EEVblog #723 - Keysight 34470A 7.5 Digit Multimeter Teardown

[EEVblog]

Are the front ends in these new meters still partly discrete? (ie. discrete matched FETs as inputs on the initial differential amplifier.)

The Keithley DMM7510 7.5 digit meter uses a lot of matched paid FET's

[Marco]

The only ones which could be it on Keysight are Q301/302/401/402 ... but there are no guard rails on them, so it seems that Keysight at least have ditched them.

The diversity of opamps and multiplexers Keysight uses is almost silly BTW.

[Marco]

Unless I'm missing something I don't see anything which could be matched (J)FETs in the 34470A, or even non matched individual ones, no metal cans for a start ... but even if they used some in plastic packages (like the ALD ones) there should still be guard rings.

[Marco]

Just like the 3458A, I think the "autocal" feature is great, but it's no substitute for ultra high quality components.

Hard to say without doing the math (which I can't, I still don't understand how they even deal with dielectric absorption). In the end if you can keep it steady enough to keep the errors below noise while throwing the occasional cycle of autocal measurements in between the normal ones that's enough.

[Dr. Frank]

Just like the 3458A, I think the "autocal" feature is great, but it's no substitute for ultra high quality components.  ..The Fluke 8508A is accurate to full specs +/-5C from Tcal, and this is *without* any "autocal" feature. 

I think, i already posted this earlier in this thread..

The AUTOCAL of the 34470a is really nothing compared to the 3458a. It  also has to be calibrated on each mode and each range, whereas the 3458a only needs 10V and 10kOhm references.
Simple reason: The 34470a only is linear to 1ppm, but the 3458a to 0.02 ppm, so precise 10:1 Transfers are possible.

The 8508a only lives from its many ultra stable Vishay bulk metal foil resistors.
Its linearity is not as good, as the 3458a, so it needs the usual > 34 Cardinal Point referencess from a calibrator, plus additional ratio standards..

If you would use better resistors for the 3458A, it would outperform the 8508a, of course.
Stable thermal environment is needed in any case for analogue metrology.

To my opinion, Keysight is not capable anymore to design such an improved sophisticated instrument..brain drain.. Otherwise the 34470a would've been better.. worth the 7 1/2 digits..

Frank

[plesa]

Interesting comparisons Keysight 34470A with Keithley DMM7510 http://imgur.com/gallery/Wbx7q

[saturation]

Very nice, when the Keysight is configured to match the Keithley, you get more with Keithley for your dollar.  The test equipment for the evaluation can't be beat.

Interesting comparisons Keysight 34470A with Keithley DMM7510 http://imgur.com/gallery/Wbx7q

[HighVoltage]

Interesting comparisons Keysight 34470A with Keithley DMM7510 http://imgur.com/gallery/Wbx7q

The results of the 34470A, shown in this comparison, match my own findings of my 34470A.
As if the readings are always a little low. (uV and nA)
But, who knows, may be the Keithley instrument was just fresh calibrated.

Would be nice to see Dave do a comparison between the instruments.

[Vgkid]

Looks like the Keithley has a fair bit less std dev in the measurement results.

[twobeseen]

Hi would somebody be so kind and explain to me why DMMs always have 4.5 or 7.5 displays? What about that half display? why not 4 or 7...
Also I don't see a "half" display, the last number always goes from 1 to 0

I just don't get it.... Am I just stupid? This must be something so easy to understand that nobody explains it in the data sheets.

[twobeseen]

somebody was kind, the half digit seems to be the MSB, that only goes from 0 -> 1, because the whole range is divided into multiple ranges e.g. 100mV, 1V, 10V, 100V