tracker design & assembly overview - stanford university

Tracker Design & AssemblyOverview

Tracker Design & AssemblyOverview

Robert P. JohnsonU.C. Santa Cruz

December 1, 2004 Tracker Production Review 2

Tracker Design OverviewTracker Design Overview• Stiff composite panels (>500 Hz)

– Allows small gap between x-y SSD layers

• Tungsten foils on panel bottom• SSDs on top & bottom faces• Electronics on panel edges

– Minimizes the gap between towers (1.59 cm Si to Si)

• Carbon-fiber walls for vertical support– Very stiff box structure– Passive cooling to tower base

• Flexure attachment to Grid– Decouple from thermal expansion– Lowest frequency >150 Hz– Greatly reinforced attachment to the

bottom tray.– Thermal straps couple sidewalls to the

Grid (not shown)Readout Cable

Multi-Chip Electronics Module (MCM)

2 mm gap

19 Carbon-Fiber Tray Panels

Titanium Flexure Mounts

Carbon-Fiber Sidewalls (Aluminum covered)


Tower-Grid InterfaceTower-Grid Interface• This figure shows a stack of the 3 lowest trays in a tower.• The bottom tray is made of heavier, stronger material to carry the full

load of the tower interface to the Grid.• These figures are old and do not show the conical mounting holes.


Tracker TrayTracker Tray


Tracker Production OverviewTracker Production Overview

Readout CablesUCSC, SLAC (Parlex)

SSD Procurement, TestingJapan, Italy, SLAC (HPK)

Electronics Design, Fabrication & TestUCSC, SLAC (Teledyne)

Tracker Module Assembly and Test Italy (INFN, Alenia)

Tray Assembly and TestItaly (G&A)

SSD Ladder AssemblyItaly (G&A, Mipot)

Composite Panel & ConvertersItaly (Plyform)

1440 / 2304

10800 / 9216

21 / 304

~250 / 576

0 / 16

SidewallsItaly (Plyform)

37 / 304

16 / 128

8 / 64

# complete # required for flight


Tracker Readout Architecture• Nine MCMs on each of the 4 tower sides.• Two redundant cables connect the 9 MCMs to each other and to the TEM.• Each of the 24 GTFE chips on each MCM can be controlled and read out

from either of the two cables.

24 64-channel amplifier-discriminator chips for each detector layer

2 readoutcontroller chipsfor each layer

Con

trol s

igna

l flo

w Control signal flow

Data flow to FPGAon DAQ TEM board.

Data flow to FPGAon DAQ TEM board.

Control signal flow

Data flow

Nine detector layers are read out on each side of each tower.

GTRC

GTFEGTFE

GTRC

GTRC

GTRC

GTRC

GTRC

9-998509A22


MCM Readout Module Configuration• 8-layer polyimide PWB• Top edge thickened and machined to a 1.0

mm radius• 1-layer flex circuit (“pitch adapter”)

bonded over the radius• Fully encapsulated wire bonds• Conformal coating• 2 Omnetics nano connectors

Readout IC

Machined corner radius with flex circuit bonded around the curve

TMCM, attached by screws

Detector

Tray Structure

Bias circuit

High-thermal conductivity transfer adhesive

Readout IC



Detector

Tray Structure

Bias circuit




Detector

Tray Structure

Bias circuit


24.58mm 18.0mm

359.0mm

Grounding Screws3 Total

Mounting Screws, 1 of 8

Connector, 1 of 2

GTFE, 1 of 24GTRC, 1 of 2

TMCM, bonded to tray

3M 2216 epoxy bond


Pitch Adapter Flex Circuit• 1-layer Kapton flex circuit with

½-oz copper traces• Ni (150µinch) + Au (10µinch)

plating for aluminum wedge wire bonding

• Precision tooling holes (not shown)

• Circuit & traces are trimmed to length after bonding to the PWB

SS

D S

ide (228 µm

pitch)

AS

IC S

ide“ground”

Bias HV


Pitch-Adapter BondingPitch-Adapter Bonding1 2

3

4


Tracker ASICs

64 amplifier-discriminator channels.

4-deep event memory (addressed by TEM)Custom layout 2 custom DACs

Trigger and Data mask registersStandard-cell auto route

Control logic, command decodersStandard-cell auto route

Cap

Calibration mask and capacitors

I/O pads and protection structures

• ASICs are purchased from MOSIS and tested on the wafer at UCSC.

