Temp Sensitive Boot Problem

[Jongoleur]

In the interest of getting some basic temperature figures from a Pi that appears to be working in a sensible manner, I've waved an (admittedly fairly cheap and inaccurate) infra-red thermometer over the various components of a Pi that has been running for several hours. It has been rebooted several times during this period and restarted flawlessly.

CPU/RAM 36C/96.8F
LAN/Ethernet 40.9C/105.6F
Big voltage reg by large capacitor 38.3C/100.9F
SDHC card 40.8C/105.3F at connector end, 25.6C/77.3F at other end

SDHC card is a "SiliconPower" 8GB, Class 4 device sourced from 7DayShop (cheapest they had!)
PSU is the one sold by RS. 5v@1.2A.

Ambient temperature of the board before initial power-on was 21C.

Do the problematic Raspberry Pis exhibit different temperature profiles?

I'll get hold of a infra red thermometer and ill post my figures. I have successfully fixed the issue by using an almost identical memory card it definitely something to do with memory cards and temperature though.

Ok, I`ve done some tests. Here is what I found.

The problem with the temperature is from the SD card.

I tested 4 cards:

1) Kingston, 4 GB, class 4
2) Kingmax, 4 GB, class 2
3) Kingmax microSD in SDHC adaptor, 4 GB, class 4
4) Sandisk Ultra, 4 GB, class 4

All cards had a fresh copy of the latest debian image available. Same settings, etc etc.

I booted the systems, started LXDE, opened a terminal , run apt-get update && apt-get upgrade and also ran rpi-update. This caused the Pi to be under heavy load.

1) The kingston card rebooted everytime no matter the load - SUCCES
2) The kingmax class 2 card rebooted everytime no matter the load - SUCCESS
3) The kingmax microsd card won`t reboot when it warms up - FAIL
4) The Sandisk Ultra won`t reboot when it warms up - FAIL

I am safe to say that the thermal problem is from the SD card and(or) software and not a hardware issue.

Playing with these cards I also ran some performance tests except the microsd Kingmax.

I measured the boot time:

1) Kingmax class 2, 4 GB - booted in 2:02
2) Kingston class 4, 4 GB - booted in 2:05
3) Sandisk class 4, 4 GB - booted in 1:48

Also measured the output from hdparm -tT /dev/mmcblk0

1) Sandisk Ultra cl4,4 GB - Cached reads 136 MB in 2.01 seconds = 67.54 MB/sec
Buffered disk reads 14 MB in 3.16 seconds = 4.43 MB/sec
2) Kingston class 4, 4 GB - Cached reads 134 MB in 2.00 seconds = 66.89 MB/sec
Buffered disk reads 14 MB in 3.25 seconds = 4.31 MB/sec
3) Kingmax class 2, 4 GB - Cached reads 116 MB in 2.00 seconds = 57.95 MB/sec
Buffered disk reads 10 MB in 3.23 seconds = 3.10 MB/sec

Conclusion, in the real world.. the class of the card isn`t that important.

P.S. The boot time is so big because after you change the keymap from gb (default) to whatever (us) it hangs a lot at "Setting preliminary keymap". I think the Pi will boot in about 30-40 seconds when that bug will be solved.

I would also like to mention that apt-get upgrade on the class 2 Kingmax ran almost twice as fast as the Kingston and considerably faster than the Sandisk Ultra.

Got round to my 'ice pack' test earlier this evening. I posted about it in another thread but I realised I should have reported back here (too). So...

Pi shuts down soon after powering up at random points

As a last-ditch attempt to diagnose a temperature-related problem, I strapped a (suitably wrapped) ice pack to my Pi this evening. This seemed to cure the random shutdowns and the Pi booted and remained stable/functional for much longer than it ever did before. In fact it didn't hang. Bearing in mind that I had not changed anything about the set-up from my last attempt(s) this seems to have proved that, when warm, my Pi was unstable as it was stable when cooled. The next step was to unstrap the ice pack, reboot the Pi, and see if reverted to random shutdowns once it had reached normal temperature. It's now been running for a good hour and hasn't hung yet even when I elevated its ambient temperature with gentle heat from a hairdryer!!

I'm still inclined to believe that there's an intermittent fault with my Pi and that the 'ice pack' fix is only temporary. I guess time will tell. But it might be worth others conducting a similar exercise and chilling the Pi with a poly-bag wrapped ice-pack to see if it becomes more stable.

