Hello! I have been working on a design which uses 4 total AD7173-8 devices. Each AD7173-8 is implemented in the same manner, in 8 channel mode, set to differenial inputs on all channels, and with an external crystal oscillator and external 2v5 reference voltage. We are using each of them with a split supply (i.e. AVSS = -2.5V, AVDD1/AVDD2 = +2.5V). The design is quite mature at this point, and we have gone through thermal chamber testing on a few units and not noticed any issues with performance, and we evaluate each unit for a rather precise DC Accuracy specification.
However, as we have started rolling out more units, we have noticed that we seemingly have a pretty common infant mortality problem. We perform a "room temperature" burn in on each unit where we let it run for ~100 hours to age some of the components and catch issues with marginal components before this unit is being used in the field. The issue that we have noticed on 2/5 of the units in our last batch is that, before this burn in, all ADCs are nominally functional, and after this burn in, one or two channels on one ADC (but not the same ADC on each unit!) seem to be getting pulled to + or - 2.5V. We know that this is not an issue with our AFE, since the AFE is just a resistor network to scale down input signals, and all resistors in the network are reading a nominal resistance when we probed them with a multimeter. My best guess here is that the specific ADCs are being damaged in a manner that causes a short between the supply rails and the analog input channels that are exhibiting this behavior. As to what is causing the damage, I am much less sure, because we include TVS diodes on all analog input channels at the unit top level connectors, and also monitored temperature during this test to make sure nothing was getting too hot inside the unit. So far, out of 320 ADC channels across our 5 units, we have found 3 channels with this error so far, which is a rather high rate of failure for this application.
Has AD seen this issue of infant mortality causing just one or two channels to fail in this manner on the AD7173-8? Is there other data I can provide to narrow down the solution space here?