• Grinding, dicing, inspection by GDSI.• Two such ASICs have been developed:

Digital readout controller chip64-channel amplifier/discriminator chip


MCM AssemblyMCM Assembly• All assembly is done at Teledyne Microelectronic Technologies• Parts are furnished by SLAC• Summary of the assembly flow:

• Pitch-adapter flex attachment, trimming, and inspection (MIP-1)• PWB bakeout (new)• SMT component pick-&-place and reflow soldering• Connector mounting and soldering• Die attach• Wire bonding• Functional test and inspection (MIP-2)• Encapsulation and conformal coating• Final test and inspection (MIP-3, MIP-4), incl. new pitch-adapter test


X-Ray Cross Section of an MCMX-Ray Cross Section of an MCM

Flex Circuit

Internal Cu Planes

ASIC and Conductive Glue

Wire Bond

Encapsulation FillEncapsulation Dam

Fiberglass


MCM Test FlowMCM Test Flow

GTFE wafer probing (UCSC) LAT-TD-00247

Pitch-Adapter & PWB inspection and screening (SLAC)

Wafer lapping, dicing, & inspection (GDSI)

LAT-PS-01321

Polyswitch procurement screening, & qualification

(Raychem, UCSC) LAT-SS-01116

MCM Assembly MIPs (Teledyne)

LAT-PS-01971LAT-TD-00249

1. Flex attach inspection2. Pre-encapsulation visual

inspection and electrical test3. Final electrical test and

visual inspection4. EIDP review

Environmental acceptance test, burn-in, final test (SLAC)

LAT-TD-023671. 20 unpowered thermal cycles, -30°C to 85°C2. Functional test at –30°C, 25°C, 60°C 3. Power-on burn-in at 85°C for 168 hours4. Final electrical test (incl. new PA test)

GTRC wafer probing (UCSC) LAT-TD-00248

EEE PCB parts approval, procurement,

receiving/inspection

• LAT QA does inspections and tests at 4 MIPs at Teledyne

• Environmental test and burn-in at SLAC

• Final electrical testing and inspections at SLAC before shipping to Italy

• Receiving functional test in Italy


New Pitch-Adapter Test FixtureNew Pitch-Adapter Test Fixture

CONNECTOR SAVER

TEST FIXTURE

ZEBRA CONNECTOR

GROUND LEAD

STORAGE CASE BASE

MCM WITHPITCH ADAPTER

SOFTWARE/ELECTRICAL FIXTURE

Provides a quick, positive test of pitch-adapter connections to all 1536 input channels.


Mounting MCMs onto TraysMounting MCMs onto Trays• The MCM is clamped with the pitch-adapter

pushed against a straight jig, to ensure a straight top edge for wire bonding.

• Custom pins clamp the rest of the board to the jig and also have a thin washer to guarantee the adhesive bond line.

• The jig precision aligned to the tray, gluing the MCM to the tray with a venting pattern of 3M 2216 epoxy.

MCM

Tray Panel


Electronics Test FlowElectronics Test Flow• Receiving test of MCMs delivered to INFN Pisa.• Test of the bias connections at G&A after mounting MCM to tray and

wire bonding the bias from MCM to tray.• Test of the completed tray assembly at G&A. This is the first test of

amplifiers connected to sensors.• 4 tray thermal cycles (−30C to +55C).• Test of stacks of trays (cosmic rays).• Functional testing of assembled towers.• Vibration testing.• Thermal-vacuum testing.• Pre-shipping function test.• Receiving functional & performance tests.• EMI/EMC acceptance test at SLAC.


Tower Assembly & CablingTower Assembly & CablingThe cable “cactus arms” are pre-bent before mounting.


Tower Assembly and CablingTower Assembly and Cabling

Mating the connectors Cabling Completed


Tower Assembly and Cabling Tower Assembly and Cabling Sidewall Installation Functional Testing

These photos are rotated 180 degrees to orient the tower right-side up.



Bending the cables over the top corner brackets.



Cable storage for vibration and shipping.

Cable bend through Grid

Bottom Tray

Heat Strap

Cable

MCM IssuesMCM Issues



Electronics IssuesElectronics Issues• NCR 104: glitches in GTFE configuration register readback.

