Technical discussion for CERN/MIT pixel R&D

Tuesday 4 Feb 2025, 09:00 → 10:30 America/New_York

Gian Michele Innocenti (Massachusetts Institute of Technology), Ivan Amos Calì (staff@mit.edu), James Kelsey (staff@mit.edu), Jelena Lalic (Massachusetts Institute of Technology), Patricia Smith (MIT Bates Lab for Nuclear Science)

Technical discussions for the ITS3/SVT detector R&D.

ZOOM link

Dropbox paper for round-table discussion

09:00 → 09:15 News

Speaker: Gian Michele Innocenti (Massachusetts Institute of Technology)

- many UROP students are interested in joining the group on the testing/AI part and mechanical developments. Dominic will join us for AI developments in Spring (quantization of the DNN) and will be at CERN during the summer to do some first digital design studies for the AI on CMOS. 

- No need to worry about the funding "crisis". We have enough resources to move forward without trouble.

09:15 → 09:35 Report on mechanics

Speakers: Holger Witte (Massachusetts Institute of Technology), Patricia Smith (MIT Bates Lab for Nuclear Science)

• Tricia gave an update on learning COMSOL 
  • It has applications that make it easier to change parameters in the final solution
  • It is much more intuitive and user-friendly than ANSYS
  • The physics equations it uses are plain to see 
  • It is simple to add boundary conditions
• There is a meeting with Rosario and Domenico Elia tomorrow at 11 am EST to go over the SVT FEA work that Rosario has done
• Action Items:
  • Request to have slides for next week on a brief overview of how COMSOL works (Tricia)
  • Forward the Rosario meeting information to Camelia, Ivan, and Gian Michele (Tricia)

09:35 → 09:55 Report on readout and testing

Speakers: Ivan Amos Calì (staff@mit.edu), Leyre Flores (Massachusetts Institute of Technology)

09:55 → 10:10 Update on sensor design

Speaker: Jelena Lalic (Massachusetts Institute of Technology)

Periphery

CTS optimization in the periphery https://gitlab.cern.ch/mosaix/mosaix_rtl/-/merge_requests/1481 
 -  enabling early clock flow, changed slack targets, added pre_cts opt
 -  this lead to the stable flow 
 - before the changes: biggest problem were significant fluctuations in the timing with small RTL changes; with changed power grid for metal fill impact on timing was 700ps and this was not optimzied
 -  after the changes: timing is stable, even with quite dense power grid
This optimization was targeting setup optimization as this was the biggest issue.

Currently there are a few hold reg2reg violations, a few moe improvements clearly to be added for hold as well.

Metal fill: https://gitlab.cern.ch/mosaix/mosaix_rtl/-/merge_requests/1297 

• new iteration on the latest RTL (timing violations solved + flow improvements and optimizaition)
• currently there 2 100x100u windows with ~24% density - the rest seems fine 
• there is a corner on the per_matrix_dnw macro with gaso, filling is done over this area; if not done in such way there are many violations

Without M2 power grid (required for meeting min density) flow is taking ~5h (at route there is -5ns slack); with changes required for metal fill (flow takes ~8-9h, slack at route is -12ns)

Need 1 more iteration to increase M2 density for 1% . Nt sure if any are to speed up the flow??

Very unstable env since yesterday, flow is often hunging (checking on this with Wojciech)

Tempus - Innovus signoff analysis:  https://gitlab.cern.ch/mosaix/mosaix_rtl/-/merge_requests/1507

Signoff timing in tempus didnt match innovus results (after applying derate);
Innovus signoff now adds the signoff extraction level - result timing matching the analysis in tempus wiht 10-20% diff ; currently doing opt in design finishing and check of violations in tempus are solved


Other small cleanups:
 

• antenna at inputs https://gitlab.cern.ch/mosaix/mosaix_rtl/-/merge_requests/1480  done, to be reviewed
• https://gitlab.cern.ch/mosaix/mosaix_rtl/-/merge_requests/1480,  https://gitlab.cern.ch/mosaix/mosaix_rtl/-/merge_requests/1434, https://gitlab.cern.ch/mosaix/mosaix_rtl/-/merge_requests/1435 (merged; cleanup for DRC at RSU level)

10:10 → 10:30 Report on CMOS AI framework

Speakers: Abraham Holtermann (Massachusetts Institute of Technology), Gian Michele Innocenti (Massachusetts Institute of Technology), Jelena Lalic (Massachusetts Institute of Technology)