Report on TAB Bonded 256 Channel Pad Sensor

Report on TAB Bonded 256 channel Pad Sensor

A 1 mm thick pad sensor, designed for the Compton prostate Probe, with 256 pads of 1.4 x 1.4 mm2 area has been fully assembled in Kharkhov and brought to CERN about a month ago.

This module is constructed in the following way (Fig. 1):

Fig. 1: Schematic drawing of TAB bond cables used for module assembly.

There are 3 separate cables with thick Aluminum traces allowing to TAB bond the big cable to the 256 pad silicon sensor. The chip cables are TAB bonded to the input pad rows on each VATA-GP3 chip. Finally the chip cables are TAB bonded to the end of the detector cable.

Figs 2 and 3 illustrate the challenge. The pictures were taken with a micro cable detector assembly made previously. No chip connections were made in this prototype run..

The assembly with all 3 cables and the 2 bonded chips (unfortunately we did not test the 2 chips before sending to Kharkhov) was then mounted to a Compton probe PCB designed for the Compton probe prototypes by Enrico using a bare PCB as mechanical support.

This assembly was then connected to the VME read-out (Enrico version).

Results from these tests are shown below.

Both chips worked electrically with one small problem: the register could not be read back from chip number 2. Thus we could not be sure that the register was loaded on this chip properly. This maybe the cause that we can read in sparse mode only chip2. However since we tested the module first with only chip 1 connected and chip 2 bridged we had chip 1 working perfectly well in sparse mode.

While all channels on both chips work electronically but quite a big number of channels do not respond to the Am source we believe that not all the TAB bonds make good connections. On chip 1 we observe around 80% of all channels with good TAB connections but on chip 2 there are only ~65% with good bond connections. We suspect that the faults occur in connecting the chip cable to the inputs of the chips (see fig. 3)

Fig. 2: Micrograph of the bond connections on the detector

Fig. 3: Layout of aluminum traces in the 50 micron pitch region. The openings in the polyimide allow the connection of the traces to the underlying pads on the chip input side

The next figures show the results obtained with different gamma and X-ray sources.

In general noise and amplitude of signals are very good for all working channels.

Fig. 4: Pedestals and sigmas of pedestals for first chip

Fig. 5: Hit map with Am source placed in the middle of detector

Fig. 6: Hit map with source on top of area covered by second chip

Fig.7: Mean ADC for Am source chip 1

Fig 8: Mean ADC for Am source 2nd chip

Fig. 9: X-ray spectrum of Mo (17.5 keV) in sparse readout mode of channel 205

Fig.11: X-ray spectrum of Mo in serial readout mode of channel 205

Fig. 12: X-ray spectrum of Tb in sparse readout mode, channel 205

Fig. 13: X-ray spectrum with Ag source, channel 205, sparse readour

Fig.14: -ray spectrum with Am source, channel 205, serial readout

Fig. 15: Spectrum Am source sparse readout, another channel

Fig. 16: Energy calibration using 5 different sources, red is sparse readout, black is serial readout

The slope of these curves gives an average calibration of 76 e-/ADC count.

If we take the Mo spectrum sigma from a Gaussian fit the noise of the system is ~160 e- ENC, after subtraction of ionization fluctuation.

Conclusion: the TAB bonded detector module works extremely well with noise performance very comparable to a very good Compton Probe Demonstrator module assembled in conventional wire bond technology.

The only problem is the reliability of the TAB bonds, most likely on the chip cable ( to be investigated)