Which 8.5digit Multimeter should I get?

[mendip_discovery]

Well, I know this will bring out the hardcore VoltNuts.

So work is allowing me to buy a new meter and after seeing the price for tired second hand unit the boss said to buy a brand new one. So its either the Sexy Black edition of the Keysight 3458A with the 001 and the 1kV AC options or the Fluke 8588A with its fancy Colour TV.

This will be for a small 17025 Lab. Depending on the difference in price I think I might lean towards the Fluke partly as it's got the 30A capability built in. We have a Transmille multiproduct unit that does up to 22A so would be great if I could use the Fluke to characterise it.

[Olivia]

I am not an expet in the 8.5digit realm nor in the calibration realm so take this with a grain of salt.
I have looked at the price difference between the 3458A and the predecessor to the 8588, the 8508.
What Keysights websight says is that the price of a new unit is 10500€, i would presume that the 1kV AC option
would push the price to 11000€.
The only hint towards the price of the 8588 is a used 8508 which was listed for 12500€ iirc.
Concidering the 8588 is quite a massive upgrade i would estimate the price to be around 17500-20000,
but thats just a guess as Fluke doesnt tell the unit prices for the high end equipment on their website.

My personal Idea would be to build a high current adapter with a say 5A and a 25A range via individual terminals.
At this point the self heating of the shunts will be an issue so I would try and compensate that via a thermal sensor on each shunt.
That could be read via a AIO-Chip like an ICL7106 or maybe via another Multimeter.

I dont know about the availibility and precission of high current shunts for the 5-25A Range. If I had to make the decission it would depend on the usage type, current or ohms.
Is the DMM going to primarily be utilized to characterize high current equipment? What would be the Price of the 8588?

[bdunham7]

Get the prices and ask your boss.  I think the 3458A with your options is about $14K USD, the 8588A about $16K.  Put it to him this way--one is a PB-Free rehash of a 30-year old design with a new paint job, the other is much more recent and modern, in addition to having your current range.   I suppose it depends a bit on what prices you get in your area.

[mendip_discovery]

Get the prices and ask your boss.  I think the 3458A with your options is about $14K USD, the 8588A about $16K.  Put it to him this way--one is a PB-Free rehash of a 30-year old design with a new paint job, the other is much more recent and modern, in addition to having your current range.   I suppose it depends a bit on what prices you get in your area.

Yup getting prices I am hoping it won't be too much extra for the Fluke. We did have some prices for the Transmille and its the same as the KS so we were leaning towards the KS then I found the Fluke so waiting on a price for that so I can get an idea of the cost. Fluke looks to have some good specs but I know with VoltNuts that claimed spec isn't all the same. Might see if I can get both on loan to compare them.

[bdunham7]

Fluke looks to have some good specs but I know with VoltNuts that claimed spec isn't all the same. Might see if I can get both on loan to compare them.

How would you compare them?  Assuming we're talking about the specifications, tolerances and actual vs claimed performance, not just UI and aesthetics.

As far as claimed specs, you'd have to look at all of the aspects of those.  I know the 3458A has excellent specified linearity on the 10V scale, which they make a big deal of, but Fluke doesn't really specify or emphasize that.  The Fluke 8558A might actually be the closest competitor to the 3458A, you'd have to scour the specs for what matters to you the most.

[martinr33]

1) The Keysight 10K / 10V adjustment process is wonderful, but of course you need great standards and you blow the cal cert when you do it.

2) The 8508A is much better at resistance than the 3458A. But that improvement makes calibration a more complex process. The front/rear resistance comparison that the 8508A offers is a neat trick if you need it. The 3458A can sort of do it, but I've not had good results from that mode.

3) Make sure that your quotes include:
  - annual calibration cost
  - maintenance contract option if available. 

I think that KS calibration is about 1/3 the cost of Fluke calibration.

[e61_phil]

1) The Keysight 10K / 10V adjustment process is wonderful, but of course you need great standards and you blow the cal cert when you do it.

In my opinion one of the biggest disadvantages of the 3458A. You don't have the chance to calibrate a single range by its own. If one of the ranges is out you have to recalibrate the whole DMM. And you cannot characterise ranges better than specified, because the ACAL will not allow any range to be better than the ACAL transfer. Yey, you can completely avoid ACAL, but the specifications are written with the use of ACAL.

2) The 8508A is much better at resistance than the 3458A. But that improvement makes calibration a more complex process. The front/rear resistance comparison that the 8508A offers is a neat trick if you need it. The 3458A can sort of do it, but I've not had good results from that mode.

Yes, the Fluke meters are way better than the 3458A on low and high ohms. But the real game changer is the software controllable switch. That allows direct comparisions of your DUT against better standards than the DMM.

