Andreas: You might have considered this already but - Another "Gotcha!"
Just something to check.
Hello MisterDiodes:
You are right I started with the 1K resistors without buffer.
(with 500 ohms resulting input impedance this is not a issue with the LTC2400)
But since I had some unclear offset issues due to rectification
of the input noise of the LTC2400 at the protection diodes
of the MAX4052A multiplexer I decided to use a ADA4538-1 buffer
between multiplexer and LTC2400. Of course the ADA4538-1 buffer
has a output filter 825R + some pF (carefully adjusted for linearity)
to prevent rectification of the LTC2400 noise on the output of the ADA4538.
So since 03.02.2015 I am using the ADA buffer.
I have never, ever measured drift like that on any UltraOhm PWW (measured using a real bridge) over a year,
So we have to regard the differences in measurement conditions:
You are using a clamp to contact the resistors.
I am soldering. I try to keep away the soldering heat from the body of the resistor
with 2 crocodile clips, but I cannot guarantee that there is no influence at all.
see also:
https://www.eevblog.com/forum/projects/vishay-bulk-foil-drift-after-soldering/msg445297/#msg445297I do not know wether you have a room with controlled humidity.
I have none. So humidity changes from around 40-50% in winter to 60-70% in summer.
With several % change possible over a few days.
Perhaps with "precision lab conditions" the devices would perform better.
But since I intend to solder the resistors in the final cirquit
(because most users of resistors will not use any clamps in their cirquits) I will not change this.
By the way: where are your measurent results on T.C. and drift?
Up to now I have seen none from your side.
ADC noise on the LTC2404 part you are using is not always perfectly random white noise - as you will soon be learning. Also- watch your Vref. It seems the drift data noise has gone up with increasing resistance - highly suspicious. That means something is probably wrong in measuring technique in your measuring jig itself.
By the way I am using the LTC2400 (not the LTC2404 which relative similar) together with a MAX4052A/MAX4051A multiplexer.
Noise picture I have added below.
VREF is irrelevant in ratio measurements.
Ratio stability at half input voltage over a >8 hour period is shown to be around 0.1uV (giving 0.04 ppm error at 2500 mV).
Another item: you are not sure how well your reference resistor is temperature stabilized.
Stay on the carpet:
A few days before I wrote that I have improved temperature stability from below 0.05 deg C to below 0.01 deg C.
I monitor each resistor with 2 temperature sensors.
Please read more carefully.
maybe measure a 10k/ 10k LTC5400 ratio part for a reality check (these will get you a known ratio in ppm area guaranteed). LTC5400's will have more shot noise than PWW but are a cheap way to do a basic reality check on your system.
Don´t make yourself ridiculous the LTC5400 has worse specs in datasheet than a pair of Z201
or a typical pair of UP805 PWW resistors from the same batch.
And the HP3458A is still one of the best multimeter.
The only part that I trust on around ppm ratio stability is a LTC1043 capacitive divider.
Sanity check data attached:
Ratio check VREF/VIN at VREF divided by buffered LTC1043 2:1 divider.
1) Raw data of 8 hours = 170000 values (in mV)
2) Averaged values 1 minute as I use them for measuring to reduce noise. And 25 minute averages to show drift.
3) Classified raw data showing nearly perfect gaussian distribution.
4) Allan deviation showing stability over time in ratio mode. (better than 0.1 mV for large averaging periods)
compare that with the absolute voltage mode of a HP3458A:
https://www.eevblog.com/forum/projects/project-kx-diy-calibrator-reference-sourcemeter/msg592144/#msg592144By the way there you can see: my ADCs (including compensated VREF) are even in absolute voltage mode
only around factor 5 less stable than a HP3458A if the 1 minute averaged values are used.
(please note that the measurements of 3458A are in "volt" whereas I measure in "milli volt").
With best regards
Andreas