How to find something that isn’t there

    Xray_700x350px

    Fast track to failure detection

    Sometimes things do not work out as planned. At DELTA we usually allocate a fire-brigade team among our designers, test engineers and FA specialists to investigate the real cause of chip failure. Such a multidisciplinary team often leads to fast failure detection.

    A case study on failure analysis

    This case study is about an ASIC design DELTA made for a customer. Following the normal procedure, it was taped out to a foundry and the wafers were processed and shipped back to us. When we tested the chips, neither functionality nor performance met our expectations in any way.

    Time-to-market is important and we usually get it right the first time – our motto is after all “First Time Right”. So we were very surprised to receive wafer material from a skilled foundry which did not live up to our design specification.

    DELTA’s “fire-brigade” team in action

    First we checked if we had handled the foundry tape-in procedure correctly regarding Design Rule Check and Layout Versus Schematic check, related foundry documentation, etc. Verified! This means that all development QA-gates were green-lighted from specification phase over design to layout and the further DRC/LVS-check.

    However, the chips on the wafers did not function. Consequently, we set up our failure analysis task team who – step-by-step – evaluated the measured and observed information. Based on this, they conducted further investigations in order to come to a conclusion.

    Step 1 – Electrical characterisation

    First step in a failure investigation is to perform an electrical characterisation of the device and then additional electrical measurements.

    For simplicity the circuit can be shown as a diode array with capacitors.

    FIG. 1. The circuit between RFIN1 and RFIN2. Modulator and demodulator circuits are not shown for the sake of simplicity.

    FIG. 1. The circuit between RFIN1 and RFIN2. Modulator and demodulator circuits are not shown for the sake of simplicity.

    Step 2 – Electrical measurements

    Next step was electrical measurements. The current-voltage curves (DC-VI) were measured across the CMOS circuit. Based on the knowledge of the circuit, we had expected a linear curve as the circuit should act as a resistor at different levels – presumably in the range of 100 kΩ. Instead, the chip acted as a diode (see the measurements in figure 2).

    FIG. 2 Obtained DC-VI-curves.

    FIG. 2 Obtained DC-VI-curves.

    Step 3 – Physical analysis

    Based on these measurements and observations, our ASIC design team discussed root causes and explanations. One particular question popped up: Why did the circuit behave like a diode instead of a 100 KW resistor?

    They got the idea to simulate the circuit as if the MIM (metal–insulator–metal) capacitors were short-circuited. After this simulation it was clear that one of the root causes probably was that the MIM capacitors actually had been short-circuited.

    FIG. 3 Optical micrograph showing part of the MIM capacitors.

    FIG. 3 Optical micrograph showing part of the MIM capacitors.

    The next step was to perform a microsectioning of the chip and then examining it in our Scanning Electron Microscope (SEM). See the result figure 4.

    FIG. 4 SEM micrographs showing a cross-section of the MIM capacitor.

    FIG. 4 SEM micrographs showing a cross-section of the MIM capacitor.

    In figure 4, the SEM picture shows a cross section of the MIM capacitor. On the upper side picture of the centre capacitor, you can see that the metal layer is shifted to the right. On the lower picture, you see an enlargement of the MIM capacitor terminals. Here you can see that one terminal is in direct contact with Metal 3 and therefore short-circuited the MIM capacitors.

    The conclusion

    The conclusion is that even though all design rules and QA procedures were followed and we taped-in to a foundry we have a long-term partnership with, it is inevitable from time to time come across failures – also failures initiated by professionals. In this case, the foundry’s quality system green-lighted the shipment even though the wafers were not properly processed.

    Therefore, we believe that it is important to have internal capability to detect and document failures quickly and to suggest and communicate corrective actions.

    This case study is simplified as an appetizer on how to trouble-shoot problems quickly and save time. You can read the whole story in detail here.

    If you are hungry for more, you are welcome to visit our website and read other case studies on for instance how to inspect failure modes non-destructively by use of X-ray.

    This is a unique website which will require a more modern browser to work! Please upgrade today!