C - log value - thermistor resistance to temperature conversion

[arul]

Hi,

I wrote a thermistor resistance value to temperature conversion program by the combination of PIC18F57K42 & MPLAB x IDE. Inside the conversion, I am using log calculation. I got a value using c compiler, but that value is not same as the calculator value. Here i attached one simple c code. 

Code: [Select]

#include <stdio.h>
#include <math.h>

void main()
{
    float Rth=29498.553;
    float Rr=50000;
    float a;
   
    a = log (Rth/Rr);
   
    printf("a = %f", a);

}
  ////////////////////
a =  - 0.527682     ////after compiling i got that value. But the calculator value is -0.22916929139 

I am also changing data type float to double, but i got only the same value. Please let me know, what is making wrong...

I am using the following formula for conversion: Tkelvin = (beta * room_temp) / (beta + (room_temp * log (Rthermistor / Rroom_temp_res))) . Is it correct or else please suggest other than that one.

Advance in thanks...!

[IanB]

When you do log(x) on a calculator you are getting log₁₀(x). However, when you do log(x) in your C program you are getting log_e(x), which would be ln(x) on a calculator. You need to decide which kind of logarithm you want.

[arul]

Thankyou so much #IanB

[mark03]

Also, if you only need single-precision floating point as shown (the "float" type, not "double"), use the library calls with an "f" suffix, like logf.  That will save a conversion from float to double and back again, and the library call is presumably faster too.

[georges80]

If you only need say 1C resolution and your thermistor is likely only 1% tolerance (or worse), a simple lookup table to convert from ADC reading to degree C is fast and compact (compared to dragging in a floating point library and log functions) especially with an 8 bit uC.

cheers,
george.

[MarkF]

float logf (float f); and float log10f (float f); single precision log functions are NOT defined in the XC8 compilier.

You must use the double precision versions:

   double log (double f);       // natural log
   double log10 (double f);     // base 10 log

[mark03]

float logf (float f); and float log10f (float f); single precision log functions are NOT defined in the XC8 compilier.

You must use the double precision versions:

   double log (double f);       // natural log
   double log10 (double f);     // base 10 log

What?? 
Yet another reason to switch to ARM, or any other architecture really, that doesn't tie you to a single vendor's tools.

[Nusa]

I second the recommendation to use a lookup table, with whatever precision (number of data points) you require for your application. Interpolation between data points gets you a good enough result.

[langwadt]

Also, if you only need single-precision floating point as shown (the "float" type, not "double"), use the library calls with an "f" suffix, like logf.  That will save a conversion from float to double and back again, and the library call is presumably faster too.

on AVR it doesn't matter they are the same, 32bit single precision