The perpetual PCB design/build/help thread!

Drizzt321

Ars Legatus Legionis
28,408
Subscriptor++
So since I'm doing a bit more with making my own PCBs these days, I thought I'd start up a thread!

For myself, I'm using KiCAD for design, and JLCPCB for making them.

My current project is to take a ESP32 and get a board with a 3.3v LDO and wide range of inputs (e.g. 4.5-20v) so I don't have to worry about trying to solder to the castellated edges, and can fit it into smaller places than a full board with USB->Serial/etc all embedded. And without the full 0.1" pin spacing, so more compact. Going to use it for locations that are space constrained for WLED usage.

Future might be creating my own Buck or Boost or Buck/Boost onboad for a custom form-factor for width constrained spaces. But that's the future, seemed like significantly more...challenging to do safely.

Except I'm realizing that getting an ESP32 put onto it is far more expensive (relatively) than I realized. It needs Standard rather than the cheaper Economic, and ESP32 apparently is recommended doing baking (reduce risk of moisture with solder reflow process).

So now I need to rethink that. I might redesign the board to avoid any solder through-holes on the side, and instead bring out most of the GPIO into a grid, and that way I can make it even slimmer, and not bring some of the pins out and available, to help with size. And not get the ESP32 put on, just be careful with a small soldering tip, high temp, and I think I can make it work. Unfortunately means I won't get proper soldering for the pad on the underside of the ESP32 to the bottom/GND plane to help with temperature control. Most likely fine for this application though.

Future, I really want to be able to charge (4s2p LiIon) from USB-PD source, so 5-12v with boost, or 18-20v with buck with CC/CV to charge the battery. Seems like there are chips out there, but...serious design work.

EDIT: Wait...the 0201 sized LED wants to be baked? Not the ESP32? Huh. No, I'll accept the risk to do it without baking.
 
Last edited:
For reflow and as a hobbyist more, i kinda wanted one of those miniware hotplates for surface mount soldering. (have had the TS80 for 4+ years now and really happy with it.)
Havent gotten one yet since the pandemic hiked prices (specifically of the MCU) and locally they havent dropped/normalized yet for my liking.

 

dmsilev

Ars Praefectus
5,410
Subscriptor
I've been using Fusion for PCB design (mostly because I originally learned/suffered through Eagle, and that's what underlies Fusion's PCB module) and for one-off designs usually buy from OSHPark. Generally I do assembly myself, using either regular solder or syringe-dispensed solder paste (Digikey etc. sell tubes of the stuff for a reasonable price). I don't bother with stencils, but instead squeeze a bit out into a plate, then dab it onto the pads using a toothpick or similar. I usually use a pencil hot-air rework system like the one linked a couple comments up to actually do the soldering, though a hot plate is certainly a good alternative.
 

dmsilev

Ars Praefectus
5,410
Subscriptor
Works pretty well, though it obviously is best suited for boards with not-huge numbers of SMD pads. The nice thing is that even if you’re not super-neat and get a bit of paste smearing between pads, the solder mask coating will tend to eliminate bridging between the pads once the paste is melted. The size you can work with depends on how steady your hands are and how good your eyesight is. I can manage 1 mm pitch without a magnifier and 0.5 mm pitch with a modest binocular scope. Finer than that and positioning the chip over the right pads becomes a nuisance.
 

redleader

Ars Legatus Legionis
35,019
Except I'm realizing that getting an ESP32 put onto it is far more expensive (relatively) than I realized. It needs Standard rather than the cheaper Economic, and ESP32 apparently is recommended doing baking (reduce risk of moisture with solder reflow process).
For one off and prototype boards, solder it yourself. If you don't design tons of boards good odds the first one won't work anyway or you'll have to rework it because you made some simple mistake, so might as well avoid paying to have a bunch of boards made and then having to rework each by hand. Plus its faster turn around time, which again is nice if you are going to need to iterate/fix things.

And yeah, definitely agree with skipping stencils and just dabbing down paste onto pads with a syringe. If you're going to do a lot of prototyping, buy a pneumatic paste dispenser, but you don't even need that here.

For actual reflow, on single, small boards without BGA, you can use a hot air gun or something like this:

View: https://www.youtube.com/watch?v=ORVPBt1WP1A


If you're making batches, using lead free or doing more complex parts, you want to buy a reflow oven that can do more controlled reflow profiles. The T962A is less than 300 on Amazon Prime, and can be hacked to be half good.

If you want to get that part properly soldered you might want to consider something like this:


View: https://www.amazon.com/ML-851K-Hot-Pencil-Rework-Station/dp/B003UCR332


We used to have one where I work.

that's kind of a mid priced one. There are cheaper ones.


FWIW those are in the awkward gap between the really cheap generic Chinese hot air stations (which are honestly fine for a lot of things) and the better stuff. I'd say spend either more or less. There are lots of good mid priced hot air stations coming out of China these days. This channel reviews tons of them:

View: https://www.youtube.com/playlist?list=PLZzwMlLVLdOCgT7rYau-CqcDyeZ3Rn5qV


I picked up the Atten ST-862D as a home station based on those reviews for well under 200 and its way more hot air than any hobbyist would ever need.

I've had to fix/replace a few tiny SMD resistors, always a huge pain. Some paste and hot air pen probably would have been so much easier.

