Project review: DC/DC switching converters and LED driver

[Clear as mud]

I took on the task of helping someone improve their design, and I think I'm mostly done, but still willing to make changes from any feedback I get here.

Originally he had put together a couple of series strings of 18650 cells, MOSFETs for switching power from one battery string to another, a 300W 12V converter module, a few outputs for those WS2812-type LED strings (I guess they call them neopixels nowadays), and a Teensy 4.1 board to control it all.  The original problem, he said, was that there was too much noise on the voltage measurements.

So I added filtering as per the data sheet of the DC-DC converter module, and used an entirely different switching regulator chip to provide the 5V power.  I posted previously while I was designing the 5V regulator, and was encouraged to post my board layout too, but I decided to wait until I had the complete board mostly laid out, and not just the 5V switching converter portion.  My previous posts are here and here, and they have some details on part numbers and characteristics of the components incorporated in the 5V regulator section.

I'm attaching the schematic first. It's called "TeensyBatt.pdf", and also has some text explaining the functionality.

Image 0001 is the 3D model showing the back of the board.  Those modules go on there, but I don't have 3D models for them or the switch.  The battery string inputs are toward the bottom of the board, the 12V outputs on top.  The two main high-current ones right where it says 12V in big letters, and the one above is intended for lesser current.  J6 through J11 on the opposite corner are the LED strip outputs.

0002 is the front of the board.  Again I am missing some 3D models - the relay, the fuse, the 10 uH inductor.  The Teensy board will be mounted on male headers, soldered to both boards.  I don't show the headers, but I show a 2.5mm space where they will be.

I still intend to add some zones and fix some silkscreen text, and make sure my mounting holes are squared up (I think the top right one got moved out of line somehow).  And, I didn't think of this until I got the board to this stage, but I think it would be better if I completely flipped the two modules and the two big capacitors, left to right as if turning the page of a book, so that the power would come in on the right and out on the left.  Then I wouldn't need those big high-current traces going all the way across the middle of the board.  And I was thinking, as long as I am moving those biggest electrolytic capacitors to the back, might as well move the other three there also.  They are the tallest components, and it would probably be good to have them all on one side.  Picture 0003 is a side view of the 3D model so you can see the heights of the capacitors better.

0004 is a picture of the entire 100mm x 100mm board with zones not filled.  Copper layers are KiCad standard colors, red for top, yellow for 1st inner, pink for 2nd inner, and green for bottom.  I changed the "margin" layer to a light blue color (the light blue around the board edges).

On the top part of the board, I put the net GNDPWR on the mounting holes.  This is the middle ground in DC-DC filtering schematic.  It is connected to that middle pin on the FLT007 module, and in that data sheet it specifies it as "chassis ground."  The board is actually to mount into a plastic box, so there is no chassis ground as far as the case around the box, but my thought was it would be good to connect it to the mounting holes anyway, and provide fill zones on that net around the entire filter module and DC-DC converter module, on all copper layers, wherever I don't need the layers for other purposes.  Is that a good idea?  0005 is a zoom of just this area of the board.  I don't have the GNDPWR zones done on all layers yet, and zones are still unfilled in this picture anyway.

0006 through 0009 are the same region of the board, with each of the four copper layers brought to the forefront.

0010 through 0013 are the same thing for the other half of the board.

Picture 0014 is zoomed in on the 5V regulator, zones unfilled and front silkscreen brought to the foreground.  The purpleish circle is the front adhesive layer, which was included as part of the inductor footprint.  The three vias under the input capacitor are my net-tie, which consists of the vias (right one on PGND and left two on GND) and copper polygons surrounding them on the front and back layers.  And 0015 through 0018 are the four copper layers in that area with zones filled.  I wasn't sure whether to put a ground zone on the inner layer under the chip as shown, or leave it as just the two GND traces, with PGND zone surrounding them.

One thing I did that was probably unneccesary was to increase the clearance to 0.5 mm in most places, for my incoming battery voltage, which could be up near 60V.  For some reason I thought one reference I found was telling me that the clearance for 60V should be 0.9mm, and it didn't seem practical to go that high, so I went to 0.5.  Later I saw another reference which did not recommend more than 0.1mm for voltage that high.  Other clearances are generally 0.2mm for tracks, 0.25 for zones (they're not different on purpose).

[Clear as mud]

I went ahead and switched the two big modules and their capacitors to the other side of the board and routed the power from right to left like I was talking about.  Surface-mount components remain all on the front side of the board, but now the modules are also on the front (and cover over the top of some of the surface-mount capacitors) and the two big electrolytic capacitors are on the back.

It allowed me to take 5mm off the board and make it 95 x 100 instead of 100 x 100.  But to do so I have to take the two mounting holes out of the middle and leave only the four outside mounting holes.  If it's better to have all six, I can put the 5mm back into the board.  Without it, the distance between mounting holes is 3 inches.  Possibly more in the up-down direction, I'll have to see if I can get it back down to 3 inches.  It was 3.5 before, but with another hole in between.

[Clear as mud]

I kept seeing improvements I can make.  Attached is a 3D model of what I currently have (still need to make or find 3D models for the components that are missing them).  I moved most of the through-hole electrolytics to the back side, probably still will move that last one to make it match the others.  I moved the relay to the side of the board where there's a natural division between GND and PGND, instead of bringing a trace for 5V and one for GND to a little zone on the other side of the board.

The client does not want a NC relay contact pad.  I put one in anyway, but covered it with soldermask, so if sometime he changes his mind, he could just scrape off the soldermask.  That's easier than making another board, right?

[Clear as mud]

Update: added stitching vias, couple other changes.  It's pretty much done.  2 pictures of the 3D model attached.  Feel free to comment if you see anything you would have done differently.

I still need to double check that C1 and C2 can fit under the filter module, and move their silkscreen text.