[Arm-netbook] eoma68-a20 ethernet testing

[Henrik Nordström]

lör 2013-08-31 klockan 12:36 +0100 skrev luke.leighton:
> oh btw, warning warning warning, as you can see from chris'ss
> experience and also mine here as well as phil's (thank for reminding
> me phil), you must* get a decent RS232 converter.  there are lots of
> "duff" USB-to-RS232 converters out there: not only with this but also

It's not really about the cable, it's about power-off state tolerances
at the device RX pin. Even the best 3.3V UART cables can cause major
problems.

If you connect RX pin with the device powered of then this will feed
3.3V into the device which leaks in various directions. On many
Allwinner based devices this will make the AXP fail or behave badly.

One solution is to add a diode on the RW wire and device-side 3.3V
pullup resistor, making sure that no current enters the board via the RX
pin. Serial communication is active pull-down to indicate a bit.

Another is to use a buffer, which serves the same purpose (separating
the power domains of the UART cable and the SoC).

> as a "workaround" if you run into this problem then power up the
> device by USB-OTG *before* plugging in the USB-to-RS232 converter.
> bit of a pain, but it works.

Yes, once the SoC is powered then there is no problem to connect 3.3V to
the RX pin. But it is generally not OK to provide power to an SoC I/O
pin with the I/O bank powered off.

Regards
Henrik