LabView 7.x and FFTW (DLL)

[Volvo JazzHouse]

Has anyone tried to use FFTW DLL in LabView by calling node function?

If any, would you like to share the basic VI wih me?

Thanks

[KileenC]

I was not able to find any VIs that used the FFTW library in LabVIEW. However, in looking at the FFTW API, you will probably want to write some sort of wrapper code that converts some of the special datatypes that the FFTW code uses to the datatypes that LabVIEW will recognize.

On a side note, is there any particular reason you are trying to use the FFTW dll in LabVIEW rather than using the built-in FFT VIs available in the LabVIEW development environment? The LabVIEW FFT VIs are pretty extensive and should give you all the functionality that you need. If there is a particular function in the FFTW package that is not available in the LabVIEW VIs, then please do let us know either by responding to this post and/or entering it as a Product Suggestion.

Thanks!

Kileen C.
NI

[mseb]

Well, this question is not that young, nevertheless, It remains unanswered 😉

I'd like to do that (use fftw3 shared library to compute FFT's in LV), can anybody help me here ? I really have no idea of how to do it, exepted that I need the DLL and I must find my way using the Call Libray Function Node.
My motivations for this are that I don't know HOW NI libraries do the computation and that I can't find the documentation for the functions involved in computing FFT's
.

Any help would be gratly apreciated.

[Jimmie_A.]

Hello!

Here is a general tutorial:

The Fundamentals of FFT-Based Signal Analysis and Measurement in LabVIEW and LabWindows/CVI

http://zone.ni.com/devzone/conceptd.nsf/webmain/c045a890751303a6862568650061ea98

From the LabVIEW 8 Help:

FFT (Not in Base Package)

http://zone.ni.com/reference/en-XX/help/371361A-01/lvanls/fft/

FFT Fundamentals

http://zone.ni.com/reference/en-XX/help/371361A-01/lvanlsconcepts/fft_fundamentals/

[mseb]

Well, maybe I wasn't clear enough 😞

I _may_ need to have a look at the algorithm which is implemented for computing the FFT. This is why I'd like to use a function whose source code is available. I do not - for now - need help on what is FFT or how to use it, I need help on trying to interface my LV code with a DLL which implement FFT (FFTW3). I'm a little short on time, did your answer mean that LV functions are FFTW-based ?

[Charlie S.]

Hello mseb,

Unfortunately, we do not have any built-in VI's that utilize FFTW (as indicated above). If you would like to call a DLL that does implement such an FFT, there are several good resources on our website to help you get started:

Configuring the Call Library Function Node

Writing Win32 Dynamic Link Libraries (DLLs) and Calling Them from LabVIEW

Using External Code in LabVIEW

You can also submit a product suggestion (see the link above).

Thanks!

[mseb]

Thank you for this unambiguous answer.
Should I dig any further this way, I would keep you informed.

[pwebster]

Greetings,

I'm also very interested in the appropriate wrapper code to run the FFTW .dll . I'm mostly interested in its superior performance. If you look at the benchmarks on their site, you can see that they have significant advantages over many other implimentations.

Additionally, does anyone know of an efficient interpolation algorithm to convert unevenly sampled points in frequency to constantly spaced points for an FFT? I know that LabVIEW contains an unevenly sampled FFT VI, but it is painfully slow.

Best regards,

Paul Webster
Department of Physics
Queen's University

[Brian_K.]

Hi pwebster,

I couldn't find any example code where the FFTW.DLL was used.  If it is a standard DLL, then the DLL example in the LabVIEW example finder should be able to help.

I looked over their benchmarking notes, and I was wondering what kind of performance you were getting using the built-in FFT functions?

[pwebster]

Hi Brian,

The FFTW .dll needs three functions to be callled from it for a successful FFT calculation. The problem with using it is that custom datatypes need to be fed from one function call to another. This is not supported by LabVIEW from what I can see. FFTW has been accepted into the Intel Math Kernel because of its performance. It also offers special optimizations for special cases of FFTs (all real input etc.) and their "wisdom" function that allows the algorithm to optimize itself based on system hardware and past executions.

I'll have to get back to you on the benchmark for the FFT. I'm currently rewriting my code to execute multiple FFTs in parallel.

Best regards,

Paul Webster