FFT Coding Challenge

[Greg_McKaskle2]

I just wanted to bring a bit of attention to the current coding
challenge. I understand that the deadline is going to be extended by a
week or so, meaning everyone has more time to try your hand at it.

Unlike the earlier challenges which focused more on programming, this
one is much more of an analysis and mathematical focus. Hopefully this
will allow a different group of people to show their stuff.

Greg McKaskle

[Harold Payson]

It would have been a lot more interesting if LV didn't already do such a
fantastic job of FFTs...

Just my 2 cents,

-- Harold "definitely NOT asking NI to slow down the FFT routines"


In article <3E681C36.3070208@austin.rr.com>,
Greg McKaskle wrote:

> I just wanted to bring a bit of attention to the current coding
> challenge. I understand that the deadline is going to be extended by a
> week or so, meaning everyone has more time to try your hand at it.
>
> Unlike the earlier challenges which focused more on programming, this
> one is much more of an analysis and mathematical focus. Hopefully this
> will allow a different group of people to show their stuff.
>
> Greg McKaskle
>

[Greg_McKaskle2]

> It would have been a lot more interesting if LV didn't already do such a
> fantastic job of FFTs...
>

Thanks for the complement, but the point wasn't to try and beat NI's
FFT. Rather, it was to see what the algorithm looks like as a diagram.
I've seen a couple internal implementations over the years. It is
such a fundamental thing that it also seemed like a good concept to
learn or brush up on.

So even if your implementation isn't faster than NIs, go ahead and
submit it.

Greg McKaskle

[Mikael_Garcia]

Harold,

Good point! Yes, LabVIEW comes with a great library loaded with mathematical functions with good performance.

I have been following the challanges here on DE. It seems that the first challange attracted most users so far. Perhaps because it was the first. However, I am not so sure about that. I think that the type of problem defines the "success". The bit twiddler was easy to understand for the majority of users regardles of your theoretical background. But also, in my opinion, it is more fun to experiment with the effectiveness of shiftregisters and so on than to go and find that old book about Fourier analysis and see if I can find a short-cut for FFTs. It takes longer time and the end result is not as interesting. In the end, most of u
s are here on DE because we code. Not because we are experts on FFTs. Those were my two cents. /Mikael

[Greg_McKaskle2]

> I have been following the challanges here on DE. It seems that the
> first challange attracted most users so far. Perhaps because it was
> the first. However, I am not so sure about that. I think that the type
> of problem defines the "success". The bit twiddler was easy to
> understand for the majority of users regardles of your theoretical
> background. But also, in my opinion, it is more fun to experiment with
> the effectiveness of shiftregisters and so on than to go and find that
> old book about Fourier analysis and see if I can find a short-cut for
> FFTs. It takes longer time and the end result is not as interesting.
> In the end, most of us are here on DE because we code. Not because we
> are experts on FFTs. Those were my two cents. /
Mikael

We will keep that in mind. We just wanted to mix things up a bit.

Greg McKaskle