Thanks to everyone who picked apart my schematic a while back, I've taken that feedback and got the board itself laid out now. It's version V0.1 of a function generator built around the STM32H723. The USB-C port talks to the microcontroller, which makes the waveforms either with its own DAC or with an external DAC over SPI, and then the signal goes thru filtering and a discrete op amp driving a buffer stage before it comes out the 50 ohm BNC connector. Power comes in from a 24V barrel jack and I'm getting -12V, +12V, +3V3 and +3V3A rails out of that. The goal is 100kHz of bandwidth and 10Vpp into a 50 ohm load, so I want to be reasonably confident before I send this off for fab. I had a bunch of spare pins left on the chip so I put some of them on a header along the edge. The test pads are scattered all over too - square ones are ground, round ones are signal. Honestly I'm second guessing a few of the placement choices, so if anything looks off or there's something I've completely overlooked, please tear into it.
STM32H723 function generator, first board layout - anything I missed before I order?
Nov 22, 2026
Last reply: 6 hours ago
4 Replies
First reply
D5 is reverse biased and that needs fixing before anything gets ordered. If that part is backwards on the board the way it is on the schematic, the first time you plug the 24V jack in you're going to have a bad day. Sort that one out first, then worry about the rest. On the…
Read the full replyD5 is reverse biased and that needs fixing before anything gets ordered. If that part is backwards on the board the way it is on the schematic, the first time you plug the 24V jack in you're going to have a bad day. Sort that one out first, then worry about the rest.
On the internal DAC side, have a look at ST's application note AN4566, it's about extending DAC performance on STM32 parts. I actually tried the techniques from it on an STM32H745 Nucleo recently and managed 15MSPS with roughly 6MHz of analog bandwidth. Against your 100kHz target that's an enormous amount of headroom, so it should be well within reach if you follow what the note suggests.
The signal chain you describe sounds sensible on paper, the filtered output going through an op amp gain stage into that discrete transistor buffer, then out to the BNC. The thing I'd be double checking before fab is the output stage under a real 50 ohm load at your 10Vpp target, since that's where these designs usually fall over rather than in the low level sections.
On the placement you're second guessing: the way you describe it, the signal path runs from the DAC area through filtering to the BNC on the left edge, so I'd want everything along that path kept short and away from the power supply section near the barrel jack. Is there a reason the BNC ended up on the opposite side from the header and USB area? Not saying it's wrong, just asking how you arrived at it.
The spare pin header and the test pads are a nice touch, honestly. I wish I'd put as many test points on my own boards early on, it saves a lot of poking around with a scope probe wondering where to land it. Keeping ground pads square and signal round is a good habit too, you'll thank yourself when you're probing in a hurry. What layer count are you planning? That'd affect how much you need to worry about the routing around the analog section.
Back when we laid boards out on tape, the first thing everyone guarded was the analog supply, and that habit has not dated a bit. You keeping a separate +3V3A tells me some of the old rules still hold, and quiet analog supply separation is one of them. One thing I learned the hard way: silk the revision number on the board. I once had two spins sitting in the drawer with no way to tell them apart, and that cost me an afternoon of head scratching.
Where does the external DAC sit in your plan, is it going on this first run or saved for a later spin?
The question about the output stage under a real 50 ohm load is the one I'd chase first in this thread. Everything before that point can look tidy on paper and still fall short right at the connector, and that matches the way you laid the signal chain out in your write up. When you measured your targets, did you budget the buffer stage's drop from the rails, or is that still an open question? On the external DAC side, I'd also ask how the SPI lines are planned to route relative to the filter section. On my own boards I've regretted letting digital and analog traces wander near each other more than anything else.
The two ways of making the waveform confuse me a bit. The way you put it, the main chip can do it by itself, or a seprate chip can do it and they talk over that serial bus thing. Which one is the plan for the real version, and is the other one just there as a backup or for trying stuff out?
Also those spare pins on the header along the edge, what do you actually end up hooking to those when testing? I always wonder what people connect to headers like that, since it seems like alot of extra pins just sitting there.
Join the Discussion
Have something to add? Share your thoughts — no account required.
Didn't find your answer?
Ask the community — no account required