Sometimes you create beautiful projects, and they don't work. Then there are times you make something that has absolutely no right to work, but it does. This little project falls into the latter category.
I was experimenting with different power schemes for my Raspberry Pi to reduce its power consumption. I wanted to see how much less power was consumed if I powered the 3.3V rail with a switcher instead of the on-board linear regulator.
First I wanted to test whether it was possible to feed in 3.3V on the GPIO header with the linear regulator in place. I thought maybe the pass transistor in the regulator might just pinch off if there was a voltage on the output. This did not work at all, the Pi ended up in some hickup mode. Then I lifted a leg on the 3.3V regulator and tested the same thing, and this worked much better. Lifting the leg broke a corner of the chip's package though. When I soldered the leg back down, I found the Pi did not work anymore. I scoped the 3.3V rail and found it was not stable, the voltage would sag once the board tried to boot. Whether I killed the regulator electrically or mechanically I don't know, but it was dead.
So here I was with a dead Raspberry Pi. Any sane person would either have tossed the board and bought a new Pi, or replaced the regulator with an identical part, but I decided that was too easy. Since I had broken my board while testing what could be done to make it more efficient, I decided to replace the power wasting linear regulator with an efficient step down converter.
The problem is that a step down converter requires more external parts than a linear regulator, at the very least a fairly large inductor. I decided the best option was to build a little circuit board for my switcher that had the same footprint and roughly the same size as the original linear regulator chip. The result was decidedly not pretty.
As you can tell, I have an obsession with making things small, but this was pushing it. Part of the problem was that I only had access to small diameter lead-free solder, and that stuff is horrendous to work with. You need to apply way more heat and it still doesn't flow nearly as nicely as leaded solder. The bottom line is that in the process of cobbling this together, I cooked that poor little chip as I've never cooked one before, my PCB lost a pad, a corner of the chip broke off which caused a pin to become loose, and part of another pin broke clear off. I was fully prepared to throw the chip away and start over, but when I tested it by itself, it seemed to work!
So I put it on my Pi, and behold: it's back alive! My insanity paid off. :-)
In case anyone would like to be as insane as I am, here is the recipe:
- One FT441, I used the variable output version, but using the fixed 3.3V version should make things much simpler since you don't need to add the resistors and compensation capacitor.
- I followed the datasheet's schematic and used a 3.0uH inductor for L1, 10K for R1, 45.3K for R2, and 47pF for C2.
- The inductor I used is a Taiyo Yuden NR6028T3R0N.
- I did not add any input and output capacitors, but relied on those on the Pi. This seems to work!
Happy cobbling! :-)