How to make raspberry stable ?

[MORA]

Hi,

I have been working with the Rpi for a while now, to make a stable sensor platform.
At first I hit the SD corruption problem, so after some guides I got my root mounted as read only, only to find the network didnt stay up for more than 1.5days or so.
That was on a 1A supply without x running or even a monitor attatched, no monitor also means I cant see the error since I didnt have hdmi force option, so I had to force boot it by pulling power.
This happened 2 times, with no problems, its readonly after all.

On 3rd go it didnt boot up again, says kernel panic, cant mount root fs
I reinstall SD card today, boot it up and it does not detect the USB keyboard that I usually use...

I connect it to a lab supply to see how low voltage can go to make problems, at 4.7 there are some indications of problems, at 4.6 the screen goes off and starts flickering on and off again.
I power off, restore 5V and turn on again, still the screen flicker...
The supply showed a usage of about 250-300mA so the 1A should have been more than enough.

I will reflash the SD again and see what happens, maybe try one of my other raspberrys, but please tell me, how can I make this product stable enough to be used ?
I see many problems have SD corrupt problems, network dropping, etc.
I am really losing hope in this project, if its this fragile my idea of a stable sensor platform is dead in the water.

[stupid_boy]

first of all, don't hurry!
don't hurry to hang up with RPI. it can be stable enought, and during time you will catch limits that you must not cross to keep it stable.

be sure you have reliable power supply. phone chargers, and all similar sized/formfactored ones usually aren't.
next you have to try moving root partition on usb flash drive. i use usb HDD, but it consumes more energy. this will eliminate card corruptions. well, not all. they are still possible when you update firmware/kernel, but at this point you will not loose everything and you will go with simple restore and re-update of non critical part of the system.
next you may want to play a little with system settings. lan, overclock, etc. during this step you eventually will change places of some peripherals and will determine their best configuration. may be you will try different peripherals, as lan vs WiFi for example, at this step too.
next you will determine which programs are and which are not suitable for your needs. some are long runners, some can cause troubles. eventually you will continue with system tuning at this step.

following these easy steps, i thing mine RPI currently is stable. i know what can cause system crash and i am prepared for it when i do something. i know how to touch system settings to overcome this crash too. but, this require some time and some wish for success. there is not a ready working and universal recipe for this.

remember, RPI is learning tool. how better you can use it if not to learn how to get it perfectly stable and predictable? all depends on you.

[leexgx]

its most likely still the power supply your using, does not matter how much the charger "says" it can do its if it can't do it cleanly,

network,keyboard and mouse issues tend to point towards your power supply (voltage needs to be around 4.8-5.2v (4.75v been the lowest) http://elinux.org/R-Pi_Troubleshooting# ... r_problems

is it an branded charger (motorola, HTC that you got with your phone) or china nock off (apple charger or random unbranded charger)

[MORA]

The keyboard not working was with a lab supply, im fairly sure thats stable, since thats kinda what they are made for.

The charger I used in the first few days was a unbranded one, I have a HTC one too, but I dont know if thats much better.

I see RS has one linked from the raspberry page, that is not too expensive, that must be good enough when they link it ?
http://uk.rs-online.com/web/p/stromfors ... 430702|rel

[leexgx]

The keyboard not working was with a lab supply, im fairly sure thats stable, since thats kinda what they are made for.

The charger I used in the first few days was a unbranded one, I have a HTC one too, but I dont know if thats much better.

I see RS has one linked from the raspberry page, that is not too expensive, that must be good enough when they link it ?
http://uk.rs-online.com/web/p/stromfors ... 430702|rel

try the HTC charger as that's more likely to be stable then the other two PSUs you have the "RS online one is likely an china copy" it only needs to be 750mA as i am using now (charger of my motorola headset)

plug an unstable power into an phone the touch screen on your phone will norm go funny (does on my phone and other phones before that when i have used bad PSUs)

the PSU could provide 5A does not mean its an good PSU

[bgreat]

I would try replacing the power supply and micro-usb cable being used for the power.

I find the Raspberry Pi to be very stable. I have three currently running 24/7. All three run overclocked with Raspbian Wheezy (Dec-2012) as the OS.

My first Pi is a Rev. 1 256MB running as a headless server on wired Ethernet. It been up for 40 days without issue. It provides VPN service, some web services, and home monitoring/control. The last time it was down was when I did a OS upgrade to Raspbian. I use an Apple power supply with it.

My second Pi is a Rev. 1 256MB running as a development and test platform. It is set up with a Motorola Atrix keyboard/trackpad/display dock. The first USB port is connected to the Atrix and the second has a USB Wireless interface. I routinely leave it up and running for over a week. It is only down when I connect/disconnect hardware for testing or update the OS. It is powered from the dock power/USB input connection.

