If I understand what you're typing, this is an old problem, and I wish Apple would fix it.
The CH340/CH341/CH342/CH347/CH348/CH9102 driver (yes, I'm keyword stuffing) in macOS is convenient but is slightly broken. In some ways, that's more frustrating than being totally broken. It loses data at high speeds under stress. (This is confirmed by your report that it works at the—leisurely, by modern standards—speed of B115200.)
The fix is to install Win Chip Head's serial driver, which should instinctively go against everything you've ever been taught. It's not a signed binary. It's in a weird language. The instructions are terrible. You have to grant it the highest possible trust levels. etc. Your spidey senses SHOULD be bristling. But install it anyway...
Now you find that your USB/serial bridge actually has TWO dev nodes. There's the /dev/cu.whateveritusedtobe that's still there, routes through the old driver, and still works badly, and the /dev/cu.wchusbserialwhatever that's NEW ... and it actually works reliably.
Any time your apps (like esptool for uploading or monitoring) try to talk to the device at high speed, they HAVE to use the /dev/wch dev node in order to route through the driver that really, really knows how to communicate with those bridges. If they use the old node, they only sort of work...and at slow speeds.
This thread is polluted with other issues, but see:
https://github.com/espressif/esptool/is ... 2257272462
TL;DR: The macOS driver is bad. The WCH driver (/dev/cu.wchBLABLAH) is good. Really.
This whole process is probably the biggest black eye about this specific intersection of the ecosystem. The ESP32 with an external USB/bridge (which largely hasn't been required for any ESP32 released in the last 6-8 years, but here we are) falls down with obscure symptoms some small(ish) percentage of the time, but with fatal symptoms. |
That's very much the opposite of the "easy" experience that we buy Macs for. It's the 1990s level of Windows sucking, and that's most unfortunate. It's pretty easily the biggest black eye for this combination, and I'm sure that if the right Jr. engineer... If Apple would talk to the right Jr. engr. at WCH to authorize the transfer of an updated driver, this could be solved in an afternoon. The solution _exists_; it's just getting the attention of multiple $$$$ companies to talk to each other at the right level and getting it shipped so that dozens of Mac-using embedded people don't have to rediscover this every day.
Citations:
https://github.com/WCHSoftGroup/ch34xser_macos
https://discussions.apple.com/thread/25 ... ortBy=rank
https://github.com/WCHSoftGroup/ch34xser_macos/issues
Signed,
One who develops extensively for ESP32s using this dumb bridge AND is a (mostly) happy multi-Mac user. This has plagued me from my 201x Intel systems on through my latest Tahoe images.