FPGA Modeling in MATLAB/Octave

I’m looking to model the rx_chain of the USRP using MATLAB’s fixed
point toolbox and I’m wondering if anyone has tried this yet in MATLAB
or Octave.

Thanks,
Kyle

You mean as in simulink (matlab only)? Or as a standalone (.m file)
matlab/octave program?

Is there a fixed point rough equivalent for Octave?

Bob

ARRL SDR Working Group Chair
Member: ARRL, AMSAT, AMSAT-DL, TAPR, Packrats,
NJQRP, QRP ARCI, QCWA, FRC.
“Trample the slow … Hurdle the dead”

On Wed, Oct 01, 2008 at 02:53:40PM -0400, Kyle Pearson wrote:

I’m looking to model the rx_chain of the USRP using MATLAB’s fixed
point toolbox and I’m wondering if anyone has tried this yet in MATLAB
or Octave.

Thanks,
Kyle

Excuse me if this is totally off the wall, but since if you run the
verilog under a simulator (e.g., icarus), it will generate traces that
give you all the state, on every clock. What’s to model?

Eric

On Wed, Oct 1, 2008 at 3:43 PM, Eric B. [email protected] wrote:

Excuse me if this is totally off the wall, but since if you run the
verilog under a simulator (e.g., icarus), it will generate traces that
give you all the state, on every clock. What’s to model?

Verilog simulators can take a very long time to run for large data
sets. It might also be helpful in making sure signals sent down the
chain don’t overflow or clip within the USRP.

Lucky for Kyle, there isn’t much going on within the USRP. The
upconverters and downconverters are pretty standard and should be
easily modeled if there aren’t models already in Octave.

Octave does have a fixed point toolbox. A quick google search revealed:

http://wiki.octave.org/wiki.pl?CategoryFixedPoint

A simple m-file should be able to accomplish the modeling.

Brian

On Wed, Oct 1, 2008 at 3:43 PM, Eric B. [email protected] wrote:

give you all the state, on every clock. What’s to model?

Eric

I want to model the algorithms in MATLAB to see if the performance of
the USRP can be improved via increased bitwidths, more cordic stages,
changes to the filters, etc. Since MATLAB code is easier to change and
simulate (for me at least) and the timing doesn’t matter, simulating
with MATLAB seemed like the better way to model.

I am open to any suggestions you have on how to go about this.

Thanks,
Kyle