AVR-Chat

I know I'm coming in late on this subject, however I've worked an  
awful lot with thernocouples so I thought I'd add a few items...

>
> I am probably wrong, but I thought cold junction was needed to make  
> the
> circuit accurate ??
>
> -- 
> Regards,
>
> Kat.

If you're using a second thermocouple as a reference junction then  
all that's required is that you know the temperature of the second  
junction. I usually use either 0 degrees or 150 degrees (C) as a  
reference temperature since these are standard temperatures for which  
there are plentiful tables. For most of my designs where I'm not  
hooking up to a purchased ice point reference I use either a Dewar  
filled with icewater and reference junctions (stir occasionally) or a  
block of aluminum heated to 150 degrees.

For a homespun electronic reference you can use some other  
temperature sensor that tells you the ambient temperature at your  
instrument. Then you can calculate the correction voltage that a  
second TC would have provided at that temperature, use it to offset  
your measurement TC voltage as recorded, and then use your TC's  
linearization equation to get your temperature value. The way I  
usually do it myself is have an actual second TC junction at the  
point where I'm measuring temperature with the thermistor, RTD or  
whatever.

Some further tips, some applicable to this project, others not:

I recommend putting a decent bit of thermal mass on your reference.  
This will not only provide better readings, but allow you to skip  
checking it on every TC reading. You can just use the same offset for  
several consecutive TC readings and save some CPU time.

Keep your TC's P and N connections at the same temperature. Normally  
this isn't a problem, but I once had a tech who had a Bright Idea. He  
wired the P and N sides of about 300 TCs to different terminal blocks  
inside a NEMA box, all the Ns on one side of the box, all the Ps on  
the other. The temperature readings exhibited a shift through the day  
as the sun moved. Cloudy days gave us good results. I was tearing my  
hair out until I inspected the box for myself. I then explained the  
importance of placing the P and N terminations on consecutive screws  
on the terminal strip to the tech. Once it was rewired, all was well.

If you use a hot point reference, drill holes well into the thermal  
mass and bury the TCs in them, don't just screw them onto the  
surface. I find sinking them in with a blob of epoxy works pretty  
well. Give the mass plenty of time to come to temperature. For single  
junctions, I use a chunk of aluminum about 3cm on a side when I want  
something fairly small. Another thing you can do is drill the holes,  
cut the mass in half through the holes for the TCs, spot weld the TCs  
at the bottom of one side of the hole, screw the two halves together  
again, and place a blob of epoxy over the top of the hole to cap it off.

When in doubt, test your setups with a Dewar filled with ice water.  
They don't lie, whereas electronic reference points lie all the time-- 
even commercial ones from reliable suppliers. Lots of ice, just  
enough water to let things flow. Stir every so often and keep it in  
the shade if you're outside. Keep the TC junction off the wall.  
Freshen up the ice every so often.

If you've got a fixed installation where you use TCs a lot I  
recommend building an ice bath with a stirring mechanism even if you  
normally use electronic ice point references. I've solved a number of  
vexing problems by switching a bank of TCs from an electronic  
reference that had gone on the blink to a reference junction in an  
ice bath (and managed to keep some schedules that wouldn't have been  
kept otherwise.)

-Mark G.