[USRP-users] Using Xilinx CORDIC 4.0 on USRP N210 FPGA
ianb at ionconcepts.com
Mon Apr 27 12:33:49 EDT 2015
Using a CORDIC is a great approach to a accurate polar<-> rectangular coordinate transformation without needing a hardware square root processor, and in addition to the Xilinx IP, you could also use the Ettus provided CORDIC code to do the same thing. Note that the CORDIC algorithm has a non-unity gain, and the exact gain value to compensate and normalize the results is dependent on how many iterations of the CORDIC algorithm are used. There's also another well known old trick or two to get a good estimate at very low hardware cost…rather than me write it up just take a look at this: http://www.dspguru.com/dsp/tricks/magnitude-estimator
w.r.t your question about bit weighting, its very unusual to see well designed fixed point data path use a representation of +1 >= x >= -1 because it requires an extra sign-bit to represent solely the +1 value. It's common to saturate the +1 to 0.9999 and use the hardware bit you just won back to extend dynamic range and deal with the asymentric bias this operation introduces elsewhere in the system (often for free). Beyond that comment I'm not sure what exactly you are asking. Get in the habit of always thinking of fixed point representation (for DSP) as always having a fractional representation (i.e. a sign bit and n bits of fractional magnitude) and it will make tracking word growth and weighting manageable. This is the so-called "Q Format" which you can google for further information.
On Apr 27, 2015, at 6:12 AM, Patrick DaSilva via USRP-users <usrp-users at lists.ettus.com> wrote:
> On the N210 FPGA, I'm passing the baseband samples through a Xilinx FFT IP core and am looking to take the magnitude of that output. I've found the CORDIC 4.0 core. This core requires the input samples to be between the range of -1 and 1 with 2 bits for the integer portion of the fixed point samples. I was wondering if anyone has had any experience with this and if so what a recommended scale back factor is OR if there is another way to capture the magnitude of the FFT without using the CORDIC core.
> USRP-users mailing list
> USRP-users at lists.ettus.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the USRP-users