NI Scope vs Keithley 3458A - Accuracy

[JW-JnJ]

We have a calibration fixture that historically used the Keithley 3458A to measure Vrms and Irms at 50KHz and 500KHz. (spec sheet)

I have a problem using this because we have to measure the Voltage and Current at two different activations because the DMM is incapable of measuring both voltage and current at the same time. It slows down the process and introduces extra relays into the system which I don't like due to contact resistance and reliability over time.

An engineer refuses to budge on using this DMM with this argument. He says we need high AC accuracy across the entire bandwidth. The AC accuracy of the PXI-5922 is 0.06% and is only measured at 1KHz. He claims the DMM (in synchronous subsampling mode) has high accuracy across all of the bandwidth.

I'm not an expert on this, but I get the feeling he is comparing apples and oranges (even if they're named the same). The DMM is reporting AC accuracy after a Vrms calculation is done. The oscillscope is reporting sample AC accuracy. I would need to then calculate Vrms over multiple cycles which would then increase the accuracy. Am I right about this? I don't know how to answer why it's only reported at 1KHz and how much worse it gets at higher frequencies.

Are there any other advantages or disadvantages to using this DMM vs a high accuracy scope? I don't want to counter this guys argument unless I can build a pretty good case.

(side question is, is there a more accurate scope in the lineup?)

[HSD · Accepted solution]

The 5922 (and most digitizers/oscilloscopes) don't have flat responses, so the accuracy is limited (the specific accuracy is dependent on the product.  The 5922 has pretty good flatness.  The spec sheet calls out ~0.01 db @1MS/s.  Look at the frequency response graphs).  Averaging increases precision (repeatability) but not accuracy.  If accuracy is important then a DMM is a good choice.  If you need both V and I then use an SMU (the NI PXIe-4139 is a very popular option).

Sounds like a 5922 or SMU can accomplish what you need.  The 5922 has limited voltage and current protection so you'll need to be careful about your application (most SMU's are designed for high voltage and current).

[JW-JnJ]

Thanks for the answer.

It is high voltage (could be hundreds of volts), so unfortunately we have to divide down and use a current coil. That limits the option of using a pass-through option like the SMU. The high voltage, up to 500KHz, and up to 3A current tends to really limit our options. Hence the $10k DMM. 

[Henrik_Volkers · Accepted solution]

We use NI-5922 and [HP,AG,KS] 3458 and even a Fluke 4700 in our lab, and now look for known accuracy in the 100 kHz to < 1 MHz region too ...

However we handle only low voltages (<= 10 V) ... 

The 5922 needs a dedicated AC calibration.. @>100 kHz with 10 MSPS you already have systematic deviations due to the internal filters.   And you need to calibrate all the signal-conditioners ...BUT you gain measurement speed.

When it comes to accuracy, it's hard to beat the 3458 , and subsampling is different to AC measurement .. different specs ...

(and the new black edition, even by claiming the same spec, has a even  better linearity, as far as I understand my colleages)

So, I can understand your engineer..

And I try to get my hands on a Fluke 5790  (another beast, even more $$ ) for some weeks to validate my setup 😄

[JW-JnJ]

---

@Henrik_Volkers wrote:

We use NI-5922 and [HP,AG,KS] 3458 and even a Fluke 4700 in our lab, and now look for known accuracy in the 100 kHz to < 1 MHz region too ...

However we handle only low voltages (<= 10 V) ... 

The 5922 needs a dedicated AC calibration.. @>100 kHz with 10 MSPS you already have systematic deviations due to the internal filters.   And you need to calibrate all the signal-conditioners ...BUT you gain measurement speed.

When it comes to accuracy, it's hard to beat the 3458 , and subsampling is different to AC measurement .. different specs ...

(and the new black edition, even by claiming the same spec, has a even  better linearity, as far as I understand my colleages)

So, I can understand your engineer..

And I try to get my hands on a Fluke 5790  (another beast, even more $$ ) for some weeks to validate my setup 😄

---

This is exactly the type of thing I wanted to hear. Thanks.

I always hesitate to provide a custom calibration for a piece of equipment in this case because it is being used to calibration other equipment. The paperwork trail turns into an utter nightmare when it fails. At this point it's a "right tool for the right job" thing and I'll concede.