Mmm, sort of doubt this. EMI would only be causing serious problems with a return path - in this case the RX line, which wouldn't have much of an effect on the

Mmm, sort of doubt this. EMI would only be causing serious problems with a return path - in this case the RX line, which wouldn't have much of an effect on the transmitted byte stream since relatively little is coming back over the UART_RX. Since this is all at 3.3V logic levels, I doubt there'd be much crosstalk anyway. If by "EMI" you mean some sort of signal quality issue due to transmission line effects, I'd say that's probably closer to the truth, but still not likely to be the case. At 1MBaud, your UART Tx will be switching in the range of 12-15MHz. As long as your physical UART connection is under six inches or so (and it probably would be if you were using a USB FTDI cable or something like that), it'd be pretty hard to get seriously bad transmission line effects at that frequency or most of the important harmonics in that square wave. This sounds a little more like an additive error from a baud clock mismatch. At some point, the FTDI is going to be generating a clock at some frequency that isn't compatible with the ESP's receive oversampler. I'm inclined to believe his explanation if he was able to show JTAG captures of the ESP's UART error register.