USA Cal Club: Round 2

[TiN]

Thanks for photos. I'd wish they would be bigger size, hard to see any details.

[rhb]

I'll upload a higher resolution image to your website.  This camera does not offer a reasonable image resolution for posting.  Eventually I'll get around to fixing the battery compartment door on my old camera and go back to using it.

Questions? 

The leads are all from the same piece of CAT 5, length variation is under 3 mm. 

Banana plugs are cheap Chinese with gold color plating.  No idea if it's gold or not.  The 34401As are on the bench surface because there is no way to reach the GPIB socket with them on the bench. Wires are secured to the banana plugs by a loop through the wire ring.

PSU is an Instek GPE-4323.  I'm using 1 & 2 for the FX set at 12.04 V and 4 set to 12.20 V feeding the regulator you supplied for the PX. 12.04 V is what the PSU claims.   I was so flustered I didn't stop to think about the regulator drop feeding the PX.   I'll check both after the run finishes.

My major complaint about the Instek is that adjusting the output  voltage closer than 50 mV is almost impossible.  I've had it apart to see if there was room to change the pots, but it's too cramped.  Though I might bring out leads for fine adjustment pots in a box on top.

I made a shorting plug for the FX from a dual banana plug with a piece of 12 AWG wire and a tag attached to the wire ring.  Inserting a bare copper wire into the FX banana jacks risks an ESD.  As the banana plug can be inserted without human electrical contact it reduces that risk.  It also makes it less likely that someone with vision impairment might power up the FX with a dead short across the output.  The tag reads "ANTI-STATIC SHORT"  on one side and "REMOVE BEFORE USE" on the other.  I plan to put heat shrink over the 12 AWG short  wire and fill the sockets with hot melt glue before I ship it out.

I'm sending out my Uno with the FW hacked to just spit T & H without the display working.  The original Tempduino board is untouched.  I think that should stay that way until we resolve the issues @bitseeker ran into.  The one I'm sending out is on an acrylic base with plastic feet,  so very easy to identify.

"Work Horse" at address 22 is on top with the "Closet Queen" at address 10 on the bottom.

I have a stock of Cinch banana sockets and should be able to find a suitable plastic enclosure, so if it meets with general approval I'm willing to repackage the PX so that all the connections are via case mounted  banana jacks before I place the kit in the post on Monday.  General idea is to hot melt glue the PX and VR to the case, trim the existing leads,  solder  and label the jacks. If I do that, I'll make another shorting plug like the one I made for the FX.

A transit temperature and humidity logger will take some time, so I'll send that to @vindoline separately when it is ready.

[rhb]

I see the size limits have changed. 

[hwj-d]

Are there something about verifying resistor values?

[rhb]

I am currently logging the 10K resistor on "Work Horse".  I plan to log both resistors on both meters up until I take it to the PO.

I'd like to propose some changes to the cal kit:

1) box the PX reference with the voltage regulator TiN furnished as follows:

 dual banana jacks for power in and voltage out
 dual banana jacks for the resistors, one pair for each resistor

2) add the following cables

 a pair of one dual banana jack to two dual banana jack cables
 a ribbon GPIB cable to operate a stack of two 3478A/34401A meters

3) add a 2nd shorting plug like the one I made for the FX

I don't have the required female Centronics IDC connector, but can make up the cable with the males I have and let the next person add the female. Or I can remove the Centronics connector from the AR488 and box it with the cable coming out the side.  Without a regular GPIB connector to space the AR488 out from the 34401A there is no way to connect it and the IEC power cord.  I think that placing the AR488 in a box with the pair of male Centronics connectors would be the best, but it would delay mailing until tomorrow.

The PX changes would allow logging both resistors to two meters simultaneously in addition to doing voltage and ratio measurements.  So in five logging runs with two meters the following tasks could be performed:

measure FX voltage on 2 DMMs
measure PX voltage on 2 DMMs
measure PX/FX ratio on 2 DMMs
measure 10K resistor on 2 DMMs
measure 100K resistor on 2 DMMs

Everything needed to do the above would be in the kit.  I still need to rewrite the logging software and clean up the Tempduino code.  But it would make the kit completely turnkey for the measurements listed above.

If the group wants me to make the changes I've outlined I need to hear by 4 pm CDT today as that is when I plan to post it back to @vindoline.  If I make the changes it will go out tomorrow instead.

Have Fun!
Reg

[hwj-d]

The quality of hollow bananaplugs are lightweight goldplated berillium-copper ones. So they are not on the cheapest side. The little plugs at the other side too. Keep in mind, that the mechanical quality of the precision socket is actually not suitable for others than quick cautious measurements with this lightweight cables and shortly plug them out after that. So I suggest we leave it at that.

