Use the power adapter. Don't power from USB port.

This is an option, but the board itself is powered from the USB's +5V line, so I would need break this line in order to use external 5V supply because I want to talk with Carambola using serial port (this connection is realised using onboard FTDI converter).

I am googling and reading USB driver's (ftdi_sio) source code but I can't find anything related to my problem. It is not a big issue (since it is always possible to press reset button after connecting ant the board boots normally) and it is not related to Carambola's software, so it may be moved into another folder of this forum.

EDIT: I solved it in a dirty way: Made a program which sends "\nboot\n" message to /dev/ttyUSB0 device and created bash alias called carambola_boot. I call it whenever I plug Carambola into my Linux machine and want to play with it:) I think it could be also automated if there is some way to determine that currently connected FTDI chip is the one on my Carambola (maybe some unique ID).

The clean way is to modify&precompile u-boot , but carambola2's u-boot source code still missing in github/8devices .

C1 Uboot countdown start menu has not interrupted by plug-in USB HOST activity.
C2 Uboot menu countdown has interrupted in this case. I think a reset RX/TX buffer in Uboot source code will solve the problem. But i'am not sure.

I also have this problem with Ubuntu 10.04 on multiple machines, as well as later versions [have to check which].

Observations
(1) There are two sets of 'random' strings that can be seen as supposedly sent - you can see those as an error command when you actually open the serial console ('screen /dev/ttyUSB0 115200') and press enter.
(2) If you open up the serial console immediately after power on then there is no problem with the boot process.
(3) Resetting via reset button has no problem.

Thoughts
(1) Carambola2 board has the #RESET line of the FT232RQ tied to 5V, even the FTDI references don't have this connected even though they say it can be connected... based on (3) above
(2) Can it be a buffer overrun thing? If the C2 bootloader sends too much data before the 'Press any key to halt boot' with the serial console not yet opened... based on (2) above
(3) It would be nice if the bootloader looked for a specific key or a key sequence to stop boot...
(4) Would be interesting to power a Carambola2 from another Carambola2...


I can probably test the (1) Thought if gedass or someone from 8devices can confirm which is the 4.7kOhm resistor connecting #RESET to 5V... I presume it is the 'big thing' next to U1/FB1/J14...

Would be better if 8devices tests it though

Idea with buffer overflow s interesting. Do you have any FTDI chip to test this behavior? I don't have my equipment with me now.

I tried to connect it to Windows machine and it booted normally. Connecting it to Ubuntu machine ends up with stopping the autoboot process. Both connections (Windows, Ubuntu) were tested without opening the terminal - Windows sends nothing to stop the autoboot, Ubuntu sends. Do you think difference in functionality between Windows/Linux FTDI drivers would be that big?

BTW. my observation is that tthere is only one "random" string sent (therefore it is not very random, is it? ). I personally think it is some configuration message from the OS (something like "Dear FTDI chip, welcome to the system. Wouldn't you mind switching baudrate to X and data format to Y? Thank you and have a nice day!" but probably in less polite way). But I stoped the "investigation" (wanted to read source code of drivers first) and ended up with simple program that sends "boot" command as this is not an important issue for me. The dev board is just for testing some principes, real device will not be powered from USB port hooked to FTDI chip.