Author Topic: Advice / Feedback on buck converter /PSU layout  (Read 81 times)

0 Members and 1 Guest are viewing this topic.

Offline Kryptychon

  • Contributor
  • Posts: 5
  • Country: de
Advice / Feedback on buck converter /PSU layout
« on: July 20, 2021, 03:30:11 pm »
Hi there,

I just finished my very first layout. I appreciate any feedback and/or advice.

Short description of the project:

I want to convert the voltage of a LiFePO4 battery (8S, ~30V) to 9V (for further usage on a MCU board) and 12V to supply other features (like control panel and sensor boards and so on).
As this will be the PSU for a small EV, I opted for 12V and chose to generate 5V or 3,3V locally if needed. These requirements lead to the shown design: Two buck converters, one feeding a 9V LDO regulator and one feeding two 12V LDO regulators. Additionally there is a simple comparator circuit to sense the temperature on the heatsinks.

This is a amateur project, cost or space optimization is not an issue.

Things I'm aware of:

- The reverse polarity / over-voltage protection circuitry can be considered a little archaic. The main supply will always provide enough current to melt the fuse, eventually ending the reverse polarity or over-voltage condition in any case. I know that there will be ~2V reverse voltage in such a case. It looks tolerable to me.
- The bypass / filter caps for the LDO voltage regulators might be omitted as they're placed "near" the PSU filter.
- The free-wheeling schottky diode might be a little oversized, I opted for low forward voltage, thus I chose this generously sized part.
- Heatsinks that large migh not be necessary to meet thermal specs, but (hate me for that if you like) they look cool :-)
- I only need three temperature sensors but I decided to implement all four channels of the LM339 for temperature sensing.


- It feels like I messed up the layout around the LM339 thus "needing" a second layer. Can anyone see a more elegant way of routing and maybe avoiding using the second layer for the "12V_OUT2" track/net?
- As I'm a little insecure about proper low impedance grounding, I chose using two ground planes and (excessively) stitched them together.
- The data sheet of the LM2678 advises to keep the feedback track away from the inductor field. On the other hand, the current loops should be as short as possible, this leads to compromises in the layout. Is the feedback track far enough from the inductor?
- It felt like an good idea to spare the ground planes under the inductor, so it cannot induce currents in the groundplane. On the other hand, it disrupts the ground plane in a large area of the board. What do you think?

Attached you'll find the component layout, the two layers layout and the schematic.

Best regards

Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo