How I finally got my first STM32 board programmed - steps that actually worked

I had a rough time with this exact setup when I was prototyping my sensor interface last fall. The part about just hooking up four wires seems simple, but getting the voltage stable is where most people run into trouble. My board had a slight power dip when the chip started processing, which caused wierd resets during the flashing process. I ended up wiring up a seprate rail just for the chip and made sure the power was regulated down to exactly 3.3 volts. If you skip that step and just try to power it from the main supply rail, you risk the programmer losing its grip on the chip mid-upload. It is annoying because you lose all your work.

For the SWD connection itself, beyond the headers, I always double check that my resistor values on the pull-up lines are reasonable, usally just 10 or 47 ohms. It prevents alot of unnecessary confusion when I am trying to find out why the connection is dropping under load. It saved me hours of trying to debug the logic flow instead of the electrical connection.

Yeah the separate rail thing matches what I ran into with my first board build. The flashing kept failing halfway thru and I blamed the software for a week before I realized the chip was booting my own code before the debugger could grab it. Once I ticked the option to connect under reset in the programmer and ran a wire to the reset pin, uploads stopped failing. So if anyone gets connection drops like that, it might not be power at all, the chip is just already running when the tool shows up.

Gotta push back on the pull-up resistor bit. 10 or 47 ohms is way too low for a pull-up, those numbers sound like series resistors you'd put inline on the clock and data lines. Actual pull-ups on SWD are usually in the kilohm range, and if you actually strapped a line high through 10 ohms you'd be fighting the debugger's own driver and probaly making the connection worse, not better. I run my boards with nothing but the inline resistors and a proper 10k pull-up on the reset line, and connections have been rock solid. Maybe that's what your board actually had and the labels just got mixed up in your notes.

This clears up a lot for me. I kept thinking you had to unsolder the chip off the board and stick it in some special socket, like when I pulled a chip off an old appliance board to read it. So those little pads by the chip edge stay right where they are and you just clip on. Does it matter which way round the two signal wires go, or does the computer software sort that out on its own if you get them swapped?