Logarithm

andypugh wrote: The thermistor comp runs in userspace.

Yes, I saw that but still it has no declaration in the header files for log(). The other user space components are quite liberal with standard #include files. Makes me wonder if log() should not be added to rtapi_math.h

I found a code. I will try it before post it. ASAP hope tomorow

I don't know if ther is a link but I found some people who get trouble to compile log in C with linux. They had to compiled the program with the -lm option. And it worked
like : www.unix.com/programming/63585-what-wrong-log-function-c.html
By the way, no clue what does it meens.

For simplicity, reliability and portability I think that "lincurve" is the best answer.
It was designed for exactly this sort of job, with inputs inside a well defined range.
We use the same approach extensively in my day job as it limits the computational overhead. And automatically limits outputs to sane values regardless of inputs.

I will work on lincurve, but i need some result. I use the easy way...
So : if somebody need a natural logarithm function in C, here it is


double log(double x){
int i=1;
double logx = 0 ;
double ty = (x-1)/(x+1) ;
double tty=0;
do
{
logx = logx + ty / i;
tty = ty ;
ty = (ty * ((x-1)/(x+1)) * ((x-1)/(x+1)));
i = i + 2 ;
}
while(tty - ty > 0.0000005 );
logx = 2*logx;
return logx;
}

log(2) will give 0.6314

Its great that you have a result but if you are doing this in real time rather than in user space, I would exercise caution looping in the servo thread as you are blocking every other component that is running on the thread. Maybe make a new debug pin that publishes the number of iterations and observe it for a while to see if it is excessive. You can also see the time taken to execute by the period and fperiod variables in the component so publish one of these to a debug pin. (even better publish the largest values seen)
See linuxcnc.org/docs/html/hal/comp.html#_implicit_parameters

Be careful!!!!

The previous code is not really good. It not work with number smaller than 1
This one is better.

double Ln(double x) {
int negatif = 0;
double fois = 1;
double ajout = 0;
if(x<=0.0) return 0;
if(x<1.0) {
negatif = 1;
x = 1.0/x;
}
while(x >= 10.0) {
x /= 10.0;
ajout += 2.302585092994046;
};

while(x>=1.1) {
x = SquareRoot(x);
fois *= 2;
};
x--;
double savx = x;
double i = 2;
double xp = x*x;
double quotient = (xp/i);
double dl = x-quotient;
while (1.0E-15<quotient) {
i++;
xp *= x;
dl += (xp/i);
i++;
xp *= x;
quotient = (xp/i);
dl -= quotient;
}

dl *= fois;
dl += ajout;
if(negatif) dl = -dl;
return dl;
}


For me I give up!! That working good but using a thermistor ntc 100K to calcul temperature between 200 and 300 °C don't give me enough accuracy. Not at all.

I will try to generate a sqare signal with a micocontroleur and a thermocouple K....