Further refinement of the power subsystems are making life easier. I'm tweaking things as the BOM and initial part selection are finalized and slimmed down to reduce the number of unique components and that includes moving to a fixed output LDO for the 3.3V rail, larger copper flows for the 14V LDO and the 5V SMPS.

Previously I have been rating the subsystem to be able to operate at full load when the ambient air temp is 50˚C, aka a little more than 120˚F. This is a bare minimum because though I'm not in Arizona, here in SoCal it is common to get 100˚F days in summer and then some. Anyhow...

With the more regular polygon for the 3.3V rail I can estimate it's surface area and copper volume more accurately. The number comes out to ~13.25 mm^3 split about 45/55 between the back and front copper respectively. Looking up the thermal resistance of a SOT-223, assuming 65˚C/W for the aforementioned surface area seems reasonable. Well, anyhow, since I'm powering it from the 5V rail now, this gives me ~675 mA on the 3.3V rail at 50˚C w/ a T-sub-j of 125˚C.Thermal cutout limit is 150˚C which equates to ~900 mA @ 50˚C ambient temperature.

As for the 14V rail, I'm dealing with a double-edged sword. In order to function without the engine running, I need to boost the voltage input to at least 14.4 V going into the LDO. However, I do not have a way to selectively disable/bypass the charge pump doing this job so I'm stuck with a 18.25-20V input on the 14V LDO when the engine is running. That being said, the surface area I'm presented with for this rail is about 60% the suggested size and would change my performance to about 85˚C/W, which gives me a limit of about 160 mA @ 50˚C with the charge pump enabled and I only need 95 mA. If I get rid of the charge pump though, I'd be able to output ~1.2 amps and it's limit is 1.5 A.

This next bit was something I didn't anticipate. I had previously run numbers for the MPS2209 when it was the 3.3V rail and it was plenty capable of the output desired at the heat levels I needed. However, I forgot to do this when I bumped it to being the 5V rail. Compounding the problem, the 3.3V rail is powered from the 5V rail by an LDO for space reasons. Well, running the numbers puts me at an output current of 520 mA at 25˚C. Though I'm not sure how much the PIC32MX795 will draw when its fully operating but it's rated for 300 mA and I know I'll be utilizing a lot of its architecture. But added to that, the SRAM chip and it's latches are capable of a 280 mA draw.

I have two options if my current draw is too high:

• Switch to a SMPS chip that dissipates heat more efficiently
• Start looking at [SMT] heatsinks

As a patch, the latter is likely, though I'm thinking that I'll have to switch to a different SMPS. If I do that, I'm actually more likely to move the power subsystems off-board. D-DAQ is size sensitive and to keep costs "down" I rather shrink the mainboard. Though ultimately more expensive, moving the power subsystem off-board means that I can dedicate the proper amount of space to heat dissipation for the LDOs and the SMPS. Based on how much time I've spent going over surface area vs thermal dissipation, I estimate this board will be about 2/3rds the size of the mainboard. 

Though it's more complex and undesirable to layer two boards on top of each other, it'll simplify power routing on the main board a great deal. Also, two layer boards is more desirable that increasing from a 5cm x 10 cm footprint to a 10 cm x 10 cm footprint. In fact, I'd probably trim the mainboard's width down to about 7 or 8 cm. If I do move to a different SMPS chip, a cursory look on mouser show the MIC24052/24053/24054 or ZSPM4022 as viable alternatives. Each of these chips share identical pinouts and footprints and offer a 28˚C/W juntion-to-ambient rating over the 48˚C/W rating of the MPS2209.