Spehro Pefhany wrote:
On Sun, 12 Dec 2010 13:26:25 -0800, the renowned Robert Baer
wrote:

Glen Walpert wrote:
On Sun, 12 Dec 2010 00:17:16 -0800, Robert Baer wrote:

Start with the tables available from the Minco.com site.
Pick and print a table for Platinum:PE TCR=3.85e-3, R0=100,temp range
20C to 220C inclusive.
Minco indicates A=0.0039083, B=-5.775e-07, C=-4.183e-12. Put the data
in a spreadsheet and do an XY error curve, tabular data
VS equation.
Note that i had to fiddle with B and C to get a reasonable fit, and
one is challenged to get two digits of significance for C.
See attached pic of error giraffe.
Interesting giraffe. Were you bragging about the goodness of fit or
complaining about it?

I see that your ±.00005 ohm deviation is well within the tolerance for
even IEC 751 Class A 100 ohm Pt RTDs, at ±(.06+.0008?T??2E-7(T^2)) ohms -
less than 0.1% of the allowable tolerance.

It looks like you either have measurement data in your tables, with
measurement noise (rather than tables reconstructed from equations LMS
fitted to measurement data), or perhaps just round-off error.

I have used the equations (or as many terms of it as were useful for my
range), and then done a two point calibration against two temperature
standards near the ends of my measurement range with the actual RTDs to
be used, for final equation adjustment on a per-RTD basis, when setting
up temperature monitoring for test purposes where better than 1 C
accuracy was desired. The accuracies expected without individual sensor
calibration make the bounce in your giraffe seem pretty much irrelevant.

Regards,
Glen
1) That deviation is *after* the fix in the standard equation. In fact,
the fit is better if C=0.

Well, unless I misunderstand what you are saying, I see _no_ error
other than round-off, no "fix" required.

Umm... note that coefficient C is
only non-zero for the range -200°C = T = 0 :

For 0 = T = 850°C, the equation is

Rt = R0* (1 + A*T + B*T^2)



Best regards,
Spehro Pefhany
I Pete again (repeat): Minco indicates A=0.0039083, B=-5.775e-07,
C=-4.183e-12.
Nobody states that a C=0 is to be used in any temperature region.
I found the error at 220C which is slightly less than 850C; and for
the limited temp range from 20C to 220C the third term (as given) gives
an R value that is increasingly incorrect when compared with the given
table.
Now it might be that the error will eventually go toward zero at 850C
using all 3 terms.
It is possible that more terms would be needed, assuming the table is
precisely correct, in order to have a curve that fits better.