HomeRoast Digest


Topic: Thermocouple Compensation (3 msgs / 115 lines)
1) From: David Iverson
Thanks for the feedback on thermocouples.  I do have a
cold junction compensator (LT1025) and the
thermocouple is run through an OpAmp tuned to output 1
volt at 100 deg. C.  Perhaps I didn't hook up the
compensator properly (should that make the output
linear, or does it just provide a constant cold
junction reference?), or perhaps I'm misinterpreting
the OpAmp output.
If you check pages 3 & 4 of the National
Semiconductor's Temperature Sensor Handbook
http://www.national.com/appinfo/tempsensors/files/temphb.pdf),there's a discussion of nonlinearities in the
temperature to voltage transfer function of the
thermocouple.  The curve they show seems to match what
I'm seeing in thermocouple vs. thermometer (Cooper 550
deg. thermometer from SM).
Or maybe I just need to recalibrate.
thanks,
  Dave
Do you Yahoo!?
Friends.  Fun.  Try the all-new Yahoo! Messenger.http://messenger.yahoo.com/

2) From: Jeffrey A. Bertoia
Dave
The cold junction essentially provides an offset, but that will look 
like a gain
error if your software 'thinks' it is there and it is not.
As far as the nonlinearity it is actually not quite as tough as Matt 
suggests.
The equation is T = a0 +a1*E +a3*E^3 +...+an*E^n
T= temperature (C)
an = type dependant polynomial coefficients
E = thermoelectric voltage (microvolts)
(after cold junction compensation)
n = polynomial order
NIST publishes the coefficients for J as:
Thermoelectric-to-Temperature
Conversion, Type J*
a0 = 0.0
a1 = 1.8843850x10-2
a2 = 1.2029733x10-6
a3 = -2.5278593x10-10
a4 = -2.5849263x10-14
(-200C to 0C with error range -0.4C to 0.5C)
a0 = 0.0
a1 = 1.9323799x10-2
a2 = -1.0306020x10-7
a3 = 3.7084018x10-12
a4 = -5.1031937x10-17
(0C to 760C with error range -0.9C to 0.7C)
*NIST Monograph 125, Table A6.2.3
You obviously want to use the second half of this table.
jeff
David Iverson wrote:
<Snip>

3) From: Johnny Kent
At 06:20 PM 6/9/2004 -0700, Dave wrote:
<Snip>
Dave,
I think where National have messed you up is that their straight line is
for zero error at 0 and 750C
"b) Approximate error in C vs. a straight line that passes through the
curve at 0C and 750C"
Which just makes the error least at those two temps and quite bad, as you
said, where you want to measure.
If you recalibrate for a temperature closer to what you want to measure the
problem should mostly disappear, by moving the error offset. That's
provided that the op amp gain is correct...
From the data and graphs at omega,http://www.omega.com/temperature/Z/pdf/z021-032.pdf in particular figs 16 and 17 and the polynomial terms beneath them it
seems that a J type thermocouple is actually _more_ linear than the K type
within the region you mentioned 400-450F.
Neither type should need correction for non-linearity for roasting temps.
It is not that obvious from the scale they have used so here it is=
 magnified:http://members.cox.net/gusty/Johnny


HomeRoast Digest