It's worth emphasising that I doubt it's power-related - ditto the SD card - since I'm using the same PSU and SD card that were used when the Pi was hanging before. Perhaps, the 'chill' has helped close a minute fracture. Who knows!?

I'll leave the Pi running for now and will report back if it starts acting up again. Any suggestions for the easiest way to stress a Pi without user interaction? E.g. a simple script that'll impose a continuous 100% load?

I managed to get my Sandisk Mobile Ultra 8 Gb that gave me mmc0 timeout to work with the latest firmware copied through a live linux cd but won`t reboot like the Sandisk Ultra SHDC 4 GB class 4 due to the temperature problem.

The only cards that work for me are the kingston class 4 and the kingmax class 2, both 4 GB..

I too had issues with a SanDisk card. Maybe they have compatibility issues? I bought a no name class 10 16gb memory card from ebay for £6.70 and it works flawlessly now

seems to me power supply and SD card compatibility are two of the big issues so far with the R-Pi. If you have anything specific to add to the "Known SD Cards List" either on the "Working" or "Not Working" side, it would be helpful for others.

http://elinux.org/RPi_VerifiedPeripherals#SD_cards

(maybe there should also be a "sometimes working" category?)

Edited the wiki, thanks for the heads up.

I have read most of the suggestions by others, and have tried to eliminate the known issues.
I bought a 2.1Amp USB power supply to rule out any power supply issues. I don't have anything but a USB keyboard and USB mouse plugged into my raspberry pi, and I am also using a powered HUB, so the USB ports aren't even pulling power from the PI's power supply anyway. The power is not the issue.
And the USB Fuses are not the issue. I bought an 8Gb SD-HC Kingston Class 4 card and loaded the latest Debian Squeeze onto it. I believe the problem is related to a cold solder joints in the POP (Package on Package) stack. Stacking the CPU and RAM on top of one another and getting them perfectly aligned has to be a tricky operation. I suspect that the manufacturing process in China has some issues with this soldering process and that is causing poor solder joints to the CPU or RAM.

Since I am not the only one experiencing these problems, I suspect that there are probably hundreds of bad Raspberry Pi's out there. I have no idea what the good/bad ratio is, but if the Raspberry Pi foundation cannot correct this production problem soon, I fear it will spell doom for the Pi as an ADD-IN product to other commercial devices. Perhaps students, can afford to order two or three units go get a good one, but professional businesses, can't tolerate failure rates at these levels.

I thought this too but I recommend trying as many possible different sd cards that are 2gb or bigger I had to try three before It work. The one and worked and one that didn't work were seemingly identical. Both genuine SanDisk cards bought from a reputable store. Try every memory card you can it seems to have solved most people's issues. The Pi seems ridiculously picky with SD cards I hope this is a software issue and not hardware otherwise there's a serious problem like you said businesses nor even the main target audience, schools will be able to buy them in if their so unreliable. Once I got it working though it's worked perfectly. I got a no name 16gb class 10 card from ebay. It seems good cards/brands don't always mean the pi will like the card.

My copy of the Raspberry Pi will not operate for more than 5-minutes without locking up completely.
(Unless I cool the CPU down with cold spray every 20-seconds or so). This kind of reliability is completely intolerable. I am going to talk to Premier Farnell to see if I can get a replacement.

Looking at the OK light I suspect that my particular problem is related to communications with the SD card. Program loading from the SD card causes RAM corruption, and data saved to the flash gets corrupted as well. The symptoms are that after about 5 minutes of operation the programs seem to slow down, and lock up or stop responding.

When you reboot your pi does the red led and no others come on? If so this is identical to what happened to me

Have you seen this thread? Does the new firmware & kernel today help your SD card problem?

"There was a new kernel pushed to github today. It includes another patch for failing sdcards. [...]"
viewtopic.php?f=2&t=6093

Apparently it has fixed some issues at least.

Have you seen this thread? Does the new firmware & kernel today help your SD card problem?

"There was a new kernel pushed to github today. It includes another patch for failing sdcards. [...]"
viewtopic.php?f=2&t=6093

Apparently it has fixed some issues at least.