My third Pi is a Rev. 2 512MB running as a headless development and test platform with USB Wireless interface. I routinely leave it up and running without problems for over a week. It is only down when I upgrade the OS or test new hardware interfaces. I use a 2.1A 5V adapter by Everready.

When I started using my first Pi, I did have stability and SD corruption problems. The problems were isolated to two issues.

The first issue was a poor choice of power supply. I had an HTC 5V 1A supply that did not provide stable power. It was the primary contributor to the instability.

The second issue was due to the mouse, keyboard, and wireless interfaces drawing too much power over the USB connection. I found this a little difficult to isolate, as when I tried a USB hub the problems persisted. It turned out my hub was also a poor choice. I bought a cheap mouse and keyboard ($7.50US) and the problems went away. I also picked up a newer Wireless interface ($9.99US) and solved the remaining problem when running headless.

I have worked with and designed computer systems and controllers for many years (>25) and can appreciate the power supply requirements for system stability. Still, I was caught by surprise when working with Raspberry Pi as to the sensitivity my first board. My ultimate test to convince myself that power was the issue was to connect the Pi to a lab bench supply and no USB peripherals. When the problems went away, I then tested with various combinations of my USB devices. It turned out that the "old" devices I pulled out of my scrap box were just too power hungry for the Pi and were causing the problems. The average currents measured as being acceptable, but apparently the instantaneous/switched demand was not compatible with the Pi design.

The above sounds complicated, but I identified my issues over a weekend using input from the forums and my own experiences. Hopefully you can successfully resolve your problems also.

Enjoy!
Bill

[stupid_boy]

be a china one does not mean it not work. i use one like this:
http://www.ebay.com/itm/25W-DC-5V-5A-Re ... 27b9ca5323
it is china made, yes! but it is great! currently it powers all i have connected to my RPI. i do not have any other power supply in entire system. well, i have to rewire power line of my USB hub too.

while you change your supplies, try with different cables. i test all that i have with no success. now i use regular A-A usb with converter to micro usb. even not all of my AA cables work for me. if you know what you are doing, you can test with your lab supply directly to GPIO pins.

RS are simply sellers. they recommend what they have in warehouse and what they want to sell.

[leexgx]

be a china one does not mean it not work. i use one like this:
http://www.ebay.com/itm/25W-DC-5V-5A-Re ... 27b9ca5323
it is china made, yes! but it is great! currently it powers all i have connected to my RPI. i do not have any other power supply in entire system. well, i have to rewire power line of my USB hub too.

while you change your supplies, try with different cables. i test all that i have with no success. now i use regular A-A usb with converter to micro usb. even not all of my AA cables work for me. if you know what you are doing, you can test with your lab supply directly to GPIO pins.

RS are simply sellers. they recommend what they have in warehouse and what they want to sell.

that looks quite good built PSU,
i just mean the cheap 3-8v or 5v USB chargers that you can get for £/$5-10 (i guess that counts as Unregulated PSU) only used Motorola phone and Motorola headset Bluetooth charges and they seem quite happy with that (my dads Sony phone seems to hate every charger apart from moto charger, the screen goes all wacky and does not charge if i use any other charger)

[obcd]

Measure the TP1 - TP2 voltage on the board. It should be at least 4.85V for a stable operation.
A micro usb cable can lose between 0.2 and 0.4V
The polyfuse on the board also can lose some voltage.
I use the RS power supply. It's a 1.2 amp supply, and it has the wire fixed connected to the supply.
(not much cable losses) I soldered another polyfuse over the already present one.
It also increased the voltage with 0.15V
I haven't noticed wired ethernet failures, but wireless was going out untill I disabled the power managment in the etc/network/interfaces file.

[MORA]

Thanks for all the responses.
I have switched to the HTC charger and it runs for now.

The testpoints are however showing 4.7V so its a bit on the low-side, and thats sitting in a terminal, although the CPU is probaly peaked.

Even if there were 5V at the other end of that cable it would drop quite a bit.

Since I am going to be doing an expansion board anyway, I am wondering if it would be easy to implement a proper power circuit to supply a stable 5V supply for the raspberry.
I have done linear regulators before, but with 1A+ thats not really such a good idea

I really wish they would have spent those 2$ on a proper regulator instead of everyone having to spend 10-30$ on a supply that will deliver near-perfect voltage across 1-2meters of usb cable.

[pluggy]

Make sure you've got a quality lead between the PSU and the Pi, a cheap (read thin, high resistance) lead can cause a large voltage drop. You can also check the polyfuse by checking the voltage drop across it on the bottom of the board. An alternative with a bench supply is feeding it via the GPIO pins. You have a bit more control of the gauge of the cable and it takes the polyfuse out of circuit. The CPU is a minor player in current consumption, on a 'B' the lan9512 chip uses more than the SOC, and the GPU is the lions share of the SOC. How much voltage you need to get stable varies by the Pi, 4.85v across the test points should be OK on most.

