USB bus seems to be shutting down after 30seconds

[Helpme1986]

oh and i have done a ArchLinux card as well, seems to last longer on that (and when im not using it doesnt seem to stop)

[Chromatix]

That does sound like a dodgy power supply to me.  A lot of cheap "phone charger" type supplies are actually pretty weak - all too often the rating on their label is, charitably speaking, optimistic.  Dropping to 4.8V under "normal" load is certainly suspicious.

On a desktop PC, randomly cutting out under load is a common symptom for a bad PSU.  Because the R-Pi consumes so little power overall, USB's power consumption looks that much larger next to it, so it can also trigger the problem.

[Helpme1986]

right found a more stable way

with keyboard and mouse (via powered USB, each only 100mA anyway) 4.84v

when ethernet is plugged in 4.79v

when launching LXDE it drops to as low as 4.73 but once loaded goes back to 4.79

all of a sudden mouse n keyboard went out now at 4.93v (ethernet still on)

ethernet dropped 20secs after now a constant 5.05v

[abishur]

Can we get some more people to test on their pi?  It does sound like a questionable PSU especially since we aren't hearing about this happening to others.

RazorD said:


I think i've got the issue too - After a little bit, ssh just returns "Connection refused".

I'm powered by a USB port from a powered hub. Nothing appears on error-ry on the HDMI output...



You have it powered through *A* USB port on a powered hub?  Is it one of those super special yellow ports that provide more than 500 mA?  Otherwise 500 mA will not be enough to power the pi   You could use one of those "Y" cables to get 500mA from two ports for a total of 1A to power the Pi.

[jbeale]

with the voltage rising as things cut out, it sounds like the thing that stopped working has stopped drawing power, and with the reduced load the voltage comes up. That makes some sense, but it's hard to tell what is the cause and effect- ideally you'd like to know what was the voltage immediately before the device quit working- might need a scope instead of a voltmeter if it's a quick glitch dropping the voltage for some reason.

I don't actually know if 4.8 V is "too low" for reliable operation on the RasPi. The USB spec says no more than 5.25 V and no less then 4.75 V according to http://en.wikipedia.org/wiki/U.....Serial_Bus

so I guess, if you're at least 4.75 V it should still be a "legal" voltage per USB spec.  However if some momentary power draw glitched down to 4.5 V for 10 microseconds, that might upset the apple cart, and a multimeter won't tell you that, you'd need an oscilloscope.

This post indicates that some cheap out-of-spec USB cables can cause trouble from high resistance, too-thin wires:

http://www.raspberrypi.org/for.....usb-cables

[Helpme1986]

Chargers ive tried, all at idle after booting into LXDE with keyboard and mouse via powered hub, ethernet and HDMI connected

Official HTC 1000mah - idle 4.72v

Official Blackberry 700mah - Idle 4.84v (as soon as in LXDE all dropped out giving 5.01v)

2nd Official Blackberry 700mah - idle 4.85v (4.82v under heavy CPU usage)

Powered USB Hub (via samung usb cable i got with my phone) idle 4.8v

Taking ethernet out tends to give me 0.05v back and taking the powered usb out gives me 0.01v back (meaning with only power and HDMI in its 4.86v put if i was to leave it to fail it would be 5.01v

Anyone got a power supply they know works that I could order from Amazon?

[Helpme1986]

yes would be good to see what others get

All the stuff ive tried SHOULD be good quality as they have all been official stuff

[Helpme1986]

Another update (SORRY!)

Stuck with the Blackberry 700mah charger with just the keyboard/mouse in via powered hub but for some strange reason over the past hour the voltage has been slowly creeping up

Originally at idle it was 4.86v but it is currently sitting at 5.05v, when i plug the ethernet it it drops slightly and seems to work until i do anything involving data transfer or CPU usage (so basicly anything)

Any one know why the voltage is going up? (hasnt been used for a long time so could be the reason)

[bredman]

Very interesting that the voltage changes when the load on the processor changes. This would imply that there is resistance somewhere between the power supply and the processor.

