Mysterious Board Failure

[sdschulze]

I got my A20-OLinuXino-LIME2 board yesterday.  After a bit of effort, I managed to get my Debian system running and it was all fine.  Then I put it in the chassis and left it running for a bit.  When switching back to its HDMI output, the screen was all green.  I rebooted by reconnecting power, but then there was no HDMI output any more at all, and the green LED no longer lit.

I retried this a couple of times, but no success.  Eventually, the yellow LED stopped lighting, too.  Now, when I attach power, there is only the red LED, plus, when I attach network, the green RJ45 LED lights and sometimes the orange RJ45 LED blinks in a "natural" manner.

What happened to my board?  Is it fried, and how is that possible?  What exactly does a non-lit yellow LED indicate?  Could it simply be SD card read failure (like, dirt in the slot), or is that unrelated?  I remember it happening once before, but that time I could fix it be reinserting the microSD card and reapplying power.

In other news, upon closer inspection, the leftmost pin inside the RJ45 connector looks a bit bent.  I don't know if it was like that when I got it, but it worked in the meantime and I didn't do anything to the connector other than attaching a cable.  But that's probably unrelated, right?


Thanks,
Sören

[olimex]

We got your board back and it pass all tests and HDMI etc works correctly.

I guess the transport back mysterious healed it?

[Lurch]

... or the power supply is/was crap?
Expensive way to test.

[sdschulze]

The board was working for the first couple of hours when it arrived back here, but it failed again after that.  Now I left it unpowered for a couple of days; after that, it first booted but froze after a few minutes and failed to reboot.

Fortunately I now have a USB-serial adapter, so I can provide new information.

When it fails to boot, the error message is:

Code Select Expand

U-Boot SPL 2015.01 (Feb 12 2015 - 00:20:35)
DRAM: 0 MiB
### ERROR ### Please RESET the board ###
This is with U-Boot 2015.01 configured for LIME2.  A similar error occurs when using the old uboot-sunxi fork:

Code Select Expand

U-Boot SPL 2014.04-01572-gea1ac32 (Feb 06 2015 - 22:24:03)
Board: A20-OLinuXino_Lime2
DRAM: 0 MiB
### ERROR ### Please RESET the board ###
And even with the bootloader from the official Debian image, I get an error:

Code Select Expand

U-Boot SPL 2014.04-10706-g36080eb-dirty (Oct 08 2014 - 11:45:32)
Board: A20-SOM
DRAM: 0 MiB

I also tried the new 2015.04 U-Boot, which manages to get a little further

Code Select Expand

U-Boot SPL 2015.04-dirty (Apr 27 2015 - 23:57:23)
DRAM: 1024 MiB
CPU: 912000000Hz, AXI/AHB/APB: 3/2/2
But sometimes, it also freezes right after "DRAM:".

Apparently, I am not the only person experiencing this:
https://www.google.com/search?hl=en&q=u-boot+%22dram+0+mib%22+%22please+reset+the+board%22

What surprises me is that the same error also occurs on other Allwinner boards from different manufacturers (CubieTruck, Banana Pi).  Maybe this is a problem with the SoC?

I tried 3 different power supplies (including USB) and 2 different SD cards, so I am now pretty sure that the error is on the board.

While this sounds like a thermal problem, I do not think it is one because the SOC does not really get hot (plus it has a heatsink) and the timespans are too long.  It takes days to weeks for the board to "recover", and after it came back from Olimex, it worked for hours, while it now only works for minutes.  Is there anything on the board that has this kind of "memory"?

What I imagine is that there is a spurious capacitance on the board/SoC which accumulates charge within hours while the board is running and discharges with a half-life of about a week while the board is powered off.

[JohnS]

Probably each of the power supplies is poor.  So many bad ones are out there

John

[bitwelder]

I also experienced one of these 'DRAM: 0 MiB' errors.
https://www.olimex.com/forum/index.php?topic=4337.msg18286#msg18286
I wasn't able to resolve the issue, and the amount of similar unexplained failures with boards running Allwinner CPUs doesn't inspire much confidence, unfortunately.

[JohnS]

Obviously it's vital to boot exactly the right software for the LIME2 (else all the RAM timings etc will be wrong) - but hopefully you are doing.

John

[martinayotte]

About 8 month ago, I had a board that started to give such "DRAM: 0 MiB" errors after running fine for about 3 months. Olimex support suggested to me to use a hair dryer to fix possible bad soldering on the DRAM. It worked, but last only 2 more weeks. Then, they told me to send back my board so they can fix it. Since then, no more problems.

So, maybe your board has also bad soldering under the DRAM ...