What is the next best ratio measurement device after the 3458A?

[e61_phil]

However, the stack of voltages of not precisely known(although they are stable and roughly equal in value) cannot be used to check the linearity of Fluke 5440B.

I thought it is a minor problem, if i connect the HP 3456A with the LO input to the 5440B. The 5440B should be a low impedance source and will drive the LO input an the guard of the 3456A. If neccessary one could also connect the guard from the meter with the guard of the 5440b instead of the output. The HI input has very low input current, so it shouldn't be a problem on a resistor chain.

Is my assumption wrong?

[TiN]

A very common failure of 3458A is the 'big bad chip'(U180 on A3 board).

That photo with U180's is real pain to see, at least for me. Still scares me at night 

used 3458A is a good buy, though a 34470A is probably better, if you sacrifice some accuracy, and get it "half the price", more features.

Big difference of 3458A is it's linearity and resulting ACAL feature, to significantly remove temperature/drift of working references.
34470A nowhere close in that area, making it nearly useless for ppm-level stuff, as it's challenge to distinguish DUT drifts/changes from meter's. But I'm sure you know that already, just to be clear for other readers

[zlymex]

However, the stack of voltages of not precisely known(although they are stable and roughly equal in value) cannot be used to check the linearity of Fluke 5440B.

I thought it is a minor problem, if i connect the HP 3456A with the LO input to the 5440B. The 5440B should be a low impedance source and will drive the LO input an the guard of the 3456A............

I agree with this. LO input of a DMM should be connected to a low impedance source but should not be "floated" such as connecting to the middle of a resistive divider. I tried this with my 3458A(when I connected to 10k standard resistors in series to an 20V source) and got a very noisy result with the guard terminal connected to earth and LO,

[wiss]

I connected the low end of the divider to the divider case that was also connected to mains-earth-ground and all MEG of multimeters to that same MEG power outlet as the case, for unsyncronous meters (my PREMA and Keithley 192 for example) you get a very characteristic beating between line frequency and conversion clock frequency when you screw this up, with a synchronous meter I just got an offset...

[e61_phil]

One more option to get a sub-ppm linearity check for not much money is to use the AD5791 evaluation board. My sample is below 1ppm INL over 0-10V scale (measured by HP3458A) and there is a space on the board to put LTZ1000 reference on (which I did). As the stability of the DAC output ratio is better than 0.2ppm, it should be possible to calibrate at least some points using HP3458A to about that level of non-linearity (or even compensate for it with a secondary correction DAC). Also this DAC can be used in a bridge configuration to measure the ratio directly using a null-meter output to control the DAC for balancing the bridge.

Cheers

Alex

Just to get it right: If I would build a bridge with a AD5791 (specified INL 0,19LSB ~ 0,2ppm) and measure the residual voltage. My uncertainty in a voltage transfer (for example the voltage measured on a 10k resistor (well known) and a 1k resistor (unknown) 1:10) would never be smaller than 2ppm? Is it that simple?

[Dr. Frank]

Good advice. A very common failure of 3458A is the 'big bad chip'(U180 on A3 board). Here is another photo of collection of the replaced U180, most of which result in halt of the meter.

Hello zlymex,

can you please unfold, which error messages are related to the failure of U180?
What do you think, is the real root cause inside U180 to let the 3458A halt?
Maybe excessive drift of its internal resistors, so that certain ratios of these are outside limits?

Is error 202 'Slave Test: Convergence' also related to such a failed U180?

Thank you
Frank

[Dr. Frank]

Just to get it right: If I would build a bridge with a AD5791 (specified INL 0,19LSB ~ 0,2ppm) and measure the residual voltage. My uncertainty in a voltage transfer (for example the voltage measured on a 10k resistor (well known) and a 1k resistor (unknown) 1:10) would never be smaller than 2ppm? Is it that simple?

Sorry, the AD5791 is specified as 1ppm INL plus 1ppm DNL; both have to be taken into account.
It's the stability of the INL only, which is 0.2ppm.

Please check hpj4/89, page 22 and following, about the 10:1 transfer calculation related to DNL and INL errors.
Therefore, you can't expect better than about 11ppm 10:1 transfer from that AD5791.

If you characterize some ratios with a 3458A, I really doubt, that this would be a stable solution, due to varying code, noise, temperature influence (specified???), and so on.

So, if you need a 10:1 ratio only, instead of random ratio numbers between 1.0 and 0.1, better build yourself a simple, fixed 10:1  Hamon  divider, with self-adjust capability. (6 stable PWW resistors required, only)

This will probably give better 10:1 transfers (at low voltage levels) than even a 3458A, <= 0.2ppm of output is possible.

Frank

[zlymex]

Hello Dr. Frank,

There are 11(out of 13) chips were marked "??" in Chinese which means dead/halted in the photo I posted. This is the most serious error that U180 may cause to an 3458A. The halt may be occasional or only on DC range(according to the hand writing on the chips).
There is one marked as "AC measurement Ok", which probably means DC will be halted as well.
There is only one saying "workable but large tempco", I assume there is no error message associated with it.

I've just asked the one who provided the photo(thy888), he said they were all halted with error message 114, and the auxiliary message being either Multi-slope ADC error or Time Out error.

There are two root causes, one is the severe mismatch of resistors, the other is the bad input switching JFETs of the ADC. There are two paths for the signal, one is the high resolution one(50k resistor) the other is the high speed one(10k resistor).

I once try to fixed an 3458A with large daily drift but failed, there is no error message, just like mine. At the end, I suspect it was due to the U180, had it replaced, and the meter is Ok. There was a thread at 38hot that record my work: http://bbs.38hot.net/thread-436-1-1.html  http://bbs.38hot.net/thread-4104-1-1.html

For that error 202(Slave Test: Convergence), it is caused most probably by open JEFT(cannot be closed) of the high resolution switch for 50k resistor, but the high speed JFET switch will probably work, this can be verified by setting the NPLC to a very small value(so that only high speed path is used) and see if the meter will work. And yes, those 10k/50k resistors and JFET switches are all part of the U180 chip that cannot be replace alone easily.

Edit: add a pic from HPJ 1998-04

[Kleinstein]

Likely they will not reuse the old U180. It's failure prone and likely quite expensive. Today would likely not us a single converter for high speed and super good INL, as there are rather good fast converter chips (LTC2378.. and similar) that could be used for the very fast tasks and AC.

They may still end up with a new similar chip though, but if they don't need high speed this could be a simpler version or than could get away with separate resistor array and switching chip, maybe even of the shelf parts. My guess would be more like continuous integration, integrator with AZ OP and than an ADC chip for the residual charge - so more like in between the 34470 and the Datron 1281. One thing they should really add is a temperature controlled fan and better RF shielding (way more cell phones and WLAN today).