this will save you memory, but i don't think you are tight on memory. i don't think it will save you much of other stuff....
dan On Fri, Nov 29, 2013 at 6:26 PM, Andres Alvear <[email protected]>wrote: > thank you Dan, I'm going to check it out and i'll do that tomorrow. Thanks > again! > > Andres > > > 2013/11/29 Dan Werthimer <[email protected]> > >> >> Hi Andres, >> >> Below is link to software and documentation from Hong Chen >> on how to unscramble the FFT output in software >> (saving resources in the FPGA by removing the FFT unscramble block): >> >> >> >> Hi Ashish, cc Dan, >> >> May I ask which unscramble block(s) would you like to remove? I'v only >> tried removing the fft_unscrambler block and doing the unscrambling in >> software. And the fft blocks in current Casper library has an option >> "Unscramble Output" for adding/removing that block. As for the other >> unscramble blocks such as bi_real_unscr_4x, I've never try that. >> >> What I did with doing unscrambling in software was basically do what the >> unscrambler block would do to the output data streams - square transpose >> and then reorder. I scratchsome quick >> notes<https://docs.google.com/document/d/13TU2CqZM3NziDy4LkQwGHzmg3azJlARHUGoPtQA80SU/edit?usp=sharing> >> on >> this (google doc, page 5). A memo and some code I wrote back in Summer 2010 >> were linked at the end of the doc as well. I would attach a test model file >> but it seems it doesn't open with current casper library (though being very >> simple) and I'll make another one very soon. >> >> Please let me know if there's any question. >> >> Cheers, >> Hong >> >> >> On Fri, Nov 29, 2013 at 12:00 PM, Andres Alvear <[email protected] >> > wrote: >> >>> hello everyone, >>> >>> I'm working in a Speed Optimization in a couple of spectrometers >>> implemented in the ROACH 1 with the FPGA Virtex-5 xc5vsx95tff1136-1, with 2 >>> ADC083000 boards. These both spectrometers have a bandwidth of 500 MHz and >>> 2048 channels each. I'm absolutely desperate to save as much RAM and logic >>> as possible so I just want to leave the unscramble option unchecked, >>> because with the actual design my compilations have an utilization of >>> slices, slice registers and and LUT flip flops are extremely high - 90, 95, >>> 99%. So i'm trying to cut my design down. Perhaps cut the number of >>> channels from 2K to 1K, but is kind of not good lose spectral resolution. >>> So another way is like Ashish Soni's idea: >>> >>> http://www.mail-archive.com/[email protected]/msg04651.html >>> >>> So following this idea i'm trying to implement FFT unscrambler in >>> software for FFT wideband real, I would like to know if there is a python >>> script or the way to make a software unscrambler to save as much recourses >>> as possible in my designs? >>> >>> Cheers! >>> >>> Andres Alvear Cabezon >>> >> >> >
