Bench meter with most accurate current range?

[EEVblog]

What is the 6.5digit bench meter that has the best DC current accuracy on uA and mA ranges?
Preferably 1 year spec.
There doesn't seem to be any standouts in the major players.
I'm looking for substantially better than 0.05% for a year, and preferably up to the A range with 0.05% or better.
My old 3457A seems to be the best 90 day spec.
Even the Keithley 2001 7 1/2 digit can't do any better than 0.04%

[jpb]

Here are the specs for the ADCMT 7461A - they are similar, probably not better but I thought I'd add them to the mix.

[KJDS]

Keithley 2015THD is better on some ranges and worse on others

[cyr]

I think you may need the big boys (8.5 digit class) to get *significantly* better than 0.05% for one year DCI.

Fluke 8508A claims "Best 1 Year Absolute Specification" of +/- 12ppm for DC current, although you may need to be a metrologist to even understand the full spec.

[quarks]

Hello Dave,

if it does not have to be a bench meter, the handheld Gossen 30M seems to be specwise one of the best, but is only going to 100mA (see att.)

If you are not looking for datasheet specs only, I can not praise enough the TEK DMM 4050 = Fluke 8846A!

Mine performs way better then specs tell. Robrenz will probably be able to confirm that for his own 8846A too.

bye
quarks

[ecat]

+/- 150ppm +/- 4 digits any good for you?

If I have this right, % = ppm / 10,000 so 0.015% ?

6 1/2 digits Datron 1071.

Edit:
<Grrr> looks like current reading is only 5 1/2 digits

[sync]

Hello Dave

This is a list of specifications for many multimeters: http://electronicsandbooks.com/eab1/kb/k0zfca.htm

Unfortunately they don't have an index by DC I stability. Thus i made an excel sheet from it (attached). DC I stability is Column P.

According that site Rigol DM3061, DM3062 and DM3064 have 50ppm DC I stability for 1 year.
Edit: Unfortunately the Rigol entries are wrong. According the data sheet, they have at best  750ppm + 60ppm fs for DC I.

[robrenz]

This is the one year on paper specs normalized to percent of value error.  Yellow highlight is the more accurate meter for that value.

Edit:
IMO it is a waste of time to compare specs without doing a percent of value spreadsheet for all instruments considered. It is important that the sheet includes the nominal max. of each range for all meters to be fair. In this case the "3's" were included for the 3457 and the "1's" were included for the 8846 and the 1 Amp max. of the 3457.

Added the spreadsheet with graph.  Feel free to check my formulas, I mess them up frequently.

[jeroen74]

Is there any special about measuring current that makes it difficult to get it long term stable? Or is just that manufacturers simply don't bother with it?

[edavid]

Is there any special about measuring current that makes it difficult to get it long term stable? Or is just that manufacturers simply don't bother with it?

Current measurements depend on the absolute value of the shunt, where voltage measurements mostly depend on the voltage reference and resistor ratios.

If you had a precision current reference, you could reverse this.

[eevblogfan]

hey

Dave , can you shear with us the need for such accurate measurement ?

I can tell you that all of these specs are just on paper , agilent HP fluke keithley and such high end companies tend to write " max 100ppm " while when measuring 1A , the meter will be much better then 100ppm , and will show " 1.000040A " .

tell you something ? my HP34401 is not that great , but has 6.5Digit res ,  within 1 Day after ADJUSTMENT I get like times 10 better specs then written on the data sheet ( yes I know not to many ppl can adjust they're multimeters as often as I can _ BUT Dave has the access to Trio smart cal's instrument , hence he may be able to adjust he's 3457 )

Note , my 34401 can show spot on only on the 5,5 digit mode , while on the 6.5 it is wrong by few digits ( Ie few ppm's ) , and that's straight after adjustment   

so my conclusion - use you'r meter one digit less , Ie , if it's 5.5 , use it as absolute accuracy of only 4.5 , 

although my two 3478A's are preforming absolutely amazing ! , one shows 10.0000V when the fluke 5700 set on 10.00000 ... V .... so on that meter I'll count for the fully 5.5 resolution ( ON THE DC voltage ONLY ! )

[jeroen74]

Is there any special about measuring current that makes it difficult to get it long term stable? Or is just that manufacturers simply don't bother with it?

