Roberts II, Charles K. wrote:
 >> and if you wanted 10% accuracy, you'd need them to match and track
within +/-10ppm.
 >  So the mismatch in resistance is drowning out my measurement, and it
can't be accounted for as an offset?
 >  So what type of tolerance would that be to get it to work?

That works out to a combined tolerance of 0.001%.  This includes initial
tolerance (probably 1% at best), tempco (typically 100ppm/degC for
"good" dividers), voltage coefficient, phase of the moon, light in the
side office over there ... you get the idea.  As has been pointed out,
it's not going to work this way.

 >  So it will not work as it is done now? My boss designed this baby,
if his design won't work how should I tell him and
 >  convince him it won't work?

So, your real problem isn't technical but political.  Presenting a
problem is always easier if you can simultaneously present the
solution.  Any attempt at showing that the circuit doesn't work will
inevitably force you to conduct the tolerance analysis so, you may as
well do it ahead of time.  Start with one of the dividers perfect at
1000:1 and show what happens when you alter the other divider by its
nameplate tolerance.  The argument will then be made that you can
calibrate the error out.  Then show what happens if that tolerance is
reduced by a factor of ten (say from 1% to 0.1%) but that the value
changes (use real numbers and work through the math, assume ideal op
amps) from before the offset adjustment to after the offset adjustment.
This is comparable to the effect of tempco over a normal operating range
unless you have a fairly good ratio tempco (great ones can be 10ppm/degC
or better but note, this is per degree C).  Then show that all the
problems go away if you can put the sense resistor in the ground leg of
the HV output section (then it becomes principally an initial tolerance
problem).  If your boss is reasonable, he'll appreciate what you've done
and work with you to implement the solution.  If he's not, he'll spend a
lot of time, effort and money trying to get the wrong approach to work
and eventually, natural selection will take over.

Good luck,

Dave

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads