[E3-hacking] possible fix for dead serial/usb port

[en4rab]

Hello,  I realise that this might be a little pointless. I have just spent 2 hours fixing the fault with my E3 and I thought it might be of interest to one or tow people.
I dragged my emailer out of the box of stuff a few days ago as the constant stream of cold calls about PPI were really beginning to annoy me.
There had been an article on http://www.theregister.co.uk/2013/02/08/bt6500/ about BT selling a phone that will check the caller id and go straight to answerphone if its international or withheld and in the comment someone suggested robodialer could be fooled into thinking a number was out of service if you played the no circuit SIT tones (the dee doo daah you hear before the voice telling you the number you have dialed is incorrect)

This all got me thinking could you build such a setup with a usb modem and a raspberry pi, turns out someone already did:
http://jcblock.sourceforge.net/

Which obviously then lead to the thought perhaps i could finally do something with my E3 (although i can barely manage "hello world"), after alot of frustration and with alot of thanks due to Janusz Krzysztofik both for keeping support in the kernel and his posts on setting it up I finally managed to get openembedded to build a minimal image, I rashly configured it to build a 2.6.39.4 kernel and angstrom 2010 distro, since i had no idea what i was doing i thought it would probably be broken in the 3.x versions, im guessing though that was stupid thinking on my part.

I did encounter problems installing this on my phone though, i couldnt get pbltool to talk to the phone at all, yet i was able to see kernel boot messages fine, I eventually got round this by using the uboot shell and Tera Term (but any terminal client that supports kermit transfers should work) which seemed to work ok over serial, a few hung sends but got there eventually, as my e3 was already flashed with the v2release the commands for this are:
loadb 0x11d00000             {then send uImage by kermit}
loadb 0x11800000             {then send install-initrd.ext2.gz by kermit}
setenv bootargs mem=32M console=tty0 console=ttyS0,115200n8 root=/dev/ram0 initrd=0x11800000,8M
bootm 11d00000

at this point I was pretty happy, apart from id left out the driver of my usb ethernet adapter, I was just about to install a rebuilt kernel with the driver included when serial comms and usb both died, the phone would still boot but i couldnt do anything with it.
I know i should take this as a sign to throw it out but its personal now so I took the E3 apart and took a multimeter to it :)
When I checked it I was only getting about 2.1v across the power pins of the USB port and the serial port shares this 5v rail, which is why both usb and serial died.
I couldnt find a 5v regulator but alot more poking led me to the 8 pin chip U17 next to the blue 2 pin socket, which is a DC/DC converter controller chip from ST:
http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00001232.pdf
Page 11 of that datasheet is pretty much the schematic for the 5v power on the E3 as far as i can tell, some more poking and it turns out the current shunt resistor  was open circuit, i think its marked as R223 you can find it between the ST chip and te 3 electrolytic caps just to the side of the crystal.
If you were to look at noodles picture here http://www.earth.li/~noodles/images/e3-logic.jpg and imagine a line fron the bottom left of the blue 2 pin socket to the middle of the crystal just below it the black and silver chip (marked 3R3) the line goes over is the one at fault.
Unless i have totally stuffed up the markings it should be a 3.3ohm resistor I havent got a suitable replacement to hand so at the the moment the board is on my desk with a 100 ohm throughole resistor bodged across the dead one, but that was good enough to confirm im now seeing 5.1V on the usb socket and serial comms is working again.
Switch mode psu's are a mystery to me so im guessing you should probably get a 3.3ohm resistor if you encounter this problem rather than bodging it like i did.

I hope this is of use to someone :)
Robin Bradshaw