Red LED on Faint Green LED

[jhasler]

Half an hour in the freezer and it boots.

The Pi, the sdcard or both? Can you try to narrow it down?

Both. The card alone is in the freezer now. I'll try it in a few minutes and then try chilling just the Pi.

Ok. Chilling only the SD card does not help. When I chill only the Pi it boots.

[mahjongg]

Could be one of the three polyfuses, or any number of other component that is unexpectedly temperature sensitive, or it could be mechanical, for example the cold shrunk the PCB which closed a bad contact. Could be the SD-connector which seems a cheap variant.
If you have "freezer spray" (AKA "kalte spray") if the problem returns you could try to give the polyfuse on the back a blast of the spray, then try to boot, also try the SD-card and al the chips especially the SoC, and the LAN9512, also try both crystals.
Maybe you can find what is causing the problem this way. With a hair dryer you can quickly warm up the board again.

That is, if you are up to trying all this...

[jhasler]

I'm able to log in via ssh now. The board keeps working after it warms up and I can read and write to the SD card.

There are lots of possibllities: marginal timing is one of them. Cooling will (most likely: check the data sheets) slow down the crystals. It will also shorten on-chip propagation delays. Both effects will increase timing margin. The fact that the device has to be chilled to boot when the GPU is reading the card but works ok warm when Linux is running and the ARM is reading it is suggestive.

I'll pick up some freeze spray Monday. I used to design computers as complex as this one (but built of parts large enough to see with the naked eye). Unfortunately I have neither a logic analyzer nor a decent scope. Still, there are more things I can try if I have time.

[sharix]

You could try cooling it with butane lighter refills

[applekeith]

I had the exact same problem red LED on Faint Green LED - No Video. However, I have two raspberry pi's. The odds of both Pis being bad for the same thing would be very rare.

I had a power supply that was borderline, so I purchased a cellphone charger that has 2.5a. I also had several different types of SD cards. I even tried a micro SD card in an adapter... It got really hot really fast. Not sure why either.

Nothing was working. So, today I was looking for a known working SD card, but I couldn't find one. So, I settled on a class 4, 4g PNY card. I got a two pack from WalMart.

After imaging the card, it booted the first time!!!. I used the other SD card to try another image and it worked too.

So, this issue could easily be caused by an SD card.

[jhasler]

applekeith writes:
> So, this issue could easily be caused by an SD card.

The problem is that many if not most Pis only work with certain cards. As the cards that don't work with the Pi work with everything else in the world, the problem clearly lies in the Pi.

[Paul_L]

It is clear for me that the problem lies into thi Pi. On the behavior I was thinking that the problem might be hardware so form my experience the series resistors between processor and sd card socket of 32 ohm sometimes are to high for SPI so I tried to put some 10 ohms in parallel . Since my 10 ohms smd resistors were bigger I only was able to take out the 32 ohm resistors and shorten the links. All the cards that were working are still working and my feeling is that are working faster but I can be subjective. The non working card is still not working.

[jhasler]

Paul_L writes:
> On the behavior I was thinking that the problem might be hardware so form my
> experience the series resistors between processor and sd card socket of 32
> ohm sometimes are to high for SPI so I tried to put some 10 ohms in parallel .

Well, considering the short lines and low loading I don't think those resistors can cause significant signal degradation, and they will protect the Pi against various card faults. Still, they aren't in the spec. You might try jumpering them all since you have the tools for working on SMD. It seems to me that there should be pullups on those lines, though, if only for static protection (see
http://www.flashgenie.net/img/productma ... 2final.pdf).

Do any cameras and/or phones use the SPI mode? It occurs to me that the Chinese clone makers might not bother to implement it properly if it is not used in common consumer products.

I know that the boot code is proprietary but does the foundation have access to the source? Since the card interface seems to be pure bit-banging that may be where the bug is.

It would be interesting (though difficult for obvious reasons) to learn how well the current Linux driver works with the troublesome cards (I know it works fine with mine).

[Paul_L]

It is indeed a question if Pi uses spi mode or sd card mode. I think it is using sd card mode . I also was thinking about some pull ups . In this moment I am not able to make some tests anymore since I am leaving for one week to a project abroad. Maybe some other can test some pull-ups as recommended by specs. Anyway it is only a card, capacity is low and quite short lines. Pull-ups might not improve much the signals. Jumpering the R46 to R50 does not helped a card that was not working. That issue with cooling the Pi and starting booting is quite interesting thus. Should be investigated more

BTW The name of the topic has to change since with the new firmware launched a few hours ago we have red AND green led lightning when is not booting.

[jhasler]

Paul_L writes:
> Pull-ups might not improve much the signals.

I don't think they'd improve the levels or the timing but in their absence floating lines might confuse the card as the spec implies that it expects inactive lines to be pulled high. "Open drain mode" is mentioned. I may try adding pullups.

I haven't seen the official SD card spec from sdcard.org: that's "secret".

[Paul_L]

I can confirm your cooling tests. I was able to boot a never booting card with the new firmware "raspberrypi-firmware-51840bd " AND cooling the processor with liquid propane. I was able to boot three times. The temperature has not to be very low . I am cooling it up to propane ice is over the processor. It is not booting since the propane ice disappear.

