Author Topic: Nanovoltmeters Performance (Read 21053 times)

Alex Nikitin · « **on:** May 17, 2018, 11:34:43 am »

Starting a separate thread, here are some graphs from my Keithley 181 .

Data recorded by the K34465A from the analogue output of the K181, vertical scale is in microvolts.

Cheers

Alex

maxwell3e10 · « **Reply #1 on:** May 18, 2018, 02:16:01 pm »

Here are the data for HP34420A measured overnight. I used the proper low-noise cable with the copper lugs shorted. The meter is set to 10 PLC and no filters. On 1 mV scale this gives a reading at 2 Hz. It's read over RS232 in talk-only mode.

To compare more precisely, we need to define the measurement bandwidth. For the Keithley 181, was the damping on or off? For the Keithley 2182A, I am not sure how the settings for analog and digital filter translate into a bandwidth?

Another way to look into it is to plot the Allan deviation. For HP34420A it crosses 1 nV at about 10 sec. The best resolution is about 0.4 nV at 100 sec and the long-term drift is about 2 nV.

The input noise density is 5 nV/Hz^(1/2). This is only a factor of 2-3 better than typical chopper-stabilized op-amp noise. Can one build a nano-voltmeter just by averaging 4 LTC2057?

guenthert · « **Reply #2 on:** May 21, 2018, 11:21:22 am »

Quote from: maxwell3e10 on May 18, 2018, 02:16:01 pm

[..]
The input noise density is 5 nV/Hz^(1/2). This is only a factor of 2-3 better than typical chopper-stabilized op-amp noise. Can one build a nano-voltmeter just by averaging 4 LTC2057?

Only? I think the fact that it is (including input protection) better than a low-noise op-amp is remarkable. How do they do it? Some funny business with unobtanium matched FETs, discrete chopper or even transformer?

Kleinstein · « **Reply #3 on:** May 21, 2018, 11:38:43 am »

The 34420 uses a pair of large, very low noise, expensive (though not unobtainable - at least a replacement) JFETs together with the normal auto-zero function of the ADC to do a kind of chopper mode. So nothing really magic more like brute force. The performance might even get slightly better with less than 10 PLC and averaging afterwards.

maxwell3e10 · « **Reply #4 on:** May 21, 2018, 03:30:57 pm »

The noise spectral density is the same on all speeds. There is a small change in reading speed on 1 mV scale vs. 10 mV scale, but the spectral density remains the same still. They maybe using internal autozero at a different rate than the measurement PLC setting. On this meter one cannot turn off autozero.

The real king of chopper amplifiers is here:

TiN · « **Reply #5 on:** May 21, 2018, 03:41:40 pm »

Today Fedex birdie brought some reinforcements (two more brand new A10s). On the right my old A10 (mfg date 2010) which somebody (before me) butchered by soldering on input posts.

.

Will need to fire my X1801 setup with K2002's and test these.

egonotto · « **Reply #6 on:** May 21, 2018, 05:13:06 pm »

Hello,

I try to control the Keithley 181 with a Keysight 82357B GPIB-USB Interface, but without success.

With Keysight Interactive IO I got:
"
* Connected to: GPIB0::2::INSTR
-> R4
! VI_ERROR_TMO: A timeout occurred
Visa ErrorCode: 0xBFFF0015 (-1073807339)
"

Sometimes i can read the voltage from the 181, but my bootdisk crashes and now it gives only error.

How I can control the 181?

Best regards
egonotto

lukier · « **Reply #7 on:** May 21, 2018, 05:16:42 pm »

Quote from: TiN on May 21, 2018, 03:41:40 pm

Today Fedex birdie brought some reinforcements (two more brand new A10s).

How much for a new A10 amplifier if you don't mind? Unfortunately, no pricing info on the EM Electronics website $:-\$

I'm asking as I would like to have a nanovoltmeter, but the big boy devices on eBay are quite expensive and I thought just the amplifier would suffice.

branadic · « **Reply #8 on:** May 21, 2018, 05:52:24 pm »

Hi Manfred,

nice to read you here. You can set GPIB address with the switches at the back of the device. Do they match with your address in Keysight Interactive I/O?

-branadic-

egonotto · « **Reply #9 on:** May 21, 2018, 06:38:22 pm »

Hi André,

I set the GPIB address to 2 and the connection expert find the 181.
And now I can read the values. But i can not write a command.

The input is open at 2V so it drifts.

Supposedly the software from Keysight can only speak with newer equipment that understand *IDN?

In the manual of 181 the GPIB is not good understandable for me, but in the manual from Keithley 195A there is a good description about GPIB.

Now i search a program that can speak with the 181.

Best regards
egonotto

HighVoltage · « **Reply #10 on:** May 21, 2018, 07:52:28 pm »

Isn't one of the biggest limitations of the nanovoltmeter the Johnson noise, as the noise of the thermal motion of electrons.
I think I read in some older Keithley books that this is the ultimate limitation of nano volt measurements.

branadic · « **Reply #11 on:** May 22, 2018, 07:21:07 am »

Quote

Supposedly the software from Keysight can only speak with newer equipment that understand *IDN?