Current measurements depend on the absolute value of the shunt, where voltage measurements mostly depend on the voltage reference and resistor ratios.

If you had a precision current reference, you could reverse this.

But a shunt is nothing but a simple piece of wire, preferably made of a suitable material with near zero tempco. I can't imagine a piece of wiring changing that much in resistance over time. The stable voltage reference is already there, so... where's the real problem?

[edavid]

But a shunt is nothing but a simple piece of wire, preferably made of a suitable material with near zero tempco. I can't imagine a piece of wiring changing that much in resistance over time. The stable voltage reference is already there, so... where's the real problem?

I don't think you will find a 6.5 digit resistor for sale, but I could be wrong.

P.S. Here's a paper I found on the subject of shunt resistor stability:

https://accelconf.web.cern.ch/AccelConf/p07/PAPERS/MOPAN071.PDF

[eevblogfan]

But a shunt is nothing but a simple piece of wire, preferably made of a suitable material with near zero tempco. I can't imagine a piece of wiring changing that much in resistance over time. The stable voltage reference is already there, so... where's the real problem?

the real problem ( I am assuming ) is the difference ,  say when adjusting the device the resistance was " 0.1000142 ohm "  at 20.000C , and when you does measure that accurate measurement the temperature is 23C , at 40ppmC , you'll get total of 120ppm out , assuming you'll need say 50ppm measurement , then you'll have to know the exact temperature the device was adjusted at , in that case you'll have to be half a degree within that 20.000C , which is pretty difficult ,

also , when current passes through the shunt , he heats up , and due to thermal mass , it takes time , so lets assume it will heat up to say 30C _ when 1A flow through it , then you have  to know how much did the operator of the adjustment waited , ideally you'd like him to wait 1HR in order to settle things down as good as possible , so then , you can achieve 10 to say 60 times better reading ,  instead of 0.1% , you'll get 0.01 or even better , ONLY if the operator of the adjustment waited 1HR or so when he adjusted the device ( I am pretty sure he didn't , time cost money you know , and there are several ranges of current and so on , so 1 HR per rang ? no way !  )

hope I managed to explain ( and didn't make any mistake(s)  )

Cheers !

[jeroen74]

The paper was an interesting read. It seems mechanical stress is a factor too for long-term stability.

[EEVblog]

I can tell you that all of these specs are just on paper , agilent HP fluke keithley and such high end companies tend to write " max 100ppm " while when measuring 1A , the meter will be much better then 100ppm , and will show " 1.000040A " .

Yes, but unless you have proven historical data on your particular meter, the max spec is all you have to compare with.
Beyond that you get into the whole metrology thing.

tell you something ? my HP34401 is not that great , but has 6.5Digit res ,  within 1 Day after ADJUSTMENT I get like times 10 better specs then written on the data sheet

Yes, of course, that's why most meters have a 24 hour stability spec as well.

[Fsck]

I don't believe (/know of) there are any 6.5d mm's with significantly better than 500ppm accuracy. You'd really need to jump to the 8.5d mm's to make a dent in that number.

At it's outermost edge, a fluke 8508 just clears 500ppm at 20A (overkill, yes.) and but both the 8508 or 3458 will give you your 100-200ish ppm accuracy jump you want.

The price difference between the typical 6.5d to 8.5d meters (~600$0 jumping to >2000$) is kind of self-explanatory, they've hit a wall at producing a cheap high accuracy measurements.

[ivan747]

Any reasons not to use a high precision voltage source (you have one) and a high presicion resistance substitution box (you got two) and a giga-ohm input impedance multimeter (you must have heaps of those  ) to measure the voltage drop?

Or are you simply avoiding this hassle? It should work for the µA range without issues. Maybe in the mA range you start to get problems due to any contact resistance errors you cannot eliminate with 4 wire measurement.

I assume you are working on some high accuracy version of the µCurrent. Maybe you can set up a test gig.

[EEVblog]

Any reasons not to use a high precision voltage source (you have one) and a high presicion resistance substitution box (you got two) and a giga-ohm input impedance multimeter (you must have heaps of those  ) to measure the voltage drop?
Or are you simply avoiding this hassle?

