Re: [Discuss-gnuradio] Basic Bandwidth math

[Jeffrey Karrels]

Thanks Eric. So my math was right. Yesterday I seemed to be having
issues because I was running out of memory. I since have upgraded
boxes.  Now I think I am down to 1 issue and I believe it is a basic
RF one that everyone is going to jump over me for!

Let me give a picture tutorial of my problem.

1) If I transmit a modulated signal with two additional signals mixed
with -400kHz and 400kHz, I get what I expected. (see image 400.png)

2) If I transmit a modulated signal with two additional signal mixed
with -400kHz and 800kHz, I get the +800kHz showing up with some
additional power down at  -1200kHz. (see image 800)

3) If I transmit... mixed with -400kHz and 1200kHz, I get something at
+ and - 1000kHz.
(see image 1200)

4) If I transmit... mixed with -400kHz and 1600kHz, I get the same as
#3.  (see image 1600).

 I was messing with samples per symbol, but it will not go any higher
than 7 without an error.

Anyone have any suggestions on where to start looking or things to
try? Filters?

Thanks all for your help!
Jeff


On 9/6/07, Eric Blossom <address@hidden> wrote:
> On Thu, Sep 06, 2007 at 11:56:30AM -0400, Jeffrey Karrels wrote:
> > Ok. I am having a world of problems today, the main one being
> > determining if it is my issues causing the headache or the headache
> > causing the issues. ;)
> >
> > I have a basic bandwidth limit that I am hitting and for some reason I
> > cannot do basic math today.
> >
> > I am trying to see how many carriers I can transmit on. I am mixing a
> > single gmsk modulated source with LOs in software and seeing how many
> > times I can do that. A single channel is 250kHz wide and the channels
> > are 400kHz apart center to center. In other words my LOs are 400e3,
> > 800e3, ...
> >
> > How many carriers would you expect that I could transmit on?
> >
> >
> > Thanks
> > Jeff
> > P.S. Sorry in advance, I am nuts today.
>
> In theory, 20.  In reality probably about 16, unless you run out of
> CPU first.
>
> You can get 8Mz of IF bandwidth across the USB.
> This is 8MS/s * 2 * 2 = 32MB/s (16-bit I & Q).
>
> 8e6/4e5 = 20 "bins"  (ignoring roll off)
>
> Allow for roll off because of CIC interpolator in FPGA.  Say that 80%
> is usable (depends on your app, whether you preemph signal etc).
>
> 20 * .8 = 16 bins.
>
> There's a fence post problem here ;)
>
> So, 1 carrier needs 250kHz
>
> 3 carriers needs 2 * 400 + 2 * 125
>
> 5 carriers needs 4 * 400 + 2 * 125
>
> N carriers needs (N-1) * 400 + 250 (for odd N)
>
>   15 -> 5.85MHz
>   17 -> 6.65MHz
>   19 -> 7.45MHz
>
> Put the middle carrier at 0 Hz,
> then arrange the others symmetrically about it.
>
> Eric
>

images.zip
Description: Zip archive