– This was found to be 100% cured by changing the internal MCM clock termination from 100 ohms to 75 ohms.

– All MCMs produced with 100-ohm clock termination were reworked by Zentek.

– Teledyne production was changed to incorporate the new resistor value.• NCR 107: scrambled readback of event data on every other trigger.

– This was first noticed on a few MCMs when testing at +60C.– The symptom can be reproduced in all MCMs by raising the clock duty

cycle sufficiently far above 50%.– The errors occur internally in the GTRC chip memory access.– The root cause is poor clock transmission from the TEM to GTRCs, but

some GTRC chips are much more sensitive than others.– We found that we can guarantee proper operation of almost 100% of

MCMs over a duty cycle range of 45% to 55% if we lower the termination of the flex-circuit cable clock bus from 100 ohms to 75 ohms.

– All test and flight cables are being modified in this way.


Short Circuits, NCR 128 Short Circuits, NCR 128 • Short circuits between layers 7 and 8 of the PWB, from the AVDDA

plane (1.5 V) to the Bias plane (120 V).• 10 MCMs from 7 panels clustered tightly in time (panels 90 through

164 and MCM Serial Numbers 559 through 737).• We rejected all good MCMs from the same panels (now in use by

the electronics and flight software groups).• All shorts occurred either before burn-in or in the first day of the 7-

day burn-in, except one case where the short occurred about 2/3 of the way through the burn-in.

• The effect of a short circuit in the completed instrument would be a loss of one SSD ladder (1/4 of one of the 36 SSD layers in a tower).

• This issue generated concerns about moisture, prompting the addition of a PWB bakeout before SMT soldering at Teledyne.

• However, the root cause appears to be metallic contamination of the prepreg, based on DPA done at GSFC.


Short-Circuit Smoking GunShort-Circuit Smoking GunCopper particle between layers 7 and 8

Short circuit location

DPA by Diane Kolosof GSFC

MCM SN 612


Conformal Coating IssuesConformal Coating Issues• 154, 201: lifting around

encapsulation and components• The conformal coating (Humiseal)

does not adhere to the cured encapsulation epoxy. The masking allowed it to overlap slightly.

• Conformal coating tends to be worked loose around the mounting screws.

• Some small bubbles appeared in the conformal coating around components.

Coating lightly pushed back here using a probe


Conformal Coating IssuesConformal Coating Issues• NCR 202: voids in conformal coating.• In some cases one side of several components was not covered,

presumably by spraying in only one direction.• Part of this problem resulted from trying to correct the bubble

problem by spraying a thinner mix into a thinner overall layer.


Conformal Coating IssuesConformal Coating Issues• NCRs 190, 196: conformal coating on the connector mating surface.

– This photo is probably the worst case.– The severity correlates strongly with the looseness of the connector

mounting. MCMs with no coating on the connector also were found to have the minimum (or zero) gap between PWB and connector.

• Corrective actions:– Existing MCMs were

cleaned by hand at SLAC per a released rework procedure.

– Connector savers could be removed and the mating surface masked before conformal coating.

– Elimination of the gap between PWB and connector.


Conformal Coating CorrectionsConformal Coating Corrections• SLAC

– Rework of existing MCMs. This is well under way, with 102 out of 181 completed.

– Improved source inspection.• Teledyne:

– Masking around screw holes. Teledyne has made some plastic washers for this purpose.

– Coating on connector surface. No final decision yet. SLAC has recommended removing the connector saver and masking the connector face.

– Void on components. Ensure that the spray is applied from multiple directions.

– Other improvements already implemented prior to the shutdown.• Improved masking around the encapsulation.• Thinner coating with less bubbles.


Misc. Production IssuesMisc. Production Issues• Silicone Adhesive on Masking Tape

– Jim Lohr noticed that the masking tape that Teledyne was using in the production used a silicone pressure sensitive adhesive.

– Teledyne has removed this tape from the GLAST cell and has procured new tape with acrylic pressure sensitive adhesive.

– The new tape was tried out successfully but still needs to be validated in a larger run of parts.

• Localized damage to PWB from mounting screws.– Plastic washers will help.– Check that the screws are not being over torqued.

• Rusting of mounting screws.– This is GSE. Black steel was used to allow magnetic handling.– Changed to stainless steel.

