From e92456457456cf7d674416add4b1c79f0d2b7cd9 Mon Sep 17 00:00:00 2001 From: Daniel van der Schuur <schuur@astron.nl> Date: Wed, 15 Apr 2015 13:14:07 +0000 Subject: [PATCH] -Added development remarks in comment block. --- .../src/vhdl/apertif_unb1_correlator.vhd | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/applications/apertif/designs/apertif_unb1_correlator/src/vhdl/apertif_unb1_correlator.vhd b/applications/apertif/designs/apertif_unb1_correlator/src/vhdl/apertif_unb1_correlator.vhd index 225c152e9a..41fc741248 100644 --- a/applications/apertif/designs/apertif_unb1_correlator/src/vhdl/apertif_unb1_correlator.vhd +++ b/applications/apertif/designs/apertif_unb1_correlator/src/vhdl/apertif_unb1_correlator.vhd @@ -48,12 +48,15 @@ USE bf_lib.bf_pkg.ALL; -- . The incpoming beamlets (3x 10GbE from fn_beamformer) are distributed among the 8 FPGAs per board; each FPGA correlates -- 1/8th of the beamlets. -- Remarks: +-- . The BG version (g_use_bg=TRUE) functionally works. +-- . The 10G-input version (no BG) needs the DDR3 transpose in fn_beamformer to be in place. +-- . Does the WPFB require SOP/EOP at its inputs? If so, we need to (re)set packet boundaries in the incoming stream of beamlet samples. --FIXME +-- . fn_beamformer now outputs 176/256 beamlets but this might be increased to 192/256 (9.6Gbps) again as its 10G issue is solved. -- . URGENT: -- . We need 9b filter coefficients before we can synthesize! --FIXME --- . How are we going to (re)map fn_beamformer's 176-word beamlet blocks onto the 128-word blocks wanted by the WPFB? --FIXME +-- . Note: the 18b coeffieciens *do* fit and meet timing with only BG (without 10G RX stage). -- . Other: -- . Synthesis incl. mesh terminals went OK but used 96% of the logic resources. --- . We need to add data buffers to monitor the incoming 3 (times 4 bf_units) streams (determine which bits are toggling etc). --FIXME -- . Keep an eye on the FIXME marks. ENTITY apertif_unb1_correlator IS -- GitLab