Thermocouple signal amplification

I hope your comment refers only to the ice water bath -- you have to have a reference junction compensation of some sort regardless. "Universal T/C front ends" work by bringing the thermocouple alloys to a pair of terminals of accurately known temperature, then calculating the "cold junction" temperature correction. You can't get a valid thermocouple reading without it. The "universal T/C front ends" use calibrated thermistors or diodes to measure the reference junction terminal temperature.

jp

"Spehro Pefhany" a écrit dans le message de news: snipped-for-privacy@4ax.com...

But you can still measure the LM335 temperature with a pyrometer.

Over the years, I've developed a fear of that sort of thing, at least in the kind of EMI-rich industrial applications where thermocouple temperature controllers typically live. Possibly if the impedances could be kept down in the ohms range, the extra bandwidth could be kept out of trouble without having to spend more on shielding and isolation.

Best regards, Spehro Pefhany

Spef,

I've often wondered why process control often relies on a single sensor. Why not use two sensors and compare the readings? This will help catch RFI/EMI problems, allow you to monitor sensor drift, and provide a hot backup if the main sensor fails so it can be replaced when convenient.

These people seem to have the right idea:

My comment refers to old-fashioned linear analog electronic compensation methods, not some lab icewater slurry setup.

Incorrect.

Among many other methods.

Best regards, Spehro Pefhany

A man with two temperature sensors is never sure? ;-)

Yes, if you decide to put the electronics out in the field (which has many consequences) then a dual sensor makes some sense (at least in process control, probably not machine control in most cases). A lot of the cost is in the wiring anyway, and of course downtime is typically extremely expensive.

Best regards, Spehro Pefhany

Thanks, Spef. Can you elaborate on the problems with remote electronics?

It seems transferring the information digitally would be much better than having long analog lines with millivolt signals.

Not only that, think of who gets to go out at midnight in a blizzard to replace the sensor:)

Mike Monett

Hello Spehro,

That's why I don't understand all the fuss about opamps. If he has a micro he can use any old amp and just generate a regular pulse that clamps the input to ground for a few milliseconds. Clamp, measure the offset, back to the thermocouple, measure voltage, subtract offset, back to clamp and so on. I'd probably use a couple of jelly bean pnp transistors, whatever is currently cheapest, some other jelly bean parts and a BSS123 to clamp.

Regards, Joerg

A more harsh environment generally, and even if they are sealed, and even if the seal works, there are often temperature extremes that affect the accuracy (especially in "hockey puck" head-mounted signal conditioners). Also the desire for intrinsic safety and two-wire supply means that there is less power available for the electronics and for sensor excitation (where required) so that may not be optimal.

Industrial Ethernet is gathering a lot of interest lately.

Best regards, Spehro Pefhany

OK, educate me. In 30 years of building data acquisition systems that acquire, among other things, thermocouple data, I've never seen one that didn't work exactly as I said, whether with ice, thermistors, or whatever. How would you get an absolute measurement out of the inherently differential measurement without a known reference point to remove the ambiguity?

jp

Okay. You do NOT want a temperature correction- superposition does not apply to a nonlinear transfer function. Think about how you would deal with the variables manually, if you wanted the highest possible accuracy, not the way a 30-year-old instrument design works.

Best regards, Spehro Pefhany

Spehro, the point is that you can't get an unambiguous voltage out of a thermocouple without a known reference temperature. You can't know the measured junction voltage until you've determined the interfering voltage from your measurement circuit. The simplest possible case is

Vo = (Vt1 - Vt2)

You have no access to either Vtx -- that's as true today as it was in

jp

Of course the reference junction temperature must be known.

Yes, exactly. A *voltage* correction is what's needed.

Best regards, Spehro Pefhany

....which you can't get without knowing the reference junction's temperature. As I've been saying all along.

jp

Can you afford US$97.00 for one off-the-shelf?

Good Luck! Rich

There never was any question about that.

The correction, however, is a *voltage*, which is a function of the reference junctions'/junction's temperature.

In the case of constant temperature (an ice-water slurry can come close) it reduces to a constant.

In the case of your 30-year-old instruments, it probably has been reduced to a straight-line approximation, which (in general) is accurate at only two arbitrarily-chosen points.

Best regards, Spehro Pefhany

Incorrect. This started when you said my assertion that the OP had to measure the reference junction's temperature with diodes or thermistors was incorrect.

...which in no way conflicts with my assertion that you objected to.

I haven't used 30-year-old instruments in, well, 15 years at least. I've _never_ used straight-line approximations. I've _never_ used an ice bath.

I _have_ used two-controlled-temperature reference junctions, thermistor-monitored isothermal reference junctions, RTD-monitored isothermal reference junctions, and diode-monitored isothermal reference junctions. I've not had occasion yet to use the new band-gap based IC-monitored isothermal reference junctions. (Note the emphatic repetition of "isothermal reference junction", which is an unavoidable prerequisite to getting the unknown voltage/temperature :-).

I have also used (for quick checkout -- never for data acquisition) the analog compensation circuits still sold in some handheld temerature meters, and which _still_ have to start by measuring the reference junction's temperature.

I have always done data acquisition with the NIST (formerly NBS) formulas _AFTER_ measuring the reference junction temperature.

jp

Nope, look again:

What you said, and I objected to, was:

Because it's not a "temperature correction" that is calculated.

You also said in the same post:

On which my only comment was "Among many other methods".

RTDs and various non-diode semiconductor methods are rather common. I've used temperature-sensitive resistors (not thermistors) in some high-volume product designs.

Best regards, Spehro Pefhany

Ah, now I see. Your one-word dismissal of my whole paragraph because you inferred something I didn't imply was very confusing. We agree again.

Of course.

Now that's interesting. In what cases do you not need a reference?

jp