The GLAST LAT Tracker a Large-Scale Application of Solid-State Detectors in Space SLAC AIS October 31, 2007

Size: px

Start display at page:

Download "The GLAST LAT Tracker a Large-Scale Application of Solid-State Detectors in Space SLAC AIS October 31, 2007"

Alan Owen
5 years ago
Views:

1 The GLAST LAT Tracker a Large-Scale Application of Solid-State Detectors in Space SLAC AIS October 31, 2007 Robert P. Johnson Physics Department and Institute for Particle Physics University of California at Santa Cruz

2 Abstract Large-scale detector systems based on silicon-strip sensors have been operated successfully in terrestrial particle physics experiments for many years, but applying the same technology in orbit brings several new challenges. A space-based system must operate on an extraordinarily tight power budget, must cool without the aid of air or liquids, must survive relatively severe mechanical and thermal environments, and must operate reliably without any possibility of maintenance after launch. This seminar will present how these challenges were addressed and met by the LAT Tracker design and will also focus on some of the lessons learned from fabrication problems that resulted from various design choices. October 31, 2007 R.P. Johnson, SLAC AIS 2

3 Outline Pair-Conversion Telescope Concept GLAST LAT Instrument and Tracker Overview Basic Tracker Requirements Measurement Process Silicon-Strip Sensors Readout Electronics Electronics Integration Mechanical Structure and Assembly Tracker Testing Tracker Performance W.B. Atwood et al., Design and initial tests of the Tracker-converter of the Gamma-ray Large Area Space Telescope, Astropart. Phys. (2007), doi: /j.astropartphys (SLAC-PUB-12406) October 31, 2007 R.P. Johnson, SLAC AIS 3

4 Pair-Conversion Telescope gamma Veto counters e- e+ Calorimeter Converters Position-sensitive detectors Tracker/Converter: heavy metal converts the photon to a positron-electron pair. The measured tracks point back to the astronomical source. Calorimeter: measures the photon energy Veto counters: a signal indicates presence of a charge cosmic ray, instead of a photon. Main limitation: the electron and positron scatter in the converter and detector material, limiting the angular resolution to the order of 0.1 to 1 degree! October 31, 2007 R.P. Johnson, SLAC AIS 4

The GLAST LAT Si Tracker 8.8 10 5 channels 160 Watts 16 tungsten layers 36 SSD layers Strip pitch = 228 µm Self triggering γ ACD Segmented scintillator tiles 0.

5 The GLAST LAT Si Tracker channels 160 Watts 16 tungsten layers 36 SSD layers Strip pitch = 228 µm Self triggering γ ACD Segmented scintillator tiles efficiency Minimal self veto Grid (& Thermal Radiators) CsI Calorimeter Hodoscopic array 8.4 X bars cm cosmic-ray rejection shower leakage correction e + e Data acquisition 3000 kg, 650 W (allocation) 1.8 m 1.8 m 1.0 m 74 m 2 of Si in the flight instrument! About $8 per square centimeter. October 31, 2007 R.P. Johnson, SLAC AIS 5

Tracker Assembly Overview Module Structure Components

Italy (Plyform): Sidewalls SSD Procurement, Testing Japan,

Module Assembly and Test Italy (Alenia Spazio) 18 10,368

UCSC, SLAC (Parlex) Electronics Fabrication, burn-in, &

and Bias Circuits Italy (Plyform): fabrication SLAC: CC,

6 Tracker Assembly Overview Module Structure Components SLAC: Ti parts, thermal straps, fasteners. Italy (Plyform): Sidewalls SSD Procurement, Testing Japan, Italy (HPK) SSD Ladder Assembly Italy (G&A, Mipot) Tracker Module Assembly and Test Italy (Alenia Spazio) 18 10,368 Tray Assembly and Test Italy (G&A) 2592 Readout Cables UCSC, SLAC (Parlex) Electronics Fabrication, burn-in, & Test UCSC, SLAC (Teledyne) Composite Panel, Converters, and Bias Circuits Italy (Plyform): fabrication SLAC: CC, bias circuits, thick W, Al cores October 31, 2007 R.P. Johnson, SLAC AIS

