Transcript Jet/Energy Module Prototype Tests and Test Vectors

ATLAS Tracker Upgrade:
Hybrid Production and
Tests at Birmingham
John Wilson, Simon Pyatt,
Sam Edwards, Juergen Thomas
(Birmingham)
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Hybrid Production and Test Setup at Birmingham
1
Outline




Work in the Clean-Room (Simon):
 Hybrid Production
 Tests for Heated Curing
Measurements of Hybrids (Sam)
Work on the Test System (Juergen):
 Test Results from Hybrids
produced at B’ham
Acknowledgement: We have a lot of support, both h/w and advice, from UK
Tracker Upgrade collaboration, namely RAL, Liverpool, Cambridge, Glasgow
(Peter Phillips, Bruce Gallop, Tony Affolder, Ashley Greenall, Bart Hommels,
Andrew Blue etc.), and also from the B’ham Astro group.
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Hybrid Production and Test Setup at Birmingham
2
Hybrid Panel heated plate study (Simon)






Current procedure: Curing of glue at room temperature, for 24h,
during ASIC placement session. Only two hybrids at a time (tools
don’t fix when trying two neighbour hybrids). Information from
glue data sheet: ‘Cure Schedule: 24h at 25C, 1-4h at 65C’
Study for heated curing: Can’t use oven due to necessity of
constant vacuum connection. Instead use: Flat heating element.
First test: Test samples of glue in oven: ‘Hard shell’ forming
(pokeable) after 2 hours at 30C,40C,50C. OK, would that work ?
Flat heating element type: “Watlow Silicon Heater 060100C2A7B1B”, purchased one for 63GBP. Sandwiched between metal
block, underneath vacuum panel.
First glue Glass ASICs, then real ABC-N250 ASICs.
Note: The panels used for this are ‘Rejects’, some have been in
drawers for years, poor coating and such.
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Hybrid Production and Test Setup at Birmingham
3
Hybrid panel heated plate study (2)
Based on oven tests a target temperature above 30C is needed.
Thermocouple on the heater and another one on the panel surface (On an ASIC pad)
is connected to a temperature measurement unit.
A dual power supply is connected and turned up to the maximum of 70 volts (370mA
drawn)
After over 2 hours the temperature settles at 34.4C,
not as fast as hoped but it got there. Time to glue one glass ASIC hybrid.
Bare stencil used to put an additional glue
sample directly on a hybrid as a
“Prod and stir” sample
Hybrid panel heated plate study (3)
Ok, some real tests now. First test, temperature stability, how much control is needed.
A proper temperature monitoring system is now used, a HP Data logger connected to a PC
to capture the data. One day short test and one three day long test.
(Windows 95 PC!)
Hybrid panel heated plate study (4)
Third test. Produce samples for full testing. Real ASICs.
Three heated hybrid cures done in a single session. Three room temperature hybrid cures
for comparison (Done later). A timer is added to start the heater early in the morning so the
setup is ready to work on at 10am. Timer is from RS components, stock number 329-676
£10.69. Timer is only to turn heater on, heater is turned off manually.
Hybrid panel heated plate study (5)
Procedure goes smoothly.
Hybrid Panel heated plate study (Simon)
Tests with Glass ASICs were promising
But big problem with panel with real ASICs:
Six hybrids in total glued and cured.

On the three hybrids which underwent ‘heated curing’,
ASICs are fixed nicely, ie can only be removed using
some force.

BUT the ASICs on three hybrids from ‘room
temperature curing’ did fall off or could be removed
easily !
Proper investigation ongoing…

Microscope: Glue barely touched ASIC (no flat top)
-> Measure all items used (Sam). Some gaps ?

Influence of ‘reject’ hybrids: Poor surface, bent ?

Did the heated curing help to prevent the chips falling
off the panel ?



21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Hybrid Production and Test Setup at Birmingham
8
Measurements of the Hybrids (Sam)






Sam is using the ‘Coordinate Measurement Machine’ (CMM) owned
by the Astro group to precisely measure the hybrids.
Precise to 5μm.
Initial challenge: Find a very good reference point. Used mainly ‘Gold
pads’ for ‘touch-down pins’ on hybrid. All measurements relative to
this reference (due to eg dust on surface under panel).
All measurements made using vacuum (hybrid held flat on panel).
Rather big variations from one chip to another on a hybrid (~40μm),
chips also slightly ‘tilted’.
Investigating the tools used for assembly (which hold the chips in
place during curing: ‘Pick-up Tool’). They may not be completely flat
or wear-out over time (needs Quality Assurance (QA) procedure ?).
We have two such assembly tools, checking consistency between
chip height (ie glue thickness) and tool dimensions/flatness.
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Hybrid Production and Test Setup at Birmingham
9
Measurements: Inaccuracy in Pick-Up Tool (Sam)
Height of Pick-Up Pad from the Touch
Down Pin level (microns)
Pick-Up Tool Profile
440
430
420
410
Top Row Hybrid - Far Side
400
Bottom Row Hybrid - Near Side
390
380
Pick-Up ‘Pad’
370
360
1
Measurement
shown on
graph is this
gap.
Touch
down
pin
Variation between
highest and lowest
‘pad’ is 43 Microns
2
3
4
5
6
7
8
9
10
Variation in Panel Surface Height (Sam)
Panel surface measured from the
gold disc’s used for the Pick-Up
tool touch down pins
Variation of Surface Height Below ASIC's
10
Height in Mircons
5
0
1
-5
-10
2
3
4
5
6
7
8
9
10
Bottom Row
Top Row
-15
-20
-25
Measurements taken on the ‘pads’ below where an
ASIC would be glued. There is a variation of 25 Microns
between the high and low areas.
Measurements: First Results (Sam)

The highest ‘pad’ on the pick up tool is 429.25 Microns

The lowest ‘pad’ on the panel is -19.25 Microns

This could lead to a possible total gap of 448.5 Microns

The thickness of an ASIC is 325 Microns

This means a glue thickness of 123.5 Microns would be
required for even partial contact with the ‘pad’ on the
panel. Larger thickness of glue required to get a good
amount of contact and therefore a strong bond.
ASIC Tests (Juergen):
ABC-N Functionality in a Nutshell
Tasks of the ABC-N chip (from specifications document):
“The chip must provide all functions required for processing of signal from
128 strips of a silicon strip detector in the ATLAS experiment employing the
binary readout architecture.” (ie: Output is a yes/no for each strip)
1.
2.
3.
4.
5.
6.
7.
8.
Charge integration
Pulse shaping
Amplitude discrimination.
Threshold value for the amplitude discrimination is provided as differential
voltage either from internal programmable DAC or from external source.
Outputs of discriminators must be latched either in edge sensing mode or
in the level sensing mode.
At the start of each clock cycle the chip must sample the outputs from the discriminators and
store these values in a pipeline until a decision can be made whether to keep the data.
Upon receipt of a L1 Trigger signal the corresponding set of values together with it's neighbours
are to be copied into the readout buffer serving as a derandomizing buffer.
The data written into the readout buffer is to be compressed before
being transmitted off the chip.
Transmission of data from the chip will be by means of token passing and
must be compatible with the ATLAS protocol.
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
13
Test Setup Status: HSIO



Reminder: HSIO (High-Speed Input Output) is Xilinx-4-FPGA-based board
designed and produced by SLAC, it is the standard for testing Tracker ASICs,
Hybrids and Modules. Firmware developed by UCL. Controlled by SCTDAQ
s/w package (main developers: RAL), from Win7 PC.
HSIO ‘mimics’ real control systems and can capture and store output data
from the ABC-N chips. Connection between HSIO and ABC-N chip is via a
very small number of LVDS channels.
Initially connected a Hybrid from Liverpool: Connection goes in stages:

When starting SCTDAQ, S/W will report if hybrid responds

Connection test: ‘st_nmask’ (command line) produced ‘saw tooth’
triangle pattern, one triangle/tooth for each chip connected

Then can move to the serious tests (started from GUI):
StrobeDelay and 3PointGain