This resistance could be within the power supply (if you are running the power supply close to its max rating) or within the cable (if you are using a cheap cable).

Does anybody know enough about poylfuses to say if the 5v polyfuse could behave in this manner?

[jbeale]

I have no knowledge of the RasPi hardware specifically, but a slow drift in voltage usually has something to do with temperature- something heating up, or cooling down. For example, a voltage reference could heat up and drift, a bad connection could become better or worse due to temperature change shifting contacts ever-so-slightly.

A polyfuse (resettable fuse) which has been "tripped" due to overcurrent recovers gradually over time, as it cools off and allows more current to pass, see for example http://en.wikipedia.org/wiki/R.....table_fuse

Hopefully the boards do not have out-of-spec fuses installed, but I'd think that is one possible explanation of the symptoms reported here.

[Helpme1986]

jbeale said:


I have no knowledge of the RasPi hardware specifically, but a slow drift in voltage usually has something to do with temperature- something heating up, or cooling down. For example, a voltage reference could heat up and drift, a bad connection could become better or worse due to temperature change shifting contacts ever-so-slightly.

A polyfuse (resettable fuse) which has been "tripped" due to overcurrent recovers gradually over time, as it cools off and allows more current to pass, see for example http://en.wikipedia.org/wiki/R.....table_fuse

Hopefully the boards do not have out-of-spec fuses installed, but I'd think that is one possible explanation of the symptoms reported here.


are you saying i should let the device cool down for a few hours as I might have tripped it and not give it time to self heal? (its been on since i got home 6hrs ago!)

Its late anyway so going to bed, will report back tomorrow!

[Helpme1986]

Well I thought it would work today after giving the polyfuse some time to selfheal and I found a 5v 2amp USB charger but still the same...

Could it go as far as being my house wiring!? At the moment its being plugged in by wall socket to 4 way gang with just Pi supply and powered USB hub (belkin £15 so not cheap)

[jbeale]

My only slightly informed opinion (I don't have a RasPi myself, or even the schematics) is that the self-resetting polyfuse is one possible source of a low-voltage problem that gets better when left to rest "off" for a while. Another possibility is the charger itself. If it behaves the same way with a different type of charger and USB cable, that tends to point towards the polyfuse idea.

One way to determine it, is to measure the voltage drop directly across the polyfuse. I don't know where that device is on the board electrically or physically, or what it looks like- supposedly a schematic will be published Real Soon Now, though.  If your total current draw is within spec (< 700 mA) and you have more than 0.2 V across the polyfuse, then I would say you've found a problem.

Again, I have no idea what actual part they used, but just for-instance: if it is a LittleFuse 06R075B (0.75 A pass rating, 1.3A trip) it should have at most 0.23 ohms of resistance in normal operation, which would be 0.16 V at 700 mA.

example data: http://www.littelfuse.com/data.....C_USBR.pdf

[jbeale]

I would not suspect your house wiring, that would be last on my list. The RasPi takes only a few watts, so it is one of the lowest-powered devices anyone would power from a wall outlet.  House wiring is supposed to be able to handle a toaster or a microwave, which is in the 750 to 1200 watt category, so powering the RasPi shouldn't be even a tiny bit of effort by contrast.

[Helpme1986]

yeah it was just a long shot,

Just tried a different SDCard on the off chance it was that causing the issue (even though everything looks to be working fine) but had no affect,

Any more suggestions? I would buy another power adaptor but no-one has posted any that I could buy online and dont want to waste money buying loads of adaptors for them not to work (especially when most of mine are offical "High Quality" ones)

It seems longer I leave the Pi plugged in the higher the voltage goes and more becomes more stable (until I do something with it then it just cuts out)

[jbeale]

If you've got your voltmeter handy, check the voltage across F3 on the bottom of the board near the microUSB power input.  This is a green rectangular part at the lower right-hand corner of this image:

http://upload.wikimedia.org/wi.....bottom.jpg

[Helpme1986]

well im putting a prong each side and im gettin 0.11….. sounds like im doing it wrong!!!

on the multimeter I have it on 20 in V Direct Current,



Red in middle hole and black in bottom

[jbeale]

Actually, I think 0.11 volts is a reasonable value for the voltage drop across the input polyfuse. Say you had 5 V input, then you still have 4.89 V after the fuse and, everything should still work.   The real test is to check what the voltage is when you trigger the problem (CPU activity, network access, whatever causes the dropout). Probably you need three or four hands to do that test, though!  If the voltage across the F3 component is still under 0.2 volts at that point, then the problem must lie elsewhere.