[MrBunsy]

I've got one which has been running for over 200 days, and I'm not the only one who's been running them for long periods of time (there was a thread of competing longest uptimes, can't remember where). I dare say there are some faulty pis out there, but I don't think there's any inherant reason why they can't be stable.

As an idea, I wonder how much of a impact different makes of SD cards have.

[rpdom]

As an idea, I wonder how much of a impact different makes of SD cards have.

They may have some effect. I believe some of the Class 10 cards have issues.

However, not as much as a good, stable, power supply (don't always believe what it says on the label!), and a good quality micro USB power lead.

A cheap digital voltmeter/multimeter is a good investment for checking the voltage supplied to the Pi, and is often useful around the home for testing batteries and things.

[remsnet]

Hi,
..
.

Don´t juge the PI for the Linux kernel.

My self had only issues when using the 3.1.x/3.2.x kernels and rarly used featgers.
at the start at nov 2012 i had issues with the kingston / sandisk 64G sd cards class 10 due the 3.2.x kernels

With Current 3.6.11 kernel and up ( asume you updated you raspberry ) the SD card issues be gone due the SHDC / SHDX Fixes made past 20 weeks..

The RPI Coders do theire best - and i thanks them .

Otherwise there been an option to move the the betta Firmware .


PI´s are rock solid - make NO Noise ..
since Nov 2012 i transported Mulible Terrabytes thru it - with using fuse zfs i.e !!

[jbeale]

I have a Model B Rev.1 R-Pi with uptime now about 144 days. It is running a small python program continually logging some data to CF card and uplinking it every 30 minutes. There is the usual mess of background processes, including ntpd to keep time synchronized to the network. The data is coming in through GPIO so it is not using a USB-Serial port. It has no wired-ethernet link and just one USB device, which is a Realtek RTL8188CU WiFi device from ebay, for the network link.

I have another device, Model A R-Pi plus a Rosewill brand 4-port hub, containing a keyboard, two USB-serial devices (1 Arduino, one generic FTDI serial device) and a RALINK RT5370 WiFi device. I'm feeding it 5V through the GPIO pins so there is no drop across the input fuse. TP1-TP2 measures 4.9 V. The R-Pi itself works OK (current uptime only 2.5 days due to recent reconfigs), but the Ralink RT5370 quits every day or so, and has to be removed and reinserted to restart.

I have a third R-Pi, which is sometimes used as a media center running XBMC, it seems to work fine but I don't leave it on 24/7.

I do not see CF card corruption on any of my Pi devices.

You can see some people with long uptimes posted here: Longest Raspberry Pi uptime known? http://www.raspberrypi.org/phpBB3/viewt ... 26&t=18080

[minaya]

They may have some effect. I believe some of the Class 10 cards have issues.

And also class 6 (the one that was burned by the Pi), class 4, class 2 (with both I get stability issues). One friend of mine also managed to kill a NEW class 10 card with his Pi. After looking at the PI and raspbmc forums, it doesn't seem to be a isolated problem of some units ...

The PSUs I tested so far were different kinds of phone chargers (like 4-5 of different brands) a PSU taken from an external disk (5V/2A), with a format similar to those we find in laptops and so. The PSU I use in the SheevaPlug is the cheapest I found in ebay at the time and it worked like a charm.

The PI I have is a RS 512mb one. Maybe it's stability improves if I put some kind of filtering stage (with capacitors of different values) at power input?.

However, not as much as a good, stable, power supply (don't always believe what it says on the label!), and a good quality micro USB power lead.

A cheap digital voltmeter/multimeter is a good investment for checking the voltage supplied to the Pi, and is often useful around the home for testing batteries and things.

Well, I've already tested the supplied voltage and all seems correct. In fact, I've already tested these PSUs with devices like disks, phones and so.

[remsnet]

They may have some effect. I believe some of the Class 10 cards have issues.

And also class 6 (the one that was burned by the Pi), class 4, class 2 (with both I get stability issues). One friend of mine also managed to kill a NEW class 10 card with his Pi.
.

Well , I killed 2 new 64G sd ´s class 10 due my OpenSuSE RPI experiments past weeks on my Model B ( 512MB Pi´s ) .

IT was NOT the PI killed it - it was an bady kernel release.
so i Joined GitHUP again after +10j absense at the kernel scene ..

