LabVIEW
How to use an FFT with decimal I32
Hi David,
as said several times before: you need to collect your data samples in an array!
I nice and easy way is using Point-By-Point functions:
@GerdW wrote:
as said several times before: you need to collect your data samples in an array!
I nice and easy way is using Point-By-Point functions:
I have mentioned before that even easier would be the use of FFT ptbypt, It manages the collection and transform at once. Just set the sample length to 125.
Why 125? Is that number somehow related to your fundamental signal frequencies to get an integer number of periods? If not, you probably want to use the Windowed FFT PtByPt VI to avoid artifacts and spectral leakage by selecting a suitable windowing function.
How familiar are you with signal processing theory?
@DPJones wrote:
The reason for the 125 samples is that it is the sample rate of the accelerometer
....
I have tried the FFT ptbypt and that works as well but looking ahead I want to try the wavelets and other signal processing ...
...
As for signal processing theory not much.
OK, so you want to operate on 1s worth of history data.
Instead of telling us what function you want to use, tell us what information you want to get out of the signal.
For example to look at the frequency distribution over time, you could just take the magnitude spectrum and chart it on an intensity graph to create a spectrogram.
There is even STFT Spectrogram PtByPt VI and similar functions. There is a rich collection of other ptbypt tools, even for wavelets (example).
(To create a brown waveform wire, you just need to build a waveform by adding timing information, but I don't see why you would want to do that)
Do you mean a human heartbeat, e.g. of the driver?
You are basically talking about vibration analysis. If the car is in idle, the fundamental engine frequency is ~700rpm (~ 12Hz, depending on the car, of course) and there are quite a few other harmonics, e.g. the cam shaft is at half that frequency, so you have all these frequencies that are more than 5-6 Hz. A human heartbeat at rest is slightly above 1Hz (probably with some known higher harmonics that define the non-sinusoidal shape), i.e. in a different frequency range and it should be detectable by proper frequency analysis even if it is many orders of magnitude weaker than the engine vibrations unless there are also car vibrations in that range, e.g. some resonances of a fender or door panel, who knows?
Of course there are many technical limitations, for example to detect such a weak component, you need a sensor with very good resolution (e.g. 24bit) to see these small signals without overflowing the detector by the much larger vibrations of the engine. It is also safe to say that 1s worth of data is NOT sufficient to detect something around 1Hz.
This is all not really my field of expertise and if it were that simple, NI would have no reason to produce specialized tools for sound and vibration analysis.
So overcoming the LabVIEW programming problems is just the tip of the iceberg. You need to solve the instrumental, technical, and theoretical problems. Do you have a good set of test data already?
You can get bins by using "array subset" on the fft transform. You probably want to get the magnitude data because phase is irrelevant. There is a ptbypt for that too.
