Your goal is low cost--skip dual slope for now...1/2% is way above that...

What strikes me from my lab bench experience with sigma delta front ends and
from some DSP code for
a sigma delta interpolator is the way this method always has a
part in the signal loop that drives as well as measures.  Look at the
way the input terminals of the comparator and/or amplifier  affect the
quality of the voltage on the interpolating capacitor and the quality of the
voltage
of the measured signal.  Either loading, or noise, or input current mismatch
could cause a lack of accuracy.  The 1/2% is likely an analog design
problem.

To keep your costs in the $US0.16 in volume for a quad op-amp
range it is hard to add resistors and
such...but that may account for the 1/2%.  See if you can think
of a way to use
the extra amps in the package as mirror image or chained pieces in the
signal path,(even if they cause an equal delay).  Use them to
create equal input offset currents--they are a matched set.  Use them
to reduce load on the measured signal.  Create noise immunity.
Where is their linear input operating voltage relative to the power
supply voltage available?--can their function as comparators be improved
economically by powering them from a separate voltage than the uP? (for 2
kinds of improvement potential--random noise, and DC operating point).

This economic balancing againxt performance
is the toughest and the most rewarding
part of the art--do not expect much help from any cookbook anywhere...
just trial, err, and observe observe observe...

John G. Austin TX

> -----Original Message-----
>Anybody had better than about 1/2% accuracy results from a Sigma
>Delta built discretely?
>Any thoughts on the process?
>This is a stunningly cheap and simple method for A2D if it would only
>work :-) - once sorted it would/will prove a boon to other PIC users.
>
>regards
>
>Russell McMahon

An LMC660 comes close to the bargin price, but the 324 should do. Some more
information is required, are you using a split supply for the OPAMP, if not
how close to 0V do you get. When you say 0.5% error at what voltage input
level is this? Is the error at the same input level or over the entire
range?
The 324 has a 2mV offset that's up to 1% error in 200mV! Also are you using
dual slope conversion or single slope?
The number of counts not altering the value indicates that this is what the
converter is reading, so I would dare say that the reference is also ok.
But a Z8! Bad boy!
Please tell me more

Dennis