7.5 digit bench DMMs comparison

[Dragony]

Hello,

I want to purchase a good 7.5 digit DMM for my bench. The problem is that for a hobbyist the price tags of this devices make it impossible to just buy and hope you got the right one. Furthermore a hobbyist won't look into 8.5 DMMs, since, well, hobbyist, you know.

I am surprised I were unable to find a good comparison between just 7.5 DMMs combined with price and year of release. When I decide to spend $2000-$4000 for a bench DMM, for many years to come I want to say "thats the best thing I can get for the money spent." Of course, who doesn't?

As far as I have found out so far the only available rather recently released 7.5 DMMS are Keithley DMM7510 and Keysight 34470A. Do you know any other ones?

[Vgkid]

Your other option for 7.5 digit dmms are: datron 1071, solartron 7075, solartron 7071, keithley 2001, hp 3457. There are most likely some from advantest, and prema.

[Dragony]

I found a suspiciously good comparison of DMM7510 vs 34470A. The DMM7510 is the clear winner. Can someone say if its plausible, or is it Keithley Propaganda?
http://imgur.com/gallery/Wbx7q

[Kjelt]

Why do you need/want 7,5 digits and have you looked at the cost for calibration each and every year?

[zlymex]

The Advantest one is R6871, the Prema one is 5017. There are also another Datron one 1081, another Keysight 34420A, another Solartron 7061, and 8071 from Transmille. I am a hobbyist and I got three 8.5 digit DMMs(1281, 3458A and R6581T)

[TiN]

DiligentMinds.com

I prefer this over the Keithley, because as I understand it you cannot calibrate just one range on the Keithley (it's all or nothing)-- I don't have one of these so I may be wrong, someone please correct me if I'm wrong.

Correct, even more of that - full manufacturing-type calibration is required via GPIB for K2001 to get 200mADC,2ADC ranges calibrated. That require high-voltage AC source access, as levels like 200VAC 30kHz are needed.

...rather nice user interface upgrades that you don't get with older equipment...

Downgrades too, as 30-year old K2001 have option slot for scanner card , making it 10-channel DAQ system in one box.

Dragony
You must define your primary use of 7.5-digit, and go from there, based on your requirements.
There is no "best" meter, it's much more like right tool for right task kind of thing. If you need remote onsite box with graphing and logging without access to computer - surely DMM7510 is great for that. If you need scan 10 sensors for your project - old K2001 is the option. Looking for something cheaper and decent - used 34465 or Keithley 2010 might be good pick. Need low level measurements or voltage comparisons? Keithley 2182/HPAK 34420A hard to beat on this area.
As you can see, these are rather different tools.

If your budget scales up to 4-5K$ then worth to think really carefully, as that amount of money can open the door to 3458A, which is different league instrument (but bit less useful for everyday poking and have large hidden costs risk). I had collected many datalogs including some comparisons here.

Just another 5c from hobbyist with 3 8.5d DMMs and 4 7.5d ones.

[Jay_Diddy_B]

Hi,

What are you doing that requires 7.5 digits?

In a lot of meters this extra digits give you resolution, but not more accuracy. There has been digit inflation. I am not going to call them 'marketing digits' because there may be some instances such as transfer accuracy when they may be useful. Transfer accuracy is very short term accuracy.

One of my favorite meters, and under-rated not talked about very much is the Fluke 8842A. This uses the LTFLU or SZA263 reference and has the following specifications:

1 year accuracy on the 20V DC range is 0.0035% (35 ppm) percentage of reading +2 counts.


The HP34401A is a 6.5 digit meter but the 1 year accuracy on the 10V dc range is 0.0035%  + 0.0005% of the range

It is the same accuracy but one more digit of resolution.


The HP3457A 1 year accuracy is 0.0040% (40ppm)  +20 counts (at 6.5 digits) when averaging over 10 power line cycles.

This is a 6.5 digit meter with a high resolution 7.5 digit mode.

The Keithley 2001 does a little better at 0.0024% (24ppm) + 4ppm of range on the 20V dc range (1 year)

A lot of these meters use the same LM399 reference or a reference that has similar performance, and the accuracy specifications reflect this. Now there are other contributions from resistors and the ADC.

A good indication of better performance is a meter with the LTZ1000 or LTZ1000A reference.

The Keithley 2002 is 10ppm +0.15ppm of range on the 20V DC range at 1 year.

The HP 3458A is 8ppm on the 10V range at 1 year, 4ppm with the high stability option.

The Datron 1281 is 6ppm + 0.1ppm of range on the 10V range at 1 year. They also suggest a 2.5ppm calibration uncertainty.


To get the ppm level accuracy you have to take special care not to introduce thermocouples in the leads, warm up times etc.

I chose 1 year accuracy and the 10V or 20V range, because this is a typical use case.

I just wanted to show you that resolution doesn't mean accuracy.
Calibration doesn't imply accuracy either. Calibration is not adjustment. Calibration is confirming that meter is inside the specifications. Calibration improves confidence.


You want to consider other features when choosing a bench meter, data logging is one feature.


My collection includes:

Datron 1281, K2001, 3x 3457A, Solatron 7081 (rarely used), 4x 3478A, 2 Fluke 8842A and a Fluke 732A DC voltage standard to check them.

Regards,

Jay_Diddy_B

[Dragony]

I don't really feel like buying a DMM for every application, but maybe this explains why some of you folks have half a dozen DMMs...

Right now I have one bench DMM, the Rigol DM3068, but the device was not able to fulfil the following tasks I actually encountered:

- Measuring current flow of 10nA failed. Too noisy.

- Measuring current consumption based on cap discharge hardly works. I don't trust the results. The problem is DC Bias. I also encountered the phenomenon that the unconnected DMM set to 10G input impedance rises on voltage 1V/s. Looks like DC Bias as well. Of course this does not happen with 10M, but I can't use 10M for this task as it would bias the measurement. I use 10G for this.

- Annoying fan (although on this topic I won't find happyness with both above metioned DMM candidates...)

- Differential measurements my oscilloscope can't do without Math. This is more a bonus. When I create DC/DC converters, for most non-critital diagnostics I could use the Keithleys 1M/s as a "cheap" Osc. If it really is able to display them, its a 1us resolution. Good enough for my requirements.

- Data logging to PC for time based calculations.

And of course it has to be much better(tm) than my Rigol, because I love having some new cool stuff

[Faith]

You might be interested in taking a look at the last page or two of the Keysight 34465A & 34470A thread (linked below). As already mentioned while the extra digit does give you extra resolution it may not necessarily translate into extra accuracy.

And with the Keysight 34465A it is easy to use math scaling to see well beyond the six and a half digits which you would normally get on the display without the need to export the data to a PC.

https://www.eevblog.com/forum/testgear/keysight's-new-34465a-(6-5-digit)-and-34470a-(7-5-digit)-bench-multimeters/msg1008179/#msg1008179

[TiN]

Dragony
Seem that your needs are towards low-current high-impedance stuff. In this case all abovementioned 7.5digit DMMs and 8.5d ones will fail as well. You need electrometer for such job, such as Keithley 6514/6517 with guarded triax cabling, shield boxes and all related stuff.

[Kjelt]

Calibration doesn't imply accuracy either. Calibration is not adjustment. Calibration is confirming that meter is inside the specifications. Calibration improves confidence.

As discussed in another topic with someone from a cal lab if you do not ask explicitly for adjustment and the meter is within a one year accuracy you are right.
If you ask explicitly for readjustment with the calibration they will do this but since they have double work they will charge you double. For a cheap lab with cal instruments suitable for 6,5 digits dmm this will already cost you the price of half the new meter (400-500) , i do not want to know how much an official 7,5 or 8,5 adj +cal will set you back for, probably the price of a brand new 6,5 digit dmm.

[Dr. Frank]

I found a suspiciously good comparison of DMM7510 vs 34470A. The DMM7510 is the clear winner. Can someone say if its plausible, or is it Keithley Propaganda?
http://imgur.com/gallery/Wbx7q

Clear winner in which aspect?
These tests are not well-grounded.

The absolute readings / uncertainty will depend on the passed time of usage, after last calibration, and this is not defined in his 'test'. The 7510 may be freshly calibrated, and the 34470A may have had last calibration one year ago. So that's an invalid comparison. In the end, both instruments are still well inside their 24h specs!

If you look closely to the StD statistics of the different DCV ranges (100mV, 1V, 10V),  which is in practice identical to rms noise, you will see, that in contradiction to his final noise testing, the 34470A and the 7510 always  show nearly identical StD.