7 Basic Tracker Requirements X-section area: sized to fit inside of a Delta-II rocket shroud Height: small relative to width, to maximize the field of view Gamma-ray conversion efficiency: ~60%, requiring ~1.1 radiation lengths (~4 mm) of tungsten foil (in 16 layers) Transparent structure: minimum photon conversion and electron scattering in the support material Thin, efficient detectors: measure the electron/positron trajectory immediately following the gamma-ray conversion No consumables, other than electrical power Low power consumption: < 200 W (160 W actual) Minimal dead areas in the aperture Important to know the location of the inevitable inefficiencies Hi-speed, low-deadtime readout Sensitive to new events during readout of previous events Environmental requirements (launch and on orbit) Cost October 31, 2007 R.P. Johnson, SLAC AIS 7

8 Basic Tracker Requirements X-section area: sized to fit inside of a Delta-II rocket shroud Height: small relative to width, to maximize the field of view Gamma-ray conversion efficiency: ~60%, requiring ~1.1 radiation lengths (~4 mm) of tungsten foil (in 16 layers) Transparent structure: minimum photon conversion and electron scattering in the support material Thin, efficient detectors: measure the electron/positron trajectory immediately following the gamma-ray conversion No consumables, other than electrical power Low power consumption: < 200 W (160 W actual) Minimal dead areas in the aperture Important to know the location of the inevitable inefficiencies Hi-speed, low-deadtime readout Silicon-Strips are ideal from these points of view. Sensitive to new events during readout of previous events Environmental requirements (launch and on orbit) Cost October 31, 2007 R.P. Johnson, SLAC AIS 8

9 Measurement Process Except at high energy, almost all of the direction information comes from the first two points measured. Ideally, the conversion should occur only in tungsten, close to the detector layers. Measure the trajectory close to the conversion point, to minimize lever arm for multiple scattering. If one of the first measurement layers fails to register, the resolution is drastically degraded! Measure a second point on the track as close as possible following the tungsten, so that MS in that layer contributes little error. October 31, 2007 R.P. Johnson, SLAC AIS 9

10 A few highlights: Silicon Strip Sensors 10,368 sensors installed in 18 towers. Specifications were prepared specifically to facilitate production, enhance reliability, and address GLAST-specific requirements: Single-sided (vs. double-sided typical in spectrometer experiments). 400-micron thick (vs. 300-microns typical in HEP experiments). 6 Si wafers (largest sensors available). Depletion < 120 V. (Relatively low voltage is another huge advantage of solid-state detectors for use in space.) Leakage < 500 na/sensor (<200 na on average). Excellent diagnostic for selecting reliable sensors. Dicing from wafers to 20 micron precision, to allow rapid assembly using mechanical jigs. AC coupled, to facilitate readout with low-powered electronics. Undiced sensor from Hamamatsu Photonics: cm micron strip pitch 56-micron strip width 384 p+ strips (n-instrinsic substrate) 2 wire-bond pads at each end of each strip October 31, 2007 R.P. Johnson, SLAC AIS 10

11 A few highlights: Silicon Strip Sensors 10,368 sensors installed in 18 towers. Specifications were prepared specifically to facilitate production, enhance reliability, and address GLAST-specific requirements: Single-sided (vs. double-sided typical in spectrometer experiments). 400-micron thick (vs. 300-microns typical in HEP experiments). 6 Si wafers (largest Spectacularly sensors available). good results: Depletion < 120 -V. Bad (Relatively channel low rate less than 1 in 10,000! voltage is another huge advantage of solid-state detectors - Average for use leakage in space.) current ~110 na! Leakage < 500 na/sensor - All specifications (<200 na on met or exceeded. average). Excellent diagnostic for cm 2 selecting reliable Specifications sensors. were worked out and vendors Dicing from wafers researched to 20 micron very early in the program. In precision, to allow retrospect, rapid assembly nothing using to be done differently. mechanical jigs. AC coupled, to facilitate readout with low-powered electronics. Undiced sensor from Hamamatsu Photonics: 228-micron strip pitch 56-micron strip width 384 p+ strips (n-instrinsic substrate) 2 wire-bond pads at each end of each strip October 31, 2007 R.P. Johnson, SLAC AIS 11

12 Ladder and Tray Assemblies All assembly is done with mechanical jigs. Optical CMM head is used afterwards for verification. (G&A Engineering, near Rome Italy) Each vacuum jig holds 1 ladder. 4 jigs position ladders onto the tray according to alignment pins. Ladders of 4 sensors aligned and assembled on this vacuum jig. October 31, 2007 R.P. Johnson, SLAC AIS 12

Multilayer flexible circuits connect to the data acquisition system.

