Just got my first board back - an RS232 sniffer that actually works

Congrats on a first board that works, that's rarer than people admit. My first one was a level shifter for talking to an old PLC, and I made almost the same mistake with the DB9 footprints. I pulled the dimensions from a mating connector I had on my desk instead of a datasheet, and even then the plastic shroud hit a silkscreen legend and the plug sat crooked. Now I always print the layout 1:1 on paper and physically push a real connector against it before ordering. Caught two footprint errors that way on later boards, cost me nothing but ten minutes.

The test point thing gets everyone. Mine ended up under a big electrolytic cap and I had to hold the probe at a wierd angle, readings were fine but my wrist hated it. On the next revision I moved them all to one edge in a row with 100 mil spacing, so a standard probe hook clips on without a fight.

For the LED idea, a series resistor around 1k on the TX line through an opto worked well for me since it kept the sniffing side fully passive. Didn't load the line enugh to bother the link even at 9600 baud. That repair jumper under the IC is a proper rite of passage. Mine was an address trace I had to cut and bodge with wire wrap, still running two years later.

Nice that it came up working first shot. I went a slightly diferent route with my serial tapping boards and skipped the pass-thru arrangement entirely, just wired a spare DB9 shell off the existing link with a single line to a buffer chip, so the sniffer coud be unplugged without dropping the connection between the two devices. Saved me a lot of grief with connector alignment because the main link never got touched.

On the LED idea, I'd put two of them, one on each direction, and size the resistors so the LED is dim enough not to load the line much. I got that wrong once with a bright red 3mm part straight off a spare reel and it actually skewed the waveform enough to cause the odd framing error at 115200. A high-efficiency LED with a bigger resistor fixes it and still blinks visibly. Also agree on test points but my habit is to line them up along one edge with clear space, then label them in copper, not just silkscreen. Silkscreen wears off or gets covered by a connector and you're guessing which pad is ground at 10pm.

One thing that saved me on my serial tap board was adding a row of 0.1in header pads for one of those cheap usb logic analyzer sticks. Scope is fine for eyeballing a waveform but being able to actually decode the bytes made debugging way faster. Just a few pads wired straight across the lines, cost nothing extra on the panel. Also worth considering a second ground test point on the far side of the board. I only put one near the first connector and spent a whole afternoon chasing what looked like noise, turned out my ground clip was picking up junk from the long lead.

Gotta push back on the opto suggestion in that earlier reply though. A single resistor feeding an opto on a serial line only lights the thing on one polarity, and these lines idle negative and swing both ways, so you'd catch half the bits and stare at a dark LED during idle stretches. The reverse voltage across the input side bugs me too, most of those optos don't love being driven backwards all day. I'd at least want two back to back channels or an anti-parallel arrangement before trusting it as a traffic indicator. Curious if the original poster's link even tolerated it, some older gear gets cranky when you hang extra stuff on the line.

One thing that bit me on a similar passthrough board was the labeling, or lack of it. With DB9s you never quite know which end is the DTE and which is the DCE untill you've buzzed it out, so I now print TX and RX arrows on the silkscreen for both sides of the tap. Saves ten minutes of head scratching every time I pick it up again six months later. I'd also suggest adding a couple of mounting holes near the connectors if the revision has room, because those plugs get yanked around in real use and unanchored pins eventually lift a pad or crack a joint. Ask me how I know.