mass test and calibration of sampa...gain calibration was run sigma 0.9% most within +/- 0.2 i.e....
TRANSCRIPT
Mass test and calibration of SAMPA
Lund Team: Jonathan Adolfsson Ulf Mjörnmark Anders Oskarsson David Silvermyr Lennart Österman + GBT/TRORC experts (Torsten Alt) and Experts involved in SAMPA-FEC and MuCh work
Outline: • Experience from first 700 SAMPA MPW2 (manual test) • The robotic test station • Plan
2
700 v2 SAMPAS have been tested in a systematic way by operator • Same equipment as for SAMPA characterization • PCCA test card with socket (from ALICE Muon Chambers) • Readout with ALTERA development board • Testpulsing by remote control function generator (Tektronix AFG3000)
Operation at full speed (not fully TPC-adapted yet): • 320MHz readout, • 10 MHz sampling • 11 Elinks
Large scale experience from early-on testing • Time estimates • Yield indication • Root Software for analysis • Scripts for pulser control • Socket reliability
Overview: Experience with SAMPA testing so far
3
ALTERA development board
Setup for SAMPA characterization and test of 700 SAMPAs
Inside of socket Pitch 0.65 372 spring loaded pins
PCCA socket version to be used in mass testing
SAMPA
SAMPA test board with socket
5
Fab Tests DC measure DC voltages and currents Tests built into SAMPA JTAG test – checks most digital IO Memory test checks all memory cells DFT test checks ~25000 flip flops Pedestals by filters BC2 and BC3 (amplifier faults are giving zero ped.) Data integrity same checks as in final data taking External testpulses pulse every second channel. Bondwire and BGA substrate input. crosstalk Calibration Taking calibration data for each channel ADC test Noise Gain calibration risetime-falltime
Test procedure in final testing
odd-even effect on v2
Problem in the high end
Better in the middle
Worse again at the low end
6
ADC test Testpulse arbitrary time relative to sampling clock
Populate whole ADC range Very large number of pulses
Shall be smooth - not like below
7
Affected by finite ADC resolution and ADC problem Spread partly due to ADC problem - Hope to be better on v3/v4
Noise test
Will be done with TPC det cap (18 pF) and series resistor (100Ω) in final mass test. Can also choose MuCh parameters
Gain calibration example (one channel)
70 fC signal in 1 TimeBin = 100 ns Γ(4)-fit with fixed τ (170 ns)
Amplitude = Max – Baseline used in later slides
8
Testpulser synchronized with sampling clock. Delay adjusted to capture peak max.
Can also be done with Muon amplifier settings
0
10
20 30 40 50
60
70
80 90
100
1000 testpulses per amplitude . Numbers refer to the injected amount of fC.
ADC bit error
2017-08-30 A. Oskarsson, Lund University TPC upgrade meeting 9
Distribution of max values
Calibration results for each chip and for individual channels
Fit par 1 = slope
10
All 32 channels at the same time
Slope parameter all channels, for 198 chips
2017-08-30 A. Oskarsson, Lund University TPC upgrade meeting 11
12
Slope parameters for 603 SAMPAS for which the Gain calibration was run
Sigma 0.9% Most within +/- 0.2 i.e. +/-2.2%
Includes some long term variation
Some systematics underneath to take away
Gain calibration: Fit slope
13
all channels on ca 700 chips over 3 months
+/- 1%
Have to correct for long term effects +/- 1% within reach after correction
Gain calibration: Slope vs chip number
14
pulse
Delay in ns
1 20000
2 23510
3 27020
4 30530
5 34040
6 37550
7 41060
8 44570
9 48080
10 51590
1 event is a pulse train of 10 pulses of same amplitude Triggered by synchronized trigger pulse. 1000 events gives 1000 pulses on each 10ns sample
Steps 10ns longer than multiple of 100ns
Risetime, Falltime
2017-08-30 A. Oskarsson, Lund University TPC upgrade meeting
Can also be done with Muon analog settings
15
Γ(4)-fit
Fake 100MHz sampling by phase shift in 10ns steps
High resolution pulse sampling
2017-08-30 A. Oskarsson, Lund University TPC upgrade meeting 16
Time sweep of pulse train
10MHz sampling, 100 microsecs full sweep Induce opposite charge
IMPORTANT. No manual handling of individual chips Manual tray handling. For SAMPA - full table 756 chips. Reload every 16 hours. 17
Robotic production testing in clean room (ISO7) • automatic mass testing was done for 80000 ALTRO and 80000 PASA • CHIP handling – chip sorting - database procedures will be re-used • New test card and readout
PASA/ALTRO robotic test setup
Robotic testing
2017-08-30 A. Oskarsson, Lund University TPC upgrade meeting 18
• Use socket PCCA which has all signals pulled out to a connector • Read out with VLDB board • VLDB: GBT +SCA + optics as on FEC backend • Datataking identical to FEC. Should take data after very small FW modifications • PCCA needs no modification (just some passive components to leave out) • Connect PCCA and VLDB with an interface card • Fast signals from VLDB to/from PCCA on mHDMI cables – flexible routing • Slow signals and DC: plug in with board to board connector. • Connect to pin headers on PCCA for DC measurements and programmable
configurations by flat cables. • Protoyping of intermediate card cheap and quick.