GPIB commands are more or less universal, that means you can also control a device from a competitor with Interactive I/O with the correct commands. The GPIB commands can be found in Operator's Manuel page 4-1 ff. For example R1 to R7 sets the range from 2mV to 1000V, B0 and B1 the resolution (5.5 Digits or 6.5 Digits).
Please note, that you can set the mode of the device. In talk only mode the device will ignore bus commands (page 4-2, chapter 4.3 and figure on page 4-3).

-branadic-

Echo88 · « **Reply #12 on:** May 22, 2018, 10:38:17 am »

I use my K181 in talk-only-mode, since i also couldnt get it to answer on requests in GPIB.

@HighVoltage: Indeed, Johnson-noise limits the measurements. See for example: http://www.emelectronics.co.uk/a20.html -> "

Noise.
Equivalent noise resistance less than 10 ohms. Noise voltage depends on bandwidth e.g. rise time constant 10 seconds gives peak to peak noise voltage of 250 picovolts."

"With source resistances below 10 ohms, the noise is substantially
constant down to zero ohms. Above 100 Kohms, the noise voltage
increases over the thermal noise characteristic by about 10 dB per
decade of source resistance."

Alex Nikitin · « **Reply #13 on:** May 22, 2018, 11:17:49 am »

Here is the noise of K181 (input shorted by 10 Ohm) measured over 10h period at a relatively stable room temperature, and the resulting Allan Std Deviation graph.

Cheers

Alex

maxwell3e10 · « **Reply #14 on:** May 22, 2018, 01:53:10 pm »

Something is funny about this Allan deviation plot. From the time domain data one can see a lot of 1/f noise on the time scale of a few hundred seconds, yet the Allan deviation looks as if the noise is white. Also on long time scales the drift is more than 1 nV.

Kleinstein · « **Reply #15 on:** May 22, 2018, 03:08:49 pm »

There are only a few data-point on the Allan deviation curve. For the time domain data I am not so sure there is significant 1/f noise - I looks a little there might be some narrow band signal superimposed like power line aliasing of some kind. This might get lost with only a so few points in the Allan deviation curve.

branadic · « **Reply #16 on:** May 22, 2018, 04:30:18 pm »

Maybe Alex can share his original data to check that this is not a problem of maloperation of Allan Deviation plot? Voltage data has to be taken as frequency information not phase information.
If someone out there with software skills would like to reprogramm Stable32 or Plotter for voltnuts?

-branadic-

Alex Nikitin · « **Reply #17 on:** May 23, 2018, 07:12:08 am »

Quote from: branadic on May 22, 2018, 04:30:18 pm

Maybe Alex can share his original data to check that this is not a problem of maloperation of Allan Deviation plot? Voltage data has to be taken as frequency information not phase information.
If someone out there with software skills would like to reprogramm Stable32 or Plotter for voltnuts?

-branadic-

Here is the data I've used for the plot. Values are in microvolts, sampled at 10s intervals, NPLC100 on the K34465A .

Cheers

Alex

branadic · « **Reply #18 on:** May 23, 2018, 07:39:41 am »

Hi Alex,

I can reproduce your Allan Plot.

-branadic-

maxwell3e10 · « **Reply #19 on:** May 23, 2018, 12:56:24 pm »

That is consistent with my calculation as well. I didn't notice that in the previous time domain plot the red line represents a moving average. This appears to introduce long-term correlations which are not there.

branadic, can you calculate the Allan variance for the earlier data with Keithley 2182 that you posted on the other thread.

branadic · « **Reply #20 on:** May 23, 2018, 01:49:55 pm »

Quote

branadic, can you calculate the Allan variance for the earlier data with Keithley 2182 that you posted on the other thread.

Did that already: https://www.eevblog.com/forum/testgear/keithley-2182a-digital-nanovoltmeter/msg1531166/#msg1531166

Alex Nikitin · « **Reply #21 on:** May 23, 2018, 02:01:26 pm »

Here is the k181 noise AD plot for 1s sampling on K34465A. The increase over 1000s is purely the room temperature effect.

Cheers

Alex

maxwell3e10 · « **Reply #22 on:** May 23, 2018, 02:04:43 pm »

wow, that's a big bump in the Allan variance in both K2182A and K181. Looks like these Keithley meters have the same problem as many others with the autozero algorithm.

Alex Nikitin · « **Reply #23 on:** May 23, 2018, 02:16:09 pm »

Quote from: maxwell3e10 on May 23, 2018, 02:04:43 pm

wow, that's a big bump in the Allan variance in both K2182A and K181. Looks like these Keithley meters have the same problem as many others with the autozero algorithm.

The "bump" around 10s mark is just the result of the meter's integration time. I am sampling the output faster than the meter time constant. If I use the additional internal filter (Filter ON), the time constant increases and the "bump" moves to ~60s mark.

Cheers

Alex

branadic · « **Reply #24 on:** May 23, 2018, 04:34:19 pm »

Quote

wow, that's a big bump in the Allan variance in both K2182A and K181. Looks like these Keithley meters have the same problem as many others with the autozero algorithm.

Well, the measurement I made was without the ambient temperature being stable. Maybe in a controlled lab environment results are better? Will test that maybe some day.

-branadic-


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Nanovoltmeters Performance (Read 21053 times)

Share me