Temperature measurement

vince expressed precisely :

In 1953 I had to research this very problem but applied to a jet engine test stand. State of the art then and no PICs or of the shelf data loggers OR GOOGLE.

Probably you should do some research before asking us for help.

I can still remeber mostly how it went.

--
John G

Google NE555. It'll do all you need + more.

We don't do people's homework for them for free.

Good Luck! Rich

....unless they've had a go and need a nudge.

Answer: Use a PIC.

The usual routine is to put the thermistor in series with a resistor which has the same resistance at the thermistor at 32.5C, and drive that from a reference voltage, probably a 1.23V bandgap - though you do have to keep in mind that it is not a good idea to dissipate much more than 10uW in the thermistor, since NTC thermistors tend to become unstable at higher power dissipations.

You then digitise the voltage across the thermistor using the same reference voltage as the reference for your A/D converter. NTC thermistors typically have a a temperature coefficient of about -4% per degree Celcius, so the voltage at the junction swings from about

25% of the reference to 75% of the reference - you can work out the exact value from the thermistor data sheet, which may even give you the parameter for the Steenhart-Hart relationship between temperature and thermistor resistance.

You can do most of this in a small single chip microprocessor, and turn the A/D converter output into numbers than you can display on small liquid crystal display - LEDs would eat up too much current in a battery powered device.

The people who sell these devices make them in very large volumes and can afford to develop an application specific integrated circuit for the job. How they put in the calibration data for each thermistor - this sort of application is a bit too price sensitive to let you use "interchangeable" thermistors which can be accurate to +/-0.2C (or

0.005F if you spend enough money) - is left as an exercise for the reader. A trimming potentiometer would work, but isn't the solution usually chosen.

-- Bill Sloman, Nijmegen

vince does not need a nudge, he needs a kick in the pants !!

Or he needs to stop partying and get to work on his own assgnment (assignments). Or he needs another career.

hamilton

How 'bout. Use a DMM?

George H.

How 'bout, use a freakin' thermometer! ;-P

Cheers! Rich

How do you insure accuracy without a standard. ?? How do you calibrate without a standard ?? How do you insure good connection on the skin ?? How do you compare that reading on the skin with another device ??

??

greg

You run down to Radio Shack and pick up one of their digital thermometer modules (if they still sell them).

I've had one for over 10 years, still running on the same single AA battery.

--
Paul Hovnanian  paul@hovnanian.com
----------------------------------------------------------------------
Have gnu, will travel.

I often use mine to calibrate things, since I know it accurate. Indoor/outdoor min/max. The hardest thing to do is know how to take readings of different things. I got a couple thermocouple backups, and I quit using mercury thermometer. I also use the IR.

greg

Do your own grade school assignment..

No need for a voltage reference. If the ADC uses the supply rail for its reference, as most uP ADCs do, and the measured voltage is a resistor-thermistor divider from that same rail, it's ratiometric.

John

Actually, the usual approach (by volume produced, not by number of "designs") is to make an oscillator with the thermistor and a capacitor. It's serially ratiometric when you throw in a $0.001 reference resistor and appropriate switching.

First thing, find out what that accuracy specification means. Is it one percent of the stated range (20 to 45 C is a 25C range, so that'd be +/- 0.25C), or one percent of the value as-measured (so it has to be +/- .20 C at the low end of the range, but can deviate by +/- .45 C at the high end of the range), or is it in percent of the absolute temperature (which would give you about +/- 3 degrees C), which is completely sensible in terms of Kelvin absolute temperature.

Then, you need to know what the thermistor characteristic is. The problem might be impossible, if someone foists on you a characteristic curve that isn't monotonic in the range of interest.

In 'real' engineering, you might not get paid if you don't get the full assignment done in accordance to the contract requirements. Negotiate those requirements now.

A whopping 15mA?? Thats a lot. 1.5mA would be more like a challenge :-) Google MSP430 or Cortex-M0.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------

On a sunny day (Mon, 15 Nov 2010 13:30:41 -0800 (PST)) it happened whit3rd wrote in :

That last thing is very unlikely, as this is obviously for measuring fever.

3° C difference is between undercooling and an extremely high fever. (About 37 °C would be normal, 40 °C would be really sick).

But the rail tends to be noisy. As long as you are using some kind of integrating A/D converter, this doesn't matter much, but with a successive approximation A/D it could be inconvenient.

-- Bill Sloman, Nijmegen