Linear fit

[Kossmann]

Hi people.   How i make a VI with Linear Fit running in real-time?

[johnsold]

Icaro Kossmann,

Please be specific about what you mean by "real-time." Are you running LV on one of the NI provided real-time OSes? Are you running on an FPGA? Or does "real-time" simply mean that it keeps up with what a human operator can observe?

How many data points are you trying to fit?

Some of the fitting routines have been improved substantially since LV 8.6. A 1000 point data set can be fit in ~40 us on a recent version of LV.  For 100000 points it takes about 2.5 ms.

Lynn

[Henrik_Volkers]

(in some cases) a linear fit can be done with a FIR filter 🙂

So samplerate (&looptime) , number of points, kind of fit???

FPGAs can do FIR filter very fast 😄

[Kossmann]

So, this is my job: want to continously do a linear fit over acquired data sets that has to be able to keep up with the incoming data. I will acquire temperature in max 40 minutes, and i want to make a linear fit with each new temperature value durig acquisition. In real time.

[Kossmann]

And build equation in real time

[johnsold]

If you are working with temperature over a 40 minute time scale, you should have no problem doing a linear fit as you acquire the data. As I pointed out before fitting the numbers of points you would likely have takes only a small fraction of a second.

Lynn

[Kossmann]

So, i got it :), but my next problem is build and show de equation from the data collected. Exist a VI for this?

[johnsold]

I am not sure about LV 8.6 -  that is pretty old. In the newer versions of LV the Linear Fit VI has slope and intercept outputs. Those are the coefficients needed for the standard equation of a line of the form y = m*x + b.

The fit VI also has an array of points of the values of y for each of the values of x in the input array. This can be used to plot a line of the fit without any extra work.

Lynn