging a capacitor through
the "nanovolt zener diode" in forward direction. After reaching some critical voltage (that also fluctuates around some mean value of about 1.5nV) the breakdown occurs and the capacitor discharges. Then it continues like this again indefinitely.
I checked the web and indeed it might be relevant somehow :
https://www.sciencedirect.com/science/article/abs/pii/S0167577X13015899
https://www.sciencedirect.com/science/article/abs/pii/S0925838810012284
...Copper oxide (CuO) is a p-type metal oxide semiconductor with a narrow band gap (1.2 eV). It has received much attention because of its various applications in optoelectronic devices, gas sensors and catalysts...
1.2V is not 1.5nV, but who knows. It might be that some unwanted / unwelcome HiTech science sneaked secretly into the nanovolt circuitry of this old box.
Looks like cleaning is unavoidable >:( It may help or not, we'll see later.

Whitefoot:
I'm not very familiar with the Keithley 148. I haven't touched it for decades. I pulled it off the shelf, turned it on and let it run over night. Then I scoped the output for a couple 25 minute runs. The 148 is set to 10nV full scale. The scope is set to 200mV/div (2nV/div at the 148 input) and 100S/div timebase. One trace shows the 148 without zeroing - the other trace shows it zeroed with the zero suppression controls.

Whitefoot:

Navigation

[0] Message Index

[#] Next page