incredible PREMA current accuracy

[e61_phil]

Hi,

a colleague of mine brought his private PREMA 6000 DMM (6.5 digit) to our workplace today. Nothing special in the voltage ranges (50ppm/Year), but the current range (up to 2A) is specified with 100ppm/year within 23°C+/-5°C. Which is better than the specifications of the 1281, 3458A or 8508A (1A). I also had a look at the PREMA 6048 (8.5 digits) and the specification of this meter is only 30ppm/year!

There was nothing special in the schematics of the PREMA 6000. 100mR Shunt, thats it. A few pictures from google image search looks like a manganin 100mR resistor (only PREMA 5000 pictures found).

The 200mV Range is specified with 60ppm/year. Therefore, there is only 40ppm left for the resistor. Without any aging, there shouldn't be more than 8ppm/K for the shunt to fulfil the specs (without selfheating). Are there any manganin resistor with such good TCs?

Are these PREMA meters really that good in the current range or is it a very optimistic specifiaction?

Best
Philipp

[branadic]

Well, I had a look into the manual of my Prema 5000 and Prema 5017 SC and found:

Accuracy:
Prema 5000: 0,05% reading + 0,005% fullscale, 1 year 23°C +/-5°C (up to 1A)
Prema 5017: 0,03% reading + 0,002% (200µA - 20mA) / 0,004% (200mA - 2A) fullscale 1 year 23°C +/-5°C
Prema 6000: 0,01% reading + 0,005% fullscale, 1 year 23°C +/-5°C (up to 1A)
Prema 6047/48: 0,003% reading + 0,001% fullscale, 1 year 23°C +/-5°C

So maybe this was some kind of development process they did with each new version. And probably the resistors where preaged and selected too like they did with the voltage references.

[e61_phil]

Hmm, I would think the 5017 is much newer than the 6000 (~mid 80's)?
And the voltage specifications of the 6000 aren't very impressive compared to a HP 3456A with similar age (the HP is twice as stable).

[branadic]

What I wanted to say is, the 5017 is probably the successor of the 5000, while the 6047/48 are the successor of the 6000.

The 100mohm shunt in the 6047/48 seems to be a part made by Isaballenhütte (Zeranin):

https://farm8.static.flickr.com/7460/14118058402_ba92d10f69_b.jpg

[e61_phil]

Very interesting! This PDF here http://www.isabellenhuette.de/pdf/WIDER_LEG/DE/ZERANIN%2030.pdf says special types can have 3ppm/K. Which will fit to the TC specification from the 6048 manual.

Hmm, and I thought my Vishay VCS332s are good shunts

So, perhaps the current range is really that good. It would be interesting to give it a try and compare it to one if the 3458A..

[Pipelie]

interesting, I believe I have a PREMA 6000 siting in the corner, perhaps  it's time to put it in the test? 

[Dr. Frank]

Very interesting! This PDF here http://www.isabellenhuette.de/pdf/WIDER_LEG/DE/ZERANIN%2030.pdf says special types can have 3ppm/K. Which will fit to the TC specification from the 6048 manual.

Hmm, and I thought my Vishay VCS332s are good shunts

So, perhaps the current range is really that good. It would be interesting to give it a try and compare it to one if the 3458A..

The 3458A is specified 1A DCI, (110 +10) ppm @ 1 year, and the PREMA 6000 is (100+50) ppm @ 1 year.
So what's the deal?

And how would you like to compare both instruments, against which third 1A standard / shunt?
That'll be the classic man-with-two-clocks problem.

Frank

[e61_phil]

The 3458A is specified 1A DCI, (110 +10) ppm @ 1 year, and the PREMA 6000 is (100+50) ppm @ 1 year.
So what's the deal?

And how would you like to compare both instruments, against which third 1A standard / shunt?
That'll be the classic man-with-two-clocks problem.

The deal is, that this "cheap" Prema has nearly the same specs as the 3458A and other 8.5 digit DMMs.

I will compare it to two HP 3458A with several years of calibration reports.

[quarks]

unfortunately I do not have a PREMA DMM, but I have compared most 8.5 digit DMMs and if I remember correctly the PREMA 6048 looked especially bad in DCI comparison, mainly because it seems to have only a 2A range.

