Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Also (because I feel like it/am procrastinating on other stuff), some first-glance review on the design at https://git.sr.ht/~aparrish/abc-pcb/tree/main/docs/abc-pcb.p... :

- U6 and U8 need nearby decoupling. LVC logic has horrid power consumption during transitions, so best to feed it well if you expect any kind of transition rate out of these gates. It doesn't have to be explicitly dedicated to these parts but it needs to be physically close.

- That goes extra for the big fat WideBus 16-wide level translators. They have multiple power pins for a reason. They are vicious little devices. Decouple every last power input pin with its own individual capacitor.

- U6 outputs appear to have wrong bus directions? Probably not important (also maybe not a problem at all, I usually do Altium not KiCad)

- VBUS is not logic level, so you shouldn't use it as a logic signal if you want reliability (it will work... mostly... but also be weird sometimes), but you absolutely cannot drive multiple things with it and expect them to work (their input thresholds will be slightly different and you will be sad when they switch at slightly different times because it's moving slowly). Clean it up with a Schmitt trigger (1G17 etc).

- No ESD protection on the USB port. Do you want it to live? Try something like ECMF02-4CMX8, nice and simple to implement (mildly annoying to solder though).

- Whatever is going on with Q1 is poorly enough drawn that I can't understand it at a glance. Best to just draw these parts as two ordinary MOSFETS with A and B suffixes, then the circuits can be legible.

- On IC2 (why not U2?) lines 4, 5, 6, 7 are being cross-driven. Do not ground both sides, you will be sad. Ground the input sides and let the outputs be unconnected, because they're strongly driven by the chip. Or use resistors to pull things places if you must satisfy someone somewhere who loves that kind of stuff (medical! medical? medical.).

- U7 SENSE pin draws not-much current (~25nA) so no reason to burn power in that resistor divider

- Consider stuffing a big fat electrolytic or two down somewhere and making your PDN nicely damped.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: