Other Parts Discussed in Thread: LM25117, LM5116
Let me preface this with I'm more of an embedded systems guy then a power guy, but I know enough to be dangerous.
I've implemented a synchronous Zeta converter using the LM25117 for an automotive application. The design is based off the TI Design PMP9581_REVB. The Zeta is operating as a UPS and is powered off a bank of super capacitors. Rough Specs:
Vin: 1-12V
Vout: 5V @ <=5A & Ripple <= 50mV
Efficiency >= 85%
While the TI design uses the LM5116, I've elected to use the LM25117 because of its auto qualification. I've done a fair amount of simulations of the supply in LTSpice as well as built a bench version for real world evaluation and characterization. For the most part the supply works well, doesn't exhibit any subharmonic oscillation, and is close to meeting the ripple spec.
The problem comes when I go to load it down (in this case I've got it setup to draw 1A). Under some conditions the supply works fine and I get a nice clean 5V output, but at some operating points I will see the output voltage droop down by about 0.4-0.6V and never recover even after I unload it. To get the supply back up to a 5V output I have to completely remove power and restart the converter. Admittedly my spice model is lacking parasitics, but it doesn't display this droop phenomenon. Any ideas on what could cause this and how to fix it?
Here's a schematic of the test board as it sits now:
