Keithley 2000 TRMS converter noise

[jpb]

A better test would be to see what level of input signal can actually be measured within a certain % maximum error. I don't have a suitable source to check...

My 2000 was calibrated just last month so I had a look at the AC results and they are both below and above the calibration standards depending on the frequency as well as magnitude so the rms noise is hidden by calibration constants presumably. The values are all well within spec so unless my meter has suddenly changed, having a relatively high (on my second measurement) count of 149 is hopefully not a problem.

For example, the lowest level calibration standard was 10mV at 1kHz and my meter read this as 9.9993mV. At 100mV on the same range (100mV range) it read this at 99.9972mV for the 10kHz signal but at 100.0072mV for the 50kHz signal so it is difficult to deduce anything.

[jpb]

The mystery of getting different values comes down to measurement rate (as per one of the posts above)

Range    Fast    Medium   Slow
100mV   ~20      162        162       
1V            40      154        155
10V          39      155        155
100V        39      154        155

So it seems you get less noise at the faster rate (or at least less offset)

[robrenz]

My previous post here was with the 3Hz filter on. These are with the 200Hz filter on and are significantly lower.

Edit: Very uniform 34% +/-1% reduction from the 200Hz filter.  The 8846A high med and low resolution modes in AC only change the normal display resolution and do not affect the stat mode resolution or readings at all.

100mV range was 12.48081µ with 3Hz filter


1V range was 106.8218µ with 3Hz filter


10 range was 998.5290µ with 3Hz filter


100V range was 11.89725m with 3Hz filter


1000V range was 110.9135m with 3Hz filter

[macboy]

Off-the-shelf, it is just about as good as it gets, and has been for decades.
Better can be done discretely. I have a HP audio distortion analyser that has several AD637 in it, but not to do the actual readings. They only to do the auto-ranging and other side functions that are not directly part of the measurement. The measurements are done with a discrete RMS converter that is presumably better.

Could you state which kind of discrete RMS converter you use? What is the advantage over a fab trimmed bipolar ADI part?

You can find the circuit and a very good description of its operation in the 8903B service manual, available at Agilent's website.

[rf-design]

Off-the-shelf, it is just about as good as it gets, and has been for decades.
Better can be done discretely. I have a HP audio distortion analyser that has several AD637 in it, but not to do the actual readings. They only to do the auto-ranging and other side functions that are not directly part of the measurement. The measurements are done with a discrete RMS converter that is presumably better.

Could you state which kind of discrete RMS converter you use? What is the advantage over a fab trimmed bipolar ADI part?

You can find the circuit and a very good description of its operation in the 8903B service manual, available at Agilent's website.

Sorry, could not find the difference. I assume that HP is using similar matched quad NPN. Otherwise the performance lacks because of mismatch and non-thermal tracking. I think that the AD637 could be much better because it could use internal trimming methods to improve over an already matched quad.

Could you state what is the specific advantage of the HP8903B rms detector over ADI637?