While the analysis can become fairly complicated, the theory is very simple. The idea is to probe, using a mechanical microprobe in this case, starting from the failing output node and work your way backward into the IC until the node in the wrong state is found. With the complexity of modern ICs, this may quickly become an impossible task for failing nodes that are not in very close proximity (electrically) to the failing output. Multiple levels of interconnects may also limit access to many ideal test locations without the use of FIB deposited probe pads.
For cases where this is practical, there are significant benefits. Most important is the ability of the mechanical probe to not only measure signals on a given node, but also to inject signals into the next gate downstream. This is only possible with direct contact probing. While voltage contrast can analyze signals on nodes, it cannot force a node into a given state.
Mechanical probe signal tracing is fairly straightforward to do. Once the IC has been characterized electrically and the packaging material has been removed, the signal should be traced from the failing node back into the IC. At some point, gate causing the incorrect output can be found. While this sounds simple, with the complexity of modern ICs and the high fan outs found on many nodes, this process can get very tedious very quickly. Other non-contact methods, such as voltage contrast, are better suited to handle complicated ICs, but if they are not available, mechanical signal tracing is a very inexpensive alternative.
The first problem to overcome is the passivation on the IC. Contact probing implies that provisions are made for the probe tip to contact a metal or poly layer. There are several ways the passivation can be prepared for probing. First, and easiest is to completely remove all of the passivation from the IC. This process does leave the metal unprotected, so probing must be done very carefully to avoid scratching the surface and ruining the sample. Several methods are available to selectively remove passivation to make probe windows. Focused ion beam systems can make precise, repeatable windows. Lithographic techniques have been successfully used for may years in failure analysis to selectively remove passivation. If the passivation is silicon nitride, or sometimes silicon oxynitride, a xenon laser cutter can be used to cut holes in the passivation without damaging the metal underneath.
After the IC is placed in the failing electrical state and contact can be made to the metal or poly layer of interest, the rest is easy. Start from the failing output and trace the signals backward until a node is found in the wrong state. A common way to determine if a node is in the wrong state is to probe the same node on a comparison IC that is in the same electrical state. If possible, placing the two ICs side by side under the same probe station will help facilitate this process. Care must be taken at all times not to damage the IC with the probes.
The most cumbersome part of this process is dealing with complicated branching that takes place on many circuits. Remember that these ICs were not laid out for ease of signal tracing for failure analysis.
Mechanical probe signal tracing is usually done under two circumstances. First if the laboratory does not have other equipment such as electron beam probers or other non-destructive evaluation methods. Mechanical probing may be the only method available to trace the failure down to the failing gate. If a non-destructive method has been used to locate the failing node, mechanical probing can be used to further characterize the defect, i.e. is it a gate rupture?, is it a bridging metal short?, etc. Voltage contrast can be used to monitor the signal propagation, but cannot easily measure the electrical characteristics of the defect, this is a result of the electron beam's inability to force a signal upon a node. The e-beam can only read voltages. Other techniques, such as resistive contrast imaging (RCI), can image the resistance of a given node to the rest of the circuit.