Increasing RTD accuracy

[leitmotif]

If your RTD/IO card system is repeatable you can compensate for the input card issues.

Use an ice bath, boiling water, and a known high accuracy thermometer. (The thermometer is needed because boiling and melting points vary with barometric pressure.) Even better if you can use a medium close the temperature you will be measuring.

At any rate, if you can take the output of the temperature of the IO card and create a correction factor. This will probably be an offset to the reading to correct to the "actual" temperature. This will have to be done for each RTD and IO card, and if the RTD is changed in the future will have to be repeated.

Tom
Looked at your statement re boil point and freeze point of water re pressure. I thought the freeze point was a constant and not affected by temperature change. So I looked it up. I think we can agree the effect of pressure change on boil point is much more than that on the freeze point. The bottom line is the effect on freeze point is nearly neglibible and only in the case of looking for or calibrating for 0.1 C tolerance is there any need to measure it.

" When are the boiling temperature and freezing temperature equal?
For all substances, as we lower pressure, the boiling temperature falls much more rapidly than does the freezing temperature. (For water, the freezing temperature rises slightly at low pressure.) Hence the obvious question: Are the boiling temperature and freezing temperature ever equal?
The answer is yes. At the low pressure of 611 Pa (only 0.006 times atmospheric pressure), pure water boils at 0.01 °C, and it also freezes at 0.01 °C. The combination of conditions (P, T) = (611 Pa, 0.01 °C) is called the triple point of water because, at this pressure and temperature ice, liquid water and steam can coexist in equilibrium. This point is used to define our scale of temperature: by definition, the triple point of water occurs at 273.16 K, where K is the kelvin. 273.16 K = 0.01 °C

Full article athttp://www.animations.physics.unsw.edu.au/jw/freezing-point-depression-boiling-point-elevation.htm#pressure

Dan Bentler

 

[Tom Jenkins]

We definitely can agree, Dan, that the impact is much more pronounced for boiling. Since the original poster seems like he is going to be picking fly specks out of pepper, so to speak, I figured I'd get a little pedantic about it.

 

[leitmotif]

Sounds to me like he is in a primary standards application where very close measurements are needed.

For most of us +/- 1 degree is more than close enough.

Dan

 

[mutabi]

Why do I need such accuracy? We are testing heat transfer efficiency. There may be only a 3 degree difference between tests so tenths of a degree play a big part. It can mean the difference between a customer accepting or rejecting an order.

It appears that I need to determine if I can establish repeatability. And if so, I can add an offset. Which kinda brings me in a full circle. In order to add an offset, I need the response of the RTD to be linear. And while the response of an RTD is almost linear, its not quite. Hence the need to use a 2nd order quadratic equation to curve fit each RTD.

 

[Timbert]

Trying to measure a mere 3 degree difference isn't probably something you want to do with RTDs, especially hooked to an industrial PLC.

Also it doesn't seem that you would need absolute accuracy; rather you would only to be able to accurately measure the relative temperature difference between two sides of some medium.

The first thing that jumps to my mind is a couple of thermistors and a Wheatstone bridge.

 

[leitmotif]

Why do I need such accuracy? We are testing heat transfer efficiency. There may be only a 3 degree difference between tests so tenths of a degree play a big part. It can mean the difference between a customer accepting or rejecting an order.

It appears that I need to determine if I can establish repeatability. And if so, I can add an offset. Which kinda brings me in a full circle. In order to add an offset, I need the response of the RTD to be linear. And while the response of an RTD is almost linear, its not quite. Hence the need to use a 2nd order quadratic equation to curve fit each RTD.

Instead of temperature to do your heat transfer calculations consider using ice - well established BTU content and to put it simplistically when ice is melted you know how many BTU were trasferred. Way oversimplified and just not all that easy but something to consider.

I dont think your issue lies with the detector but lies with the "readout" Stay with Platinum RTD - you may have to go to wheatstone bridge to achive the very small resolution you desire. You need to start talking with makers of primary standard instruments. Most of us are industrial guys who think +/- 1 degree is fantastic.

Also I caution you not to write your specs too tite. Precision can get very expensive very quick.

Saw a case at Boeing where they demanded +/- 5 degree tolerance on a 1,800F heat treat furnace. Upon research they only needed +/- 50.

Dan Bentler

 

[Tom Jenkins]

You are out of the realm of standard industrial applications, which is what I and many of the people here are most familiar with.

This kind of lab accuracy testing isn't usually done with PLCs. I guess I've usually seen LabView, PCs, and GPIB / IEEE-488 communications links on this kind of application. I suggest you start talking to a lot of different suppliers. If you start with the industrial guys that you know they can probably start pointing you at the right suppliers. Be prepared to ask lots of questions and don't worry about looking stupid - this will be a learning experience!

Let us know what you end up with.

 

[danw]

Siemens clamp-on flow meter for chilled water BTU measurements uses matched pair of RTDs for the inlet/outlet temperature measurements.

People like Minco offer special high accuracy (matched) RTD calibrations:

http://www.minco.com/products/temperaturedevices.aspx?id=80

I would be partial to Modbus from the temperature transmitter because it avoids the subsequent D/A and A/D errors. Here's Laurel's RTD xmtr spec, it isn't clear what units 0.02 are, but I'm sure a phone call would clarify.

http://www.laurels.com/transmitter-modbus-rtd.htm

Profibus or Foundation Fieldbus would also provide loss-less digital comm.

 

[killer]

To muddy the water further, I'm not sure why temperature detection by ICTD isn't more popular than it is. Its certainly cost effective. We use the Analog Devices AD590 a lot in industrial refrigeration applications.

http://www.analog.com/static/imported-files/data_sheets/AD590.pdf

 

[Timbert]

I agree they are nifty little devices.

If I had to hazard a guess why they ain't more common, I image a lot of people don't know that such devices exist or are simply more trusting of old school technology.

Still you need an external signal conditioner to use one with a PLC. Do you know of any off-the-shelf ones? We installed them in custom electronics so adding the necessary components to our own PCB was trivial.

 

[shooter]

Again, use the simple algorithm and use a offset for each sensor.
As it is impossible to use one sensor you must use an offset
and yes you can make your own equation (however simpler is a table and it is more accurate) keep in mind that you will have errors as you wil read in 0.1 or better .
for this high precision check for a differential analog signal like the XTR105 gives. it is direct in mA.
I do a lot of this high works as i control temps in holds better as 0.02 Celsius) i measure 0.002 C and control is 0.1 but the proces is not stable.

 

[mutabi]

Dang guys, thanks for the last bunch of posts. I have a lot more research to do now. Definitely some real good prospects!