AVR-Chat

--- In AVR-Chat@yahoogroups.com, "kernels_nz" <kernels@...> wrote:

> Wow, I see to be generating a bit of interest ...

Although I share Roy's distaste for allowing the wrong skill set to 
control a design approach, I can see reasons why what Hein describes 
could be optimum.  I think we're looking not only to resolve 10,000 
steps in 15 mV, but to have comparable total accuracy also.  Or, if 
we're going to calibrate, at least have sources of inaccuracy that are 
constant and stable in between calibrations.  So, to amplify the 
signal, and thus allow a less accurate A-to-D to be used, we'd need 
around 0.01% accuracy in the amplifier.  I'm not saying that can't be 
done, but that's not the question.  Can it be done more cheaply that 
using a mass-market delta-sigma A-to-D converter?  The accuracy of 
these converters is pretty much guaranteed by the conversion 
technique.  Using the load cell excitation voltage as the reference and 
having no analog gain eliminates a whole raft of error sources.  I 
think this is the right approach.

Also, I think the thermocouple effects are a red herring.  Indeed, you 
can't mitigate with low impedance because the source is low impedance.  
But it is normal to use wiring with the same material on the signal and 
ground side (or the two sides of a differential connection), not the 
differing materials of a thermocouple junction.  You're also not going 
to use materials chosen to produce the maximum thermoelectric effect.  
Probably the only part of the current path that isn't copper will be 
the load cell itself.  As long as both connections to that are at the 
same temperature, there should be no problem.

Graham.