[rhb]

I'll leave it as is.  I'd been logging the 10K on one meter and just moved it to the other meter before switching to the 100K.   I only proposed the changes out of concern that the connectors are not very robust.  I was a bit surprised when I saw the connectors on the PX.

I should have time to collect about 15,000 samples on each meter for each resistor before the trip to the PO.

[hwj-d]

I'll leave it as is.  I'd been logging the 10K on one meter and just moved it to the other meter before switching to the 100K.   I only proposed the changes out of concern that the connectors are not very robust.  I was a bit surprised when I saw the connectors on the PX.

I should have time to collect about 15,000 samples on each meter for each resistor before the trip to the PO.

Thanks.
The socket isn't that robust. The cables and their connectors are. You have a 1k and a 10k resistor.
 

[rhb]

Well, I'll actually have 5000 samples for each on the "Work Horse" and 1000 for each on the "Closet Queen".  Much too late I spotted a typo in my script and had "RESISTANCE" instead of "FRESISTANCE".  So I got lots of 2 wire data, but not a lot of 4 wire.

On the bright side, I should have an excellent measurement of the resistance of the wire leads ;-)

I'm just waiting on the 10K and will be heading to the PO in about 15 minutes.

BTW I started a thread on suggestions for round 3.

[rhb]

It's in the post and I've sent vindoline the tracking number. 

I fitted a space for the 2nd Uno with a note on the acrylic base what it is for.

The Tempduino FW on the board I added is hacked so the display does not show time T & H values.  I had to comment out a bunch of stuff to get it to respond quickly enough to a query.  It's right around 2 seconds to read time, temperature, humidity and two 34401As set at 100 NPLC using my scripts.

I reattached the LCD to the original Uno which has a cardboard base.  So everything is as received, but with a 2nd Uno added.

The one I added is just for backup in case the problem bitseeker saw reappears.  That way we don't need to change the original FW to test fixes.

Thanks to all for their contributions and patience.  I've shut down that computer for maintenance.  Once that is done I'll post the full data sets to xdevs.com.

Reg

[rhb]

Here is the full data set.  I'll let TiN transfer it to xdevs.com if he will be so kind.  I'm going to delete the earlier copy of the FX run data.

As I botched the script for the resistor data and did not get much data I am just postnig the voltage reference data.

[rhb]

I have started on the data analysis.  For the benefit of other members I should like to offer a few suggestions and comments.  It is unclear what data I actually got and how usable it is.

1)  If you do not have a GPIB interface already, make one using an RTC, Si7021, an Arduino and AR488 *before* the kit arrives.

2) Test your data collection setup with a 9 V battery if you have nothing else prior to the kit arrival. If you need help ask.  I'll be working on this over the next few weeks.

3) For characterizing noise, NLPC=10 is preferable to NLPC=100 if the instruments will respond to a query fast enough and you are logging automatically.  If you're doing it manually, use NLPC=100.

4) Synchronous sampling will aid in suppressing EMI.  I think that can be done, but am not yet sure if  or how.

Gross errors in my data acquisition procedure which I have identified so far.  There are likely more.

1) Failure to query setting after sending the instrument a command to set a parameter.

2) Stacking instruments and failing to shield them and the reference from room air currents.

3) Use of 1 second timing resolution in logging.  I'm not sure yet how to address this as my timing was the concatenation of the system, Tempduino and DMM response times.  There is a lot of variation in the interval between my samples.

So far, the only thing I can say about my data is that the noise on the Work Horse FX data is Gaussian distributed.

Have Fun!
Reg

[Conrad Hoffman]

How important is the consistency of the sampling?

[rhb]

How important is the consistency of the sampling?

I don't have the foggiest idea.  That's one of a long list of questions.  At the moment the list is getting longer rather than shorter.   So far I have only looked at Work Horse.  Closet Queen is next.  Work Horse has a nice linear ~0.3 ppm/C TC except at around 20 C there is a ~0.25 ppm jump above the mean based on an eyeball estimate.  IIRC that roughly corresponds to hysteresis in the FX reference as measured by TiN.

The Closet Queen TC is quite non-linear and much steeper.  I'm very curious about the ratio of the two DMMs  vs humidity.  Work Horse got a thorough wash with isopropyl.  The Closet Queen did not that I recall.  I'm pretty sure I opened it and looked at it for evidence of repairs, but I don't think I did anything other than put it back together.

