Differential of noisy response curve

[Ellen_Ward]

I am measuring the temperature from an RTD to produce a heating (or cooling) curve. The response is roughly sigmoidal, however it is noisy, and thus differentiating the data in real time does not give me a 'heating rate' of any sensible appearance.

Does anyone know how i could do this in real time rather than resorting to post processing?

One option I tried was to watch the data for the correct type of change, then only use that. I.e. when recording a heating process, the code watches for a rise in temperature. When this happens the new temperature and the current time is stored in an array. This is then compared to the previously recorded value, and the gradient of the increase is calculated and displayed. Of course when r
ecording a cooling process, a decrease is stored instead. This still gives me a very fuzzy differential, however it appears to show a 'real' response (has a maxima/minima and finally settles to 0).

Is there a 'better' method for doing this in real time?

Many thanks

[unclebump]

Do you have the signal processing toolset?? There is a denoise vi that results in a very nice signal from an array of data. Can you post a file with your data and graph in it?? What LABVIEW version are you using??

[JuanCarlos]

Hi,

In general noise will have a higher frequency than your temperature signal. You could take a look at the point-by-point filtering VIs they allow you to filter each point that is generated. You can have a low-pass filterremove some of the unwanted movement.

You can also increate your acquisition rate and perform some sample reduction via averaging. Instead of acquiring 1 point, you acquire 50, calculate the average of this 50 points and report this average as your sameple. This method also helps reducing noise.

I hope this helps, let me know if you have any further questions.

Regads,

Juan Carlos
N.I.

[Ellen_Ward]

Juan,

Thank you for your response.

Unfortunately our server has been down and I've only just read this, but your suggestion is pretty much what I did in the end.

I altered the VI to allow me to control the sampling frequency from the front panel, and tried different values (chosen pretty much at random). I've found several that work very well (the mean value is taken each second, giving me a much smoother responce).

Thank you again.

Ellen