Another case of unable to boot from SD

ehj666 · January 15, 2013, 03:09:45 AM

That is all there is, in fact "Allwinner Technology" does not even appear on the first line.

That is followed by:

DRAM: 512MB
SUNXI SD/MMC: 0
sp1: mmc blk read err - 0

jwischka · January 15, 2013, 03:18:58 AM

You may have mentioned this before, but what are you using to power the board? I am wondering if you have enough power to drive the nand, but not enough to drive the SD and the nand.

Quote from: ehj666 on January 15, 2013, 03:09:45 AM
That is all there is, in fact "Allwinner Technology" does not even appear on the first line.

That is followed by:

DRAM: 512MB
SUNXI SD/MMC: 0
sp1: mmc blk read err - 0

ehj666 · January 15, 2013, 03:56:24 AM

At the moment I am using a 2.1A 5V USB power supply. A 1.2A @ 12V just arrived today, but I have not hooked it up as yet.

jwischka · January 15, 2013, 03:57:44 AM

Quote from: ehj666 on January 15, 2013, 03:56:24 AM
At the moment I am using a 2.1A 5V USB power supply. A 1.2A @ 12V just arrived today, but I have not hooked it up as yet.

See if it makes any difference. If not let me know and I'll brainstorm some more.

JohnS · January 15, 2013, 01:59:45 PM

With that missing text, I wonder if it could be dropping characters at the host end?

jwischka · January 15, 2013, 03:08:43 PM

Quote from: JohnS on January 15, 2013, 01:59:45 PM
With that missing text, I wonder if it could be dropping characters at the host end?

I would think dropping characters, or a bad cable would be a more intermittent problem, but given that the text is at the beginning, I suppose it's possible.

One other thing might be to recompile u-boot to use a different UART by default and try that.

ehj666 · January 16, 2013, 10:46:02 PM

Just to update status, Olimex got back to me and also concludes that there is probably something wrong with the SD port. I did not want to wait to go through the whole RMA process to get a replacement, so I ordered a second board and will return the first one after the new one arrives, if necessary. Fortunately they had one A13-Olinuxino-SD in stock so snagged that as well.

Thanks for all the input everyone.

ehj666 · January 19, 2013, 01:03:02 AM

Well I am really perplexed now. I just received another A13_Olinuxino-Wifi and this time I got the A13-Olinuxino-SD (the SD itself is a Samsung). It boots just fine to Android, but again if I insert the SD while Android is running, I get damaged SD. If I try to boot to the supplied SD I get:

U-Boot SPL 2012.10-rc1-03956-gd7ea23d (Oct 11 2012 - 15:48:58)
MMC: SUNXI SD/MMC: 0
Loading U-Boot...
### ERROR ### Please RESET the board ###

Resetting the board results in exactly the same message.

If I try to boot to one of the SDs burned from the downloaded image, I get:

U-Boot SPL 2012.10-04266-g20efdee-dirty (Jan 08 2013 - 16:20:29)
DRAM: 512MB
SUNXI SD/MMC: 0
sp1: mmc blk read err - 0
### ERROR ### Please RESET the board ###

The latter is basically exactly what I was seeing on the first board.

I do not believe that I have two bad boards. I am still running just from the USB port / 2.1A @ 5V power. I cannot believe that is insufficient for the very slight additional draw of the SD, but am willing to try another power supply.

Any idea what is going on or what could explain the symptoms I am seeing.

jwischka · January 19, 2013, 01:24:06 AM

I would still vote for power supply. It's not really the sustained capacity, but whether it is strong enough to handle the spikes as loads switch on. Particularly since you are at 5v, you are more susceptible to dropping below the voltage needed to run the board. If you have an external hard drive, they are usually 12v 2a supplies, and should be more than enough to handle the board. Would be a good test, at least. First thing I'd try.

ehj666 · January 19, 2013, 01:47:37 AM

I agree with you on the point of voltage drop, OTOH I do not think it can be current. I have booted both with and without a USB keyboard and mouse attached, and they are certainly going to draw more current than the SD port.

I will give it a shot with a different power supply over the weekend.

JohnS · January 19, 2013, 02:00:26 AM

I've also had problems. Different than yours but could only boot Linux with a LiPo battery attached.

I've tried very different supplies including 15V 4.5A and got the same result as with the other PSUs.