High voltage is also better with the Fluke ones.

3) Make sure that your quotes include:
  - annual calibration cost
  - maintenance contract option if available. 

I think that KS calibration is about 1/3 the cost of Fluke calibration.

I don't know the calibration costs of a 3458A at Keysight. Our 3458As are just calibrated at a local calibration lab. The 8588A is a bit cheaper to calibrate than the 8508A. We paid ~1k€ for the 8588A and ~1k5€ for the 8508A in the last years.

Since we have the 8588A the 3458As are rarely used. The 8508A is also great at some points but very slow.

If you can afford it I would go with the 8588A.

[manganin]

What is the most important in your application?

Linearity and speed -> 3458A
Long term stability and ranges -> 8508A, 8588A

If you are running a cal lab, stability is more important than sub ppm linearity. In research and primary metrology it is usually the opposite.

[e61_phil]

What is the most important in your application?

Linearity and speed -> 3458A
Long term stability and ranges -> 8508A, 8588A

If you are running a cal lab, stability is more important than sub ppm linearity. In research and primary metrology it is usually the opposite.

the 3458A can do 100kS/s the 8588A 5MS/s -> speed goes to the 8588A

linearity is more or less the same. The 3458A is specified over the whole range, but you can ask a cal lab to measure the linearity of your 8588A if you don't trust it

[Kleinstein]

Both the 3458 and 8588 have nominal ranges or 0.1 / 1 / 10 V and so one, but the fluke meter actually has some 0.2 / 2 / 20 V as full scale, while the KS meter only goes to 0.12 / 1.2 / 12 V. Depending on the application (e.g. check a 19 V test point) this can make a difference.
The differente full scale can be confusing when comparing the specs.

For the higher speeds keep in mind that the good linearity specs apply to a rather low speed and not the higher speed mode. At higher speed (e.g. 10 kSPS) neither the KS nor Fluke have tight linearity specs (if at all). The 3458 changes the input path/ADC quite a bit and the Fluke meter uses a separate ADC, both with lower accuracy.

Getting a meaningful linearity measurement for the 8588A is a difficult task - so not an easy and likely an expensive way.

[branadic]

the 3458A can do 100kS/s the 8588A 5MS/s -> speed goes to the 8588A

linearity is more or less the same. The 3458A is specified over the whole range, but you can ask a cal lab to measure the linearity of your 8588A if you don't trust it

That's some broad generalisation and unfair comparison. The 8588A uses an additional build in 18-Bit SAR AD-converter with 5 MS/s, while it reaches only "1 measurement/s @ 8.5 digits up to 100.000 measurements/s @ 4.5 digits" with its integrating ADC. Compare that to the 3458A that is reaching 18bit resolution at 100 kSps with its integrating ADC. That's better for the integrating ADC, isn't it? Everything else is just cheating. For the higher cost of the 8588A you can easily get a 3458A and an additional fast sampling device.

For linearity the specman sheet only claims "high linearity" and "extraordinary linearity". What ever that means, so you might want to explain to us what we should trust in?

I do agree, that the switching between front/rear is a nice feature, but that is nothing Fluke invented and is available on many other meters too. But that's not worth the extra thousends of dollar in my opinion.

-branadic-

[e61_phil]

That's some broad generalisation and unfair comparison. The 8588A uses an additional build in 18-Bit SAR AD-converter with 5 MS/s,

Who cares about how it is done as long all my measurements are fully specified? And if I remember correctly there are more specs for the 8588A sampling.

while it reaches only "1 measurement/s @ 8.5 digits

I think we already had that discussion and I showed you screenshots which show that isn't true.

Compare that to the 3458A that is reaching 18bit resolution at 100 kSps with its integrating ADC. That's better for the integrating ADC, isn't it? Everything else is just cheating.

As already said: I don't care how it is done, as long as the specs are there.

For the higher cost of the 8588A you can easily get a 3458A and an additional fast sampling device.

Show me the shopping cart which contains a 3458A, a low thermal switch, a 30A shunt and a 5MS/s 18Bit digitizer for not more than the cost of a 8588A. Don't forget a proper 1kV divider, a 100mA current source for proper 1R measurements and the 200V source for proper GOhm measurements... Shorter cal cycles to reach the 10V stability and so on and so on.. ahh and don't forget to buy a capacitance meter.

[branadic]

I think we already had that discussion and I showed you screenshots which show that isn't true.

I was only quoting Flukes document. Every statement from my last post related to 8588A directly comes form here.

As already said: I don't care how it is done, as long as the specs are there.

What specs? "high" and "extraordinary" linearity are just words, no specs. 18 bit @ 5 MSps is no spec either, that are just marketing words. I don't get this fanboy-ism for a company that is not able to write a clear, understandable and comparable spec list to promote their product.

