Some here may remember my big RIS-767 project, the "highly-efficient low-voltage boost-converter testbed", with free testbed PCBs offered up for experimenting. I claimed we could create 98% efficient converters. The RIS-767 testbed allows creating custom converter circuits, as well as using the same platform to test the best commercial boost-converter ICs we could find.
Rob Legg has been investing his consider talents in using these testbed PCBs to test conventional boost converters and creating new ones that meet and exceed the 98% criteria. His work uncovered deficiencies in my PCB design, and even as he was spending untold hours at the bench, I've been spending untold hours taking his observations and advice to improve and fix the RIS-767 testbed. He's been working with rev 3, and meanwhile I'm well along on changes with rev 4.
But I digress. About small diodes. We're talking about voltages from 0.6 to 2.4 volts, and this means exclusively using a synchronous converter topology. Rob Legg discovered that the LT3400 and LT3402 boost- converter families require external Schottky diodes across the internal synchronous-rectifier P-channel MOSFETs, as suggested in their datasheets, so I'm cramming the parts onboard. Sorry, digressing again. Any parallel diode capacitance adds to the MOSFET switch Coss capacitance, creating P = C V^2 f loss. So we really need to use low-capacitance diodes.
Small low-capacitance diodes. Most diode footprints are rather large**, for example the common SOD-123 is substantially-bigger than the already-large 0805 footprint. In this context, the next-size smaller SOD-323 is welcome. Hence my enthusiasm for ST's '323 part, the bat54j*, compared to its larger -123 brother, the bat54z. Hmm, I previously encountered, with shock, that ST's bat46z was OK, but the smaller bat46j had disappeared from distributors. (The scene might be different for an alternate manufacturer, like Diodes, Inc., etc., but not investigated.)
- The bat54 has about 6pF of capacitance at 2 volts, and its loss is about 8.5uW at 2.4 volts and 250kHz, which is an acceptable 0.036% compared to 24mW.***
One suggestion for anybody responding to this long multi-topic post, edit out my irrelevant blah blah before hitting send. :-)