SCTDAQ writes output into histograms (ps file),
Root files and logs.
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
14
Test system with Hybrid (1)
News from the test
system, in Poynting
Basement (PB8):
New power supply
(CPX400DP):
‘Remote controllable’
via USB from
SCTDAQ S/W
(this still needs to be
set-up)
… and: The Unit
Under Test is now a
Birminghamproduced Hybrid !
(This panel, our
label ‘B2’: 5 hybrids
populated)
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
15
Test system with
Hybrid (2)
Overview of the setup (Win7
PC not on picture):
- Bottom: HSIO and Interface
Board.
- Top-left: Hybrid, small vacuum
pump for cooling (airflow) and
holding hybrid flat on panel
- Top-right: New power supply,
and Oscilloscope showing HSIO
LEMO outputs (after LVDS-toTTL conversion in FPGA).
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
16
Test system with Hybrid (3)
Picture here: Liverpool Panel and
Hybrid (‘Chip 6 not working’)
Left-hand connector: Datalines
(8 LVDS channels).
Extra power input for LVDS
drivers (3.3V, 0.26A)
(not always used)
1 Aug 2013
Right-hand connector:
Power input 3.3V (drawing
4.4-4.8A). Serial powering.
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
17
Powering-up B’ham made hybrids






Simon has produced (place&glue, cure and wire-bond)
10 hybrids with ABC-N 250 ASICs, on two panels (‘B1’, ‘B2’).
B2 has generally ‘best bonds’ (last ones made, tricks learned).
Those two panels have passive components and connectors
already in place (panels eg used in curing tests don’t).
Initial power-up: Each hybrid drew more than 5A (at 3.3V) and
one or even two wires started to glow and burn-through. Power
consumption dropped by some 0.5A after that.
BUT afterwards, they worked just as well as the Liverpool hybrid !
Investigation showed our wire bonding program has one wire
differently than the Liverpool hybrid. We assume this caused a
short and the wire burned through (ie ‘self-cured’).
As the ABC-N250 ASIC and hybrid is now quite an old system, we
won’t make a fuss about this. But need to be sure to carefully
match and update wire bonding programs from now on.
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Hybrid Production and Test Setup at Birmingham
18
Hybrid power-up: ‘st_nmask’ test
- SCTDAQ s/w setup modified to
run in ‘Hybrid’ mode (before:
Single-Chip testboard mode)
- ‘st_nmask’ is an important test
to see if chips are responding to
commands from the test system
-One fully black triangle for each
chip on the Hybrid: All 20 chips
responding fine.
-BUT big difference on how
well/fast this works depending on
supply voltage: No communication
with Vcc=2.5V, very slow at 3V,
better (faster) at 3.3V, best at
3.55V. With or without driver.
Not sure I understand serial
powering well.
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
19
Hybrid Test: StrobeDelay25
The calibration strobe sets the
timing (delay) of the injected
calibration pulse with respect to the
arrival time of the command to
actually issue that pulse.
This ensures that the discriminators,
always firing at the clock frequency,
