USA Cal Club: Round 2

[bitseeker]

I created a new subdirectory, Round2, on the Cal Club FTP at xdevs (thanks, TiN!) and uploaded my logs for the FX ref, PX ref, and some short logs of the 1kΩ and 10kΩ. I also included copies of the images I posted to this thread.

Attached is the 4-wire setup. The hollow plugs on these wires are a bit fiddly to get positioned correctly since the DMM's jacks are very deep.


For fun, I also briefly measured the PX and FX refs on the 100VDC and 1000VDC ranges. Both bench DMMs agreed with their 10V range. I also checked my Keysight U1282A, Agilent U1252B, and Greenlee DM-820A. They're all within tolerance. So, quite the happy family of multimeters.

For the resistors, both Agilent U1732B and DER DE5000 LCR meters agree with the bench meters.

[bitseeker]

And, no calibration check is complete without a little vintage Nixie fun. Looks like the resistance measurements could use just a bit of a tweak. Maybe after I restore it and clean all the switches it'll be spot on again.

[bitseeker]

OK, let me know if you guys want me to do (or redo) anything else with the cal kit. Then, I'll pack it all up and send it back to vindoline for verification.

[TiN]

34410A leaked out ppms (wrong range for 1Kohm resistor) 

Also good idea to use DCV monitoring with Hi-Z input impedance (but it's negligible error with FX).

Graphs were plot so fast! I am not used to so short sample files. My logs often span days, so I see little help of 1 hour log 

I only wish temperature/humidity to be logged along the readings to determine TC correlation. Then we can measure tempco of DUT, like DMM under test.

[bitseeker]

34410A leaked out ppms (wrong range for 1Kohm resistor) 

Regarding the range for 1kΩ resistor, I left the DMM on auto (unlike when I did DCV) and the CSV shows it chose the 1kΩ range. What should it have been?

Ah, I looked at the CSV for the 10kΩ resistor and the DMM chose the 1kΩ range. I guess that means it could've been forced to use the 100Ω range for the 1kΩ resistor. I guess auto mode was overly conservative.

Also good idea to use DCV monitoring with Hi-Z input impedance (but it's negligible error with FX).

OK. It seems some Cal Club documentation for recommended test setups, procedures, etc. would be helpful, especially for those of us with less metrology experience (*raises hand*).

Actually, TiN, is there a place on xdevs where I could at least start to document the inventory of what's in the club's cal kit, what goes where, and how it all fits into the shipping container? Then you or other experienced members could add a section of recommended settings, configurations, procedures, etc. for the most useful and consistent results.

Graphs were plot so fast! I am not used to so short sample files. My logs often span days, so I see little help of 1 hour log 

I'm going to have to write my own logger since BenchVue stops after every hour (plus it isn't all that reliable; I crashed it a couple times).

I only wish temperature/humidity to be logged along the readings to determine TC correlation. Then we can measure tempco of DUT, like DMM under test.

That's where it'd also be good to have my own logging software to synchronize the capture of the data from multiple DMMs, temp, and humidity. I can explore extending the tempduino's Python code to incorporate GPIB/SCPI.


Question for vindoline and TiN: Should I do more right now or move the kit along until I have my own logging setup for better data capture? I feel that I'm delaying others, but having good data is also important.

[TiN]

For ranges, should use fixed range for best fit value. 1KΩ resistor measure at 1kΩ range, etc. 100Ω for 1kΩ resistor wouldn't work, and 10kΩ for 1kΩ will give worse results, because it's only 10% of the scale and resolution meter otherwise have.

Usually idea is to collect as much data as possible, to determine correlations and to build up confidence.
I'm not in position to decide the flow, I'm on the wrong side of the globe  .

You could document and take notes, photos and what not, uploading into folder together with sample data, like you already have.
Then it can be taken by next member and updated/posted here in thread as well, so everyone have transparent trail of data.
It would be hard to recommend specific setups, as that likely to start limiting involved members, and ultimately defeats one of the purpose of a club?

Key point is to share everything, even something you think as obvious, as the next member in the loop may be not as familiar with setup/methods, and there are multiple ways to do same measurement. This way when all is transparent, it will be a good source to learn for everyone.

[bitseeker]

Thanks, TiN.

So, for the 1kΩ resistor, the CSV indicates that it was measured on the 1000Ω range. I still don't get where the ppms got leaked (back in your original comment). The DMM chose the correct range, did it not? (I'll be sure to use manual ranging next time, regardless.)

[TiN]

Didn't look on CSV for resistance, I looked at photo of setup in post #300. 

[bitseeker]

I see what you mean. That's interesting. What's also weird is that the CSV for both the 1kΩ and 10kΩ resistor indicate that the 34410A was on 1000Ω range. I'm really not liking BenchVue.

[MaxFrister]

I tried using bench view for logging a 34401A.  I had lots of problems with drivers and bugs.  I was also unhappy that it had no obvious way in incorporate non-HP equipment such as a temperature sensor.  There was also no (obvious) way to turn the VFD displays off.

In the end I ended up using a simple python script (I recently posted a version of it on the Datron 1081 thread).  This is simple and easily configured for non-standard instruments.  The result is a tab-deliminated file that you can then analyze with the tool of your choice: python, excel, alpha, whatever. 

