Satish - CS5520 and CS5532 - see their website.

Erik - In the US, every scale manufacturer tries to show a higher resolution than his competitor. When I worked in NJ, I proposed a random number generator for the LSDs. Some joke. NIST Handbook 44 allows this kind of stupidity. Fortunately (or unfortunately for me) OIML (the international regulatory body) forces the manufacturer to prove his displayed precision in the face of time, temperature, hysteresis (see below), EMC, ESD, barometric pressure, and the price of fish ;-)

Kai - There's much worse errors than hysteresis in load cells! Hysteresis in load cells is caused by tiny plastic deformation of the substrate material. If the load cell manufacturer got his design right, this creep should be exactly matched by the creep in the strain gauge bonding adhesive! No wonder they are so expensive... But no, that's not the problem. Good quality simulator (HBM or similar) shows precisely the same effect.

One problem I encountered some years ago (but not on these ADCs) was if I sampled at 20ms. This was precise of course. The (European) mains runs at nominally 50Hz (+/-1Hz), so this is a supposedly good sampling rate for maximum 50Hz rejection.

But it's not so. If the mains is 49.9Hz, one ends up with a beat of 0.1Hz on the ADC output. This, of course, passes right through my subsequent digital low pass filtering (Fco variable, but in the 2-10Hz range).

I now run the sample rate sufficiently far from 50 or 60Hz to place any beat outside my filter passband.

Very best,
Dave