Maybe I'll look into it fir my next round of low volume stuff. Cheaper to get the boards and just a few SMD things like the tiny resistors/etc, and on the cheap/economic assembly.
Paste is for assembling new boards. Once the board is assembled and you're doing rework you don't usually want to use paste. If it is a tiny resistor you want to change, quick and easy way to remove is to put a big blob of solder on a chisel tip, touch the blob onto the part from above, hold for 2 seconds, and then lift. The part will float up into the solder blob and you can wipe it off your iron on a sponge. Once it is off, you can usually reuse the existing solder with just a bit more flux.

Hot air is another way, but unless you do it a lot and get lots of experience theres a real risk of blowing off surrounding parts, melting connectors, etc. Tweezers are the best solution, but there aren't a lot of options for tweezers that are worth using that aren't going to cost double or more what your iron and hot air combined cost.

I bought a soldering station with hot tweezers for little resistors, but I don't do enough SMD work to know if that was the right option. I do have a hot air station too though, with nozzles ranging between pencil and moderately large. How do you folks with actual experience feel about hot tweezers?

Part of the reason that there are no value tweezers that enthusiasts recommend like there are for hot air is that they're a niche tool. They save a tiny bit of time if you're going to swap out small SMD resistors, but you don't actually need them. They're more for reworking dozens or hundreds of boards at a time where the extra time it takes to use hot air or a single iron would matter.
 
Last edited:

Shavano

Ars Legatus Legionis
59,253
Subscriptor
I bought a soldering station with hot tweezers for little resistors, but I don't do enough SMD work to know if that was the right option. I do have a hot air station too though, with nozzles ranging between pencil and moderately large. How do you folks with actual experience feel about hot tweezers?
they're excellent. I always try to use them for 2-pin surface mount packages. Not needed for soldering. It's easier to hold the part with plain old tweezers, solder one end, and then solder the other, for me.
 

Drizzt321

Ars Legatus Legionis
28,408
Subscriptor++
sigh

So let my rookie mistake be a lesson for you all. LDOs should not be used when doing large voltage shifts. Going from 4s/14.8v nominal to 3.3v is quite a large voltage drop, and that turns a lot of power into heat, so either be ready to handle that heat, or it'll keep cycling overheat-off-on-overheat-off-on/etc. Besides being a waste of power. Ugh. Ah well.
 
  • Like
Reactions: Xenocrates

Drizzt321

Ars Legatus Legionis
28,408
Subscriptor++
Nice. Except it's not a TO-220, it's SMD SOT-223. Not to say that's not pretty great, useful for sure, but a bit smaller and physically on the board would be useful. I have seen some other switch mode SMD, will eventually figure out how to do their circuits, but for now I'm just going to go off of the larger 5v buck for the LED strips. Which what I wanted to do was be able to use the enable/disable on that buck board, so I could completely shut off the LEDs, including the quiescent power. Ah well. v3 design it'll have to be, perhaps next year.
 
  • Like
Reactions: Xenocrates

redleader

Ars Legatus Legionis
35,019
Nice. Except it's not a TO-220, it's SMD SOT-223. Not to say that's not pretty great, useful for sure, but a bit smaller and physically on the board would be useful. I have seen some other switch mode SMD, will eventually figure out how to do their circuits, but for now I'm just going to go off of the larger 5v buck for the LED strips. Which what I wanted to do was be able to use the enable/disable on that buck board, so I could completely shut off the LEDs, including the quiescent power. Ah well. v3 design it'll have to be, perhaps next year.
FWIW it really is not that hard to implement a buck converter that works well enough for typical low power applications, and since you're doing the super common ~12V to ~3.3V conversion, half of the datasheets you open are going to have an example schematic you can use literally unmodified. Getting high efficiency at low cost under a wide range of loads is where expertise makes a difference, but if you're starting with a linear regulator you probably don't mind if you're 5% less efficient than you could be. Just google "buck hot loop" before you do the layout and think hard about where you put that input capacitor and you'll be fine.
 

Drizzt321

Ars Legatus Legionis
28,408
Subscriptor++
FWIW it really is not that hard to implement a buck converter that works well enough for typical low power applications, and since you're doing the super common ~12V to ~3.3V conversion, half of the datasheets you open are going to have an example schematic you can use literally unmodified. Getting high efficiency at low cost under a wide range of loads is where expertise makes a difference, but if you're starting with a linear regulator you probably don't mind if you're 5% less efficient than you could be. Just google "buck hot loop" before you do the layout and think hard about where you put that input capacitor and you'll be fine.
Actually I'm not, I'm doing up to 16.8v (4s lithium), not 12v. Close-ish to 12v, but still a fair bit above.

And yeah, ~85-90% efficient would be just fine. But then I also want to design my own 1s-4s (or at least 2s-4s) buck to 5v for 5-8a peak. And at that point I'll probably just use a LDO for 3.3v given how low current (~300mA max I think?) for something like an ESP32, less even for an ESP8266.

I also want to figure out a USB-C PD -> 3-4S Lithium buck-boost (in case the USB-PD only supports up to 12v out), to have USB-PD charging of the 1-4s.

Also, a buck-boost to to work towards a Framework 16 Battery Expansion Bay, to add some additional Wh to the laptop. Which will undoubtedly require a micro controller, might just use a RP2040 like most of the things in the Framework, since it has a nice USB2.0 <-> Serial communication built in, so I can put it on the USB bus and be able to update the firmware that way, as needed. With plenty of I/O for controller the buck-boost, disconnect the output from the battery when the laptop is plugged in and start the charging via the laptop 20v power bus, talk to the laptops embedded controller, possibly via SMBus, or just straight I2C or what not, to report battery status, etc. And of course fit on a balancing BMS in there, somewhere, somehow.
 
  • Like
Reactions: redleader