will be synchronous with the
calibration signal.
Scanning is done using a 6bit
register (ie for 50ns clk: ~1ns steps)
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
20
Hybrid Test: 3PointGain1pC
Threshold Scan:
The output of a channel for
a given input of charge is
measured systematically as
the threshold is varied.
The channel's response to
an injected charge is to
record a "1" (hit) or a "0"
(no hit).
3PointGain:
These tests are the
threshold scan with 3
different levels of injected
charge.
https://znwiki3.ifh.de/
ATLAS/Projects/SCT/
SCTUpgrade/DAQ/SCTDAQ
21 Nov 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
21
B2
Panel Number:
206
Internal B'ham Label:
Overview: Testing the two Birmingham Panels
Date of test:
B2 (as on the sticker)
24/10/2013
Hybrid #: Chip type Power-Up consumption [mA] Test consumption [mA]
0Glass
n/a
n/a
Reponding to s/w requests
n/a
1ABCN250
4760
2Glass
n/a
n/a
3None
n/a
n/a
4990 (draws 67mA when PS
4ABCN250 ch is 'Off' !)
4760Y (very slow)
n/a
n/a
5ABCN250
4300
4500Y
6ABCN250
4810
Y
7ABCN250
40204800-5500
Y
5120Y
Tester:
Test system sequence:
JPT
.x Stavelet.cpp / .x NoBCCSetup.cpp / st_nmask()
Saw tooth pattern available
n/a
Comments
Glass ASICs
Very slow reaction to s/w access, but no saw tooth
pattern. Same with LVDSDriver/3.55V. Info from
Simon: 1 chip not bonded (bonding problems). No
signal lines on one row. Not expected to work
N
properly.
n/a
Glass ASICs
n/a
Not populated (bad hybrid)
Slow reaction to s/w access. 2 Chips not
Y (18 chips, slow, 3.55: fast) responding. 3.55V: Fast reaction !
LVDSDriver: Fast reaction to s/w access, 7 chips
respond in tooth pattern. Hybrid looks good ! But
wires which should have burned. haven't. To do:
Y (7 chips LVDSDriver)
Power-up again and try to burn wire…
Only 16 out of 20 chips show in saw pattern
Y (16 chips, LVDSDriver)
(st_nmask). LVDSDriver: Fast access !
Working. SD25/40 and 3PG1/2 histos look good.
Y (all chips)
Info from Simon: Bonding etc all fine.
n/a = not applicable
Panel Number:
202
Internal B'ham Label:
B1 (as on sticker)
B1
Date of test:
Tester:
.x Stavelet.cpp / .x NoBCCSetup.cpp / st_nmask()
Hybrid #: Chip type Power-Up consumption [mA] Test consumption [mA]
Reponding to s/w requests
Saw tooth pattern ok
Y
N
Y
n/a
n/a
n/a
N
n/a
n/a
n/a
0ABCN250
5300
1ABCN250
2Glass
n/a
3Glass
n/a
4Glass
n/a
4000
5ABCN250
1900
N
N
6ABCN250
4100
Y
Y (all chips)
7ABCN250
2700
N
N
21 Nov 2013
Test system sequence:
n/a
n/a
n/a
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
Comments
Bonding problems. Not expected to work well. Test:
'Saw pattern' very slow, no use for further tests.
Bonding problems. Not expected to work well. Test:
'Saw pattern' very slow, no use for further tests.
Glass ASICs
Glass ASICs
Glass ASICs
No data connection. No further test. Info from
Simon: Hybrid to Panel bonds failed, bend hybrid.
Bond glowed away after 2min. Saw tooth pattern
ok. 2 chips noisy ? Info from Simon: Bonds not
great, somewhat surprised it works.
No data connection. No further test. Info from
Simon: Hybrid to Panel bonds failed, bend hybrid.
22
Next Steps





