Case Study: Data Channel Amp
A mixed-signal design team at a large IDM redesigned a signal amplifier IC for an electromechanical device. The designers retargeted the chip to run faster on a smaller process node. The new chip was no longer housed in a package; instead, the bare die was solder bumped and attached directly to the device in a mobile form factor. Because the previous version had been in production for some time, minimal changes were made to the functional design of the new version, and the team was confident that the redesign would go smoothly.
Worst Case Scenario: Field Failure
However, after the product was deployed, consumers reported field failures. Once the design team isolated the failure to the bare-die amplifier IC, they began emergency debug procedures.
The failure was isolated to a data channel that exceeded the maximum temperature specs. However, the design team was unable to reproduce this failure in the lab. The operational modes of the bare-die chip were the same as the packaged chip, which worked fine in high volume production. What was causing the bare-die version to fail?
The team suspected that the problem was related to temperature, so they retained the Gradient Virtual Thermography team to perform thermal simulations using HeatWave on the new amplifier IC. Together, Gradient and the design team gathered the required layout, thermal model, and die material data in order to run the thermal simulation.
Pinpointing the Failure Mode
Gradient’s electro-thermal simulation pinpointed the failure mode within hours. On sustained data transfer through a single amplifier channel, a group of matched transistors were subjected to a ΔT of over 7°C. This caused the amplifier bias currents to be mismatched, which caused the signal path to malfunction. The new circuit simulation, with device temperatures annotated, confirmed the malfunctioning behavior.
The ΔT in one matched group among many matched groups is easily overlooked. The designers did not anticipate that a sustained burst on one channel would result in such a high ΔT on this particular group of matched transistors. This failure mode was also masked in the prototype testing because the die was attached to a heat spreader bar in a package, resulting in a much smaller ΔT across the matched transistors.
The team realized that if they had performed the thermal simulation before fabricating the bare-die version, they would have caught the problem and prevented the expensive field failure.