19
PCCA socket All signals to back contact
VLDB • GBTX + GBT-SCA + Optics • Same as FEC backend • Sends data to T-RORC • Fast signals on mHDMI cables
Intermediate card
• Transfer fast signals GBTx
• ADC for DC measurements
• Configuration from SCA (DC levels for strapping)
• Level translation (a few)
Taking data shall be the same as on FEC Firmware for readout exists Firmware for built-in tests to be done
New readout schematic for robotic testing
Database
Linux DAQ
T-RORC
Comp controlled Function generators Tec. AFG3000 series
ethernet
windows
20
vldb
PCCA
vldb vldb vldb
PCCA PCCA PCCA
Setup
Pneumatic movement for fixed forces
Robot arm
PCCA
socket
Lennart Österman
Tested the chip contacting in automatic handling. Works at 80N force (8 Kg weight)
Wagon sliding on ball bearings on 12mm rods (commercial product)
21
Positioning guide
Need to know chip position better than 100um
Mechanics
Side view:
Top view:
22
Capacity with 4 independent stations (one robot serving) • 5 minutes per chip including movement • 12 chips tested per hour per station • 48 chips in total per hour • 960 chips tested per 20hr day For TPC 20 000 SAMPAS including spares needed Test 30 000 in three lots of 10 000 each Fab plan 2018: Lot 1: 10000 SAMPA to Lund Oct 1 - ship from Lund Oct 29 Lot 2: 10000 SAMPA to Lund Oct 30 - ship from Lund Nov 29 Lot 3: 10000 SAMPA to Lund Nov 30 - ship from Lund Dec 29 This will be doable with margins. 10 days per lot. Pending decision about early start with extra wafers from v3/v4 Hope to run some thousand chips in the spring to settle on acceptance criteria
Planning
23
First version (to be re-discussed after v3/v4 performance is available): Likely to have large scale experience (>1000 chips) before final production • Power consumption: nominal ± 20% (will be split up in power domains) • Peaking time: ± 10% • Input Charge to ADC code calibration ± 2% • Noise < 1 LSB for TPC settings • Pedestals < 100 for TPC settings • Crosstalk < 1%
• Built-in tests passed • Data integrity (checked by real datataking procedures) • No dead channels
• To be defined: ADC test acceptance criteria. We have no direct access to the
ADC input.
Acceptance criteria (TPC)
2017-08-30 A. Oskarsson, Lund University TPC upgrade meeting 24
Some options to consider for SAMPA muon testing
1. Use TPC test as it is but with Muon analog settings. 2. Use your own readout with dual SAMPA (one position socketed) or PCCA. Include Built in tests and DC measurements in your system 3. We make new intermediate card which allows switching between. TPC and Muon DAQ Run DFT and built in tests and DC tests from TPC system. And take data with muon setting. 4. Use your own readout with dual SAMPA (one position socketed) for normal datataking We run DFT and DC tests one day with TPC setup and the next day we run the muon datataking tests tests in your own system. How often to switch we will have to see.
2017-08-30 A. Oskarsson, Lund University TPC upgrade meeting 25
Backup slides
Effective Yield ≈ 505/694 = 72% Need to order and test substantial surplus
26
Score after 700 tested MPW2 SAMPA
judgement # of chips
Fraction comment
Fully OK 252 36%
OK with corrective ref voltage
253 36% Considered good
Dead channel (ped zero) 60 9% Some may be due to ADC problem
Dead channel (no signal) 8 1% Bondwire ?
Memory error 48 7%
DFT failed 48 7%
Sync 9 1% May be readout
Inoperable 16 2% 5 shorted
Not fully digested 6 1%