I need to track down my post on the analysis of aliasing in a multislope ADC.  Another question is how is NPLC implemented in the 34401A?  Zero crossing count with a comparator or timer?  Short term power frequency fluctuations resulting from grid level load balance changes  would result in intermittent variations in the error statistics if it is timer based.  There would also be a regular change in the error statistics as they adjust the frequency coming up on midnight.

Edit:

Unknown irregularities in sampling will lead to an erroneous estimate of the exponent of 1/f**a noise.  It will introduce spurious noise at higher frequencies which will flatten the slope and lower the estimate of the exponent.

In the case of known irregularities in sampling, an L2 DFT is unable to correct the errors.  I *think* that an L1 DFT will correct them, but have not done a test.

[RandallMcRee]

Received the cal club shipment today.

Hooked up the FX reference to +-12v and powered on. Going to wait 24 hours before logging.

It has turned into a very busy week at work and so does anybody mind if I don't touch the PX reference? My wife thinks that I am also working on her gardening projects all weekend. Sigh. 

Preliminary measurements for FX reference are
K2001 (last cal date 5/2018 by TiN): 9.999997 volts
R6581T (cal by me to Fluke731B): 9.9999725 volts
with a fair amount of noise for both...

Other meters to check, but no logging scheduled, are an HP3456a (self cal) and a K2015 last cal 2016. I won't report on these here unless there is some interest?

Randall

[TiN]

Why wait 24 hours? Log away from the start, this way you can also check how long it takes to settle..

I won't report on these here unless there is some interest?

Up to you really, but I wouldn't see any issue getting as much data as there can be. 

[rhb]

I'd like to 2nd TiN.  The startup data is very valuable for the investigations I want to pursue.  You can always throw data away.  But you can't get what you don't have.

[RandallMcRee]

Argh. After reading your messages, and after work, I tried mightily to get logging going.  However, windows did its massive os update on my laptop and my prologix adapter no longer works—or its ezgpib, not sure.  I will have to debug it or use the arduino. Very disappointing.


Sent from my iPhone using Tapatalk

[TiN]

Sounds like we need to rename USA Cal Club to USA Gpib Club.
No-one have working datalogging from the first try 

[maxwell3e10]

It occurred to me at some point that one "modern" way of reading these instruments would be to use a webcam with image capture and OCR. I even looked a bit if anyone had written a program like that. There are GitHub programs for timed image capture and separately for OCR. It should be possible to write a program like that and it would also be useful for logging instruments that don't have a digital interface. ("modern" means to transfer 1 Mb of data to read less than 10 bytes).

Edit: Googling a bit more I just found a few possible programs:
https://github.com/skaringa/emeocv
https://www.pyimagesearch.com/2017/02/13/recognizing-digits-with-opencv-and-python/
https://github.com/xyk2/Scoreboard-webcam-OCR
https://www.unix-ag.uni-kl.de/~auerswal/ssocr/
https://github.com/LazizEx/WebAcm_Number_Detect

Edit2: The last program, available as an executable for windows here https://yadi.sk/d/FGzn5L6s3EK9Xr seems to be exactly what one would want. It even makes a real time plot and time stamp for saving to Excel. The only feature I did not find is switching between webcams.

[vindoline]

However, windows did its massive os update on my laptop and my prologix adapter no longer works

Yes, very aggravating. Thanks Bill Gates!

I'm a big fan of unsing a stand alone, headless raspberry pi as my GPIB logging solution. The whole setup isn't much bigger than the little GPIB interface itself and it can be tucked away in back of your meter if you wish. I connect to it via WiFi and the built in VNC server to control it.

[RandallMcRee]

Made progress--confirmed that windows update "lost" several of my usb drivers and reinstalled the important stuff...wireless mouse included.

So now logging is getting started.

I use Linux at work, again I know why.

Thanks,
Randall

[rhb]

Glad to hear you got it sorted.  I *really* felt beat up by the time I got stuff running.

I didn't participate in Round 1.  I'm hoping that by the end of Round 2 we will have the process worked out so that we can smoothly provide participants with an annual cal with good traceablility even if not quite NIST level.

Have  Fun!
Reg

[bitseeker]

Sounds like we need to rename USA Cal Club to USA Gpib Club.

  I've got plenty of GPIB stuff to be a member of that club. I'm not sure if that's good or bad.

[rhb]

I've discovered that the Tempduino reported a bunch of 2000+ C temperatures towards the end of my PX/FX ratio run and an old research code of mine is giving me occasional NaNs.  So I have some debugging to do.  And a *lot* of cleanup in the form of adding bounds checking, etc.  But I'm getting fairly close to being able to post plots and analyses of my runs.

I read through the Mega version of AR488 and it's *really* fine code.  So it will be a real pleasure to work on extending the functionality.

Have Fun!
Reg