-branadic-

[e61_phil]

I think we already had that discussion and I showed you screenshots which show that isn't true.

I was only quoting Flukes document. Every statement from my last post related to 8588A directly comes form here.

You can find it in the tables. That means that you don't need to add addiotional aperture error from 1s on. Same with the 3458A. There is also an error for faster readings specified in the datasheet.

As already said: I don't care how it is done, as long as the specs are there.

What specs? "high" and "extraordinary" linearity are just words, no specs. 18 bit @ 5 MSps is no spec either, that are just marketing words. I don't get this fanboy-ism for a company that is not able to write a clear, understandable and comparable spec list to promote their product.

Same here. You will find error adders for the different sample rates.

[Kleinstein]

The 3458 ADC is using the high linearity low speed mode from 100 µs aperture on. The 8588 can also use the integrating ADC down to 100 µs aperture (4700 SPS if I anderstand the specs right). AFAIK the linearity specs for the 3458 are only for 100 PLC and should also apply to 10 PLC, but no direct specs for faster conversions. Some of the linearity errors are known to get larger with faster conversions.

Below 100µs the 3458 ADC switches to a different mode and this does additionally increase the nonlinearity (higher input current and simpler amplifier). It is not at all clear that the integrating ADC in the fast mode is better than the SAR ADC in the 8588.
For the specs for the harmonics in digitizing mode the Fluke is slightly better than the KS meter (-100 dB vs -96 dB).
The harmonics can be seen as some linearity specs, though they may favor an SAR type ADC over an integrating ADC, as they mainly reflect the soft INL.
The accuracy specs for the digitizing mode still don't looks impressive - kind of miss some ACAL there to like the faster ADC to the accurate one.

For a cal lab I don't think the digitize mode or fast conversions would be a main feature. High accuracy AC may be the more important case.

[e61_phil]

It is not at all clear that the integrating ADC in the fast mode is better than the SAR ADC in the 8588.

I couldn't find that much for fast sampling in the 3458A datasheet. There is a graph which goes down to 0.01NPLC (200µs with 50Hz). At this speed the meters seems to be equal. The 3458A specifies and gain error of additional 20ppm and the Fluke specifies also 20ppm for >= 100µs < 500µs. From that point on there are only specifications for the 8588A. And the meters are a quite different here. The 8588A does real-time sampling with the fast SAR. The 3458A is not that fast, but with sub sampling and S/H it can reach even higher frequencies for periodic signals. But I can't find anything about the accuracy here.

We bought the 8588A mainly because of its fast sampling to measure non periodic signals.

[NaxFM]

Sorry if i'm not answering anything but asking another question.
Does anybody knows the price for a single fluke 8558A and an 8588A? I know that "if you have to ask the price you can't afford it (for now)" but i'm just curious, because it seems a big improvement over the 3458A and i wanted to know if the price difference is worth it

[e61_phil]

Sorry if i'm not answering anything but asking another question.
Does anybody knows the price for a single fluke 8558A and an 8588A? I know that "if you have to ask the price you can't afford it (for now)" but i'm just curious, because it seems a big improvement over the 3458A and i wanted to know if the price difference is worth it

We paid 15k€ in 2019. In my opinion that is a fair price and not so much more than 3458A with 4ppm option. (and you get much more than that)

Yes, I personally like the meter. We aren't a cal lab, but we develop high precision mass spectrometers. And for me the features of a "new" meter compared to the 3458A are really helpful. I can take the meter from my electronics bench to the R&D lab and directly connect it to the instrument under development. The grahp display allows me to directly see influences of other parameter to my measured signal. If I want to log for longer time, I can log to USB or simply connect the meter to the company network and control it from my desk in another room. I understand, that GPIB is fine for most cal labs, but the 8588A can do more.

The second input is the second very important thing. I often do TC measurements in a big climate chamber. The problem here is, that the climate chamber will heat the room and the TC of the DMM becomes a problem. With the 8588A I can connect a voltage standard to the second input and compensate for that.

I don't want to say the 8588A beats the 3458A in everything. But in almost everything it is at least on the same level. The only big disadvantage over the 3458A is that the DCV transfer specs are only valid for 10% of the initial measurement. Because of that I send our 8588A out to measure the linearity.

[NaxFM]

We paid 15k€ in 2019.

Not bad, thanks! Honestly I didn't expect a price so similar to the 3458A with 002 option, being Fluke i expected at least something near 20k.
I'm starting to think that there is almost no reason at all to prefer the 3458A. Ok, there's better linearity (supposedly), but what else? The accuracy is the same if not better, it has double the range on everything and a huge 30A current range. All of this without even taking into account the screen and interface.
Are a couple of ppm of better linearity really worth it?