13 Readout Electronics Architecture is designed to provide redundancy yet minimize wiring. Redundant 20 MHz serial control and readout paths: Each chip can be controlled or read by either of two paths. Multilayer flexible circuits connect to the data acquisition system. 24 amplifier chips (1536 channels) At present, 1 of 576 MCMs in the LAT Tracker can be read only to one side. 2 readout controller chips 648 Temporary connector savers Multi-chip readout module October 31, 2007 R.P. Johnson, SLAC AIS 13

5-micron, 3-metal CMOS) 1296 Trigger and Data mask registers Standard-cell auto route Control logic, command decoders Standard-cell auto route

14 ASICs Low power (180 µw/ch) achieved with Standard commercial process. Full custom design to achieve specific GLAST requirements and no more! Two ASICs Digital readout controller chip 64-channel amplifier/discriminator chip (Agilent 0.5-micron, 3-metal CMOS) 1296 Trigger and Data mask registers Standard-cell auto route Control logic, command decoders Standard-cell auto route Calibration mask and capacitors 4-deep event memory (addressed by TEM) Custom layout 2 custom DACs Cap 15, amplifier-discriminator channels. I/O pads and protection structures October 31, 2007 R.P. Johnson, SLAC AIS 14

15 Analog Chain Simple binary output! No transmission of analog signals! Detector load ~47 pf Charge-Sensitive Preamplifier: Folded cascode input µm 1.2 µm P-type input FET, drawing ~30 µa from 1.5V reference. ~5000 ohm input impedance. ~0.2 µs rise and very slow (several ms) decay. Shaper: ~RC/CR (but no true resistors used) with ~1.5 µs peak time Slow differential amplifier in the feedback loop holds the output baseline voltage at the desired reference point. Discriminator: simple comparator Overall performance: Gain ~100 mv/fc Noise: ~1500 electrons equivalent (MIP in SSD is ~32,000 electrons) Threshold variation across a chip <9% (Threshold is settable per chip.) October 31, 2007 R.P. Johnson, SLAC AIS 15

16 Trigger The Tracker is self triggering and provides the principal trigger for the LAT. MCM generates an OR of all 1536 channels in a layer. 6 consecutive layers in coincidence triggers the LAT. This signal is used to latch the comparator outputs before the shaper signal decays. Each MCM also records the time-over-threshold of the trigger OR signal. Gives de/dx information useful for background rejection. The trigger is live during event readout. 4 event buffers in the GTFE chips 2 event buffers in the GTRC chips Great care was taken to avoid triggering on readout noise: 1. All clocking and digital activity in the GTFE chips is continuous, to keep the power draw constant. No evidence of any 2. All data transmission is LVDS. increase in noise when readout is in progress! October 31, 2007 R.P. Johnson, SLAC AIS 16

17 Electronics Parts See R. Johnson, Electronics for Satellite Experiments, SLAC- PUB for a lot more information on this subject. Generally, the Tracker had to follow NASA QA requirements and requirements on EEE parts and part qualification. Many of our parts were custom, which required us to carry out the qualification work, including radiation testing. The custom SSDs and ASICs were some of the least problematic items in the system, even though the most complex. Thorough testing and screening were essential, however. Circuit boards, both rigid and flexible, were probably the most problematic parts. We found the hard way that vendors generally cannot deliver their state-of-the-art feature sizes to the quality standards demanded by NASA. One needs to design more conservatively, use more real estate, and monitor very closely. October 31, 2007 R.P. Johnson, SLAC AIS 17

Electronics Integration Electronics have to be on the tray sides to avoid a very large inactive gap between towers. Requires right-angle interconnect (RAI).

Tray RAI X-section of tray edge Sidewall Prototype Tray MCM RAI Detailed changes, such as a new, larger connector, really squeezed the mechanical limits!

