Bench multimeter readings and drift data over 3 years

[nukie]

Hello guys,

I am sharing some data for my collection of multimeter and their readings over the pass three years(nearly). Over the years I also bought a two additional multimeter so their readings did not get recorded earlier.

I have a few precision voltage reference for deploy but unfortunately due to time constraints I have only built one and purchased one.
They are Linear Tech 1.25v LT1790A 0.05% 10ppm and Geller Labs 5.00v AD586KQ 15ppm voltage reference board. The Geller board is an interesting board, it has a jumper that enables trimpot correction that was calibrated through a Fluke 732B.

I should build the National LM4140BCM 0.1% 6ppm and the Analog Devices AD780BN +-1mV 3ppm which I have on hand. So I can have more reference. These are the voltage readings between a time span of 2 Years 11 Months 1 Day. You might want to click on the pictures to expand it.

They are all powered for 30 minutes warm up before readings were taken. I see only about 0.1mv drift. I am not going to make any conclusions on the data because I am not clever, so I let you guys decide. For all I know, I am quite happy with my very old and weary Fluke 8840A. I have fear one day it will just give up as the memory storage in the Zilog controller might just fade away...  

[saturation]

This is good to see.  Your Geller boards, what was the calibration voltage supplied by Geller for the uncorrected values, do they match your 3468a measurements?  More later while I study your good results.

[Kiriakos-GR]

Well at this point I had to ask ... are those measurements taken with the original cables - leads ?  

Just from the leads , there is possible to be some deviation.

And another factor could be the room temperature , that effects the resistance of everything.  


Here is a link from my own reference IC , and tests when it is powered with 12V and 15V ..
The results looks interesting.

https://www.eevblog.com/forum/index.php?topic=1148.msg14949#msg14949
 

[saturation]

Here's a quick and dirty stat sheet.  Puts the changes in visual form.
Raw excel sheet provided for those who wish it.

Shows what you pay for in a Fluke or Agilent.

I have one 20 year old FLuke 85i, only 3.5 digits but cal is stable for 20+years. Also a 15 year old Fluke 85i, 10 year old 85iii, ~ 5 year old Fluke 87V all in cal.

To read the graph:

2008-1790: original measurement of the LT1790
2011-1790: update measurement of the LT1790 3 years later

2008-5: original measurement referenced to the Geller 5V board
2011-5: update measurement referenced to the Geller 5V board 3 years later

2008-5u: original measurement referenced to the uncorrected Geller 5V board
2011-5u: update measurement referenced to the uncorrected Geller 5V board 3 years later

[nukie]

Your Geller SVR board, what was the calibration voltage supplied by Geller for the uncorrected values? 

My Geller calibration check is no.513
   Calibrated value - 5.00000V at 24c
Uncorrected value - 5.00129V at 24c

It has been quite a while since it left Geller labs, it's time to go back for a check up 

Well at this point I had to ask ... are those measurements taken with the original cables - leads ?  

Here is a link from my own reference IC , and tests when it is powered with 12V and 15V ..

I don't get what do you mean by "original cables". I remember using the exact same set of cables because that silly HP 3468A doesn't accept protected(shrouded) banana plugs. So I had this set of Chauvin Arnoux probes, it's similar to the second probe in the picture, I had the banana shrouds trim-off. So that's the only 'universal' cable that I have. I had a read at your decade box it's a professional job. 

[saturation]

Thanks nukie, will updated the spreadsheet.  In the current graph, the HP 3468a is considered the reference, so its values are zero.

But with the Geller values, you can now calculate the accuracy and drift relative to Geller.

3 years ago, Geller not only used the Fluke 732b volt reference, he also re-checks the accuracy on a calibrated HP3458a, both NIST traceable.

My calibration data for 4x HP3456a is 10 months old, but so stable, its unbelievable we can do this at home on the cheap.

[nukie]

Here's a quick and dirty stat sheet.  Puts the changes in visual form.
Raw excel sheet provided for those who wish it.

Shows what you pay for in a Fluke or Agilent.

I have one 20 year old FLuke 85i, only 3.5 digits but cal is stable for 20+years. Also a 15 year old Fluke 85i, 10 year old 85iii, ~ 5 year old Fluke 87V all in cal.

Nice! Always easier when there's a graph.
I am wondering if the stability is due to the age of the equipment? The Uni-T is youngest of them all, where as the Fluke and HP is pretty old. I forgot to mention, the Uni-T despite being top of range DMM it produced erratic readings especially in the 5V range, sometimes it would read 4.991 sometimes 4.993 very weird. However, lower range is fine.

HP3456a is a very nice DMM but it's too big to fit on my desk 

[saturation]

Hi nukie,

You're welcome. 

I'd give Uni-T the benefit of the doubt, but my old Fluke 80 series meters were new when I got them, they are now between 5-20+ years old and they haven't drifted from factory settings, calibrated from the Fluke factory only once. My 3456a are over 10-20 years old too, and they have not drifted anywhere as badly as that Uni-T.

With 3 years between your performance checks the Uni-T has aged at ~> 3+ years; alas aging of references must be done powered up for at least 200 hours, so as an experiment you can leave the Uni-T on for 200+ hours to insure the reference is aged, and see how much it drifts for the future.   

So the Uni-T designers may not have chosen their internal reference well, or did not age them, or their design is not precise enough and/or components are not well matched.

According to their spec sheet, basic DC accuracy is 0.025% accross the entire Vdc range.

http://www.uni-trend.com/UT71D.html

Given your data, the Uni-T is out of cal.

If you add calibration costs to keep the Uni-T in shape, versus the nearly lifelong stability of the Agilent, HP or Fluke models, it becomes a false economy, unless you can cal it yourself and end up just paying for it with your time.  Cal costs about $US 70-100, if done annually the lifelong cost would be greater that if one bought a better meter in the first place.

Nice! Always easier when there's a graph.
I am wondering if the stability is due to the age of the equipment? The Uni-T is youngest of them all, where as the Fluke and HP is pretty old. I forgot to mention, the Uni-T despite being top of range DMM it produced erratic readings especially in the 5V range, sometimes it would read 4.991 sometimes 4.993 very weird. However, lower range is fine.

HP3456a is a very nice DMM but it's too big to fit on my desk 

[saturation]

Also one thing you can see in the data:

In May 2008, the Uni-T 1V range was within cal spec, < 0.025%, but the 5V range was already off, its about 0.1%.

@Kiariakos:  cables can affect the readings, but cable issues often causes erratic values in the low range, uV resolution.  Just switch cables with other DMM being tested, and if there is a change in values, the cables are part of it.   Given how stable the values are, its unlikely but its something easy to try if you are squeezing the last uV of resolution for comparison.

[alm]

I am wondering if the stability is due to the age of the equipment? The Uni-T is youngest of them all, where as the Fluke and HP is pretty old. I forgot to mention, the Uni-T despite being top of range DMM it produced erratic readings especially in the 5V range, sometimes it would read 4.991 sometimes 4.993 very weird. However, lower range is fine.

Drift tends to get lower with age (which is why parts are often aged for better performance).

The potential issue I can see with cables is the Seebeck (thermocouple) effect, which is why pure copper wires (without plating) are usually used for precision measurements. I don't think this qualifies as that, though.