The PSEC-3 & PSEC-4 ASICs 5-15 GSa/s waveform sampling/digitizing ICs Eric Oberla 20-May-2011 LAPPD Electronics + Integration GPC Review
Outline PSEC series architecture review PSEC testing results – Submitted 11-Aug-2010, 40 parts received 16-Dec-2010 PSEC design & specifications – Submitted 9-May to MOSIS prototyping run Plans LAPPD Electronics + Integration GPC review5/20/20112
LAPPD Electronics + Integration GPC review3 PSEC series architecture Waveform sampling using Switched Capacitor Array (SCA) On-chip digitization up to 12 bits Serial data readout Region of interest readout capability Self-triggering option Designed to handle fast pulses (MCPs) Sampling rate capability > 10GSa/s Analog bandwidth > 1 GHz (challenge!) Relatively short buffer size Medium event-rate capability (~100 KHz) IBM 130 nm CMOS process Primary front-end chip for LAPPD MCPs
5/20/2011LAPPD Electronics + Integration GPC review4 PSEC-3 Design targeted to fix issues with PSEC-2 Still a prototyping version 6 channels (2 test + 4 regular) Test structures on chip Testing underway December, mm 4.0mm
5/20/2011LAPPD Electronics + Integration GPC review5 PSEC-3 specifications + results SPECIFICATIONACTUAL Sampling Rate500 MS/s-17GS/s2.5 GSa/s-17GS/s # Channels44 Sampling Depth256 cells256 Cells Sampling Window256*(Sampling Rate) -1 Input Noise1 mV RMS1-1.5 mV RMS Analog Bandwidth1.5 GHzAverage 600 MHz ADC conversionUp to 12 2GHzUp to ~10 2GHz Latency2 µs (min) – 16 µs (max)3 µs (min) – 30 µs (max) Internal Triggeryes
5/20/2011LAPPD Electronics + Integration GPC review6 PSEC-3 die photo Open-cavity ceramic package
PSEC-3 Evaluation Card (revA & B) 5/20/20117LAPPD Electronics + Integration GPC review
PSEC-3 Evaluation Card (revA & B) PSEC-3 USB 2.0 FPGA: Altera Cyclone III -Firmware developed at UC -USB readout software joint UH/UC effort 5/20/20118LAPPD Electronics + Integration GPC review
5/20/2011LAPPD Electronics + Integration GPC review9 PSEC-3 Sampling Rate Sampling rates adjustable 2.5 – 17 GSa/s Currently running at 10 GS/s, sampling lock with on-chip Delay-Locked Loop (DLL) Good agreement with data + post-layout simulation
Custom delay locked loop (DLL) design Works great, except for some issues of de-locking when ADC is running – Attributed to drop in rail voltage/digital noise ---> issues addressed in PSEC-4 Video: 5/20/2011LAPPD Electronics + Integration GPC review10 PSEC-3 Sample Lock
5/20/2011LAPPD Electronics + Integration GPC review11 PSEC-3 ADC performance Wilkinson ADC runs successfully to 2GHz (registers can be clocked to 3GHz) Firmware has ADC running in ~10 bit mode: 700 ns conversion (ramp 1.6 GHz Test structure (counter + ring oscillator) Channel clock fan- out A/D conversion main power consumer in PSEC-3 – ~15 mW per channel (only ON during 700 ns ADC period) A/D conversion main power consumer in PSEC-3 – ~15 mW per channel (only ON during 700 ns ADC period)
PSEC-3 Linearity & Dynamic Range Channel 3 Channel 4 Fine DC scan + fit in linear region to get voltage/count conversion Blue data points are raw data, without correction Very linear in mV range Deviation < 400mV not fully understood (possibly due to comparator/buffer response or transistor Vthresh) 5/20/201112LAPPD Electronics + Integration GPC review
Channel 3 Channel 4 Plot of fit residuals + cubic spline interpolation Create look up table (LUT) to correct for ADC differential non-linearity Currently, a software correction – work with Andrew Wong (UHawaii) In linear region, ΔV < 1 LSB no (very few) missing codes PSEC-3 Linearity & Dynamic Range 5/20/201113LAPPD Electronics + Integration GPC review
5/20/2011LAPPD Electronics + Integration GPC review14 PSEC-3 Noise Input noise ranges ~1-1.5 mV RMS Dominant source of on- chip noise comes from analog buffers Issue addressed in PSEC-4 design NOISE CHANNELS 1-4
5/20/2011LAPPD Electronics + Integration GPC review15 PSEC-3 Noise
5/20/2011LAPPD Electronics + Integration GPC review16 PSEC-3 AC performance 80 MHz sine 10GSa/s (time-base not calibrated) Analog Bandwidth Timing Calibrations – save for Kurtis’ talk
5/20/2011LAPPD Electronics + Integration GPC review17 PSEC-3 Analog Bandwidth Quantify observation of attenuation along chip input line -High series R of line (~160ohm) --- bad! Histogram random phase sine waves – get amplitude along line Example (400MHz): Overlay sine data on 256 cells > From histograms, amplitudes compared for 3 groups of cells:
More examples: (CW from top left 100MHz, 400 MHz, 1.2 GHz, 700MHz) 5/20/201118LAPPD Electronics + Integration GPC review
5/20/2011LAPPD Electronics + Integration GPC review19 PSEC-3 Analog Bandwidth Time-domain reflectrometery (TDR) of PSEC-3 input -> input line is much too capacitive ~11pF! (>> 2pF expected from post-layout extraction) - suspected culprits: coupling to subtrate/top layer fill metal Herve Grabas
5/20/2011LAPPD Electronics + Integration GPC review20 PSEC-3 Analog Bandwidth Bandwidth highly dependent on location along input line: – First 5 cells have -3dB ~1.3 GHz (excluding 800MHz region) – Need to take more data to confirm trend MHz – Reducing input line resistance should extend 1.3 GHz BW to later cells –> PSEC-4 --Careful layout to reduce input capacitance –> PSEC-4
5/20/2011LAPPD Electronics + Integration GPC review21 PSEC-3 summary PSEC-3 = working waveform sampling ASIC Room for improvement: ABW Dynamic flip-flops proven difficult in ADC design Overall readout speed Fixes covered in careful redesign -> PSEC-4 Still to characterize: Timing calibrations -> timing measurements (resolution, jitter, etc.) Temperature dependence Up next: Readout a detector!