Simon’s ideas for curing process optimisations under discussion with UK
colleagues. Simon and Sam investigate carefully ‘Chips falling off’ problem.
Continuing with further hybrid tests (more systematic).
Discuss results with experts (at RAL, Liverpool).
Need to fix our test system (Why do we need 3.3V or even 3.55V hybrid
supply to ‘talk to’ chips ?) BUT don’t change a working system…
John and Juergen currently running a Yr4 project on hybrid tests,
ie practical part of that project to be done early ’14.
First ‘data taking’ next week. One task: Compare test results in Root from
our two fully working hybrids with Liverpool’s.
Next Generation ABC-N chip on its way (wafers being diced currently):
ABC-N 130. In Birmingham early next year ?

Many changes: Hybrid only with 10 ASICs, assembly tools changing,
ie some plastic becoming metal.

UK colleagues are designing new testboards, hybrids, s/w interface.

ASIC runs at 1.2V, has 256 channels. Also a new HSIO (Artix-7 based).
1 Aug 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
23
Back-up Slides
1 Aug 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
24
Hybrid panel heated plate study
A glass ASIC hybrid is placed. A “Stir and prod” sample is placed which is then
checked periodically. Sample forms a hard shell after two hours as seen on previous
tests. Assembly is left for a few more hours, ASIC jig is removed, heater turned off
and left to cure overnight.
Test system with Hybrid (1)
HSIO at Poynting
basement
Top-right: HSIO
and Interface
Board.
Bottom-left:
One-hybridpanel from
Liverpool.
Flat-ribbon
cable carries
‘Clk, Control,
Gnd and
Dataout’ signals
(4 LVDS-pairs)
1 Aug 2013
Juergen Thomas:
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
26
Hybrid panel heated plate study
First test is to determine curing rates,
this is done with test samples of glue
on glass slides in an oven.
Several tests are done at 30oC, 40oC
and 50oC. Samples are prodded every
half an hour.
All temperatures give the same result,
a hard shell forms on the glue after 2 hours.
The hard shell is self supporting but not completely
cured, it can be dented with a sharp point indicating
that the glue interior is not completely cured.
But could this be good enough to support the ASICs
without the ASIC vacuum jig once it has reached
this state?
Next stage is to design a system that can be usable.
Hybrid panel heated plate study
Currently the curing of the ASIC glue for the hybrid panels are done at room temperature.
This is a study to determine if the panel can be heated to reduce the current cure time of
24 hours per ASIC placement session to speed up production.
Extract from TRA-DUCT 2902 datasheet:
So yes, the glue can be heat cured...
However the panel cannot be placed in an oven because of the vacuum hoses,
so an alternative needs to be found.
A proposed alternative is to use a heating plate that can be placed under the
current curing setup. Flat heating elements are available.
Hybrid panel heated plate study
Proposed hybrid panel heating system is as follows:
Panel vacuum jig
Hybrid panel
Metal block
Heating element
Insulating sheet of glass (Temporary, will get something better later)
The heating element is stuck (Pressure sensitive adhesive) to the metal block,
then this just needs placing under the existing setup.
Panel vacuum jig is placed on the metal block without any fixings or screws.
No modifications are needed to the existing mechanics.
Hybrid panel heated plate study
After some searching we found two companies who sell flat heating plates,
Minco (Based in France) and Watlow (Based in Nottingham):
Minco:
HK5529R101L12B
Unit price : 185.00€
Minimum Order Quantity : 25 pieces
Lead time : 7 to 9 weeks from order receipt.
4,625 Euros to buy a heater! Awful customer service, two months to get a quote.
Watlow:
Helpful on the phone, willing to sell a single item. £63.20, bargain!
Hybrid panel heated plate study
Heating element is self adhesive, quick to fit on the heating block.
A “K” type thermocouple is embedded into the heater.
Lets try it out...
Hybrid panel heated plate study
Results look good, dots are well defined, glass ASICs have not moved position.
The glue is completely cured on the sample.
Hybrid panel heated plate study
3 day test. Temperature is very stable to within a degree, lucky.
The procedure should work without requiring a controller.
Hybrid panel heated plate study
Second test. 2 Hybrid glass ASIC cure test. How stable is the temperature while the panel
is being worked on. How smooth is the procedure to change over the ASIC jig.
Hybrid “Stir and prod”
sample
Room temperature
comparison
sample
Procedure works well.
Hybrid glue sample cures as predicted.
glass slide sample still stirrable long after procedure is finished.
Hybrid panel heated plate study
There is no significant change in the
temperature during the procedure.
Swopping the ASIC jig over takes 5 minutes.
finished hybrids look good, including
the positions of the ASICs.
Hybrid panel heated plate study
Disaster!
This ASIC fell off.
Something is wrong with this panel.
Height distance between jig pads and ASIC pads too big.
This ASIC came
off easily.
This ASIC was
fairly firm.
Hybrid panel heated plate study
Conclusions (So far):
•Procedure requires some organisation and timing.
However there is a two hour gap between placements and the final jig placement can be left
overnight with the heater on.
•Total expenditure is a £63 heater and a £10 timer. The metal plate and base are easy to make.
The power supply type is already being used on the project.
•The heating system can be added to the current setup without any additional modifications.
•Positioning (Using visual inspection) of ASICs is the same as room temperature cured ASICs.
•CMM results need checking because of problems with last panel.
Next steps:
•Shear testing in Berlin (We are already talking to this group).
•A complete 8 hybrid gluing session (We have run out of panels and ASICs).
•Other groups trying out the procedure.
Next Steps (OLD)






Simon is soon moving to gluing and curing real ASICs – we have a lot of
‘old’ prototypes, ABC-N 250, cut as dies (from Glasgow)
Continuing with test system setup for hybrid, esp run more complex tests,
called ‘StrobeDelay’ and ‘3PointGain’. Discuss results with experts (at RAL,
Liverpool)
Probably connect and test one of the hybrids we’ve assembled in B’ham
Further production preparations on-going:

New power-supply units purchased by the UK groups (remotecontrollable: TTI CPX400DP), B’ham will receive one, needs to be
integrated into HSIO test system, s/w can then control voltage/current.

Simon’s ideas for curing process optimisations under discussion with UK
colleagues
John and Juergen are offering a Yr4 project on hybrid tests,
ie practical part of that project to start January ’14
Next Generation ABC-N chip on its way: ABC-N 130. In Birmingham about
end of year ? UK colleagues are designing new testboards, hybrids, s/w
interface.
Juergen Thomas:
1 Aug 2013
ATLAS Tracker Upgrade: Module ASIC Gluing and Test Setup
38