Still, I can run u-boot so can try other tests, step by step.

Try a battery if you have one.

Of course it could be some other issue: RAM timings or the like...

John

ehj666 · January 19, 2013, 02:37:23 PM

It does not appear to be the power supply. I now have a 12V power supply powering the board. The one sold by Olimex is rated at 0.5A @ 12V, or 6W. The one I am using is rated at 1.2A @ 12V. Everything powers up fine, in fact acts exactly the same as powered with the USB OTG.

The manual also states that maximum current draw at 12V is approximately 350mA. I am only driving a USB keyboard, mouse, VGA and Wifi, and have disconnected the keyboard and mouse at various times. I have not even tried to connect the LCD display.

I read the manual cover to cover to see if there was anything I was missing and have not found anything. Any ideas would be greatly appreciated.

JohnS · January 19, 2013, 03:52:11 PM

As I posted, the only way I got it to boot Linux (to a login prompt) at all was by connecting a LiPo battery.

I have much bigger PSUs and it seems either this board picks up some stray noise from them or needs some noise suppression internally or ... well, I don't know. Something's wrong. Doesn't seem to be the SD cards or my PSUs.

I'd try powering direct from a car battery if I could (guarantees no noise on power in lines) but not realistic here at the moment.

John

jwischka · January 19, 2013, 03:53:29 PM

No need to get testy, friend. You asked for help, and the best idea that fit your symptoms was a weak power supply. Being an expert in, well, power system transients, I can certainly see how a weak supply would cause what you're seeing. "12V" does not mean "12V", especially when loads are coming on, and the steady state performance of a supply or a load can be significantly different from its characteristics during transients. An air-conditioner may draw 2-3kW at full load, but may draw 10kW+ when it is starting up. It was not at all unreasonable to ask you to try another supply, particularly in light of JohnS's symptoms.

I forget the whole sequence of this thing, but have you tried disconnecting the VGA and seeing what you get over the serial console?

Have you tried to make your own card from scratch? Partitioning manually, etc?

I might also try booting with a kernel / script.bin that has most stuff disabled.

Can you confirm that the card is a Samsung, and not a knockoff (check the manufacturer ID byte)? It shouldn't matter, but then again, this thing should be booting...

ehj666 · January 19, 2013, 05:51:31 PM

I am not sure what I said that would give the impression I was getting testy? I admit my frustration level is getting rather high at this point, but I certainly would not want to take it out on people trying to help me.

I am going to combine replies to the last two posts here:

Noise:

Which port would be powered by a battery? The PWR jack I assume since a 12V car battery was mentioned. I have run into noise problems in the past and this just is not acting like a noise problem. Some noise on the power line is expected, the power supply should filter that out. If it propagates through to the board then there should be bigger problems, like the mouse acting erratically, junk characters on the serial port, etc. Noise problems tend to be inconsistent, whereas what I am seeing is consistent, repeatable and basically identical across two boards.

Power:

I do understand the concept of inrush current. The manual covers this in detail, next to last paragraph on page 9.

QuoteIf you have a standard LCD display connected to LCD_con, Android and WIFI running the typical consumption is between 150mA and 350mA depending on the current load. While the board is in stand-by mode it consumes 60mA. All the three approximate values above were taken when I applied 12V to the board.

I should be drawing even less, since I do not have an LCD display attached. If the maximum it should draw, even with an LCD is 350mA, then 1200mA should certainly provide sufficient margin for inrush current, given that the Olimex PS is rated at 500mA. Also, if it were on the harry edge, one would think something would change by increasing or decreasing the load. That is not the case, I increase or decrease the load by adding or removing components and it acts exactly the same.

SD card and Partitioning

The SD card is a genuine Samsung. I have not at this point tried to partition manually, although I have retried the other SDs created from images. I posted the differences I saw between the Olimex supplied SD and the ones I imaged above.

Very early on, before the boards had even arrived, I built a custom kernel I expected I would need, partitioned an SD and burned the kernel and Debian image to it. When that did not work I opted for the safer routes of using known good images and finally a preloaded SD. I can try that again, but it would seem to increase the number of unknowns rather than decrease them.

I hope my answer does not sound testy, I am just supplying my thought process in trying to simplify and isolate the problem in the hopes someone will tell me where my logic is faulty. At no point to I want to appear ungrateful for all the help that has been provided.