Marina Testing Status

So far we have had two batches of chips packaged.

First Batch

The first batch consisted of three dice with all pins bonded using the wedge bonding procedure.

When tested we found a VDD10-GND short observable from the package pins. Probing the die directly revealed no short when probing one VDD10 pin and one GND pin, but did (as expected) yield a short when both probes were attached to the same pin (confirming that the test equipment functioned correctly).

Second Batch

The second batch consisted of two dice with only the bare minimum nine pins bonded using the wedge bonding procedure.

We observed no unexpected VDD10-GND current, but did find quite a bit of VDD18-GND current; this is shown below on the left (CH2 should be 1.8V, but hit the current limit before it could be raised that high). On the right is a previous chip, Infinity, which used the same pad frame (and pcb), but was ball-bonded rather than wedge-bonded.

Under the microscope, we found a slightly off-center bond at one of the input pins, known as “master clear” (MC). This is shown on the left. On the right is another input pin, TMS, which was bonded with exceptional precision:

We then tried connecting only two pins: VDD18 was connected to the power supply, and MC was connected to ground. In this state (shown below on the left) we observed enough current to explain the short visible with all pins connected. On the right is the same configuration but with TMS connected to ground rather than MC; in this configuration there is no unexpected current.

The conclusion is that the wedge bonding procedure is simply not compatible with the seal ring that TSMC lays down on its shuttle chips.

Third Batch

The third batch works! All three dice powered up correctly and completed the test suite. Details to follow.

rendered from WIX