When I use CASPER vector accumulators, simple_bram_vacc or addr_bram_vacc, I often get junk which is a bunch of values grouped in either pairs or (more often) groups of four with a pattern repeating even though the values themselves don't exactly repeat. As an example, the following 64-bit values are grouped in pairs. I used "hexdump -C" on my data files to produce the following hexadecimal representation.
Address First value of pair Second value of pair -------- ----------------------- ----------------------- 00000000 00 01 1b 41 86 46 2b b4 00 00 00 4a b1 4b 6e 10 00000010 00 01 1b 26 cf 17 b8 a0 00 00 00 4a ae 6c f0 62 00000020 00 01 1b 3b 77 77 75 d4 00 00 00 4a 9d c3 85 ba 00000030 00 01 1b 32 4e 96 b7 90 00 00 00 4a b2 be 81 14 00000040 00 01 1b 27 47 93 01 68 00 00 00 4a b1 89 1c 1a You can see the repeating pattern of large value (on the left) followed by smaller value (on the right). I am fairly sure this is a latency issue, but so far my adjustments have not solved the problem. I was told the sync scheme of the PFB, FFT and accumulator blocks is not what I had thought it was. That is, the accumulators apparently don't need continued sync pulses in order to operate. What do you think the problem is? > I forgot who all asked for this message. > Thus, here it is for all. > > ---------------------------- Original Message ---------------------------- > Subject: Accumulator Problem > From: [email protected] > Date: Tue, May 28, 2013 1:49 pm > To: "Dan Werthimer" <[email protected]> > Cc: "Ron Ferril" <[email protected]> > -------------------------------------------------------------------------- > > The CASPER library I use is the > mlib_devel package I downloaded on > September 27, 2011, which was the > latest stable version on that date. > I don't know the version number. > Attached are my library, the > spectrometer (qspect07at800mha) I > use for my experiment and a test > toy model I was using to discover > the correct use of the accumulator. > These can be sent to any helpful > people. I remember you asked for > them to be sent to you. > > Thank you. > >
