| ??? 03/16/07 23:49 Read: times |
#135152 - Weighing lore Responding to: ???'s previous message |
Stefan,
Sorry to take so long to reply - have been out of office investigating 'broken' load cell ;-) Most interesting reply. For those non-weighing types out there, Stefan's '100nV', he mentions, is his Weights and Measures Authorities determined minimum load cell input per division, and it is calculated for each instrument during approval error testing. So, for example, an instrument with '1uV' approval would not be allowed in a 6000 x 1(division) with a 1mV/V cell on 5v excitation as the output per division would be 5000/6000=0.83uV per division (it gets more complicated because no-one uses the full maximum range of a cell). Stefan's 0.1uV approval is fantastic (possibly an industry first?). I work to the 'standard' 1uV which most authorities accept without lots of expensive further testing. I also don't use AC excitation and have never had problems with zero drift - as described earlier. The low frequency gain instability is a cyclical (almost - but not quite repeatable) wandering up and down of gain - not zero - in the order of, at worst, 1 part in 30,000 (15 bit accuracy!) over the course of 10 seconds, or longer (<0.1Hz). As I said, I put this down to an ADC peculiarity - but would love to be proved wrong! Filtering is a particular bugbear. I use a peculiar combination of digital filters (which of course includes the ADC's own 5 sinc jobby). My favourite is a complicated sequence of different IIR filters equating to 5 poles and a little 'black magic'. Working with 8051s is a challenge because of the very limited RAM which precludes all but the simplest FIR fiters. My filter runs to 56-bit fixed point arithmetic... It is important I have fast response in high vibration environments - it's not a lab scale. When I have time I mean to do more research into the other types of recursive filters. I think the Flintec uses FIR (not sure who's micro, but suspect 32-bits and masses of RAM). I'm not suggesting for a moment that FIR is the answer, but am interested that you say their filtering is very good. Perhaps you should see mine... Best regards, Dave |