[e61_phil]

Are a couple of ppm of better linearity really worth it?

Because of the Footnote you are in trouble as soon as you leave the 10%

Let's say you want to transfer a LTZ1000 7.15V to a 10V standard

With the 3458A it would be:
sqrt( (0.05ppm + 10/7.15 * 0.05)² + (0.05ppm + 10/10 * 0.05ppm)² ) = sqrt( 0.12² + 0.1² ) = 0.16 ppm

If you apply the 24h specs of the 8588A:
sqrt( (0.5ppm + 10/7.15 * 0.05)² + (0.5ppm + 10/10 * 0.05ppm)² ) = sqrt( 0.57² + 0.55² ) = 0.8 ppm

But that is just a part of the uncertainty. You have to add the noise of the readings and the uncertainty of your 10V source.

As long as you compare two 10V standards the specs for both meters is the same (but the question is then, if it makes sense to compare 10V against 10V in the 10V range).


We are not talking about a couple of ppm, but 0.6ppm on paper. Everything seems like the 8588A is in the same region of linearity and this could be verified for your very instrument. But as long it is not, you better stick to the specs on paper.

If the DMM is also your 10V reference, than the 2.7ppm vs. 4ppm on 10V might be also an argument for the 1year spec.

[Doctorandus_P]

I do not really understand the current ranges for bench top meters.
Sure, for a handheld it's convenient to have, but for a benchtop I'd much rather bring the current shunt to the device under test and only use sense wires towards the meter.

[NaxFM]

I do not really understand the current ranges for bench top meters.
Sure, for a handheld it's convenient to have, but for a benchtop I'd much rather bring the current shunt to the device under test and only use sense wires towards the meter.

You can with a simple script. In the fluke 8588A there's a built in function for this.
An external shunt is just more setup in a hard to control environment. If you don't care about the burden voltage it's much more convenient to use the DMM instead of a shunt.
After all it's not even that much cost to add a few shunt resistors and amplifiers for the different ranges, so why not?

[e61_phil]

I do not really understand the current ranges for bench top meters.
Sure, for a handheld it's convenient to have, but for a benchtop I'd much rather bring the current shunt to the device under test and only use sense wires towards the meter.

You can expand that to almost every measurement. Why having a 100V and 1000V range? Just bring your Fluke 752A

One real advantage of the build-in high current ranges is, that it enables you for fully automated calibration of multi product calibrators.

[maxwell3e10]

For best results on high current ranges it is better to use an external shunt. One can pick a shunt resistor value that optimizes burden voltage vs. power dissipation trade-off and mount it on a big heat sink.

Also for low currents 8588 uses low shunt resistors resulting in only 10 mV burden voltage, so it is probably pretty noisy. Not sure what is the logic of this, any current measured with an 8-digit meter would be a true current source that is not so sensitive to low burden voltage. 3458 uses larger shunts with about 100 mV burden and also goes down to lower current ranges.

I'd be curious to see the short-term noise performance of 8588. I couldn't even find in the specs a table of aperture time vs. number of digits or rms noise.

[e61_phil]

For best results on high current ranges it is better to use an external shunt. One can pick a shunt resistor value that optimizes burden voltage vs. power dissipation trade-off and mount it on a big heat sink.

That's why you can store your shunt data into the 8588A and it shows directly the current.

Also for low currents 8588 uses low shunt resistors resulting in only 10 mV burden voltage, so it is probably pretty noisy. Not sure what is the logic of this, any current measured with an 8-digit meter would be a true current source that is not so sensitive to low burden voltage. 3458 uses larger shunts with about 100 mV burden and also goes down to lower current ranges.

I'd be curious to see the short-term noise performance of 8588. I couldn't even find in the specs a table of aperture time vs. number of digits or rms noise.

I can test it against the 3458A as soon as the meter is back against the 3458A.
I assume that the 8508A and the 8588A uses some kind of transimpedance amplifier for the small ranges. That would be a benefit, because the voltage above the feedback resistor can be quite high.

In one of their 8508A whitepaper they wrote:
"By employing appropriate design techniques, the input burden voltage that the multimeter presents at its input terminals when measuring current can be essentially isolated from the voltage developed across the internal current shunts, reducing the disturbance (burden voltage) that it presents to the source of current being measured."

Something like that would fit to their other techniques. The current source for ohms is also nice on the Fluke meters and allows for proper ohms guarding.


The comment was removed, but of course is it useful to have low burden. A 8.5 digit DMM is not only used to measure outputs of a calibrator. I often use my 3458A in the 1A range to measure small currents, just to reduce the burden voltage.