1. Readout of DC coupled double sided sensors with CBMXYTER: Some first thoughts Peter Fischer, Heidelberg University

2. P. Fischer, ziti, HD: dc coupled sensor readout2 Goal of this talk  Think about consequences of dc coupling for readout chip  Start a discussion on system issues (control, readout, supplies)  This is a first collection of thoughts – to be improved!  This talk: Sensor model Connecting CBMXYTERs in dc mode Fault tolerance approach Chip connections: Power, Slow/ Fast Control, Data Out Readout options: Some thoughts (also see Walter’s talk!)

3. P. Fischer, ziti, HD: dc coupled sensor readout3 Sensor Model: Segmentation implants bottom side implants top side optional punch through bias Double sided strip sensor subdivide into diodes one diode (vertical)

4. P. Fischer, ziti, HD: dc coupled sensor readout4 N3N3 Sensor Model: Equivalent Circuit (I) P1P1 P2P2 P3P3 N1N1 N2N2 N3N3 N2N2 N1N1 P3P3 P2P2 P1P1

5. P. Fischer, ziti, HD: dc coupled sensor readout5 Sensor Model: Equivalent Circuit (I)  Observation: Input on one side has no ‘partner’ on other side, it’s a mix of all other inputs (across FE-chips!)  Thinking of a differential input, the ‘partner’ is ill defined  Detector Signal is a current pulse in one of the diodes  To be more realistic, must: Add series resistances of implants / aluminum traces Add inter strip capacitors N3N3 N2N2 N1N1 P3P3 P2P2 P1P1 N3N3 N2N2 N1N1 P3P3 P2P2 P1P1 (not all caps on n-side shown)

6. P. Fischer, ziti, HD: dc coupled sensor readout6 Connecting DC coupled chips. (1/N possibilities) more analog digital DAQ more analog digital level shift HV decoupling HV power supply digital supply analog supply Hybrid on detector CavernOutside refer single ended signal to which ground here? feedback & leakage compensation analog supply

7. P. Fischer, ziti, HD: dc coupled sensor readout7 Many questions:  Decouple what w.r. to what?  Connect which grounds? Where? and how (R, L)?  How to connect power supplies? Is it allowed to connect grounds in cavern to save cables?  These issues are not related to dc coupling  Can be tested with NXYTER. Find out how other experiments have done. (This stuff is rarely in papers..)  Be prepared for several iterations  How to do the level shifting? optical fibers – elegant! optical couplers – slow! inductive couplers – slow! ac coupling cascodes – can also be used for analog signals (currents)

8. P. Fischer, ziti, HD: dc coupled sensor readout8 Discussion: Fault tolerance  Question: Is a daisy chain architecture excluded ? Depends on fault tolerance approach  Should we be prepared to operate a module with single non functioning chips? How ‘easy’ is a repair of a bad module (access in fixed target exp. is simple) When a chip dies, does this kill the module anyway ? (for instance by shorts in supplies, noise on sensor) How often does a chip die during operation?  Is fault tolerance ‘nice to have’ or ‘required’? This needs a discussion & decision  Note: a good burn-in may catch most ‘early failures’

9. P. Fischer, ziti, HD: dc coupled sensor readout9 IO Signals required  Supply voltages Connect chip grounds where? Connect power supply ground in counting room (‘loop’) or on hybrid (more cables)?  Slow control best have this independent of readout (→ read back via slow control) avoid bi-directional signal (simplify level shifting). This rules out I2C avoid daisy chains for fault tolerance (?) (this rules out JTAG) Compromise: Unidirectional bus with geographical address: clk, din, dout, load, (reset). Very simple interface.  Fast control (‘TTC’) clock, epoch marker (DLM), timestamp fast reset (can we use DLM?)  Data outputs  Monitoring of analog signals level shifting difficult. Could use cascodes for currents. Best use on-chip ADC and digital readout

10. P. Fischer, ziti, HD: dc coupled sensor readout10 Readout. Preliminaries.  Use as few cables as possible → fill max. bandwidth  Must provide a concept to cope with different occupancies on different chips cabledata ratedistanceConnectorsdriver optical fiber>2.5Gbpsinfiniteclumsy or special Laser (clumsy), serializer coax>2Gbps>10msmallon chip, serializer CAT7 LVDS250Mbps10mclumsyon chip, clk & data Custom LVDS 100Mbpsfew m (?)smallon chip, clk & data

11. P. Fischer, ziti, HD: dc coupled sensor readout11 Optical Serial Data Transmission  Assume we have it somewhere in the path for level shift  Components required: Protocol FSM 8B/10B Serializer LVDS/CML output Laser Driver Laser Diode Connector fiber Laser Package Chip1 Chip2 Mechanics digital analog difficult Provided by SFP package! available available (?)

12. P. Fischer, ziti, HD: dc coupled sensor readout12 Readout 1: Event Builder on Module  GBT will be developed by CERN. May be suitable.  Probably a bit overkill.  Still needs many extra radhard components FE GBT Hybrid on detector Cavern Outside Laser Driver PiNTIA clock, epoch optional FPGA 80 Mbps 4.8 Gbps

13. P. Fischer, ziti, HD: dc coupled sensor readout13 Readout 2: Event Builder in Cavern  Many ‘LVDS’ cables  Number of cables = number of chips FE Hybrid on detector Cavern Outside clock, epoch FPGA ??? 80-250 Mbps 10 m (Walter)

14. P. Fischer, ziti, HD: dc coupled sensor readout14 Readout 3: Share cables  Equalize hit rates between chips with short parallel links  Number of cables limited by data rate.  May be adapted to Module Multiplicity FE Hybrid on detector Cavern Outside clock, epoch FPGA ??? 250 Mbps

15. P. Fischer, ziti, HD: dc coupled sensor readout15 Readout 4: Integrated serializer, optical  Pass all hits to one (or more) ‘master’ chips  Serializers on other chips are disabed  Still Laser Driver, rad hard Laser, Package required FE Hybrid on detector Cavern Outside clock, epoch optional FPGA ??? ser Laser Driver

16. P. Fischer, ziti, HD: dc coupled sensor readout16 Readout 5: Integrated serializer, coax  No extra components needed on Hybrid  Receiver needed (?)  Cost of cables? FE Hybrid on detector Cavern Outside clock, epoch ??? ser FPGA drv. coax cable rec.

17. P. Fischer, ziti, HD: dc coupled sensor readout17 Some conclusions  Sensor model is complicated. See what we really need.  I2C is bad for level shifting  Analog monitoring is difficult  Ground & Power connections will be big issue as usual Should try some basic configurations with NXYTER, if possible  Best readout concept depends on data rates radiation levels distance to shielded cavern (=10m)  Look at Walter’s numbers & suggestions!

18. P. Fischer, ziti, HD: dc coupled sensor readout18 To do  Evaluate speed of custom LVDS driver (HD)  Look for rad hard Laser driver / package / connector (→ CERN)  Try readout chain: Chip → Coax → FPGA  Elaborate on chip → chip protocol  Study minimum implementation of a FPGA compatible serial link (in progress @ HD)