FFTW on LabVIEW

[zep]

Hi,

I'm looking for an implementation of FFTW for LabVIEW on Win2000, either as LabVIEW code or DLL already. Does anyone know where I can find it?

Thankx...

[mikeporter]

Just out of curiosity, what are you wanting to do that LV's FFT routines can't do?

Mike...

[zep]

I feel less confident on LV's routines since I ran the attached VI to test them. If you check the VI (LV 6.0), you'll see that when I process an almost perfect signal I get harmonics. They're low, but not as low as I get from MATLAB. Since MATLAB's routines were based on FFTW I was wondering if it would be worthwhile to change the LV's DLLs by some derived from FFTW...

[mikeporter]

Have you discussed this with NI? What was their reaction?

Mike...

[zep]

No, I haven't.

[mikeporter]

How low a noise floor do you need? When I ran your example I was seeing a floor less than E-14 (!).

Mike...

[zep]

Yes, I know that both the noise floor and spurs levels are very, very low. Yet, since I cannot access the LV's routines from a certain level downwards, I cannot check what were the implementation's criteria - maybe some speed vs resolution trade off was assumed.
I'm using the MATLAB routines as my reference, so I have to have results consistent and as good/bad as I get from them. If I get different results from a virtual data, maybe I'll get larger differences on other instances...

[mikeporter]

Ah, I see your concern. The best thing would be to approach NI. They have some very knowledgeable people on staff for their DSP stuff so you should be able to locate someone who can address your concerns--perhaps even the routine's author. I know for a fact that they didn't simply code up something out of "Numeric Recipes" like some instrumentation vendors I could mention.

Mike...