• ESD– Teledyne will procure ionizers after receiving PO from SLAC.– There is zero evidence that this was ever a problem for the MCMs, so I

do not think it should gate restart of production.


Gap Between PWB and ConnectorGap Between PWB and Connector• NCR 230: most of the MCM connectors are not fully torqued, and some have

a visible gap between the connector and PWB.• This could allow conformal coating to run under the connector.• The clearance between MCM and Tracker sidewall is very small, such that

we can tolerate only a 100-micron excess in thickness beyond the 3.81mm nominal (historically the clearance was much larger but got eaten away by numerous “improvements”).

This photo is an extreme example.

• On most MCMs at SLAC we can turn the screws about ¼ turn before reaching 10 inch-oz. This pulls them to within the thickness spec.

• Teledyne process instruction:Screw in the two connectors to the PCB. Use Torque “RED” screwdriver set at 8-10 inch ounces. Make sure that the screw head seats well into the PCB without any protruding past the face of the PCB.


Gap Between PWB and ConnectorGap Between PWB and Connector• For the few extreme cases, where the MCM+Connector thickness is

greater than 4.0mm, we have not tried to pull it snug, for fear of damaging the connector or solder pins. These will require rework.

• Proposed remedy:– Monitor the connector installation closely when production restarts.– SLAC check the torque on the connector mounting screws at MIP-3.– Continue monitoring the MCM+Connector thickness at SLAC after burn-

in and before shipping to Italy, resetting the torque if necessary.


Debonding of the Pitch AdapterDebonding of the Pitch Adapter• NCR 237, up to about 20% of the

MCMs have regions on the pitch adapter where the flex is not glued down.

• Wire bonding to the SSDs in these regions is impossible without rework.

• Teledyne has reworked MCMs with this issue when found during their production.

• G&A in Italy reworked successfully 2 MCMs already bonded to trays.

• SLAC is developing a rework plan for the existing MCMs.

• We need to review the Kapton surface preparation in the Teledyne production and perhaps also the epoxy degassing.


Debonding of the Pitch AdapterDebonding of the Pitch Adapter• Note: the back edge of the PWB was masked with Kapton tape

using a silicone PSA. This will be replaced by an acrylic PSA in future production.

• Epoxy is dispensed from a double-barrel syringe with a mixing nozzle attached. It is applied to the flex by a screen printer.

• Teledyne Kapton surface preparation procedure:– Using 400 CC-CW abrasive paper, lightly abrade the backside area of

the Flex to remove the glaze. Use 3 swipes.– Wipe down flex and PCB using Alcohol/Acetone solvent.

• 3M recommended preparation for bonding plastics with Scotchweld1838 B/A green epoxy: – Solvent wipe with Isopropyl Alcohol.– Abrade using clean fine grit abrasives.– Solvent wipe with Isopropyl Alcohol.

• My suggestion: 3 swipes with abrasive paper could miss square millimeters of area. A thorough scrubbing with Scotchbrite could be more effective in getting 100% coverage.


Broken PA Traces from TrimmingBroken PA Traces from Trimming• On some MCMs the ends of the pitch-

adapter traces are cracked off during the trimming operation.

– The trimming jig runs a “pizza-cutter” wheel along the MCM, guided by rails.

• Some examples made it to SLAC and to Italy with most of the wire-bonding area missing on some traces.

– Either the wire-bonding operator has to make time-consuming interventions, or some bonds get missed.


Broken PA Traces from TrimmingBroken PA Traces from Trimming• This November 19 at Teledyne I witnessed 4 PWBs with pitch

adapters attached shortly before my visit. – The first had many such cracked ends of traces.– Before trimming the next 3, the operator changed to a new blade, and

all 3 were cut very cleanly, with zero cracking.• Need to pay more attention to the trimming tool adjustment and

blade sharpness.• SLAC will improve their inspection at MIP-1.


Cracked Pitch-Adapter TracesCracked Pitch-Adapter Traces• NCR 203: MCMs with >8 pitch-

adapter non-conductive cracked signal traces and/or >0 non-conductive cracked bias traces found during inspections at SLAC.

• About 15% of PWBs are rejected at MIP-1 because of such cracks. Either some cracks were missed, or they opened up during further processing.

• The cracks are not random. They generally occur in groups with strong spatial correlation from strip to strip.