And as i remeber i had to replace 2 bad powercords as well - not all have an good quality.
i.e mico-usb "data cables" have thiner copper .

if you Found an BUG - REPORT IT !

so again : Don´t Juge the RPI for the Software i.e kernel running on it

[hormiguero]

I have several Pi's running remotely (over 1000 miles away - headless). They are very stable - now. I had problems as described at the start. I read all of the posts about "... your problem is the power supply..." and I checked the supply with a good scope and couldn't see any appreciable noise. Still the problem persisted... I finally wired a 33000uF, 16v cap across one of the cheap supplies and the problem went away!! I now either use a GOOD power supply or automatically add the cap to the cheap supply!

Also, got some cheap SD cards at Fry's and had problems - Sandisk, Polaroid, Sony, and Kingston seem to be the most reliable that I have tried. Patriot don't seem to work well, and I have had problems with the micro-SD cards and adapters that I have tried. The unreliable cards would run for a few days and then corrupt - changing to the cards that are more reliable (and, yes, more expensive) the problem went away without changing anything on the power end.

When I use a Pi for data collection using the GPIOs, I use a coaxial 5v power receptacle and 1A polyfuse on my homemade IO board to power the Pi thru the IO connector. I really don't like the micro USB connector! I think that might also contribute to the power problem.

[jbeale]

I have several Pi's running remotely (over 1000 miles away - headless). They are very stable - now. I had problems as described at the start. I read all of the posts about "... your problem is the power supply..." and I checked the supply with a good scope and couldn't see any appreciable noise. Still the problem persisted... I finally wired a 33000uF, 16v cap across one of the cheap supplies and the problem went away!! I now either use a GOOD power supply or automatically add the cap to the cheap supply!

Just speculation on my part, but this may be a combination of two problems (1) cheap supply with too-small filter capacitor and (2) poor power line quality, for example with short brownouts / glitches that occasionally drop your supply voltage. This can happen when a motor starts up, even a small-ish motor like a refrigerator compressor or a water pump, if you've got a weak enough power connection.

[9000reasons]

I have two Pis and they are complete opposites with regards to stability.

The older Rev1 is an absolute pain. It is very flaky and when set up as a small file server frequently stopped responding. No matter what PSU, SD card, accessories or what-not this Pi just is not stable at all. It fails to last a day sometimes!

The newer 512Mb Pi with the same exact SD card has been running without fail for 50 days now and is super-fast to respond. I have had no problems with it so far and it performs quite a lot of services to multiple Windows machines and Android phones.

When I compare PCBs of the two Pis I can't help but feel that there is a difference in build quality between them.

[mikroskeem]

i ran my pi 2 weeks, and it suddenly stopped responding over ssh. good that i had ZNC running and i could access shell and poweroff the Pi. I use my router's usb port to power my raspberry pi(measured 3A and very stable, 5.1V). And if i booted, systemd segfaulted(arch linux). I have rpi rev. 2 model B, class 4 Toshiba SD card, 1,25v overvolt on cpu, sdram, 950mhz, core freq 350mhz. Now i can't even acces sd card thought computer, superblocks are corrupted. I blame that rpi gave too much volts to sd card. Any hardware modifications to sd card wiring to eliminate sd card overpowering?

[jamesh]

i ran my pi 2 weeks, and it suddenly stopped responding over ssh. good that i had ZNC running and i could access shell and poweroff the Pi. I use my router's usb port to power my raspberry pi(measured 3A and very stable, 5.1V). And if i booted, systemd segfaulted(arch linux). I have rpi rev. 2 model B, class 4 Toshiba SD card, 1,25v overvolt on cpu, sdram, 950mhz, core freq 350mhz. Now i can't even acces sd card thought computer, superblocks are corrupted. I blame that rpi gave too much volts to sd card. Any hardware modifications to sd card wiring to eliminate sd card overpowering?

I doubt its overpowering - more likely the issue that Gordon just pushed a fix for - check out the Raspi github, latest firmware, and some of the SD card corruption threads that are on here.

[mahjongg]

I have two Pis and they are complete opposites with regards to stability.

The older Rev1 is an absolute pain. It is very flaky and when set up as a small file server frequently stopped responding. No matter what PSU, SD card, accessories or what-not this Pi just is not stable at all. It fails to last a day sometimes!

The newer 512Mb Pi with the same exact SD card has been running without fail for 50 days now and is super-fast to respond. I have had no problems with it so far and it performs quite a lot of services to multiple Windows machines and Android phones.

When I compare PCBs of the two Pis I can't help but feel that there is a difference in build quality between them.

Maybe one PI has a non recovered Polyfuse, and the other one not.

[eNORm]

Are these problems solved with the new B+ modell?

-eNORm-

[joan]

Are these problems solved with the new B+ modell?

-eNORm-

Which errors? Errors caused by poor power supplies and thin cables can not be solved by the Pi.