[Kleinstein]

Just comparing 3 instruments will not tell a lot about the specs. Those specs are kind of worst case errors - so most instruments will likely be better by quite a bit. With just 3 instruments comparable accuracy (for current) it is impossible to tell which one is bad, unless one is really out of specs.

The 3458 is a little different in the current specs that long term stability is set by the 40 K reference resistor and not the shunt itself. The auto-calibration to go from the 40 K to the 0.1 Ohms shunt adds a little of uncertainty, but there is not much drift in this, more like noise. In the Prema it is the drift of the shunt itself that sets the extra error for the current.

Also the calibration quality could be am issue: not every cal lab might get a very good error limit for the 1 A current - the 10 V and 10 K refs needed for the 3458 tend to be lower uncertainty. With a very good calibration one might end with better short time specs (e.g. 30 days) while the 3458 would be better for something like 2 year after calibration.

[Marcus_S]

Hi,

it might be interesting to add the Prema 6001 to the overview above:

Specs 1 year 23 °C +/- 5 °C, DC
200 mV   0,002 % of rdg + 0,0007 % of full scale + 1 digit
20 V   0,002 % of rdg + 0,0002 % of full scale + 1 digit
and
2 A   0,007 % of rdg + 0,002 % of full scale + 1 digit (up to 1A)
2 mA   0,02 % of rdg + 0,005 % of full scale + 1 digit

In my 6001 I found an Isabellenhütte-Shunt 0,1R (Zeranin) and it should be calibrated with 1 A max to avoid self heating (according to the manual). I have no idea how to calibrate or adjust this meter at home...


Best regards

Marcus

[branadic]

In Prema 5000 and 5017 a 100mohm +/-1% shunt made by Bader (Manganin) is populated:

http://www.mikrocontroller.net/attachment/219252/prema-5000-innen-2.jpg

[branadic]

The german drugstore CONRAD has the PBV-Version with 30ppm/K in stock:

https://www.conrad.de/de/messwiderstand-01-10-w-l-x-b-x-h-22-x-4-x-17-mm-isabellenhuette-pbv-01-1-st-447331.html

So maybe the version PBV-Z (5ppm/K) is at nearly the same price level? There is also a version AZ-H (3ppm/K) available:

http://www.isabellenhuette.de/praezisions-und-leistungswiderstaende/produkte/uebersicht/

However, element14/Farnell does have a few similar products in stock too, but at somewhat higher cost:

±3ppm/K
http://de.farnell.com/powertron/pcs-302-0r100-s-1/metallfolienwiderstand-0r1-1-sip/dp/2565583 30W, 300V
http://de.farnell.com/powertron/pcs-301-0r100-s-1/metallfolienwiderstand-0r1-1-sip/dp/2565578 30W, 300V
http://de.farnell.com/powertron/usr-4-3425-0r100-d-1/metallfolienwiderstand-0r1-1-sip/dp/2565600 30W, 750V
http://de.farnell.com/powertron/usr-4-4020-0r100-d-1/metallfolienwiderstand-0r1-1-sip/dp/2565604 30W, 750V

±5ppm/K
http://de.farnell.com/powertron/shr-4-2321-0r100-s-1-m/metallfolienwiderstand-0r1-1-sip/dp/2565588 40W, 300V
http://de.farnell.com/powertron/shr-4-3825-0r100-a-1-m/metallfolienwiderstand-0r1-1-sip/dp/2565591 50W, 500V

[e61_phil]

I have no idea how to calibrate or adjust this meter at home...

I'm also thinking about creating 1A with an uncertainty below 100ppm at home. I have only "calibrated" 10k and 10V.


@branadic: Very nice list!

Any idea, how to measure this 100mR resistors with <100ppm uncertainty?

[e61_phil]

Today was the last chance to have a look at the Prema 6000. Inside the Prema there was this Bader Manganin Shunt.

We set up a little experiment: Agilent B2902A SourceMeter sources 1A through a 3458A and the Prema connected in series. The input of the Prema was than shorten to bypass the current. We let this setup stabilize a while and then we removed the short. At this point the 1A is flowing through the Prema Shunt.

The attached curve is the deviation from the current measured by the 3458A (which was quite stable). One can see, that the selfheating due to the 1A is already around 100ppm. Therefore, I think the Prema spec is very optimistic.