18 Electronics Integration Electronics have to be on the tray sides to avoid a very large inactive gap between towers. Requires right-angle interconnect (RAI). Aggressive initial design minimized the inactive gap between towers (<18 mm!). Early dimensioning of the SSDs and the LAT as a whole locked this into the design. Tray RAI X-section of tray edge Sidewall Prototype Tray MCM RAI Detailed changes, such as a new, larger connector, really squeezed the mechanical limits! It proved to be very difficult to make the right-angle interconnect work in such a tight space (alignment, planarity, gluing, trimming, cracking of traces ). Original, thinner connector. October 31, 2007 R.P. Johnson, SLAC AIS 18

Wire-Bond Encapsulation Problems The original design called for encapsulating all wire bonds To avoid electrical shorts. To avoid damage from handling during tower assembly.

19 Wire-Bond Encapsulation Problems The original design called for encapsulating all wire bonds To avoid electrical shorts. To avoid damage from handling during tower assembly. Silicone-based encapsulation (NUSIL) used within ladders and from MCMs to ladders failed during thermal cycles of first articles of heavy-converter trays and was thereafter eliminated. This did not result in any handling or electrical problems. In general, encapsulation compromises the mechanical and thermal integrity of the wire bonds and should be avoided if possible. NUSIL encapsulation on ladder wire bonds. Encapsulation of ~3000 wire bonds internal to the MCM was essential for handling. This also caused problems! October 31, 2007 R.P. Johnson, SLAC AIS 19

MCM Encapsulation Problems How to test >1500 wire bond connections from ICs to flex-circuit right-angle interconnect. Probing was too slow and unwieldy.

20 MCM Encapsulation Problems How to test >1500 wire bond connections from ICs to flex-circuit right-angle interconnect. Probing was too slow and unwieldy. We tested a set of first articles this way, found no problems, and hoped that we could get away without this test prior to connecting MCMs to SSDs. But, we found a large number of wire bond failures from encapsulation delamination in some of the first trays, following thermal cycles. Ouch! October 31, 2007 R.P. Johnson, SLAC AIS 20

Interconnect-Testing Solution The answer to our problems was simple and fast: Short all of the flex-circuit traces to ground. Run internal charge-injection test on the amplifiers.

21 Interconnect-Testing Solution The answer to our problems was simple and fast: Short all of the flex-circuit traces to ground. Run internal charge-injection test on the amplifiers. Any amplifier not connected to the test circuit will respond, while those with good connections will act dead. Clamp pushes grounded, flexible Zebra-connector against the >1500 traces. Root cause of the delamination of the black encapsulation epoxy: Silicone contamination from Kapton masking tape with a silicone-based adhesive (big NASA no-no; we were warned!). October 31, 2007 R.P. Johnson, SLAC AIS 21

Layer alignment and drill programs had to be tweaked cable-by-cable, using x-ray images.

22 Electrical Cables 4-layer Kapton flexible circuits. 84 to 98 cm long; some of the longest were made diagonal on the panels! 8-mil traces & spaces and 30- mil pads. It proved to be very difficult for 2 manufacturers to make these to NASA specs. Layer alignment and drill programs had to be tweaked cable-by-cable, using x-ray images. Became the #1 critical path in completing the flight towers. Accepted cables are beautiful, with zero failures, but a more conservative design could have saved a lot of money and time. October 31, 2007 R.P. Johnson, SLAC AIS 22

23 Mechanical Overview Carbon composites are used wherever possible to Minimize gamma-ray conversions in the support structure. Maximize transparency to charged particles. But this came at a big price Carbon composites are more expensive than aluminum to procure. They are much more difficult to machine. They are not nearly as predictable as aluminum, requiring more engineering and testing. They require a lot of manual labor for assembly. It is difficult to achieve fine dimensional tolerances with them. 2mm gaps between SSD layers. Kapton readout cables Titanium reinforcement and flexures. Carbon-composite tray panels. MCM Carbon-composite sidewalls. October 31, 2007 R.P. Johnson, SLAC AIS 23

Mechanical Structure: Trays 100. 10. 1. Chan. No. : 6 Chan. type : M Sweep type : log Sweeps done: 1 Sweeps tot.

1 ν 1 = 626 Hz Q 34 -- Testing time -- Elapsed : 0:03:25 Remaining : 0:00:00 Date : 9.10.

Marker x = 626.082 y(max) = 17.2844 Hz Stiff carbon-composite panels keep adjacent SSD planes from touching.

