Currently I am working on a thermocouple amplifier design with an approximate gain of 240. The reference circuit is from Microchip
formatting link
There they recommended to use MCP619 which is a bit expensive op amp. I would like to use cheaper op amps like LM324. I know that the offset compensation scheme, CMRR etc are different if we compare LM324 and MCP619 but here, I wonder what are the most critical parameters to look at if I were to choose an op amp for such amplifier circuits.
On a sunny day (Mon, 19 Mar 2012 03:06:42 -0700 (PDT)) it happened Myauk wrote in :
In that circuit in that pdf you refer to, the opamp offset can cause big problems. It will probably work for some simple temperature controller, but the article is hardly complete. Maybe look at this for a different solution:
formatting link
I do the corrections on the PC side and that program uses lookup one way, and successive approximation the other way using NIST thermocouple data. The Linux PC program can do the conversions forward and backward voltage versus temperature for different thermocouple types.
Thank you so much for the info. If possible could you please explain a little bit more on successive approximation using NIST thermocouple data? I only know the look up table method.
On a sunny day (Mon, 19 Mar 2012 20:17:15 -0700 (PDT)) it happened Myauk wrote in :
My program can do 2 things, temperature to voltage, and voltage to temperature.
The program uses data from
formatting link
formatting link
formatting link
formatting link
formatting link
formatting link
Those are polynomial data for different type of thermocouples
For temperature to voltage the NIST polynomials from the above site are used in the normal way, for example E = sum(i=0 to n) c_i t^i
But if you want voltage to temperature, then you can use successive approximation to find the voltage using the same equations in very few steps.
Successive approximation works as follows, say you want to know the temperature X for voltage V.
Then for a given temperature range you calculate, using the normal polynomial method, the voltage for the temperature at half that range. If that voltage is smaller than V, then you divide the lower range it is in by 2, and calculate the voltage for half that temperature range, if the voltage was bigger than V, then you divide the upper range it is in by 2, and calculate the voltage for half that temperature range, You keep dividing ranges by 2, and repeating, until you get the temperature with the accuracy you want.
A simple example, say you have numbers ranged from 0 to 255, and a secret number
114. How many questions does it take to find the secret number? Sure you are not going to try each one from 0 to 114, that would take 115 tries.
You first ask if... is it > 128? (half range), no. (range is 128) step 1 is is > 64? (half the lower range), yes (range is 64) step 2 is it > 96? (half the remaining range it is in), yes. (range is 32) step 3 is it > 112? (half the remaining range it is in), yes, (range is 16) step 4 is it > 120? (...), no (range is 8) step 5 is it > 116? (...), no (range is 4) step 6 is is > 114? (...), no (range is 2) step 7 is it > 113? (...), yes -> > 113 and < 115 must be 114 (range is 1) step 8
So now we only needed 8 steps to find the 114, in this integer example exact, value.
In case of floating point values you can keep dividing ranges until you get the precision you want, Of course in the real world the precision is set by the hardware (noise, offset, etc) so better stop after a few steps, 16 steps gives you 16 bit precision for example.
This is a long story, and doing it (writing this story) in a programming language is much faster, you can download th.c here (C code, Linux, but this should really compile on all computahs):
formatting link
Please respect the GPL license, Or else the FSF may want to talk to you.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.