Unexpected Failures of Modules on Rods Claudio Campagnari – UCSB for the US TOB group 5 December 2005

The problem FNAL 1 module w/saturated channels on a non-production DS rod 1 rod had three modules w/ saturated channels on a non-production rod 1 rod had a module w/ saturated channels on a non-production rod Groups (~3) of channels in middle of wafer Always on MRT after SRT test OK UCSB 1 rod with three modules with saturated channels and one module with high current 1 rod had a module with a saturated channel Single channels at the edge of wafers Always at the very first test (SRT) During tests of several tens of rods, last ~ 2 months

Damage on input to APV Most saturated channels show damage on APV input Silicon channel itself seems OK Works when bonded to different APV channel UCSB mod 6046 FNAL mod 9103

Corresponding Damage on Sensor Damage Location UCSB Module Damage is on n+ region, which is at HV, under bias bond, which is at ground Bias bond is lower than channel bonds Saturated channel next to sensor damage under bias bond

UCSB: more magnification, bias bond moved over to look under No visible damage to APV APV channel 1 dead Bias current not changed No visible damage to APV APV channel 768 dead Bias current not changed

FNAL Signal (Si-to-Si) bonds are lower than bias bonds See damage similar to UCSB under signal bonds corresponding to broken APV channels Bond shape asymmetric Top picture: 2nd sensor where bond lower Bottom picture: 1st sensor where bond higher perhaps passivation flaw (?)

Hypothesis HV breakdown between metal over n+ and bonds At UCSB, breakdown to bias bond, since it is lowest Then nearest channel "zapped" through capacitive coupling (?) At FNAL, breakdown to lowest signal bond Bonds to PA much higher

Obvious questions Why haven't we seen this before? Could we have missed it? Can we understand where HV breakdown happens? How high are these bonds? Why now? What to do?

Could we have missed it? We tested 1000's of modules on ARCS and in Vienna box We have never seen this before There is no way we missed it there But this problem showed up on rods Less rod experience, but we have been testing rods on-off for over a year Although emphasis had been on DAQ debug, we would have most likely noticed if it had been happening

How high are the bonds? UCSB measurements, including some FNAL modules at UCSB....FNAL measurements ~ agree TOB bond height over n+ region (not loop height)

TEC modules built at UCSB They appear to be a bit less vulnerable than TOB modules built at UCSB TEC R5 bias 152  150  TEC R6 bias 138  143  TEC R7 bias 141  new bias 247  289 

Where does breakdown occur? Increase HV up to 1000 V on UCSB TOB module. Ramp as fast as possible. No breakdown Plug and unplug HV wire at 1000 V Breakdown (but only on 5th try!) Intentionally lower the bias bond, ramp HV Breakdown at 40 height, HV=800 V

HV transients? Suspect that breakdown caused by HV transient on rod PS mixture of CAEN A132 and A332 These supplies go to 6kV All "accidents" happened on A132 This supply does not have a HV-max limiting screw Have tried to catch HV spike on PS No real success Have seen HV go from 0 to 80V when 1st establishing SW communication Have seen occasional crazy readings, up to 8kV on front panel and in SW readings In at least one case we were looking at HV when crazy reading occurred, and the DVM-monitor did not confirm

What to do? Plan agreed last week, may be revised this week Continue module production with modified bond parameters for maximum clearance above n+ region UCSB pulls all sensor-to-sensor bias bonds and remake them longer and much higher over n+ region. Rod production at UCSB suspended FNAL inspects sensor-to-sensor bias bonds and pulls if less than 200 m above the n+ region All rod test stand HV lines would be equipped with crowbars and current limits would be reduced. This has been the situation at FNAL for ~ 4 weeks, with no further accidents Rod production at FNAL continues

More Worries No sparks at 400 V  good Higher bond clearance gives more headroom  good Worry that electric field could still locally be high enough to cause ionization (corona, partial discharge) even if no sparks. What happens then over long period? We are seeing localized damage even in modules that did not spark

Damage under UCSB bias bonds After pulling bias bonds damage seen on all modules on second sensor At minimum, “dots” seen at the edge of the guard ring and n+ metal directly under bias bond wire At maximum, lines of metal (?) deposited in the area between the guard ring and n+ metal directly under wire Beginning of damage seen after 20 min on ARCS test Guard Ring Guard Ring n+ metal Guard Ring n+ metal

Damage under UCSB signal bonds Under all channel bonds, passitivation discolored at guard ring, bias ring, DC pad, and bond pad Channel bond location Channel bond location Guard Ring

Damage to Bias Wire Bonds Found bias return wire bonds pocked in region which would lie above outside edge of bias ring to halfway between guard ring and n+ metal