AVR Freaks

If my memory serves me well, the first conversion should be discarded if you want some accuracy.

Guillem.

You can still use the Excel method to generqate your table. Given values are VRef voltage, AVcc voltage (which drives the bias resistor), and the bias resistor ohms. Then I make two columns from the thermistor datasheet, entering the temperature & resistance at that temperature. (For the linear "piece" method, all values need not be entered--every 5 degrees or so will suffice.)

From the two resistance values and the AVcc value, a column is calculated for "voltage at pin". From that value and Aref and number of ADC counts (is that 1023 or 1024?), the expected A/D counts are calculated in another column.

The only time-consuming part is entering the values from the datasheet. If you are lucky you can cut-paste from the datasheet.

After that, you can make the graph and vary your choice of bias resistor value to shift the curve around. this is a very useful (and important!) step, so that good resolution is obtained at both high and low ends.

Lee

It's an NTC thermistor right? N= negative, T= temperature, C= coefficient. It's resistance goes DOWN as temperature goes UP. Pretty standard stuff. Since you are using a look up table why would you care which direction the values go?

I posted a response to a very similar question on an AVR list:

Basically one has a lookup table like
flash unsigned int ntclookup[69]=
{774,765,755,745,735,724,714,703,692,
682,671,659,648,637,626,614,603,592,580,
569,557,546,535,523,512,501,490,479,468,
457,446,436,425,415,405,395,385,375,365,
356,347,338,329,320,311,303,295,287,279,
271,264,256,249,242,235,229,222,216,210,
204,198,193,187,182,177,171,167,162,157};

And use it like
ntc2data=read_adc(2);
roomtemp=0;
while ((ntc2data < ntclookup[roomtemp])&&(roomtemp<70)) roomtemp++;

Epcos have a tool on their site for calculating all the resistance values at different temperatures.
http://www.epcos.com/web/generator/Web/Sections/DesignTools/NTCThermistors/Page,locale=en.html

The connection is from the NTC to ground and an ADC pin and a resistor from the ADC pin to V+ .The value of the resistor would be around the NTC resistance at
the average operating temperature or the middle of the desired temperature range. To have the ADC reading go up as the temperature goes up just swap the positions of the NTC and the resistor.

The Epcos tool puts the resistance values in an MSAccess database.
You can get the values for the lookup table with an Access query like

SELECT Result.t, Result.r_soll, Int(([r_soll]*1024)/([r_soll]+10000)+0.5) AS
adcval INTO adcvals
FROM Result
ORDER BY Result.t;

Replace 10000 by the resistance value connected ADC/V+

For ratiometric conversions as you desire, use AVcc both to excite the sensors and as Vref.

If using internal reference, it probably can't drive ans many mA as your rig is consuming. You could buffer it with an op amp.

Another thing you can do is to switch periodically to AVcc as Vref and do a conversion on the internal reference and then adjust all readings accordingly.

Shortest answer: use AVcc for excitation & Vref.

Lee

Yes, a bodily orifice can be a good test point. ;)

Lee

I second the ice bath suggestion. It's pretty precise.

Accuracy and precision are nice, but why would you need sub-one-degree precision and accuracy in a heating system?

Lee

Rolling boil at standard pressure is 100 deg C. No one mentioned that yet.

Most electric kettles have a switch off (bimetallic strip) that will have some largeish inaccuracy (but perfectly suitable for making tea or coffee). If you put an uncovered saucepan/pot of water on heat and bring it to the boil and keep it boiling gently (a rolling boil) it will be 100 C if this is done with distilled water at sea level and 1013.25 hPa (I am pedantic also :lol: )

[Edit] If you do this as well as the ice slurry, you will have a "two point" calibration. using your mouth will add another point :roll:

Lee