Does anyone know what kind of FFT algorithm Excel uses in its FFT function
(e.g., is it a Cooley-Tukey FFT, Sande-Tukey FFT, Winograd FFT, etc.)? I
cant seem to find this information anywhere...

xerogel wrote...
Does anyone know what kind of FFT algorithm Excel uses in its FFT
function (e.g., is it a Cooley-Tukey FFT, Sande-Tukey FFT, Winograd
FFT, etc.)? I can’t seem to find this information anywhere...

Usually safe to assume Microsoft uses the simplest, least numerically
robust algorithm.

You should also understand that if the choice of algorithm used
matters to you, you should be rolling your own in VBA, as an XLL, or
using something other than Excel.

Thanks! (I actually assumed that Microsoft would use the simplest
algorithm...) My problem is that my data has a really low standard deviation
when analyzed with Excel 2007, but not so well when I use the fancier
algorithms - frustrating. If I could find out which 'flavor' of FFT they use
it might help to implement it in our software...

"Harlan Grove" wrote:

xerogel wrote...
Does anyone know what kind of FFT algorithm Excel uses in its FFT
function (e.g., is it a Cooley-Tukey FFT, Sande-Tukey FFT, Winograd
FFT, etc.)? I cant seem to find this information anywhere...

Usually safe to assume Microsoft uses the simplest, least numerically
robust algorithm.

You should also understand that if the choice of algorithm used
matters to you, you should be rolling your own in VBA, as an XLL, or
using something other than Excel.

If I could find out which 'flavor' of FFT they use
it might help to implement it in our software...

As a side note, there are slightly different equations used for different
purposes.
If you are getting different results compared to another programs, it may be
based on this.
The equation used by Excel captures the entire signal by "not" dividing by
the size of the data.

It's funny, but Excel's FFT is listed under "Data Analysis", but the
equation is not what one would use from a "Data Analysis" point of view.
Try dividing your complex results by the size of the data for a simple way
to convert it. (Assuming this is what you need of course).
Another common conversion from Excel might be to divide each output by
Sqrt(n)
I've found Excel's FFT to be pretty good. When I want to multiply two large
numbers (say 2000 digits each) via Excel's FFT, then we need to keep the
entire spectrum, and Excel's formula is just fine as is.
Again, it depends on what you need.
Can you make a small sample of data, and list the output of both Excel and
this "Other" program?
--
Dana DeLouis



"xerogel" wrote in message
...
Thanks! (I actually assumed that Microsoft would use the simplest
algorithm...) My problem is that my data has a really low standard
deviation
when analyzed with Excel 2007, but not so well when I use the fancier
algorithms - frustrating. If I could find out which 'flavor' of FFT they
use
it might help to implement it in our software...

"Harlan Grove" wrote:

xerogel wrote...
Does anyone know what kind of FFT algorithm Excel uses in its FFT
function (e.g., is it a Cooley-Tukey FFT, Sande-Tukey FFT, Winograd
FFT, etc.)? I cant seem to find this information anywhere...

Usually safe to assume Microsoft uses the simplest, least numerically
robust algorithm.

You should also understand that if the choice of algorithm used
matters to you, you should be rolling your own in VBA, as an XLL, or
using something other than Excel.