Temperature Measurement with NTC Thermistors

[GiG]

Hello everyone,
I wanted to measure the temperature in Celsius with a 2k thermistor.
Does anyone have an idea?
The measurement range is between 0 and 45 degrees
-Maybe a sample code can be a good Help!

[Ttelmah]

There is example code with the compiler. This is for the E3 mini board,
with a KY-013 temperature sensor module plugged in. The Ky-013 has
a 10K resistor and a thermistor.
This uses a function 'thermistor_convert', which is supplied in the library
thermistor.c, and if you look at this it describes how to use it. You need the
rated resistance of the thermistor at 25C, and the beta of your thermistor.
If you don't have this, the formulae to calculate beta for a thermistor is
online.
For good accurate readings, it is much better if you use a proper Vref
source for the ADC, and the same voltage fed to the divider from the
thermistor.

[temtronic]

As Mr. T says to get accurate results ,you need a stable Vref for the thermistor. Depending on the PIC you use, it may have one, or you use an external 'precision voltage reference 'chip'.The VDD of a PIC will change depending on what the PIC is doing. Turning on an LED will drop VDD for sure ! That then changes the vref and affects the 'math'.
In the past I've used thermistors in constant current mode.You need to make a bridge from 4 resistors, feed into an amplifier, then to the ADC input pin, It will give beter results(more accurate), if designed properly but costs more in parts.
You'll also have to filter the PSU as well as the sensor pin in order to get good,repeatable 10 bit ADC results.
Another 'trick' to get good readings is to take several and average them. Using Olympic averaging, you take 10 readings, delete the highest and lowest, then average the remaining 8. It's fast and gets rid of 'bad' (out of range) readings. You should never rely on just one reading for control purposes. If there's an EMI event, that could give you a really,really bad result !

[GiG]

ok thank you helpful
but any sample code ?

[PCM programmer]

Ttelmah told you where to get it.

This file has the thermistor_convert() function:

[GiG]

[Ttelmah]

The compiler comes with that file.
No, we cannot upload it, it is CCS copyright.
If you have bought the compiler, then talk to CCS and ask why you do not
have the examples.

[temtronic]

As Mr. T says, contact CCS.
If you have an old version of the compiler, they can send you JUST the examples folder of files, once they confirm your information (compiler version, serial number, date bought, etc.).
What's odd is that you have the driver, and I'd assume, the example program would have been in the same version of compiler package.

[Ttelmah]

The thermistor example was added about 2016. I think mid 5.06xish.
Given it doesn't involve anything too mind blowing, I'd expect CCS to be
OK supplying it if he has an older compiler than this. The actual thermistor.c
driver was added about four versions earlier, so if he had for example
5.062, he could have the driver, but not the example.

[pmuldoon]

I made up my own circuit and it worked very well.
circuit:
3 fets (as switches to steer the cap to the desired resistor)
2 precision resistors
1 capacitor (held in discharge via the CCP pin between readings)
and used the CCP timer

I would alternately charge the cap with the low resistor, thermistor and hi resistor, timing the charge for each.
From the low and high result I could calculate the counts per ohm.
From that I could calculate the resistance of the thermistor.
(since a thermistor is just a resistor and I can only measure it's resistance anyway.)
And from that I used a lookup table to interpolate the temperature.

It's actually not as complicated as it sounds, and was a lot of fun to make. Much more fun than using somebody else's code.

It has the advantage of only needing precision resistors. The osc, cap and Vref only need to be stable for the time it takes to take all 3 readings.

And bigger caps give you a bigger count.

I considered using the ADC and a precision res/div but I didn't think the adc accuracy over 10-bits would be very precise. The 12-bit adc's seem to come with a lot of errata.

Of course I had the luxury of free PIC pins and enough spare time to keep everything quite during the readings.

[temtronic]

PM's way is similar to a very early (20 year ago ?) appnote from Microchip, back in the good old dayze when PICs had quartz windows on them....

[pmuldoon]

Hmm...I did read a lot of app notes back then.
It seems like every idea I have is probably a spinoff of something I've read or seen and then forgotten about.

Thanks, T.
At least my subconscious is kind enough to let me forget and think it was my idea.

Did I ever mention the interesting relationship I noticed between Voltage, Current and Resistance?

[Ttelmah]

Remember that with a thermistor, if you use more than a very little
current to measure it, you will get 'self heating' from the current, and the
resistance will change. It's a 'classic' issue working with them...