[jhasler]

Added 22k pullup to pin 2(CMD). No help. Added 22k pullups to pins 7, 8, and 9 (DAT0, DAT1, DAT2). Boots ok now without chilling. From the SanDisk data sheet for the card (http://www.flashgenie.net/img/productma ... 2final.pdf:

The extended DAT lines (DAT1-DAT3) are input on power up. They start to operate as DAT lines after the SET_BUS_WIDTH command. It is the responsibility of the host designer to connect external pullup resistors to all data lines even if only DAT0 is to be used. Otherwise, non-expected high current consumption may occur due to the floating inputs of DAT1 & DAT2 (in case they are not used).

After power up, pin 1 is input with 50Kohm(+/-20Kohm) pull-up (can be used for card detection or SPI mode selection). The pull-up may be disconnected by the user, during regular data transfer, with SET_CLR_CARD_DETECT (ACMD42) command.

CMD (pin 2) may not need a pullup but the spec seems to imply that it should have one. DAT3 (pin 1) already has a pullup in the card and an external one might interfere with card detection. Pullups are clearly required on pins 7, 8, and 9. CLK (pin 5) probably should have one as well.

[dom]

@jhasler
Interesting. The software in the bootrom applies pullups to GPIO pins 48 to 53 (CLK, CMD, DAT0, DAT1, DAT2, DAT3) before accessing the sdcard.

[jhasler]

The software in the bootrom applies pullups to GPIO pins 48 to 53 (CLK, CMD, DAT0, DAT1, DAT2, DAT3) before accessing the sdcard.

Therefor there is at least a brief period at powerup during which the pins may be in an unknown state, possibly an open circuit or a voltage in the indeterminate range. This might be long enough for some cards to lock up. Do the resistors remain connected without interruption while the lines are switching between input and output? What is the value of those "software" pullups? Is the state of the pins during and immediately after powerup documented? I would expect them to come up as inputs.

This might be a red herring. It's possible that all I've done is shifted levels and/or timing just enough to get this particular card to work. It would be good to know if this solution works for those other "bad" cards.

[dom]

Some info on GPIO pull up/downs here:
http://www.raspberrypi.org/wp-content/u ... herals.pdf

[jhasler]

Some info on GPIO pull up/downs here:
http://www.raspberrypi.org/wp-content/u ... herals.pdf

Intereresting. As I read that the pins you are using to drive the card are in a pullup state when power is off (and thus presumably remain that way during the powerup transition) and the programmed pullup/pulldown state is nonvolatile. Thus, assuming that the pullup resistor value is less than 100k there should be no need for external pullup resistors (and also assuming that the data sheet is infallible...)

[Paul_L]

Added 22k pullup to pin 2(CMD). No help. Added 22k pullups to pins 7, 8, and 9 (DAT0, DAT1, DAT2). Boots ok now without chilling.
.

You have keeped the 33ohm resistors? On which side you connected the pull-ups ? Card or CPU ?

[jhasler]

You have keeped the 33ohm resistors?

Yes.

On which side you connected the pull-ups ? Card or CPU ?

On the card side, but I doubt that it matters.

BTW both my cards (one with older bootcode) have now booted successfully ten times in a row without chilling.

[Paul_L]

Now we have to wait for someone else to test. I don't think that I will be able to test before next week. But my feelings were confirmed that is a hardware matter, even the hardware is minimal here. Has to be reconsidered the schematics. Might be a good idea to post a picture of your pullups installation so other guys to test easily .

[jhasler]

Might be a good idea to post a picture of your pullups installation so other guys to test easily .

1/8 watt resistors tack-soldered onto the connector and hanging in the air? You'd find it painful to look at.

I'll come up with something more "elegant" once I'm sure this is the solution.

[Paul_L]

I am pretty sure that the solution is there , in the connection of SD card. Your solution for testing is accessible to many users that have soldering skills and a good solder iron. The space is enough there to put also some SMD but has to connect all to 3.3 V that is not so easy .

[jhasler]

The space is enough there to put also some SMD but has to connect all to 3.3 V that is not so easy .

I'm planning on a SIP resistor package glued to the bottom of the board and hooked up with wirewrap wire.

[dom]

@jhasler
Can you use this:
https://dl.dropbox.com/u/3669512/temp/start.elf
and report what:
/opt/vc/bin/vcgencmd otp_dump
returns. (The sdcard pullups are enabled by an OTP bit - I want to check yours is programmed correctly).

[jhasler]

@dom
pi@raspberrypi:~$ sudo /opt/vc/bin/vcgencmd otp_dump
1687139347: vchiq_lib: Very incompatible VCHIQ library - cannot retrieve driver version
VCHI initialization failed

[dom]

@dom
pi@raspberrypi:~$ sudo /opt/vc/bin/vcgencmd otp_dump
1687139347: vchiq_lib: Very incompatible VCHIQ library - cannot retrieve driver version
VCHI initialization failed

You will have to update your firmware:
http://elinux.org/R-Pi_Troubleshooting# ... g_firmware

(then replace with my start.elf).