1 / 14

CMOS Status

CMOS Status. Tony Affolder Short Introduction WG1 Status/Plans Module Geometry . UK playing a significant role: 4 of 6 chairs, 8 of 20 members. Working Group Tasks. WG1: Technical Challenges and Issues Define program to prove radiation tolerance and producability (yield)

moya
Télécharger la présentation

CMOS Status

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CMOS Status Tony Affolder Short Introduction WG1 Status/Plans Module Geometry

  2. UK playing a significant role: 4 of 6 chairs, 8 of 20 members

  3. Working Group Tasks • WG1: Technical Challenges and Issues • Define program to prove radiation tolerance and producability(yield) • Sets how these can be used in a module • First gross design CMOS reticle • Define changes to ABC130 needed • WG2: Implementation Issues: • Module and mechanical systems definition • Simulation of layout • Performance enhancement from lower material and better resolution • WG3: Composition, Logistics, Funding and Scheduling of the R&D effort

  4. Summary of existing prototypes Monolithic detector – continuous readout with time measurement HVPixel– CMOS in-pixel electronics with hit detection Binary RO Pixel size 55x55μm Noise 60e MIP seed pixel signal 1800 e Time resolution <100ns Monolithic detector - frame readout Capacitive coupled hybrid detector • HVPixelMchip - frame mode readout • Pixel size 21x21μm • 4 PMOS pixel electronics • 128 on-chip ADCs • Noise: 21e (lab) - 44e (test beam) • MIP signal - cluster: 2000e/seed: 1200e • Test beam: Detection efficiency >98% • Seed Pixel SNR ~ 27 • Cluster signal/seed pixel noise ~ 47 • Spatial resolution ~ 3 m CCPD1 - capacitive coupled pixel detector Pixel size 55x55μm Noise 70e Time resolution <100ns MIP SNR 25 CCPD2 (CAPPIX) - capacitive coupled pixel detector Pixel size 50x50μm Noise 30-40e Time resolution <300ns MIP SNR 45-60 My comment: Potential has been shown to merit further investigation. No one device could show needed performance: radiation tolerance and 1 bunch crossing timing Irradiations of test pixels 60MRad – MIP SNR 22 at 10C (CCPD1) 1015neq MIP SNR 50 at 10C (CCPD2) Technology 350nm HV – substrate 20 cm uniform

  5. https://indico.cern.ch/event/298458/ CMOS reticle Architecture • Further consideration of ABCn modifications suggests that: • Digital architecture adopted as baseline • No longer strips really Hit location encoded in ~13 bits. “Strip” segment ~37mm x 800mm Most hits single, so ~11mm x 230mm resolution Most likely to be made up of 8 ~37mm x 100mm pixels OR’d together, so shape in EC can be different Individual linesto periphery 1mm pitch Periphery encode non-zero words presented to transfer buffer to ABC130 like object Simple pixel: threshold withencoding in boundary ~13 wire bonds peroutput buffer 320Mbit/s gives 8 hits/BC • At 320Mbit/s 24 hits would need ~50 wire bonds (cf 256) per ABCn and • Only half as many ABCs as for stereo options - ≈1/10th number of bonds in total

  6. In the ABCn • Analogue section removed • Same pipeline structure (synchronous) used for storage, but words • Limit is 256bits/BC ≈ 256/13 ≈ 19 hits - Is this enough for jet core? 2x2cm • [would be 38 if two ABC used for reticule] • DCL section revised • Digital RX replaces analogue section • Power Consumption • 1W/cm^2 is upper limit (driven by time walk requirement) • Clever design should reduce this by a factor of two or three

  7. Reticle Geometry • One question is efficiency of CMOS • Can start with inactive areas which are not in overlaps • Space between reticles • Inactive edges of reticles next to edge • I have no idea what these can be • Estimates given where roughly ~80-120 mm at edge of reticles and ~80-120 mm between reticles • NEED TO FIRM UP THIS ESTIMATE • From our presentation at the CMOS TF meeting, Tim Jones came up with the following sketch of reticle Active Area 512 pixels wide (~37.5 um each) x 32 pixels long (~800 um each) From Tim Jones

  8. Module Design • Layout of module critically depends on CMOS yields • If too low, will have to build out of single reticles • If high enough, can cut multi-reticle blocks • CMOS made on 8” wafers • Really inefficient for 10 cm width • Prefers 8 cm width greatly • Would have same preference if go to 8” planar wafers • So if yield high, would end up with modules 1x4 reticles wide ( ~24 x76 mm) • Otherwise, module can be made up of 24x 19 mm sections, but would need a carry material to make a module 1x5 reticle sections 1x4 reticle sections

  9. Module Architectures • Not pursuing 2d stitching • Not possible at most foundries • Could have huge cost implications (cost per wafer and yield) • With 2d info in CMOS sensor, benefit of stitchingminimized • We assuming base module element will be a 4-5 reticle wide x 1 reticle long object • Might make sense to build modules out of two of these with peripheries pointing out • Alternate modules on opposite sides of support material with overlapping active areas Stave with alternating 48 mm modules

  10. “Low” Yield Option (I) • If yield is low (<90-95%), cost benefits will hard to achieve unless modules assembled out of single good reticles • Width of module independent of CMOS wafer size • Lowest mass method to do this is to use hybrid to tie reticles together • Required spacing between die and its precision isn’t obvious • In discussions with Tim, it may be more difficult to cool middle dies • No lateral cooling between dies (in silicon) • Cooling would have to be through face sheets to pipes

  11. “Low” Yield Option (II) • A higher mass method would be to have a carrier with good thermal properties (CF, ceramics, TPG,..) which reticles are attached • Or you could extend hybrid flex to provide in-between option

  12. “Higher” Yield Option • If yield is higher (>90-95%), could dice reticles out into bigger sections. • Width of module would be set by best use of 8” CMOS wafer size • Have benefits for handling and thermal management • Hybrids could be done as now • If this option chosen, we will ask foundries if dead areas between reticles diced together can be minimized • For AMS and Lfoundry , it is ~80 microns dicing street with similar length between the last active element and the dicing street

  13. Radiation Length • Current radiation length of a stave • Cores (including tapes): 0.72% • 2 Planar Modules: 1.08% • Total: 1.8% • For CMOS assume no additional material in carrier • Drop-in Stereo and Z-Encoded (100 micron thick sensors): • 1 Drop-in Module: 0.31% • Total: 1.03% It is possible that core can be reduced with CMOS sensors Most of material improvement made with going to single modules with 2d information

  14. Open Questions to Answer • Requirements before proceeding • Is CMOS radiation tolerant? • Can large areas be made? What is the yield, etc.?? • Is both coordinates from one layer OK – otherwise CMOS is not cheaper • Are non-pointing end-region strips OK – otherwise CMOS is not cheaper • Details for implementation: • How much of a benefit is the reduced material (how thin to go?) • How will the material behave mechanically? • Will it need cooling, i.e. min operating temperature? What is its power? • Maximum number of hits in a 2 x 2cm region – pileup and physics • There is ‘no’ charge sharing, simple track hit rates is enough • 1% current design occupancy -> 3 hits only but for jet cores? 3+ year program of R&D needed to show CMOS is a viable technology has been outlined. Many potential breakpoints were program will not work for ATLAS.

More Related