24 Mechanical Structure: Trays Chan. No. : 6 Chan. type : M Sweep type : log Sweeps done: 1 Sweeps tot.: 1 Sweep dir. : up Sweep rate :2. Oct/min Ctrl strat.: Average Meas. mode : RMS Eng. unit : g Contr. mode: Closed loop.1 ν 1 = 626 Hz Q Testing time -- Elapsed : 0:03:25 Remaining : 0:00:00 Date : :34:21 C-fiber face sheets Al honeycomb core C-C machined closeout Marker x = y(max) = Hz Stiff carbon-composite panels keep adjacent SSD planes from touching. Carbon-carbon closeouts for easy machining of details and good thermal conductivity. SSDs bonded with low-t silicone (Nusil), to allow for CTE mismatch. October 31, 2007 R.P. Johnson, SLAC AIS 24

Building the carbon-composite panels to the necessary precision (e.g. ±0.

The key was to assemble and cure the adhesives at room temperature, using precision steel

Tray Panel Manufacturing Plyform, near Milan Italy Laser interferometry was used in Pisa, Italy

25 Building the carbon-composite panels to the necessary precision (e.g. ±0.1 mm in width) was a challenge. The key was to assemble and cure the adhesives at room temperature, using precision steel fixtures. Tray Panel Manufacturing Plyform, near Milan Italy Laser interferometry was used in Pisa, Italy to check for defects in the face-sheet bonding. Bonding to the tungsten foils was problematic. All foils had to be chemically etched and then primed before bonding. Bonding the Kapton bias circuits without bubbles was also challenging. October 31, 2007 R.P. Johnson, SLAC AIS 25

Mechanical Structure: Towers Carbon-composite sidewalls support the trays and form a stiff box structure (lowest modes: 130 Hz lateral; 370 Hz vertical).

26 Mechanical Structure: Towers Carbon-composite sidewalls support the trays and form a stiff box structure (lowest modes: 130 Hz lateral; 370 Hz vertical). The bottom tray transfers the load to the Grid through 8 titanium flexures and is much stronger than the other trays. Heat flows passively down the sidewalls and through copper straps to the Grid. Aluminum outer coating for EMI protection. Sidewall Alignment ball nest Cu heat strap MCM Readout cable terminations. Top tray Titanium corner bracket Bottom tray Flexures October 31, 2007 R.P. Johnson, SLAC AIS 26

Tower Assembly 19 trays are aligned to a steel fixture using pins. Tower upside down. 8 pre-bent cables are installed and electronics tested as indicated here.

27 Tower Assembly 19 trays are aligned to a steel fixture using pins. Tower upside down. 8 pre-bent cables are installed and electronics tested as indicated here. Mating the nano-connectors is always a big challenge! 2 sidewalls are installed (lots of countersunk screws). The fixture is removed and the other 2 sidewalls are installed. October 31, 2007 R.P. Johnson, SLAC AIS 27

Bottom-Tray Reinforcement Titanium corner bracket An early-prototype random-vibe test

The final design includes titanium reinforcement in the corners and closeouts made of strong

A thorough static load test verified each bottom tray before tower assembly.

28 Bottom-Tray Reinforcement Titanium corner bracket An early-prototype random-vibe test failure showed that carbon-carbon closeouts could not support the load at the tower bottom. The final design includes titanium reinforcement in the corners and closeouts made of strong carbon-fiber with carboncarbon veneers. A thorough static load test verified each bottom tray before tower assembly. Static load test fixture Carbon-carbon veneer, for machining and thermal conductivity. M55J carbon-fiber backing for strength. October 31, 2007 R.P. Johnson, SLAC AIS 28

Tracker-Grid Interface Shim Eccentric cones Nut Hex bushing

be sufficiently accurate to align towers within their tight

Large hole clearance with bolts held in place by friction was

29 Tracker-Grid Interface Shim Eccentric cones Nut Hex bushing Stud Titanium & steel parts fabricated by Advanced Machining in San Diego. Grid Bottom tray fabrication and Grid hole tolerance could not be sufficiently accurate to align towers within their tight tolerance. Large hole clearance with bolts held in place by friction was not acceptable. Solution: dual concentric cones ensure tight fit with the bolt center located anywhere within a 1-mm radius. October 31, 2007 R.P. Johnson, SLAC AIS 29

Tracker Alignment Eccentric cone orientations were calculated by computer,

75 mm smaller than stay-clear, but all of them fit.

