need a temperature sensor with .1 degree accuracy

Are you sure you need 0.1 degree accuracy?

The WMO requirement for meteorological institutes on ambient temperature measurement is +/- 0.3 degrees (WMO Guide on Meteorological Measurements and Methods of Observation)

A Pt100 Class B (IEC-60751) has an accuracy of +/- 0.15 degree at 0 deg. C and a reasonable low error in temperature coefficient, so you can have a sensor accuracy of let us say +/- 0.25 degree over a certain temperature range

You need to add the inaccuracy that is introduced by the amplifier / interface circuit / ADC. It requiers accurate (read: not so cheap) components to build an interface circuit that does cause an error less than 0.1 degree over a reasonable temperature range.

Since Pt100 elements are very stable over a long term, you can obtain a higher accuracy from a Pt100 sensor if you can calibrate your sensor against a high accuracy standard. This is one of the main questions in this story: How are you going to calibrate your system ?

Note that any system that measures temperature with an accuracy of better than 1 degree wll need serious calibration.

You will need to calibrate your sensor. How are you going to do that ?

I don't see item (3) ever meeting your accuracy requirement, no matter how much you are willing to spend. Even if you could measure the radiated power accurately, uncertainty in the emissivity of the measured objects would contribute too much error.

Mark

Yes. Getting 0.1°C *resolution* is easy and cheap (even with a relatively low-level sensor like a thermocouple).

Getting 0.1°C accuracy is not easy and not cheap, and it only gets worse the further you move from the ideal fully-immersed-in-benign-liquid "CSTR" type of situation.

Best regards, Spehro Pefhany

To know how difficult it is to get 0.1 degree accuracy one must know the temperature range to be measured and the ambient temperature range for the equipment doing the measuring. This also determines the resolution required of the ADC.

In my experience achieving this accuracy is not easy and not cheap.

Dan

-- Dan Hollands

Never mentioned if he wanted absolute accuracy. Probably not. Even if one can achieve having the instruments, there is so much variation when actually taking measurments. You could not find two places in a glass of water with the same temperature.

greg

Right!

Well, there are ways to measure temperature to millidegree accuracy, but there are caveats.

First, if you use a physical sensor (RTD, diode, thermocouple, etc.) you can never ever measure the temperature of anything but the sensor. You have to then infer the temperature of your object or process from an estimate of the thermal coupling to whatever you're trying to measure, and an estimate of the self-heating due to your probe signal, and an estimate of... (the list gets longer with increasing resolution and accuracy :-).

Second, as someone has already mentioned, it's very unlikely that anything except a really tightly controlled material of high thermal conductivity in a very tightly controlled environment will be uniform in temperature to 1 degree, much less 0.1 degree. The larger it is, the less likely it gets.

Third, the optical sensing technique can indeed measure to 0.1 degree resolution and accuracy, but only at the specific location that is being observed, and only if emissivities of the target, the sensitivities of the receptor, and the details of the transmission medium and path are very well known.

The upshot of all this is that you will be able to measure something with 0.1 degree resolution, but it'll more than likely be irrelevant to anything you want to do or observe, and you can only get that accuracy of even the things you can truly measure (which will not include what you mentioned in your original post) by spending lots and lots of bucks.

So, I agree with Graham, here. You're chasing an unattainable goal.

John Perry

OT of course but

martin

I believe there is semiconductor device from analog device or national or maxim or similar that is capable of 1C accuracy within temperature range. It worth looking into this as it very easy to use.

All you have to do is to connects to ADC and that it(!).

Some recent device has build ADC which includes all accuarcy of the ADC and temperature sensor.

Failing that, you can get PT100 type calibrated probe and then select the semi temp IC to match the accuracy and reject those do not.

If commiting to PT100 solution, you need precision constant current source involving precision voltage reference (which interface to op-amp that provides regulated current output). Some care on circuit is needed.

Good luck

Riscy

I think your 'friend' is pulling your leg !

You would be doing well to get 1C accuracy.

Graham

Most semiconductor temperature sensors are usually PN juntion types where a differential current is pulsed. The best resolution on these is

Once again, as noted, it is measuring the temperature of the sensor, rather than overall ambient temperature. Even a slight air movement will throw the readings out by far more than the OP wanted.

I agree with the previous posts that the OP is chasing an unattainable goal.

Note that Maxim does have a nice sensor (minus the PN junction - supply your own 2N3904 with base tied to collector for best results) that has a parasitic reduction / elimination mode which is extremely useful for measuring die termperatures when the transistor has been integrated on chip..

That device has accuracy of 1C (but again, only across a limited range).

Cheers

PeteS

Within 1C of What??? Describe your reference and measurement technique. Thanks, mike

"Cheap" sensors are, by definition *cheap* and have poor accuracy; no where near what you think you need.

It helps if you use AC - or at the very least, reversing DC excitation.

