overlapping square waves

[RGreg]

I'm having a problem resolving a signal that consists of separate overlapped square waves in a single trace - the attached image has two, but there could be more than that.  What I want to do is separate out those waves and determine their periods, and from that I can find the points at which the transitions occur.

I could just brute force it by going through the data, find the transition points, measure the distances between those transitions, and then compare the distances to determine the signals.  I've started writing that code.

But that's a lot of processing time, especially if the wave form is 50k to 100k long each time.  Is there a more elegant way to do this using some built in LV functionality?



Message Edited by RGreg on 03-05-2008 09:40 AM

[ShotSimon]

RGreg,

Have you tried looking at this waveform in the frequency domain vs time?  SS

[Graziano]

Hi!
   ShotSimon gives you a good advice, Frequency-domain analysis can be a good choice. 

   Just note that anyway, you'll need some computation to do this, (but maybe you'll be done with litle coding, since many DSP functions are implemented by NI), and be careful with frequency resolution: how close are these waveforms in frequency? Are they too far one each other? this can be a problem, since sqare wave has odd harmonics, so you'll find significative components at 3x, 5x... the base frequency, and you have not to mistake these components with fundamental.

graziano

[RGreg]

I had considered converting to frequency, but had to reject it once I talked with the R&D guys in more detail.  Apparently the waves are NOT a specific frequency - rather, they are a measurement of specific activation times in a digital circuit. 

But that's okay - we came up with a small circuit to subtract out some of signal so that I can easily see the transition points I have to focus on.  But thanks for the input anyways.