[Helpme1986]

The highest it got was 0.16.

I could understand if it did it on high CPU usage but sometimes it just does it when idling (like when I just tested it)

[jbeale]

Well, I don't know what the real spec is on this part, and there might yet be something going on with F3. There is a test I would do on my own board, but hesitate to suggest to anyone else- that is check the temperature dependence. The polyfuse is actuated by heat (too much current passing through heats it up, and causes a big resistance increase, which eventually drops down when current drops, or is turned off, and the device cools.)  I suppose it could also be affected by mechanical stress as well, especially if it was a defective part.  Could check that just by pushing on it with your finger, I guess.

The thermal test is to check what happens to the voltage across F3 when the ambient temperature is higher than normal, like when you get it warmer with a hairdryer or the like. But obviously I don't want to be responsible for anyone melting the plastic bits off their RasPi !

UPDATE: if you get fluctuations as high as 0.16 V while the board is idling, that actually does suggest a marginal part, which could be near the trigger threshold even at low current draw.  What is the voltage when the board is working hard?  For better accuracy you might want to use the 2V (= 2000mV) range on your meter, instead of 20 V.

[Helpme1986]

Dont think ill try any of those thank you lol.

So what are my next steps? Would be great to talk to Liz/Eben about it but dont want to waste their time :S

[jbeale]

Fair enough. Seeing a fluctuating voltage does raise questions to me (did you try the "2000mV" setting, for better voltage resolution?) -but if the voltage across F3 is under 0.2 V even when the board is working hard, then I can't claim F3 is the problem.

Pete L. is a hardware engineer who designed the RasPi board, and just got back from a vacation. He popped up yesterday on another forum in the thread below. He might be almost as oversubscribed as Liz & Eben, though.

http://www.element14.com/commu.....8;tstart=0

[johnbeetem]

Given all the symptoms listed here and at element14, I'd suspect either too much resistance in the USB cable (though you have tested two cables) or the polyfuse F3 (bad component or bad solder joint) or perhaps micro USB connector S1.  You have reported voltages at the GPIO +5V pin.  You could also check voltages at S1 pin 1, though probing is tricky.  If S1 pin 1 is steady near +5V during your failure modes, I'd suspect F3.  If the voltage at S1 pin 1 is dropping during your failure modes, it's probably a poor cable, a defective S1, or a bad S1 solder joint.

Another thing to try is shorting out F3 with a wire jumper.  If this makes the problem go away, I'd suspect F3 is bad or poorly soldered.  A bad solder joint may conduct but add resistance.

JMO/YMMV: These opinions are offered without warranty.

[Helpme1986]

Thanks for your reply, could you confirm where S1 pin is and the other pin I should connect to for the multimeter?

As for using a jump wire I dont think ill be trying that! I like to play but not when it comes to the possibility of frying the board

[jbeale]

S1 is the microUSB power input connector, at lower left in this photo:

http://itlounge.eu/wp-content/.....rry_Pi.jpg

It has 5 pins, not including the larger soldered hold-down tabs, and +5V power is at one end, and GND (0 V) at the other end. MicroUSB pinout: http://www.talkandroid.com/wp-.....png?3995d3

These pins are really closely spaced, and it will be tricky to probe them directly with standard blunt-tip multimeter probes. Probably easier to probe between the ground test point TP2, and the input end of the polyfuse F3, which should be electrically the same as the +5V input from microUSB.

EDIT: I don't know which is the input end of F3, but try both ends, the one with the higher voltage is the input end.