D17 melted my case! (the dangers of backfeeding)

[jascination]

I have a very simple/standard RaspBMC setup, like so:

USB hard drive -> Powered USB hub -> RPi <- Wifi Dongle

The hub was using a 1A/5v supply (this didn't come with the hub, is a universal power supply - the hub's own plug was 2A but couldn't even spin up my hard drive when connected to the pi).

It worked great for about 5 days, then yesterday the hard drive/hub was no longer connecting to my pi. Then I noticed a burning smell, thought it was the hub, disconnected it. Smell still persisted, even when Pi was plugged into NOTHING except the Micro USB power supply.

The TP1 capacitor was quite hot to the touch, but when I removed my plastic RPi case I found the real culprit: The d17 (capacitor? Not sure what it is) was EXTREMELEY hot, even though pi had been plugged in about 30 seconds. The burning smell was this thing melting my plastic case. Photos:




From the photo below you can see the d17 appears to have some sort of crack in it.



For completeness' sake, this is a Model B rev2. The backpower from the hub was giving me bootloop issues, so as per suggestions here I taped off the 5v pin in the USB plug coming from my hub. Sole power for the Pi was from my HTC phone charger.

So where to from here? Is my Raspberry Pi dead? Can I return it under warranty? What the hell happened?!

[jascination]

Apologies, those images aren't as large as they should be and there's no way to edit my post. Here's direct links to them:

The diode, which has a crack in it/a raised bump: http://i.imgur.com/bppwMk8.jpg

The case that has been melted by said diode (it gets super hot instantaneously, after being powered for a few seconds): http://i.imgur.com/meJvwaM.jpg

I've contacted Element14's sales department to see if I can get it replaced under warranty, but very keen to know what went wrong here so I can avoid it in the future.

[aTao]

D17 is an overvoltage protection system, in the case of "extreme" voltages its resistance breaks down and conducts large currents.
If it is "doing its thing" then you have:
a) a PSU that is outputting significantly more than 5V for at least part of the time
b) a faulty D17
c) a D17 with a wrong value that sneaked into the production chain.

To properly measure you PSU voltage you would need an oscilloscope, or you could have it charge a capacitor and measure that voltage.
Might be best to try a different PSU. But it does seem most likely that your D17 was faulty and died (they should be able to take a massive hammering), in which case it should be covered by warrantee.

[jascination]

D17 is an overvoltage protection system, in the case of "extreme" voltages its resistance breaks down and conducts large currents.
If it is "doing its thing" then you have:
a) a PSU that is outputting significantly more than 5V for at least part of the time
b) a faulty D17
c) a D17 with a wrong value that sneaked into the production chain.

To properly measure you PSU voltage you would need an oscilloscope, or you could have it charge a capacitor and measure that voltage.
Might be best to try a different PSU. But it does seem most likely that your D17 was faulty and died (they should be able to take a massive hammering), in which case it should be covered by warrantee.

Interesting, thanks for the informative response. Not much info on D17 out there, and it seems like no one else has had this issue so a strange thing to have occurred. I've emailed Element14 and will hopefully be able to send it back.

Out of interest, what kind of voltage would the D17 need to behave so dangerously? Both power supplies I've tried are 5v but curious to know what voltage could make this thing melt hard plastic.

[aTao]

Here is the datasheet: http://www.st.com/internet/com/TECHNICA ... 001366.pdf

From Table 3 (I think the RPi) uses the first one in the list, 5V, take the voltage values and have a look at Fig 1 and Fig 5. You will see that anything above ~6V will trigger the breakdown to very low impedance.

[PetervonPetrichstein]

I had nearly the same problem. I soldered a 5V power supply on an USB hub and everything worked well for two days. Today I put on the USB hub and D 17 was getting very hot and the case melted. Now I have the same problem when I plug only the micro USB on. I think it is sure that D 17 is damaged.
I tried to measure the voltage on the power supply of the USB hub but I stopped when smoke cames out of its case.
The broken power supply damaged my Raspberry Pi. Maybe in your case it is the same problem. I think I can't return it under warranty.

Sorry if my English is not the best, I am from Germany.

[mahjongg]

I have a very simple/standard RaspBMC setup, like so:

USB hard drive -> Powered USB hub -> RPi <- Wifi Dongle

I bet that in this power configuration you were actually back feeding the PI from the HUB, that is instead of feeding power from the micro-USB port you were actually applying power via its normal USB output. Normally this isn't too big a problem, but one hidden problem with it is that power no longer passes the input polyfuse. Now if for whatever reason there is a millisecond pulse that raises above ~6V the over-voltage protection diode will trigger and short the 5V supply to GND! As there is no polyfuse anymore to block current, and if the hub is also cheap and doesn't have any over-current protection itself, all the power the PSU can deliver keeps flowing through the over-voltage protection diode until it burns out completely! Which wouldn't have happened if you had fed the PI the correct way through the micro-USB port!

So this is a lesson! If you have a hub that attempts to back-feed the PI check if that hub has over-current protection on its USB ports, that is, polyfuses, or a current monitoring and power switching IC! If is so cheap that it doesn't have that, you MIGHT encounter this problem. The solution, cut the lead to the power pin (VCC) of the USB port connected to the PI, or in any other way block the back-feeding of the PI. A simple way is to put a piece of cellotape on the VCC pin of the USB connector that plugs into the PI (look the pinning of the connector up in wikipedia, make sure you are not disconnecting GND instead of VCC or the PI wont be able to communicate with the hub! Now front feed the PI from the hub through the micro-USB port, and it should be safe! If you still happen to trigger D17 then it won't burn out, and the polyfuse will recover in a few days.