[Discuss-gnuradio] Re: Multiplying blocks

[Johnathan Corgan]

Eric Blossom wrote:
> On Thu, Jun 08, 2006 at 06:01:16PM -0700, Johnathan Corgan wrote:

>   You want a block that takes in two streams: one float, one "vector float",
>   and produces a single output stream: "vector float",
>   where each element of each output vector is multplied by the scaling
>   factor of the corresponding first input.

I think you've got it, but not sure.

I want to start with a stream of floats or complex floats, and process
them in "chunks".  That is what stream_to_vector will do--accumulate a
set number of input stream items into one chunk and output one of these
chunks as a unit (a vector).

Then, I want to element-by-element multiply the first vector by a second
vector of the same length, to product a third vector.

What you wrote above I think describes this as a single block, but I'd
be happy with just an element-by-element multiply between to vectors and
use the stream_to_vector up front to create one of them.

The application here is to multiply an incoming stream of floats by an
analysis window and pass it off for further vector processing, and
ultimately send it through an FFT block.  For example, google WOLA DFT
or check this link:

http://www.epn-online.com/page/15837/efficient-digital-channelisers-for-fpgas-mixed-radix-fft-designs.html

and specifically the diagram here:

http://www.epn-online.com/common/inc/loadimage.php?name=33/10/G-09946C.JPG

> One idea has been to use the "Universal function" (ufunc) idea from
> NumPy and Numeric:

Yes, your example would work nicely for what I think is needed.

> Are you interested in building something along these lines?
> It would be a *very* useful contribution.

I might be interested but every time I try to understand what you're
doing with SWIG to bridge the C++ and Python worlds, I get a serious
migraine (that's over and above the minor headache stemming from
deciphering the autotools) :-)

-Johnathan, AE6HO