Fast thermo-couple compensation circuit

[senso]

Good evening, I'm trying to design a fast thermo-couple compensation circuit, or use an available IC if they can cope with the speed.
I want to use a K-type TC, and want to sense a temperature range between 50ºC and 700ºC(its for an EGT probe), a 1ºC resolution is perfect, but I want to read it at 1Khz, maybe 2Khz if possible, I have already used the Maxim IC's and they are great, plug and measure, but they are way to slow for my use(good for the coolant temperature, but they are pretty expensive even for that).

I have found this Microchip App-note:
http://ww1.microchip.com/downloads/en/DeviceDoc/51889a.pdf

Seems a smart idea, just plug the TC directly into the ADC, and add code to do the rest, but a fast 18bit or more ADC is going to be ultra expensive.

Any ideas on what to use?
Best regards.

[mamalala]

Hmm, does it even make sense to read them that fast? Keep in mind that there is a thermal mass involved, even the tiniest thermocoulpe will not react instantaneous. Not only does the junction have a size, but the wires close to it will also heat/cool and add some delay until the whole thing reaches equilibrium.

After all, not much to gain if you sample much faster than what the actual thermocouple needs in time ....

Greetings,

Chris

[senso]

The sensor that I will be using is an exposed junction sensor, and as its to be used in two stroke engines that can rev past 16k rpm( 266.66Hz), with lots of exhaust waves travelling, 1Khz would allow me to have 5 samples per revolution, probably a bit over-kill but I could average them and have 1 sample per revolution.

[mariush]

You can get chips like AD595 for less than 10$ a piece in quantity : http://www.analog.com/static/imported-files/data_sheets/AD594_595.pdf 

The cheaper A version only gives you +/-3c accuracy (c version gives +/-1c) but should save you quite a lot of trouble with correction tables and math for the probes so maybe it's more reasonable than using an expensive 18bit adc .

[senso]

I have considered them, but Mouser does not carry AD which is a shame and paying VAT and customs for a Digikey order will put them at more than 20€ a piece, its a shame

[mariush]

I have considered them, but Mouser does not carry AD which is a shame and paying VAT and customs for a Digikey order will put them at more than 20€ a piece, its a shame

Check Farnell ... they have a portugal version of their site: http://pt.farnell.com/analog-devices/ad595aq/amp-thermocouple-type-k-dip14-595/dp/409194

[senso]

In Portugal(and other countries) they only sell to corporates, no sell for the individual, there is a represent of them, but there is still their extra fees and VAT, its just... 
And for a max of 700ºC the output is over 5v:
700ºC 7206mV

Using a resistive divider will throw some error into the measure, and having an extra supply for that chip, but at least it is a possible solution.

[C]

Just an idea,
have you thought of using everything you do not want as your high speed temperature sensor?
Just guessing, but you may be able to use a circuit that you can read real fast that gives you just a very raw guess what the temperature did between readings. If at a slow rate you can get a this raw guess = this temp your control may be fine.
   
Thinking you could use one of the good  thermo-couple amplifiers to drive two outputs. One output to your supper accurate SLOW reading circuit. The second output to a frequency converter that you can read FAST or something like that.

C

[mikeselectricstuff]

At the sort of temps involved  in an engine you can probably just ignore the cold junction temp and assume it's constant. It then becomes a relatively simple voltage amplify+ measure problem.
Even if you need high accuracy,  you don't need to measure the CJ temp quickly as it's not going to change anything like as fast as the sensor .
 
Linearisation is easy enough using a lookup table after digitisation.

Since processing and a/d conversion has become cheap, there really isn't much reason to use a dedicated thermocouple amp these days - you can do it way cheaper with standard parts and a little bit of code.

[senso]

So, I might try the microchip idea, sensor directly to the ADC, with a little gain and do the magic in the micro.

[Jon Chandler]

I took a different approach to reading thermocouples and used a Texas Instruments TMP512/TMP513 power monitoring chip.  The ADC for measuring voltage across a shunt resistor works well to measure T/C voltage and the built-in (or remote) temperature sensor provides cold junction compensation.

Here's an article on how I did it and the program I used.  I used a table for linearizing the T/C readings.

[ConKbot]

For 1-2khz on a thermocouple youre looking at really fine wire thermocouples, wire diameter, even with a "bare junction"  of something like 28awg, youre going to be thermal-mass limited

see
http://www.omega.com/prodinfo/TC_Graphic1.HTML for a time constant.   Even with 0.001" wire, youre still going to be mass limited with sampling at 1KHz

[miceuz]

I'm currently building a thermocouple amplifier out of basic stuff - instrumentation amp, external temp sensor, compensation and linearity correction in software. So far the workbench tests look good, just finalising PCB design.

I'm not interested in speeds that high, but I see no reason why the sampling frequency cannot be increased, provided firmware avoids floating point math by using lookup tables and you use thermocouple thin enough. What I would be concerned is noise that probably is present near the engine. I have posted my schematics at this thread - www.eevblog.com/forum/projects/any-opinions-on-pcb-design-for-thermocouple-interface/msg237241