Stefan Heinzmann said:
Hi all,
I was wondering whether one can make any assumptions on how well
resistors are matched in a resistor network. It is quite convenient
nowadays to buy little SMD resistor networks with 4 independent
resistors of equal value. The tolerance usually is 5% or 1%, but maybe
the matching between resistors of the same package is actually better,
since they're manufactured together.
So the question is: Do I have a reasonable chance to have better
matching in a resistor network than in separate resistors with the same
tolerance?
Obviously, it depends on the manufacturing method whether this is true
or not. If it is similar to IC fabrication, the matching will be much
better than the absolute accuracy.
As you probably know well, the absolute accuracy is not important in
some applications as long as you have good matching.
Cheers
Stefan
Hi, Stefan. Your assumption is fairly common, and unfortunately, not
a good one. Screened thick film, even in its cured state, is not
homogeneous. When it is trimmed, the intense localized heat of the
laser leads to very significant temperature gradients, which causes
unrelieved mechanical stresses within the film near the cuts and a
phenomenon called microcracking. Microscopic radial cracks, which
extend out from the length of the laser cut, and also out from the
edge of the cut, are areas of weakness on the surface of the resistor.
These are actually one of the main limiting factors in trying to get
precise tolerances on thick film, and why tolerances better than 0.5%
are usually very expensive and custom-order only. By using premium
non-blended inks and trimming at slower rates with less power, you can
help to minimize this problem, but it will always be there. No
manufacturer of thick film is going to cut laser throughput to 1/2 to
1/5 of normal rate (pretty standard with resistor matching to 0.1%
tolerance) to meet an unspecified requirement. The microcracking
effect is usually responsible for drift up, and is usually less than
0.5%, so the trim shop will just target a little on the low side, and
run subs through as fast as possible to meet spec.
One thing you might want to try to illustrate this problem is to pick
out a few 5% SIP resistor networks having two resistors which measure
nearly identical ohms. Write down the values, then put them on rated
wattage a few times for a couple of minutes, allowing them to cool
after each time, and then measure them again. Your identical
resistors on the 5% SIP may not be identical any more, and will likely
have somewhat different values than initial readings. With thermal
cycling, the thermal expansion/contraction cycles will tend to extend
the microcracking, causing the resistance value to increase just like
extending the laser cut. (Remember that this problem is separate from
the issue of thermal coefficient of expansion differences between the
thick film and the substrate, which can be greatly reduced with
premium inks and good process engineering). Again, since thick film
by nature is not homogeneous, this will vary from resistor to
resistor, even on the same sub with resistors right next to each
other. The effect can be minimized by using very conservative loading
of the SIPs or SMT parts, and also by handling the parts with care
(mechanical shock also extends microcracking, much less of a problem
with the smaller SMT parts) and minimizing the amount of heat applied
to the parts during soldering.
Another respondent suggested using thin film precision-matched
resistors, and that sounds like a better idea, if you want a
"for-certain" engineered solution. If you want a hack on matched
resistors, grab a handful of 1% units, and cycle them at rated wattage
a few times to age the thick film. If they're high resistance and
rated wattage is impractical, set your oven for 120 degrees C, put
them on a baking pan, and put them in and out of the oven a few times,
allowing them to cool completely after each removal. Then measure
resistance, and sort out your best matches. This isn't by any means a
guaranteed method, but I guess it's better than nothing.
Good luck
Chris