Hi Arindam Thanks for letting us know. I will make this change as soon as I can.
Regards Andrew On Mon, Apr 27, 2015 at 9:43 PM, Arindam Sengupta <[email protected]> wrote: > > > Hi all, > > I encountered this issue when I first started to play around with the > fft_wideband_real from the CASPER library on Xilinx 14.6 > > The default bit-width initialization for this block is set at 18_17. > However, when we specify the 'word length' as 14 and 'binary point' as 13, > in the mask parameters dialog it returns the following error: > > > > > *The following error occured with the > filename/fft_wideband_real/fft_biplex_real_4x/bi_real_unscr_4x/mirror_spectrum/complex_conj0/imag_negate/neg1 > block: Illegal parameterization: Binary point Binary point must be less > than or equal to number of bits, 14. Binary point is currently set to "17".* > > The problem was somewhere between 'fft_wideband_real' and 'imag_negate', > as the error link points. So, I dug down, starting from > 'fft_wideband_real_init', until 'bus_negate_init', by following the > functions invoked within each other. I found out that the variable > 'bin_pt', that carries the value of the binary point we specified in the > mask parameters, *did not* traverse from 'bi_real_unscr_4x_init' to > 'mirror_spectrum_init'. This resulted in 'mirror_spectrum_init' retaining > the default 'bin_pt' value of 17. However, as the 'bit_width' was correctly > transferred as 14, it produced the error above. > > I could get the block running by adding the line *'**bin_pt_in', > num2str(bin_pt), ...* in the 'bi_real_unscr_4x_init' file (anywhere > between lines 495-502), in the current mlib_devel (locally), to pass on the > right value of the user specified 'bin_pt' to the lower layers. > > The reason I bring this up is because I noticed that this line is still > missing in the latest mlib_devel at github, which we are soon going to > adapt on Xilinx 14.7, and that might cause the bug to show up again. We > thought it would be a good idea to put this issue out to the CASPER > community to take a look at, and for the users who currently (might) face a > similar error message. If this fix seems reasonable, maybe the > 'bi_real_unscr_4x_init' file can be updated on github with the line > *'**bin_pt_in', > num2str(bin_pt), ... *? > > Thanks, > Arindam > >