The only downside is that there is no real-time logging but you can view snapshots of the logging at any time.

I know others have built much more elaborate logging systems but for me simpler is better.

[Conrad Hoffman]

Same here- simple data collection program with a field to put instrument specific codes in. It sits and collects data from my HP meters for later display in Excel. I do have a window where I display the last dozen readings so I know if anything has messed up.

[bitseeker]

Thanks, guys.

Spent some time today learning Python, PyVISA, SCPI, and how to configure, trigger and fetch data from two DMMs efficiently (i.e., integrate 100 PLC in parallel instead of waiting for each instrument to finish serially). Works great!

Next up, serial communication with the Tempduino and writing everything nicely to files.

[TiN]

If you have commercial GPIB dongle (Agilent one or more reliable NI GPIB-USB-HS) or GPIB-LAN VXI bridge there is no need to learn much.

10V logger python, that outputs values to console and saves into file:

Code: [Select]

#!/usr/bin/python -u

import sys
#import Gpib
import vxi11
from datetime import datetime

logfile_name = "test_scriptdata.dsv"

#inst = Gpib.Gpib(0,8)
inst = vxi11.Instrument('192.168.1.125', 'gpib0,13')
inst.timeout = 30
inst.write("*IDN?")
resp = inst.read(100).rstrip('\r\n')
print resp

inst.write(":FORMAT ASCII")
inst.write(":CONF:VOLT:DC")
inst.write(":VOLT:DC:DIG MAX")
inst.write(":VOLT:DC:RANG 10")
inst.write(":VOLT:DC:NPLCycles 100")

while True:
  ourdate = datetime.now().isoformat()
  inst.write(":READ?")   # read will wait the configured nplc time
  #inst.write("FETCH?")  # fetch will not wait
  data = (inst.read(100).rstrip('\r\n'))
  with open(logfile_name, 'a') as logfile:
        logfile.write (ourdate + (";%s;\r\n" % data))
  logfile.close()
  print ourdate, data


[bitseeker]

I'm using an Agilent 82357B for GPIB and USB (i.e., Python serial module) for the Arduino temp/humidity.

Yeah, not much to it once you learn the language, what modules and methods to use, SCPI commands, etc. Anyway, I have it doing what I want and will deal with the Arduino and file handling tomorrow.

[hwj-d]

@TiN,

first of all, big thanks for your great support.

Short question, your R5 (orig R2 70k) hasn't the same quality as R4 (orig R3 70k). I heard about, but don't really understand why it isn't really necessary to have (orig) R2 as VHP too. I too have two 75k VHP100 and four 71.5k RNC90Y laying arround, which I originally didn't want to mix on a pcb. But is it better to have one VHP 75k and one RNC 71.5k on one board to have the VHP's at the crucial point?

Thanks,
H.-W.

[hwj-d]

Maybe it's better to measure 4-wire the resistors without powered Ref.

[TiN]

This is what I received for USA Cal Club ref .

In reality using expensive VHP for 70K resistors is just waste of money. Sure they look nice, but there even 5ppm change of these resistors will be invisible in any typical setup, including 8.5d 3458A measurements. I have few LTZ refs with them as just 15ppm PTF56, they work just fine. Same goes to original HP 3458A A9 design. So you see fancy VHPs for these because "we can" not due to actual need.

[hwj-d]

Now I'm at the same point...   

[Conrad Hoffman]

IMO, one is far better off to do comparisons with their own personal reference resistors by current and voltage, rather than just using a meter. I have one L&N resistor (100 ohm) with a known value that I trust. Though they're not the greatest, they do well at constant temperature.

[MiDi]

@TiN,

first of all, big thanks for your great support.

Short question, your R5 (orig R2 70k) hasn't the same quality as R4 (orig R3 70k). I heard about, but don't really understand why it isn't really necessary to have (orig) R2 as VHP too. I too have two 75k VHP100 and four 71.5k RNC90Y laying arround, which I originally didn't want to mix on a pcb. But is it better to have one VHP 75k and one RNC 71.5k on one board to have the VHP's at the crucial point?

Thanks,
H.-W.

Cellularmitosis has collected the influence of resistors from several users.
The 70k have different influence on output deviation.

[hwj-d]

@MiDi

Ah, the Datasheet doesn't show. The impact of R3 is 10x vs R2.
Thanks.

But because mixing R3 75k with R2 71.5k doesn't matter?

[bitseeker]

Finished my Python logger and remote control program. PySerial was giving me some grief in Win 10 when trying to read from the Tempduino. Grrr. Then, I had to decode the goofy Python byte string. Ah, the joys of an unfamiliar programming language/environment. But it's working now.

I added some extra features such as the ability to turn off the displays to reduce wear on the VFDs or show some status info (see pic for temperature, relative humidity, and elapsed time).

Oh, I forgot to mention that this time the meters are set for 10G input impedance.

[TiN]

Impatiently awaiting for few day long CSV files to feed my matplotlib python plotter  .

[bitseeker]

Haha, yes. Waiting certainly isn't the fun part.

Each measurement takes three to four seconds for somewhere around 25K records per day. How long should I let it log before swapping the vrefs?

[TiN]

Depends on your patience and USA Cal Club schedule.
Full day or so usually good idea, so you can pick sweetspot data when ambient temp is about constant.
I like to log for a week to get decent confidence.