Yes, it's just the hassle. I could potentially get better than all those meters with a couple of 0.005% $20 resistors, and almost any 6.5digit voltmeter.
Curious to see if a nice accurate current meter exists at a sensible price, so just going through the motions on that.
There are other options like a Keithley 2401 source meter that can generate a precise known current, and also measure.

It should work for the µA range without issues. Maybe in the mA range you start to get problems due to any contact resistance errors you cannot eliminate with 4 wire measurement.

Yes, the higher current measurement is a bit trickier, but the specs requirement on that is lower anyway.

I assume you are working on some high accuracy version of the µCurrent.

 

[quarks]

I don't believe (/know of) there are any 6.5d mm's with significantly better than 500ppm accuracy. You'd really need to jump to the 8.5d mm's to make a dent in that number.

At it's outermost edge, a fluke 8508 just clears 500ppm at 20A (overkill, yes.) and but both the 8508 or 3458 will give you your 100-200ish ppm accuracy jump you want.

The price difference between the typical 6.5d to 8.5d meters (~600$0 jumping to >2000$) is kind of self-explanatory, they've hit a wall at producing a cheap high accuracy measurements.

As I wrote, I can confirm that my DMM 4050 is better and even compares very good with 8.5 digit gear.

About the 8508A I tried to buy one new and it was offered for crazy 16000€ + tax and a new 3458A is at least 7000€ + tax (and more if you add some options) in Germany.  In relation to this price differences the 4050 or 8846A I think is truely one of the best DMMs you can buy.

[Fsck]

I don't believe (/know of) there are any 6.5d mm's with significantly better than 500ppm accuracy. You'd really need to jump to the 8.5d mm's to make a dent in that number.

At it's outermost edge, a fluke 8508 just clears 500ppm at 20A (overkill, yes.) and but both the 8508 or 3458 will give you your 100-200ish ppm accuracy jump you want.

The price difference between the typical 6.5d to 8.5d meters (~600$0 jumping to >2000$) is kind of self-explanatory, they've hit a wall at producing a cheap high accuracy measurements.

As I wrote, I can confirm that my DMM 4050 is better and even compares very good with 8.5 digit gear.

About the 8508A I tried to buy one new and it was offered for crazy 16000€ + tax and a new 3458A is at least 7000€ + tax (and more if you add some options) in Germany.  In relation to this price differences the 4050 or 8846A I think is truely one of the best DMMs you can buy.

The DMM4050 hits *exactly* 0.05% base + LSD, which is insufficient according to Dave's original specifications.

Whether a particular piece of gear is better than specifications is based on the individual machine. My university has a couple(2, actually) of 3458s in one of the laser physics research labs which make the specifications look like a joke and way too broad (as far as I could measure anyways). But, specifications of equipment are given as worst case, that you will *guarantee* that every single machine you ever produce will meet to within it's confidence interval.

[quarks]

Hello Fsck,

I am sorry if I may have been unclear. Of course the specs state that it is in the 0.05% range and that is exactly what I stated in my first post. What I tried to express is, datasheet specs are not the only information to look at. If there are reports that confirm better then stated/expected performance (and you trust the source) then it think and hope this is good information to know.

Because I have my own private "little" calibration lab (with real calibration gear) I check my gear quite often. One finding of that was, the TEK4050 (=Fluke 8846A) seems just to be one of the best 6.5 Digit meters you can buy. If it is only my personal one, I am indeed a lucky man, but I do not think so, because I also have seen calibration data from robrenz.

So If anyone asks for the best 6.5 Digit meter arround (and Dave did),  I give my honest recommendation to go for this one. And also I confirmed that it is way better then the specs, because it really is. It is so good that I first could not believe it and did test it over and over to be sure I am not telling nonsense, if I confirm it to the public.

bye
quarks

[quarks]

I added the new Agilent 6.5 Digit Meter also to my DCI spreadsheet, maybe it of interest to see the chart

[robrenz]

+1 on the 8846A/4050 performance, I am very impressed and happy with mine also.  But I think we need to petition Fluke for a firmware upgrade to have trend plot, histogram and statistics be more flexible like the new Agilent 34461A