List of FEX Readout Channels to Compare ----------------------------------------- Initial Rev. 12-Sept-2018 Current Rev. 14-Sept-2018 There are differences among the 144 MGT Links that carry FEX card data to the Hub Module. These differences include: - Difference in trace length along the backplane - Difference caused by Hub MGT type: GTX vs GTY - Differenced caused by shortest vs longest traces on the Hub between the Zone 2 Backplane Connector and the Input pins on the Fanout chip. - Difference caused by the channel trace routing that sources the data on the HTM Card - Difference caused by the HTM and Backplane trace routing to Hub Slot #1 vs Hub Slot #2 The intent of this file is to identify FEX Readout channel pairs to compare with each other that from identified path differences have a maximum chance of showing differences in their IBERT test results. Note that these FEX Readout channel pairs have been selected to have signal paths with limited identified differences while the rest of their signal paths are as much alike as possible. The following table from Philippe is a compact summary that for each of the 72 FEX Readout Links associates: MGT Quad, MGT X Y, FEX Card Number and Lane > s:<>/Quad_126/MGT_X0Y9/RX:< >: > s:<>/Quad_126/MGT_X0Y8/RX:< >: > s:<>/Quad_125/MGT_X0Y7/RX:< >: > s:<>/Quad_125/MGT_X0Y6/RX:< >: > s:<>/Quad_125/MGT_X0Y5/RX:< >: > s:<>/Quad_125/MGT_X0Y4/RX:< >: > s:<>/Quad_124/MGT_X0Y3/RX:< >: > s:<>/Quad_124/MGT_X0Y2/RX:< >: > s:<>/Quad_128/MGT_X0Y17/RX:<>: > s:<>/Quad_128/MGT_X0Y16/RX:<>: > s:<>/Quad_127/MGT_X0Y15/RX:<>: > s:<>/Quad_127/MGT_X0Y14/RX:<>: > s:<>/Quad_127/MGT_X0Y13/RX:<>: > s:<>/Quad_127/MGT_X0Y12/RX:<>: > s:<>/Quad_126/MGT_X0Y11/RX:<>: > s:<>/Quad_126/MGT_X0Y10/RX:<>: > s:<>/Quad_130/MGT_X0Y25/RX:<>: > s:<>/Quad_130/MGT_X0Y24/RX:<>: > s:<>/Quad_129/MGT_X0Y23/RX:<>: > s:<>/Quad_129/MGT_X0Y22/RX:<>: > s:<>/Quad_129/MGT_X0Y21/RX:<>: > s:<>/Quad_129/MGT_X0Y20/RX:<>: > s:<>/Quad_128/MGT_X0Y19/RX:<>: > s:<>/Quad_128/MGT_X0Y18/RX:<>: > s:<>/Quad_132/MGT_X0Y35/RX:<>: > s:<>/Quad_132/MGT_X0Y34/RX:<>: > s:<>/Quad_132/MGT_X0Y33/RX:<>: > s:<>/Quad_132/MGT_X0Y32/RX:<>: > s:<>/Quad_131/MGT_X0Y29/RX:<>: > s:<>/Quad_131/MGT_X0Y28/RX:<>: > s:<>/Quad_130/MGT_X0Y27/RX:<>: > s:<>/Quad_130/MGT_X0Y26/RX:<>: > s:<>/Quad_231/MGT_X0Y30/RX:<>: > s:<>/Quad_231/MGT_X0Y31/RX:<>: > s:<>/Quad_131/MGT_X0Y31/RX:<>: > s:<>/Quad_131/MGT_X0Y30/RX:<>: > s:<>/Quad_133/MGT_X0Y37/RX:<>: > s:<>/Quad_133/MGT_X0Y36/RX:<>: > s:<>/Quad_232/MGT_X0Y32/RX:<>: > s:<>/Quad_232/MGT_X0Y33/RX:<>: > s:<>/Quad_232/MGT_X0Y34/RX:<>: > s:<>/Quad_232/MGT_X0Y35/RX:<>: > s:<>/Quad_233/MGT_X0Y36/RX:<>: > s:<>/Quad_233/MGT_X0Y37/RX:<>: > s:<>/Quad_233/MGT_X0Y38/RX:<>: > s:<>/Quad_233/MGT_X0Y39/RX:<>: > s:<>/Quad_229/MGT_X0Y22/RX:<>: > s:<>/Quad_229/MGT_X0Y23/RX:<>: > s:<>/Quad_230/MGT_X0Y24/RX:<>: > s:<>/Quad_230/MGT_X0Y25/RX:<>: > s:<>/Quad_230/MGT_X0Y26/RX:<>: > s:<>/Quad_230/MGT_X0Y27/RX:<>: > s:<>/Quad_231/MGT_X0Y28/RX:<>: > s:<>/Quad_231/MGT_X0Y29/RX:<>: > s:<>/Quad_227/MGT_X0Y14/RX:<>: > s:<>/Quad_227/MGT_X0Y15/RX:<>: > s:<>/Quad_228/MGT_X0Y16/RX:<>: > s:<>/Quad_228/MGT_X0Y17/RX:<>: > s:<>/Quad_228/MGT_X0Y18/RX:<>: > s:<>/Quad_228/MGT_X0Y19/RX:<>: > s:<>/Quad_229/MGT_X0Y20/RX:<>: > s:<>/Quad_229/MGT_X0Y21/RX:<>: > s:<>/Quad_225/MGT_X0Y4/RX:< >: > s:<>/Quad_225/MGT_X0Y5/RX:< >: > s:<>/Quad_225/MGT_X0Y6/RX:< >: > s:<>/Quad_225/MGT_X0Y7/RX:< >: > s:<>/Quad_226/MGT_X0Y8/RX:< >: > s:<>/Quad_226/MGT_X0Y9/RX:< >: > s:<>/Quad_226/MGT_X0Y10/RX:<>: > s:<>/Quad_226/MGT_X0Y11/RX:<>: > s:<>/Quad_227/MGT_X0Y12/RX:<>: > s:<>/Quad_227/MGT_X0Y13/RX:<>: Other documents that one needs to consult: ------------------------------------------ - ..../details/hub_0_ab_fpga_mgt_transceiver_usage.txt - ..../details/hub_0_ab_FEX_MGT_fanout_map.txt - ..../details/atca_zone_2_connector_tables.txt - details of the Hub pcb layout - details of the HTM pcb layout Note: The Aurora Lanes index that is used in some documents runs 0:5 where as the Streams index that is used in the Zone 2 Connector document runs 1:6. But these two indexes are intended to match up, i.e. the actual signal on Stream 1 is the Aurora Lane 0 signal. Channels within an MGT Quad run 0:3 Hub MGT Fanout Channel Numbers run 1:74 Hub MGT Fanout Channels 37 & 38 are from the Other Hub Xilinx X Y Index: ------------------- MGT Quad Channel Y Index ------- -------------------------------------------------- 0 0 4 8 12 16 20 24 28 32 36 40 1 1 5 9 13 17 21 25 29 33 37 41 2 2 6 10 14 18 22 26 30 34 38 42 3 3 7 11 15 19 23 27 31 35 39 43 Link Pairs to Compare: ---------------------- Min. vs Max. trace length between the HTM Source and the Input pins on the Hub Fanout chip: Fanout Ch 48 FEX 10 Lane 3 MGT Quad 233 Ch 3 Y 39 GTH Short vs Fanout Ch 65 FEX 13 Lane 2 MGT Quad 225 Ch 0 Y 4 GTH Long Test on Hub #2 Fanout Ch 48 "short run" is shorter in both its backplane trace length and in its Hub routing to the fanout input pins. HTM routing of Streams 3 and 4 to Hub #2 are almost exactly equivalent. Min. vs Max. trace length on the Hub between the backplane connector and the input pins of the fanout chip: Fanout Ch 40 FEX 9 Lane 1 MGT Quad 133 Ch 0 Y 36 GTY Short vs Fanout Ch 25 FEX 7 Lane 0 MGT Quad 132 Ch 3 Y 35 GTY Long Test on Hub #1 Fanout Ch 40 "short run" is 30 mm longer on the backplane (aka one pitch) but 58 mm shorter on the Hub from the backplane connector to the fanout chip input pins. HTM routing of Streams 1 and 2 to Hub #1 are not equivalent. I'm still working to find a better example where the only difference is the trace routing length on the Hub from the backplane connector to the fanout chip input pins. GTH vs GTY for perfectly middle of the road otherwise equivalent links: Fanout Ch 34 FEX 8 Lane 3 MGT Quad 231 Ch 3 Y 31 GTH vs Fanout Ch 35 FEX 8 Lane 4 MGT Quad 131 Ch 3 Y 31 GTY Test on Hub #2 HTM routing of Streams 4 and 5 to Hub #2 are almost exactly equivalent. Hub routing of these two links from FEX 8 are almost equivalent, i.e. same trace length from backplane connector to fanout chip input pins. Crossing through the Other Hub vs NOT Crossing through the Other Hub for perfectly middle of the road otherwise equivalent links: Fanout Ch 42 FEX 9 Lane 3 MGT Quad 232 Ch 1 Y 33 GTH vs Fanout Ch 45 FEX 10 Lane 0 MGT Quad 233 Ch 0 Y 36 GTH Test on Hub #1 HTM routing of Streams 1 and 4 to Hub #1 are almost exactly equivalent. Hub routing of these two links from FEXs 9 & 10 are almost equivalent - Fanout Ch 42 to FEX 9 has a slightly longer backplane to Fanout chip input trace run to compensate for Fanout Ch 45 to FEX 10 being one pitch further away on the backplane. Want to see one of the "Out of the Array" links: Fanout Ch 74 FEX 14 Lane 5 MGT Quad 227 Ch 1 Y 13 GTH Test on Hub #1 The HTM Stream #6 feed to Hub #1 has slightly worse routing than the HTM Stream #6 feed to Hub #2 - so test on Hub #1 - and testing this link on Hub #1 makes this connection from FEX 14 as long as possible and makes it cross through Hub #2. Select an MGT Link with input trace routing that is maximally close to the Ethernet trace routing that is just South of the MGT Fanout and see if the eye pattern for this MGT Link closes down when the Ethernet is running at a high rate vs when there is little or no Ethernet traffic in this Switch. Fanout Ch 1 FEX 3 Lane 0 MGT Quad 126 Ch 1 Y 9 GTY Test on Hub #2 because that is the Hub whose Enet Switch is currently carrying the IPBus traffic.