Three questions:
We have an existing product which is firmly pinned to the TPS54310 SWIFT converter, a device without internal compensation. We have recently experienced failures of this device in our system. Failed units simply fail to switch, the PH node is at a DC level (like 1/2 the target output voltage), vbias is present, SS has risen, but nothing seems to be trying. At this moment, my line is down because of this.
We performed step load testing on non-failed units, and determined the regulator was marginally stable in our design.
Would marginal stability (A solid three bumps in step response) possibly lead to premature regulator death?
We tried to make it better, but we are not experts in compensation network design:
We used both Switcher Pro and SLVA109 "Designing With the TPS54310 Synchronous Buck Regulator". Neither is perfect. SwitcherPro does let you change values and see some modification of the loop response plots, but tying those back to performance, appears to us at least, to be an imprecise exercise. SLVA109 provides an algorithm for selecting compensation components, but the reasoning behind it, and the tradeoffs made by algorithm are left to the reader. (At least it does provide some sense of dependencies).
In the end, using both sources as guidance, we played monkey; trying various things to 'improve' our design until we thought it performed better. We did achieve significant results with regard to the step response time, improving the stability, and the overall variance on the voltage rail during normal system operation.
Is there something else we should be checking?
Still we are not sure what happened with the devices that failed.
Can we return a failed device for FA?
Regards,
Hedley
