52
generation THz SLM control system has a few bad wire connections, which corre-
spond to the scattered broken pixels occuring at the same image locations on both
Figure 5.8(b) and 5.10(b). More importantly, some defects are observed under a mi-
croscope on both SLM 1 and SLM 2, which account for most of the broken pixels.
Figure 5.8(a) and 5.10(a) show the location and sketches of some cracks, scratches,
and fabrication problems on the SLM chips. The locations of defects match well with
the locations of broken pixels. For example, in SLM 1, a scratch along the right side
of the chip isolates most of the pixels in the triangular quadrant from their Schot-
tky pads, therefore cutting off these pixels from the voltage supplies. More detailed
microscope images reveal cracks in both SLM 1 and SLM 2 (see Figure 5.9(a) and
5.11) and fabrication defects in SLM 1 (see Figure 5.9(b)-(c)). During wire-bonding,
the SLM chip is glued onto the circuit board, which is then screwed tightly onto the
wire-bonding machine. The screwing force exerted on the sides of the SLM chip may
have buckled it.
For proper wire-bonding, the wire-bonding company needs to place the SLM board
at a fixed position on the wire-bonding machine at an exact height. During the first
attempt to wire-bond a practice chip to the PCB, we provide them a PCB without
cable connectors. When the company wire-bonds the two real SLM samples to the
PCB with cable connectors, they find a difference in height because of the solder joints
of the cables connectors at the bottom side of the PCB. While screwing the boards
down more to compensate for the height difference, the boards might have slightly
warped and exerted much pressure on the SLM chip, thus causing the cracks. To
overcome this problem in the future, they suggest putting the cable connectors after
wire-bonding, or cutting grooves into the workholder to accommodate the solder joints
at the bottom of the PCB, or grinding down the solder joints. However, soldering