But 3456A have much nicer warmer LED feel display, than those unholy LCDs without backlight.
That is one of the mods I'm planning to do. I saw someone added backlight to 3457A and 3478A, it is somewhere on the eevblog forum and also I saw it on bbs.38hot.net forum as well.
In my opinion, proper LTZ1000 reference cost would blow up cost of such meter way too much (unless you snag one of those 3458A's refs for <100$), and like you clearly see, rest of circuitry do not match up to reference performance.
I thought so. I think I'll just add some foam to the 20 year old LM299 and it should be fine. Hopefully, such an old instrument, if I decide to send it for calibration it will be in spec for quite a while and sort of good enough to transfer some calibration points to Keithley 2015 which I suspect is a tiny bit off.
I would send K2015 for calibration, as it is more modern, and my primary workhorse in the lab, but I doubt it is as stable and as accurate as 3457A.
Well, generator's output is divided by factor of 10000, together with it's errors. This test was not targeting absolute measurement values, but rather relative result, given particular instrument and setup at given time.
3458A A/D is way more challenging, due to it's much better linearity and sensitivity, to be sure of which JJA is required.
Ahh, I think I understand now. You were sweeping only around 100ppm (~1mV) so it was more noise and DNL performance, not standard INL measurement.
I was reading a bit recently about measuring INL of 24 bit ADCs and multimeters and the classical way is either precise Kelvin-Varley divider or JJA. Another approach is measuring repetitive sine or ramp signals and accumulating histograms for all the codes. Both approaches seem to take very long time and require top notch equipment (e.g. K-V / JJA or >140 dB THD sine generator).
I was reading some papers that claim to make this easier:
"Low-Resolution DAC-Driven Linearity Testing of Higher Resolution ADCs Using Polynomial Fitting Measurements"
"High-Performance ADC Linearity Test Using Low-Precision Signals in Non-Stationary Environments"
"Algorithm for Dramatically Improved Efficiency in ADC Linearity Test"
"ADC Integral Non-Linearity Testing with Low Linearity Monotonic Signals"
I need to read them in depth and see if the contribution is really there. My thinking goes that long time has passed since 3458A development, plenty of research out there, most multislope conversion patents are now long expired and it seems that the only special thing left that 3458A has are the matched channel length FET switches. Everything else seems to be DIY doable (does not imply it would be cheap - e.g. vishay resistor networks for slopes).
The problem then would be how to measure DIY ADC INL/DNL parameters without JJA.