5/20/2011LAPPD Electronics + Integration GPC review22 PSEC-4 Design targeted to fix issues with PSEC-3 6 identical channels each 256 samples deep Submitted to MOSIS 9- May parts May get a larger run via CERN MPW if necessary
5/20/2011LAPPD Electronics + Integration GPC review23 PSEC-4 specifications Sampling Rate2.5 GSa/s-17GS/s # Channels6 (or 2) Sampling Depth256 (or 768) points Sampling WindowDepth*(Sampling Rate) -1 Input Noise<1 mV RMS Analog Bandwidth1.5 GHz ADC conversionUp to 12 2GHz Latency2 µs (min) – 16 µs (max) Internal Triggeryes Red = redesigned using what we learned from PSEC-3 - Overall architecture the same: only small, necessary improvements made from PSEC-3 design (= confidence in design)
Comparator dominant noise source in redesign -Analog buffer removed -Replaced with effective buffering input stage to comparator (compact, low noise design) -In addition, noisy ramp analog buffer deemed unnecessary and removed (after much simulation to convince ourselves) Learned much regarding analog design from Eric Delagnes and others at the recent UChicago workshop 5/20/2011LAPPD Electronics + Integration GPC review24 PSEC-4 noise reduction
5/20/2011LAPPD Electronics + Integration GPC review25 PSEC-4 ABW improvement Move input line to top layer metal (R ~ 5 ohms) Factor of 30 reduction in resistance PSEC-3 used MQ PSEC-4 input line in microstrip configuration PSEC-3 uses balanced, coplanar line Change to unbalanced line -> less sensitive to distortion + easier to control impedance + less coupling to substrate! (=less parasitic capacitance) ABW ~1.5 GHz seen by first cells in PSEC-3 should extend to all cells in PSEC-4 IBM 130nm 8RF-DM layer resistivity -----> IBM 130nm 8RF-DM layer resistivity -----> Herve Grabas
5/20/2011LAPPD Electronics + Integration GPC review26 PSEC-4 ADC improvement Change fast, dynamic flip flop + latch configuration in ADC counter with storage d-flip flop Lose encoding capability > 2 GHz, but OK Overall architecture unchanged 12 bit +1 bit overcount with cell-addressed tri-state drivers on each for readout. Simulation: Transfer 2 GHz Storing digital values (~10 us) Readout
5/20/2011LAPPD Electronics + Integration GPC review27 PSEC-3 & PSEC-4 plans Parallel work on both chips PSEC-3: Continue characterizing PSEC-3 – work with Kurtis on time calibrations Timing measurements! Detector readout (later talk for details/opportunities) PSEC-4: Design of evaluation board (possibly using existing revA digital card) Much of PSEC-3 firmware will transfer Characterization (expect chip ~August) PSEC-4 is baseline ASIC for LAPPD readout. PSEC-3 is working 4-channel prototype and will be used extensively in meantime
BACKUP 5/20/2011LAPPD Electronics + Integration GPC review28
Charge pump Phase Compa- rator 5/20/2011LAPPD Electronics + Integration GPC review29 PSEC architecture – timing generation 256 Delay units – starved current inverter chain > Sampling window strobe (8x delay) sent to each channel’s SCA On chip phase comparator + charge pump for sample lock
5/20/2011LAPPD Electronics + Integration GPC review30 PSEC architecture -- sampling
5/20/2011LAPPD Electronics + Integration GPC review31 PSEC architecture – ADC + readout Ramping circuit Clk enable GHz Ring Oscillator Comp. fast 12 bit register 12 bit data bus Read enable Readout shift register/ one-shot: “Token” … 256x … Level from sampling cell
Bandwidth with gain=2 amplifier Comments: On-board amplifier (channel 4) unstable with unity gain – works with gain=2 -3dB BW ~700 MHz for first cells Amplifier = THS4304 5/20/201132LAPPD Electronics + Integration GPC review