He estimates the noise  from the graphic display, obviously.
In this case, he could better have used the statistical function, again, w/o disturbing the cables, as he claims to have done during the uncertainty testing.

Also, a measurement on a single NPLC number gives a limited picture only.
NPLC 1 gives always noisy readings, not worth 7 digits, so NPLC 10 or better NPLC 100..500 (or averaged equivalents) would tell the interesting facts about noise performance for 7 1/2 digits.

TiN had organized a big noise comparison test, over a broad range of NPLC numbers for each instrument, which gives much more illuminative pictures on these instruments.
https://xdevs.com/article/dmm_noise/

But as far as I can see, still no contribution on a 7510, what a pity.
But anyhow, the comparison to the superior 3458A might give you a better understanding of that parameter testing. 


To my opinion, and experience, the real performance  of 7 1/2 digits instruments in terms of noise, stability specification and linearity do not justify at all their high price, twice or four times than of comparable modern 6 1/2 digit instruments. See also my already linked 465/470 comparison tests.

Anyhow, all their user interfaces, digitizing and nice graphical features are a big pro over old boxes like the 3458A.

So, if you are crazy enough, i.e. going volt-nuts, and spend 2 .. 4k $/€/SFr, I could fully understand that 
If you decide to buy Keysight '465A or '470A, don't forget to order the digitizing option, the additional 2M memory (recommended for digitizing) currently is for free.
The 7510 already comes with  more and better features on this aspect, I think.


Frank

[EEVblog]

As far as I have found out so far the only available rather recently released 7.5 DMMS are Keithley DMM7510 and Keysight 34470A. Do you know any other ones?

When you are looking at this high end you really shouldn't consider anything else unless you have a very specific niche requirement for something.
What's wrong with 6.5 digit BTW?

[HighVoltage]

I have the Keysight 34470A for a while now and like it a lot for my VoltNuts addiction.
The calibration of this one is 1 1/2 years old

Now I have brand new Keithley DMM7510
This one was calibrated in 8/2016

In the next few days I will make some comparison measurements between the two.
But so far I see the following on my LTZ1000A reference:
The 34470A reads about 18 uV high
And the DMM7510 is about 20 uV low

What is really impressive with the 7510 is how quick it is warmed up to show a stable reading.
I have not measured it, but it is less then one minute, it seems.

Both instruments are recommended.
The handling is just very different.

If you are in to low current measurements, the 34470 has a 1 uA current range and the DMM7510 has a 10 uA current range.
And the 34470A is just amazing, in measuring this low current.
I have not tested this on the Keithley meter.

[Mickle T.]

I found a suspiciously good comparison of DMM7510 vs 34470A. The DMM7510 is the clear winner. Can someone say if its plausible, or is it Keithley Propaganda?
http://imgur.com/gallery/Wbx7q

It's a Keithley Propaganda.

[–]NorthBus ... 1 year ago
I'm an analog hardware engineer from the Keithley 7510 project team ...
https://www.reddit.com/r/electronics/comments/2t6935/keithley_7510_gsmdmm_released_75_digit_1_plc_45/

[HighVoltage]

It's a Keithley Propaganda.

I agree, most definitely propaganda by a guy that was on the 7510 design team.

But he posted a few interesting things on reddit

Well, there's the fact that our primary precision, low noise, sensing FET was pulled by the manufacturer halfway through our design process. They said it's obsolete and they're not making it anymore. Cue a crazy many-month scramble to find and design in a replacement.

Or how we learned how amazingly sensitive grounds can be, when a stray few microamps flowed through a few milliohms of trace resistance, but only under certain rare conditions. That little bit extra threw off our measurements by a few digits until we found it and fixed it.

And of course, there was the time when an early firmware bug caused some bad interaction between the program pointer and the speaker output, causing looped program data to be streamed over the speaker. Loudly. Our problem report referenced units "screaming in pain".

I think I can answer a few questions without giving away any trade secrets or anything:

Noise? Well, very careful grounding, for one. Also, our low-noise amplifiers are all made from discrete components rather than off-the-shelf op-amps, some of which components go through additional selection processes, matching, auto-calibration-biasing, etc. Also, said selection process is annoying as piss to handle form an engineer's perspective.

Front end? To keep it accurate while still providing all the functions and ranges, the unit tests and corrects various parts of the frontend in between measurements. That's the toggle-able "AutoZero" option on the unit.