• The nickel plating is inherently brittle and tends to crack on the surface, but the nonconductive cracks that pass through the copper are evidently following pre-existing flaws in the material.


Cracked Traces HistoryCracked Traces History• Original prototypes had a Kapton cover layer over the bend region

– Many traces cracked at the stress riser at the edge of the cover layer– Cover layer alignment was difficult for the circuit manufacturer– Alignment of the circuit onto the PWB was even more difficult at

Teledyne• Next iteration eliminated the cover layer and increased the trace

length, but many cracks appears throughout the bend region– Following this we increased the radius from 0.64mm to 1.0mm.– The radius was not smoothly cut, giving sharp discontinuities in slope at

the beginning and end of the curve. • We made custom form cutters to give a smooth transition.

– Glass beads in the epoxy tended to clump, and traces cracked over the clumps.

• Eliminated the glass beads.

• The next prototype trial did not show any cracking, so we geared up for pre-production– The prototype circuits were made by Tyco, Santa Clara.


Cracked Traces HistoryCracked Traces History• Preproduction:

– LAT manufacturing insisted on sending all flex-circuit orders to Parlex, and >800 pitch adapters were manufactured.

– Preproduction units did not show problems with non-conductive cracks at the MIP-1 or final inspection levels.

• 5 qual MCMs subjected to 220 thermal cycles (−30C to +85C):– Afterwards each had 7 to 35 traces with hairline cracks but still

conductive. There were no non-conductive cracked traces.• 5 qual MCMs subjected to 20 thermal cycles and 1168 hour burn-in:

– Afterwards, each had 100 to 500 hairline cracks but still conductive. One had 5 non-conductive cracks; the others had none.

– But during flight production we suddenly started seeing a large incidence of non-conductive cracks, with no change in process that we could find.

• Flight Production:– The non-conductive cracks are not increased by thermal cycling.

• We took 4 flight MCMs and thermal cycled 200 times, and we found an increase in hairline surface cracks on still conductive traces, but no increase in non-conductive cracks.


Cracked Traces ImpactCracked Traces Impact• About 15% of our PWBs and pitch adapters are lost at MIP-1.

– Besides the cost of the parts, this also slows down production.• 11/284 = 4% of final MCMs are lost because a bias trace is broken.• 31/284 = 11% of final MCMs have >8 cracked traces• 6/284 = 2% of final MCMs have >15 cracked traces (i.e. 1% of channels)• Average number of cracked traces = 3 (0.2% of channels)

Cracked Traces

0

20

4060

80

100

120140

160

180

0 5 10 15 20 25

MCMs

Non

-Con

duct

ive

Cra

cks


Cracked Traces RemediesCracked Traces Remedies• Solder mask over bend region, to prevent plating of the traces in that

region.– Conflict with alignment: placement of the solder mask on the circuit and

the placement of the circuit on the PWB add up and• The solder mask can infringe on the Italian wire bonding area• The solder mask may not cover the full bend.

– First attempt did not have a large enough solder mask, and huge numbers of traces cracked close to the end of the mask.

– Second attempt ran the solder mask well into the flat on the top of the MCM and there was no cracking.

• But INFN informed me that this would not leave enough room for them to do reliable wire bonding.


Cracked Trace RemediesCracked Trace Remedies• Alternate sources of pitch adapters.

– Different plating method at Parlex did not help.– A Swiss company is making some samples for us to try, using a

polyimide/copper material without a directionality of the grain.• The Parlex circuits have a very strong copper grain pattern oriented

perpendicular to the strips. The cracking appears to occur in places where the grain is especially strong.

– We are also ordering samples from another U.S. supplier, again with a slightly different polyimide/copper material.

• Part of the motivation for this is just to have a second vendor, because we have had very poor delivery performance from Parlex.

• In the meantime we have ~170 regular pitch adapters from Parlex and are ordering more. Production will restart with this, and we will continue to take the hit in yield.


Broken Wire Bonds to the Pitch AdapterBroken Wire Bonds to the Pitch Adapter• NCR 226, on some MCMs up to 500 or more wire bond connections

are broken between the ASICs and the pitch adapter.• This has become by far the most serious MCM issue, but the scope

of the problem was not realized until last week.• The problem was not seen during the preproduction, even in the

qual MCMs that underwent 220 thermal cycles.– However, during the preproduction we did not have the capability to test

100% of MCMs. Our probing setup was too time consuming and was only used on a sample of MCMs (showing zero broken connections).