[TiN]

Perhaps offtopic here, but 3458A's 1A shunt is some custom labelled HP 03458-82501-1.
I measured temperature coefficient (shunt removed from board) at around -5.1ppm/K linear, using 100mA test current and based on 20C ramp up/down span.

[branadic]

Inside the Prema there was this Bader Manganin Shunt.

I would have thought they used something different in the Prema 6000 than the Bader shunt. Bader seems to be only a distriburtor for such things, as you can see on their website:

http://www.bader.net/

[branadic]

The attached curve is the deviation from the current measured by the 3458A (which was quite stable). One can see, that the selfheating due to the 1A is already around 100ppm. Therefore, I think the Prema spec is very optimistic.

Well, looking at the datasheet of POWERTRON  USR 4-3425
http://www.farnell.com/datasheets/2168435.pdf

they differ between
- TC: max. ±5ppm/K (-55 to 155°C), typ. ±3ppm/K (-55 to 125°C), typ. ±1ppm/K (25 to 60°C)
- load stability, which is given as 0.01%
- shelf live stability: 25ppm / ?R after 1 year, 50ppm / ?R after 3 years

So what you see is a mixture of load stability plus self-heating and thus TC. So the question is, what is really given in the manuel of Prema 6000 that you are comparing to?

Accuracy (1 year, 23°C ±5°C ± (%of reading (%rdg.) + % of full scale  (%f.s.)) *)**)
±2mA and  ±2A ... 0.01 (%rdg.) + 0.005  (%f.s.)

Temperature Coefficient
(10°C to 18°C and 28°C to 40°C)
2mA  and 2A-range (up to 1A) ± (0.002%rdg. + 0.001 %f.s.)/°C

Temperature Coefficient
(0°C to 50°C) ... double the value  at 10°C - 40°C

*) Add ± 1 Digit. After autozero.
**) Guard connected  to  "V/Ohm-LO" input socket, sinusodial signal greater then 5% of full scale.

So nothing about load stability, only TC and stability after 1 year. And please don't miss the extra digit mentioned in *).

[e61_phil]

Well, looking at the datasheet of POWERTRON  USR 4-3425
http://www.farnell.com/datasheets/2168435.pdf

they differ between
- TC: max. ±5ppm/K (-55 to 155°C), typ. ±3ppm/K (-55 to 125°C), typ. ±1ppm/K (25 to 60°C)
- load stability, which is given as 0.01%
- shelf live stability: 25ppm / ?R after 1 year, 50ppm / ?R after 3 years

So what you see is a mixture of load stability plus self-heating and thus TC.

My interpretation of load stability is a permanent change in resistance due to the load. This is normally coupled to time. In case of the Vishay Resistor it is 1000h under load. Therefore, this isn't so important for a bench multimeter (operated only on the bench not in a system).
In my opinion this 100ppm drift was almost due to the TC of the shunt (self heating).

So the question is, what is really given in the manuel of Prema 6000 that you are comparing to?

Accuracy (1 year, 23°C ±5°C ± (%of reading (%rdg.) + % of full scale  (%f.s.)) *)**)
±2mA and  ±2A ... 0.01 (%rdg.) + 0.005  (%f.s.)

Temperature Coefficient
(10°C to 18°C and 28°C to 40°C)
2mA  and 2A-range (up to 1A) ± (0.002%rdg. + 0.001 %f.s.)/°C

Temperature Coefficient
(0°C to 50°C) ... double the value  at 10°C - 40°C

*) Add ± 1 Digit. After autozero.
**) Guard connected  to  "V/Ohm-LO" input socket, sinusodial signal greater then 5% of full scale.

So nothing about load stability, only TC and stability after 1 year. And please don't miss the extra digit mentioned in *).

The temperature was within 23+/-5°C. Therefore, the TC should be included in the spec. What I have overlooked is the second part of the spec: 50ppm of FS. This will add another 100ppm to the 100ppm of reading at 1A. So we have 200ppm uncertainty at 1A.

100ppm (self-heating uncertainty) + 74ppm (uncertainty @100mV) = 174ppm

There is not much space left, but perhaps it will fit.

Edit: If one adds 50ppm (10ppm/K over 5°C) then it doesn't fit into the spec even without aging of the shunt