You can only put a limited amount of current through a Pt resistance sesnsor before self-heating becomes a problem - 1mA through 100R is typical, giving you a voltage drop of 100mV. As the resistance of a Pt resistor changes by about 0.4% per degree Celcius, measuring to 0.1C involves resolving a voltage difference of 40uV.

You can buy interchangeable thermistors from Yellow Springs Instruments (YSI) which are good to +/-0.05C from the YSI 4604x series (where x can be 0,1,3,4,6 or 7).

Newark stock the 46046 as part number 52F9276 - and want $187.20 for each part. they are sold out at the moment and the lead time is currently 74 days. This part has a resistance of 10k at 25C.

The power you can dissipate in a thermistor is limited, both by self-heating, and - usually more important - by the nasty tendency thermistors have of giving unstable resistance values when they are dissipating too much power. I usually limit the dissipation to 10 microwatts, but people using thermistor as sensors in microdegree controllers have chosen to hold this as low as 2uW (to get measured a stability of +/-15uC).

A cheaper, lower tolerance thermistor can be calibrated to +/-0.1C accuracy with a single point calibration in a well-stirred ice bath, where the thermistor is completely immersed in crushed ice (made from distilled water) through which distilled water is being foced to circulate. When you set up one of these baths correctly, you get a stability and accuracy of the order of a few milli-degrees.

----------- Bill Sloman, Nijmegen

Seeing all the posts, I'm not sure! if he's gonna get this fabeled accuracy. I think this is a pretty complex (!) topic, and I better be spending some more time reading up on calibration and such..

Calibration? Last time I calibrated my NTC thermistor I did in Ice (0 degree C) and boiling water etc. etc. It was a 10K NTC and there was also a reference sheet with it, I remember. But nobody tells me how Accurate I can go with this 10K NTC FENWAL thermistor I have with me already?

Lets say I want As much accuracy as possible with a *decently low cost solution* (200 $ ?) To top it up, I'm in India, but I think getting electronic components is not much of a problem.

This PT100 class B appeals to me certainly.. I'll work on it.. But how do I go about calibration? can I use distilled water Ice etc. as one poster mentions?

I got hold of a Cole-Palmer scientfic catalog and they have this .1 Degree accuracy on their SAMA glass thermometers. cost around

Further reading this book, and as some other posters point out, it is clear how vain a pursuit the radiation temperature measurement at this accuracy shall be. The best models I see in this catalog give me around

Just curious, But I remember reading that CCDs have similar response to IR as to light.. is this so? if so, can there be this remote possibility that I turn a 5 Mega pixel Digital camera as a thermal imager of sorts? maybe with appropriate "IR filters" ?

Also, Can I ever use IR diodes (like those in a TV) for any kind of temperature measurement? Many thanks Suraj

Seeing all the posts, I'm not sure! if he's gonna get this fabeled accuracy. I think this is a pretty complex (!) topic, and I better be spending some more time reading up on calibration and such..

Calibration? Last time I calibrated my NTC thermistor I did in Ice (0 degree C) and boiling water etc. etc. It was a 10K NTC and there was also a reference sheet with it, I remember. But nobody tells me how Accurate I can go with this 10K NTC FENWAL thermistor I have with me already?

Lets say I want As much accuracy as possible with a *decently low cost solution* (200 $ ?) To top it up, I'm in India, but I think getting electronic components is not much of a problem.

This PT100 class B appeals to me certainly.. I'll work on it.. But how do I go about calibration? can I use distilled water Ice etc. as one poster mentions?

I got hold of a Cole-Palmer scientfic catalog and they have this .1 Degree accuracy on their SAMA glass thermometers. cost around

Further reading this book, and as some other posters point out, it is clear how vain a pursuit the radiation temperature measurement at this accuracy shall be. The best models I see in this catalog give me around

Just curious, But I remember reading that CCDs have similar response to IR as to light.. is this so? if so, can there be this remote possibility that I turn a 5 Mega pixel Digital camera as a thermal imager of sorts? maybe with appropriate "IR filters" ?

Also, Can I ever use IR diodes (like those in a TV) for any kind of temperature measurement? Many thanks Suraj

Will you take a look at this:

and this:

..... maybe...

A silicon CCD measures infrared only up to 1.1 um, not suitable for thermal detection at room temperatures.

By the way, here is a product that claims a 0.1 C sensitivity:

However, you will not get that 0.1 C accuracy if you don't know the emissivities of the different objects you'll be measuring.

Probably not for you, they'd just work for high temperature measurements. I'm not sure if these diodes are silicon (good up to 1.1 um) or something else, but it's a pretty sure bet they are NOT sensitive at 8-11 um. (Perhaps somebody else in here knows better?) A silicon photodiode might be sensitive at 250-300 C and higher temperatures, but that number is something of a guess on my part.

Mark

Did your calibration include measuring air pressure? You'll get a 1 C shift in boiling point for just a 4% change in pressure.

Mark

Depends. If he's just going to look at only one object and wants to detect a 0.1 C change in temperature, he doesn't need to worry about its emissivity. But if he is going to compare different objects, then it's a problem.

Mark