With AZ:ON and OCOMP:OFF?
how was the variation of room temperature at the time ?
AZ is always on, the meter has pretty crappy performance without it. OCOMP is not used on HP3457 for resistance ranges greater than 3 kOhm.
Temperature I am not sure, just sitting on dinning room table overnight. I am planning to setup a temperature logging system at some point.
I will try to log some data with 10k or 1k WW over night , with my 3457
Feel free to prove me wrong and show better data with another precision DMM
Here is 0.03 ppm sdev with a sample size of 64
Its a cool machine! But you couldn't measure an 8.6 kOhm resistor with it
set up my 3457 as :
NPLC 100
AZ ON,
OCOMP OFF,
DELAY .05,
OHMF,
NRDGS 100,
MATH STAT,
TRIG SIGGLE
MEAN is 10.000358e+3
SDEV is 3.0709574e-3
about 0.3ppm against the mean.
CalMachine you have so good machine
To get full performance, one has to use the hires register, which is not included for built-in statistics. I read the data over GPIB and add the hires register. The code is here
https://www.eevblog.com/forum/testgear/$5-usb-gpib-adapter-for-ezgpib/msg1650155/#msg1650155
thank you Maxwell,
but my 3457 gpib has intermittent issue on sending and receiving.
especially on receiving, it would report a "bad header" by ERROR key, and sometimes send message also reports an error.
so for now I would only use front panel key operations...
I thought 4950 can only measure around round values. But I guess if you select 100 kOhm range and 0% band, it still works pretty well.
I thought 4950 can only measure around round values. But I guess if you select 100 kOhm range and 0% band, it still works pretty well.
Nope, I'm on 100% band on 10 kOhm. I just turned off the bandwidth limitations
Interesting. So are the measurement bands just a firmware feature or is the hardware actually built around taking differences from a few discrete values?
4950 is like a mutant baby of 1281 DMM and 4920 AVMS. But that's offtopic here.
I will have something to say about this thread, but not now.
Interesting. So are the measurement bands just a firmware feature or is the hardware actually built around taking differences from a few discrete values?
I believe it is mostly a firmware feature. I also believe the stated specifications are only valid within the specified bandwidth limits.
Not sure what's being measured here. We know the Johnson noise of a wire wound resistor with high accuracy for any given value, and excess noise is near non-existent. If you're not getting that number, something else is being measured.
the setup was allowed to sit over night.
I just repeat the measurement from front panel.
took 100 samples.
MEAN: 10.000380E+3
SDEV: 0.0000000E+0
I also believe Conrad's comment, 3457 has limitations if we measure the noise only from the resistor.
You got lucky that fluctuations were entirely within the 7th digit.
The noise in resistance measurements is just voltage measurement noise plus current noise of the current source. The Johnson noise of the resistor is usually smaller, unless the resistor is approaching 1 MOhm.
I even did it again, same results.
also use RMEM to recall 100 readings, all the same 10.000380
That's impressive stability! Try it again tomorrow
Not sure what's being measured here. We know the Johnson noise of a wire wound resistor with high accuracy for any given value, and excess noise is near non-existent. If you're not getting that number, something else is being measured.
I remember the JCan project from Joe Geller.
He closed Geller Labs, but details are still available.
For anyone interested in Johnson noise see
http://www.gellerlabs.com/JCan%20Parts%20and%20Kits.htm
I built that and suspect it's actually better than he and others are willing to claim. The problem is you need FETs that aren't available anymore.
You got lucky that fluctuations were entirely within the 7th digit.
The noise in resistance measurements is just voltage measurement noise plus current noise of the current source. The Johnson noise of the resistor is usually smaller, unless the resistor is approaching 1 MOhm.
A well made resistance measurement can eliminate the voltage reference noise to a large part. So it would be the voltage noise for a measurement near zero and the current noise without the contribution from the voltage reference that would matter. With some meters there can also be a significant current noise from the voltage reading input, that can get important for high resistors.
I even did it again, same results.
also use RMEM to recall 100 readings, all the same 10.000380
If the noise is low, that can be rounding artifacts.
leave the setup to get some data last night.
I overcome my 3457 GPIB read issue by re-initialize the bus (IFC), if an issue is encountered.
so after IFC the 3457 would be *happy* again to talk to the PC which makes it be able to read back the data continuously
data was taken each 7s or about 12s if a read issue occurred, 3457 setup is 100PLC, manual range, DISP OFF.
I forgot to get extra 7 digit in HIRES register, however the 10 hours run seem to be stable with 6.5 digits, a few outliers observed.
by utilizing RMATH HIRES, one can even get *8.5* digits.
took 100 samples,
SDEV is about 0.13ppm against MEAN.
then we can see some instability.