I have tried that new firmware/kernel and my Pi still hangs after a short while. Usually it gets to the login prompt phase of booting and most of the time I get to remotely login. I also ping the Pi so I can monitor its connection. However, normally, within a few minutes the Pi hangs, the Ethernet connection is lost and the HDMI output dropped (i.e. the tv complains that the signal is lost). The 'time-to-hang' is variable but usually less than 5-10 minutes and sometimes just a few seconds. In my experience keeping the Pi 'chilled' results in reliable operation - hours of uptime.

This hang happens with all 3 SDHC cards I have - all of which are deemed to be 'good' cards according to the wiki list. However, I have noted that with the new firmware/kernel the dmesg log reports a problem with reading the SD status register. Here are the specifics:

Kingston 4GB SDHC Class 4 (SD-K04G 1150 WY2271N):

Code: Select all

Waiting for root device /dev/mmcblk0p2...
mmc0: problem reading SD Status register.
mmc0: new SDHC card at address 1234
mmcblk0: mmc0:1234 SA04G 3.63 GiB
 mmcblk0: p1 p2 p3


Kingston 8GB SDHC Class 4 (SD-K08G 1142 WY0362L):

Code: Select all

Waiting for root device /dev/mmcblk0p2...
mmc0: problem reading SD Status register.
mmc0: new SDHC card at address 1234
mmcblk0: mmc0:1234 SA08G 7.42 GiB
 mmcblk0: p1 p2 p3


Transcend 8GB SDHC Class 6 (MMBFG08GWACA-M6):

Code: Select all

Waiting for root device /dev/mmcblk0p2...
mmc0: problem reading SD Status register.
mmc0: new SDHC card at address b368
mmcblk0: mmc0:b368 00000 7.51 GiB
 mmcblk0: p1 p2 p3


Nevertheless, I still believe that the issue I have is (at least partially) Pi-related and not (just) due to the other components in the system.

I hope my posts on this issue have been of some help in diagnosing the problem. I've certainly spent far more time rebooting/debugging the Pi than I have actually using it

I wonder if this post below is relevant to the situation. USB and LAN would not work if the USB/LAN chip is not supplied with a clock! Maybe something about that clock package is temp. sensitive? But bad solder joints somewhere is also a common issue on PCBs.

viewtopic.php?f=28&t=5766&p=80624#p77576
The 25MHz clock for the LAN9512 was missing.
After applying an external clock (using a NI frequency generator), the USB and ethernet worked immediately.

PROBLEM-SOLVED!!!

When I combined the Known-good Kingston SDHC 8Gb Class4 sd card with a known-good high-current power supply with LARGE-CONDUCTOR wires, measuring a good voltage on the test points, then my problems did indeed vanish! Thanks to Raspberry Pi foundation for publishing the new "Troubleshooting Wiki" This new wiki clearly shows how to resolve the common issues with the raspberry pi.

PROBLEM-SOLVED!!!

When I combined the Known-good Kingston SDHC 8Gb Class4 sd card with a known-good high-current power supply with LARGE-CONDUCTOR wires, measuring a good voltage on the test points, then my problems did indeed vanish! Thanks to Raspberry Pi foundation for publishing the new "Troubleshooting Wiki" This new wiki clearly shows how to resolve the common issues with the raspberry pi.

Thanks for the info. This is further evidence that my Pi is faulty since I believe I've satisfied the conditions for a known good SD card and a good PSU - a nokia AC-10X 5V 1200ma - that supplies at least 4.9V between the test points.

How odd I only have 4.6v between the point and mine works. I think it's the current not voltage that's the issue. I think I shall set up a much more comprehensive SD card wiki Page so we can start figuring it out.

[mahjongg]

Actually 4.6 Volt is way below the level with which the R-PI (and most other 5V logic) should reliable work, as the lower tolerance for such devices is officially 4.75 Volt. So count yourself lucky!

Or perhaps your multimeter isn't calibrated correctly...

In any case, if I hear about "temperature sensitive and power problems", the first thing that comes to mind is "polyfuse", as these have a habit of changing their internal resistance as a function of temperature, and therefore can cause voltage drops leading to problems when heating up.

I'm not saying that in this case polyfuses are to blame, as the rest of the symptoms do not point in that direction, I'm only giving a tip. If you heat a polyfuse a little bit with a hair dryer set to low, and the problem appears then the polyfuse is a suspect. Cold spray will not "reset" the fuse immediately by the way, it has to settle over several hours.

But note the SD-cards are 3V3 powered, which means they are quite immune to a 5V drop of several tenths of a volt, as the 3V3 regulator can still function with only four volts or so at its input.