The ADC? We actually have two different ones in this unit. If you are on a "slow" function (slower than, say, 10k sample/s), you use our custom, Keithley-built 32-bit A/D. Anything faster than that uses an Analog Devices AD7982 18-bit SAR with (again) a highly customized frontend.

[Dragony]

I wonder if Keithley has addressed the complains about the fan in the meantime? How do those companies work? Is it a "fire and forget" type of development and they will just never address it, or is there a chance they will reduce the noise of the fan in future revisions? Maybe they already have done so? Thats still an annoying thing on the DMM.

And why is everyone concerned about high voltage? I have high voltage on my clothing all the time and I am fine. 

[HighVoltage]

Actually the fan noise of the 7510 is not annoying to me at all.
May be its will get louder during high current measurements but right now it is not even noticeable.

The original fan of the Agilent 34410A DMM and some Agilent power supplies are much louder.

[Dr. Frank]

The fan on the 465/470A is quite silent.
Keysight also was very agile to correct several FW bugs, (I got two fixed consecutively) so they really listen to the customer

Frank

[Kjelt]

You can always replace it with low noise brands such as Noctua and BeQuiet. First thing I do with professional network equipment is replace those airplane fans.

[plesa]

For your applications I will purchase instead of one instrument at least two.
The modern DMM with graphical LCD look nice and for basic statistic it is OK, but for normal use I needs to use PC it is more comfortable (consider BenchVue or logging with some Visual basic scripts or Python). Small factor PC (e.g Infocus Kangaroo for $150 including Win10 )Raspberry Pi or any small form factor PC and some old DMM like 2001/3441xA/34401 can be good enough.

For nA current measurement I will buy Keithley 6485, 6414/6517 is little bit overkill if you needs on sub nA measurements.

[acbern]

Also for precision gear, such as 7 and 8 digit DMMs I would only go with used ones. That always provided they are not badly worn. At least for hobbyist use, they dont run often, so the remaining lifetime is most likely not an issue at all (some preventive cap replacement on certain ones may be a good idea, as has been pointed out here, altough I do not share that experience...), and they have a well aged reference and other parts, with related low drift, and of course the price is a fraction of a new one. This is why some metrologists only buy the used 732As... I would not think a second if I had to decide between a used and warrantied 3458A and a new 7 digit DMM.

[plesa]

Also for precision gear, such as 7 and 8 digit DMMs I would only go with used ones. That always provided they are not badly worn. At least for hobbyist use, they dont run often, so the remaining lifetime is most likely not an issue at all (some preventive cap replacement on certain ones may be a good idea, as has been pointed out here, altough I do not share that experience...), and they have a well aged reference and other parts, with related low drift, and of course the price is a fraction of a new one. This is why some metrologists only buy the used 732As... I would not think a second if I had to decide between a used and warrantied 3458A and a new 7 digit DMM.

I will add that used are good in case there is schematic available, otherwise it can be painful to repair. PArts availability is important factor as well. In this segment is "better" Keithley with only few custom parts against HP/Agilent.
Used does not mean it is fully working within spec, only few customers are using all functions to their limits. So lot of issues are not know to the seller.
This was my experience mainly on low current measurement devices. There are only few applications where 6.5 digit DMM is not enough and you needs 7.5 or 8.5 digit.
I needs to return new 3458A after few months for convergence error. In case I'll buy used one for $2500 and after that return it to Keysight for repair  $2851 and calibration the total cost will be >50% of new unit ( $9778 with 3 years warranty). So the used 3458A is worthwhile only if it was recently tested/calibrated and checked for drift.

[ebclr]

I would like to know what kind of electronic hobbist circuit need a 7 1/2 digits meter accuracy , I stoped on 50000 count's meters and never need anything more than that for hobby, would be interesting to evaluate hobby application to support the investment on this class of meter for hobby

[TiN]

ebclr
High resolution DAC/ADC testing, voltage references design and experimenting, ratiometric measurements (e.g. bridges, comparators). How these areas fill into home hobby? Well, that's a question outside of the choosing DMM here. It's often not resolution what justify such instruments, but their stability and accuracy. But that's only if 7.5/8.5 meter is tested and calibrated. Otherwise calibrated 6.5-digit meter in such cases often provide better accuracy than uncalibrated unknown 8.5-digit one.