30 Top corners of 4 installed towers. 2.5 mm design gap must accommodate alignment, vibration, and thermal distortion. Tracker Alignment Eccentric cone orientations were calculated by computer, based on a complete CMM survey of the tower. Towers by design are only 0.75 mm smaller than stay-clear, but all of them fit. From the final optical survey, the maximum excursion of a tower top is 0.59-mm. Inverted prototype tower INFN-Pisa CMM October 31, 2007 R.P. Johnson, SLAC AIS 30

Fabrication Test Program Glue 26 ICs to a PC Brd. Glue 4 SSDs together. (A lot of testing of various parts and materials is omitted here for simplicity.) Glue 32 SSDs and 2 MCMs to a tray ($$$).

31 Fabrication Test Program Glue 26 ICs to a PC Brd. Glue 4 SSDs together. (A lot of testing of various parts and materials is omitted here for simplicity.) Glue 32 SSDs and 2 MCMs to a tray ($$$). Some early trays also received vibration tests. Disassembling a tower to replace a tray is We are here now, with very expensive at this Tracker still in excellent point. shape (all 13,824 GTFE October 31, 2007 R.P. Johnson, SLAC AIS chips functioning). 31

Functional Test of Flight Trays Cosmic-ray track

Full test of all trays, imaging cosmic-rays,

32 Functional Test of Flight Trays Cosmic-ray track in the stacked-tray test stand. Finished tray inside of its storage box. Stacked-Tray test stand at INFN Pisa. Full test of all trays, imaging cosmic-rays, before committing them to tower assembly. October 31, 2007 R.P. Johnson, SLAC AIS 32

33 Some Tracker Test Setups Vibration Testing (Rome) MCM Burn-in & Thermal Test (SLAC) Tray Thermal Cycles (Perugia) Thermal-Vacuum Testing (Rome) October 31, 2007 R.P. Johnson, SLAC AIS 33

EMI testing, prior to integration into the LAT

34 Completed Tracker Individual towers underwent thorough acceptance testing at SLAC, including EMI testing, prior to integration into the LAT Grid. October 31, 2007 R.P. Johnson, SLAC AIS 34

35 Selected Cosmic-Ray Events in 8 Towers Note the almost complete absence of noise hits. October 31, 2007 R.P. Johnson, SLAC AIS 35

36 Cosmic-Ray Muons in the Full LAT 16 tower LAT Actual rate: ~ 500 Hz October 31, 2007 R.P. Johnson, SLAC AIS 36

37 Two spare tracker tower modules and three calorimeter modules. Beam Test at CERN 470 MeV gamma-ray conversion shown in the singleevent display to the right. October 31, 2007 R.P. Johnson, SLAC AIS 37

$2% on average) Strip #, 1 to 1536 Hit efficiency (in active area) >99.4% Dead channel fraction 0.2% Overall Tracker active area fraction: 89.4% Noise occupancy <5 10 7 (with 0.$

38 Tracker Performance 1 example readout module Hit efficiency from cosmic-ray muons Threshold (fc) Threshold variation <9% rms in all modules (5.2% on average) Strip #, 1 to 1536 Hit efficiency (in active area) >99.4% Dead channel fraction 0.2% Overall Tracker active area fraction: 89.4% Noise occupancy < (with 0.06% of channels masked) Power consumption 160 W (180 µw/ch) 43% FWHM Muon time-overthreshold (OR of all channels per layer) October 31, 2007 R.P. Johnson, SLAC AIS 38

39 GLAST Status GLAST at General Dynamics last April Observatory Environmental Testing in Progress. Launch in May 2008? SWIFT instrument launches on the same type rocket planned for GLAST October 31, 2007 R.P. Johnson, SLAC AIS 39

A Prototype Amplifier-Discriminator Chip for the GLAST Silicon-Strip Tracker

A Prototype Amplifier-Discriminator Chip for the GLAST Silicon-Strip Tracker Robert P. Johnson Pavel Poplevin Hartmut Sadrozinski Ned Spencer Santa Cruz Institute for Particle Physics The GLAST Project