I have a little 8052 based data aquisition system whose sole role is to gather data from a micro-pressure transducer, and stuff the data to a PC at > 500Hz. That bit works.

The data stream consists effectively of a slightly noisy, sawtooth waveform, with a slow attack and a fast decay. My analysis problem is I need to extract the mean of the peaks of the signal, over some adjustable period, and report it.

I have been writing Delphi code to look at raw data and plotting the data, the first difference and the second difference. There is a pretty clear change in 1st diff and second diff which is encouraging, but I am wondering if there are any other good tricks to use to get better results ?

Here is a zoomed in image, so the trend in the raw data is harder to see, but you can see 1st and 2nd diff better