• The problem also was not seen in Tower-0 on any of the 23 MCMs connected to functional silicon-strip planes.

• In the meantime we were trying to develop a method to do 100% electrical testing of the pitch adapter prior to integrating onto a tray.– But our goal here was to test for cracked traces!– We did not expect many broken wire bonds, because the incidence on

the PWB side of the ASICs was nearly zero.


Pitch-Adapter TestingPitch-Adapter Testing

1

10

100

1000

0 100 200 300 400 500 600 700 800 900 1000

Disconnected Channels

MC

Ms

0

10

20

30

40

50

60

70

0 5 10 15 20 25

Disconnected Channels

MC

Ms

The new test fixture and software enables us to test 100% of the channels on every MCM very quickly, at a price of only a very few false negative readings.

11% of the MCMs have >15 bad connections.Average number of broken connections/MCM after rejecting those 11% is only 2.


Broken Wire BondsBroken Wire Bonds• Other observations:

– On MCMs with large numbers of broken connections, the bad channel count is often variable, with different results on each trial. We have observed that variations in the pressure applied by the fixture can cause intermittent reconnections to form.

• This indicates that the break occurred after curing of the encapsulation, and the two pieces are still very close to each other.

– The break is almost certainly at the pitch-adapter end of the wire bond, because we do not see any broken bonds on the PWB side of the ASICs.

– The highest probability of breakage is toward the center of the MCM.

0

50

100

150

200

250

300

0 200 400 600 800 1000 1200 1400 1600

Strip Number

Dis

conn

ecte

d C

hann

els

per 6

4 St

rips


Broken Wire Bond Date DependenceBroken Wire Bond Date Dependence• The problem seems to have started in July and gotten worse in August and

September.• These data are incomplete, however, as we are continuing with testing of the

existing MCM stock.

Date Dependence

0

100

200

300

400

500

600

3-Apr 23-Apr 13-May 2-Jun 22-Jun 12-Jul 1-Aug 21-Aug 10-Sep 30-Sep 20-Oct

Date

Mis

sing

Cha

nnel

s


Effect of Thermal CyclesEffect of Thermal CyclesS/N Date Visual PA

cracks Pre-Temp Cycle Post-Temp Cycle Broken Bias

511 9/20/2004 no data 7 5 0588 9/20/2004 no data 1 0 0732 9/20/2004 10 12 16 0972 7/29/2004 0 0 1 01014 7/20/2004 0 0 0 01112 7/20/2004 0 16 22 1

11011 7/20/2004 0 10 10 011030 7/20/2004 7 9 8 011080 7/26/2004 0 7 12 011120 7/20/2004 0 0 8 011358 7/29/2004 1 1 2 011368 7/29/2004 3 4 3 011377 7/29/2004 0 1 3 011386 7/20/2004 0 0 293 711419 7/29/2004 0 0 0 011428 7/29/2004 0 4 0 111443 7/20/2004 0 0 0 011449 7/20/2004 0 0 0 011507 7/20/2004 0 1 0 011510 7/20/2004 0 0 0 011512 7/20/2004 0 0 1 0

21 MCMs tested before any thermal cycles and following 20 cycles.


Crosscheck with Different Test MethodCrosscheck with Different Test Method• Marcus Ziegler developed a second test method in which he slowly

drags a wire across the 1536 traces. The wire is capacitively coupled to a pulser that injects 150 fC of charge at about 30 Hz. The MCM is read out rapidly, and his software looks for channelswith injected charge.

• For MCMs with small numbers of missing channels the results agree well with the PA test fixture.

• For MCMs with many missing channels the results can differ, because of the pressure applied by the PA test fixture.– e.g. SN-1033

• PA test fixture: 226 missing channels• Marcus: 171 missing channels


Another Observation by MarcusAnother Observation by Marcus

SN-1033

On this MCM with many missing channels, some component of the encapsulation flows out onto the traces and discolors them.


Mechanical HandlingMechanical Handling• Test of SN-1033 with the PA test fixture before and after vigorous

flexing and twisting of the MCM by hand:– Before: 226 disconnected channels– After: 233 disconnected channels

• Both measurements were repeated to be sure they were reproducible.

• The handling was far in excess of what any MCM would normally experience during inspections and so forth.


