I've been streaming the data off my 34401a's rs232 (yes, I know it is not a metrology grade instrument
) into an MCU (stm32) where I've been playing with pre-processing the data (and sending the data via BT in the .csv format out @115k2).
Now, my intention is to put inside the MCU some "useful math", such I can see some useful results without messing with excel. For example, when I hang an LCD display on the MCU I can see some useful data directly there.
While reading the older posts I saw 3roomlab's measurements where he applied stdev, skeweness, kurtosis, trimmed average. I've put some of the calcs inside, while I do moving averages on the raw data.
For example below a shot from TeraTerm with results I get at 10NPLC, from left hand side:
the measurement timestamp since start (ms), Tdmm(C), Tamb(C), raw voltage + stddev, moving average + stddev, 3xMAV multipass + stddev, skew, kurt, lin interp "a and b", and finally the processing time in ms (subtract 60ms for SD ADCs).
The skew, kurt and lin_interp is done upon 4xMAVed data here (is that the right way to do it??).
The MAVs of the voltages are made of 10 samples, the standard deviations are rolling 200 samples, temperature (16b ADCs) are MAV 10 samples, skew and kurt rolling 200 samples, linear interpolation is rolling 500 samples. All calcs are double fp. The number of rolling samples could be set differently, is the matter of free memory. The MCU runs at 48MHz (incl. BT powered off the outguard's 5V regulator).
Any hint what could be actually taken as an useful math in such an case?
@3roomlab: what formula you have actually taken for the kurtosis?