DPADPA• Done by Diane Kolos and Bruno Munoz at GSFC.• MCM SN 775, with ~180 missing connections.• Diane sectioned the MCM at the location of a few of the known bad

connections.• Bruno x-rayed the MCM.

• The results verify that the wire bonds are broken at the pitch-adapter end, at the “ankle”.

• The results also indicate debonding of the encapsulant from the substrates in the affected regions.


Example Wire Bond CrackExample Wire Bond Crack

Crack

Delaminations

Delamination

Chip 13, Channel 2


Another Example CrackAnother Example Crack

Chip 12, Channel 2


Wire Bond X-RayWire Bond X-Ray


Encapsulant DebondingEncapsulant Debonding• Albert Nguyen inspected 10 MCMs with large numbers of missing

connections to see if there was visible delamination of the encapsulation material from the pitch adapter.

• On one of the 10 he found a region with many missing channels inwhich the encapsulation could be moved slightly w.r.t. the pitch adapter, sliding along the traces, by applying some manual pressure with tweezers.

• On the other 9 he could not see anything notable.


Encapsulant Debonding in ItalyEncapsulant Debonding in Italy

SN 1100, mounted on a tray and connected to SSDs. 128 contiguous channels were found to be disconnected.

In the SLAC test there were only 2 disconnected channels.

This MCM was delivered to SLAC May 17, 2004.

Pitch-Adapter Traces

Encapsulation

Gap


Encapsulant Debonding in ItalyEncapsulant Debonding in Italy• Two other MCMs mounted on trays

had similar fates:– SN-652 (June 7) had 142

disconnected channels after mounting onto the tray. There were only 3 at SLAC.

– A 3rd MCM had 176 such disconnected channels (I don’t know the SN yet).

– Another with 36.• In all of these cases the gap

between encapsulant and pitch adapter was visible.

• On a few trays the number of disconnected channels was observed to increase during the 4 tray thermal cycles.

SN 652


Tentative ConclusionsTentative Conclusions• Source of the disconnected channels

– Small numbers are from the cracked traces, verified by visual scans– Most are from broken wire bonds.

• The cause of the broken wire bonds is movement of the encapsulant relative to the pitch adapter.– This motion would not occur if the encapsulant remained well adhered.– This problem never occurs on the PWB side of the ASICs.

• The root cause of the encapsulation debonding needs to be betterunderstood. Note that it does not happen on all MCMs.– Surface contamination? (Silicone?)– Inadequate surface preparation? Teledyne does vapor degreasing and

plasma cleaning prior to wire bonding. Abrading of the Kapton seems not possible, because of the closely spaced traces.

– Too much temperature range (+125C cure vs −30C test)?


MCM Schedule IssuesMCM Schedule Issues• All of the MCMs currently at SLAC need to be delivered to Italy

before Christmas to support production of the first 5 towers.• MCM production needs to resume next week in order to have any

chance of delivering more MCMs to Italy to continue tray production after Christmas.

• The tower production plan was increased from 2 towers per month to 3 per month in order to complete by end of June 2005.– 25 MCMs/week production will not support 3 towers per month.– It would have to be increased to at least 30/week.– Burn-in, test, and inspection at SLAC can in principle barely keep up

with 3 towers per month, but it will fall behind if we have to keep doing so much rework.

Production Status and Open Issues in Italy

Production Status and Open Issues in Italy



Flight Tray Production StatusFlight Tray Production Status• As of end of last week:

– Bare panels completed by Plyform: 140– Tray panels completed by Plyform (with Kapton and W): 36

• New machining procedures are now in place to ensure the MCM clearance, so this delivery rate will very soon accelerate.

– Trays delivered by G&A with MCMs and SSDs mounted: 21• Presently working at a rate of 3 trays/day, but this may be slowed down by

additional MCM screening requirements.– MCMs delivered to Italy: ~88 (not counting rejects being returned for

rework)• Tower-A status:

– All trays are completed (some with NCRs because of disconnected channels).

– Tower assembly and test is in progress in Pisa.• The trays have been stacked and cables attached, and comprehensive

performance testing is in progress. So far the picture is good.• All 8 cables function (full redundancy), and the TEM can communicate with

all GTRC and GTFE chips.• Leakage current and power are very good.

tracker design & assembly overview - stanford university

Documents