Pi Don't ... No.1 in a series of things to not do...

[[email protected]]

Don't accidentally (or deliberately!) plug in an Ethernet cable which is connected to an 18 volt  passive power source. (ie. power sent down the 2 unused pairs of a 4-pair 10/100 Ethernet cable)

The transciever block gets rather hot and starts to smell of that "uh-oh" burning electrical smell after a few minutes.

Fortunately no harm appears to have been done - this time.

I have such a cable + PSU combo on my workbench which I use to power up ALIX boards that I use. Now a different colour from the others!

Gordon

[john_wage]

As long as the magic smoke didn't escape you should be fine.

[Vindicator]

You means the magic smoke that steals the soul of your device and delivers it to the atmosphere I presume.

[[email protected]]

Vindicator said:


You means the magic smoke that steals the soul of your device and delivers it to the atmosphere I presume.



That's the stuff... Still firmly locked up inside!

G

[bredman]

Have you noticed how robust the RPi seems to be? Nobody has succeeded in letting out the magic smoke yet. Even using the NIC as a heating element does not do it.

For those of you under the age of 40, the term "Magic Smoke" comes from http://www.jargon.net/jargonfi.....smoke.html

[Chris.Rowland]

Slight correction - no one has admitted to letting the magic smoke out...

[BertM]

It would be quite hard to destroy the RPi using only the ethernet connector, as there are magnetics involved. The RPi and the ethernet cable are more or less galvanically isolated. You can however fry the ethernet port, but such a part is replacable.

If you really want to fry your Pi, try electrostatic discharge on the GPIO header (or CSI/DSI connectors). They are directly connected to the SoC.

[rew]

But then again, it will be difficult to break the whole thing. It is likely you'll blow up a single output driver, but the rest will remain working.

The one thing that I can think of that is likely to blow up your 'pi is to short the 3.3V and the 5V power supply rails. Then lots of 3.3V, not-5V-tolerant stuff will have to deal with the 5V.

Liz keeps saying that connecting 5V to the SOC will "shorten the lifetime". I have the impression that they blew up an alpha-pi this way. But not immediately. The reason that a chip is not 5V tolerant on its IO pins is that the transistors in the chip cannot handle voltages above in this case 3.3V. Wimpy yes. But that's what's required to get things running at near or above GHz speeds. What happens then is that NO transistor on that chip can handle the 5V. However on the BCM2835, there are a few pins that are used for the 1.2V core voltage. Those convert 5V to 1.2V. So there is a power pin (actually about 4 or 5) connected to the 5V, and there are output pins that deliver the switched 5V for the inductor to convert to 1.2V.

Son on the process used to build the BCM2835, 5V tolerant pins are possible.

[linker3000]

I had a similar experience with a 28K modem back in the days – we"d hooked it to a "Plessey" phone system in Jersey Zoo in the Channel Islands to do a remote link back to the Natural History Museum in London and although it worked fine, it started to smell "hot" and then the smoke appeared. The modem was quickly disconnected and taken outside for a quick fanning in the breeze to disperse the smoke.

After a bit of investigation, it turned out that the Plessey phone systems had a standing DC voltage on the line all the time (about 100V ISTR) and to fix our problem we dashed around the DIY shops on Jersey and bought up all the phone extension cords we could find – we couldn"t get any long reels so we had carrier bags full of 2-5m leads – and we strung them together to reach a phone socket in another building that was on a different system.

Fortunately the modem still worked although it smelled a bit baked.

[johnbeetem]

I once had a computer's RS-232 serial port die one day, without any apparent cause.  I opened up the box and the RS-232 transceiver IC package was split down the middle, with magic smoke residue left behind.  Very surprising, since RS-232 circuits generally have good high-voltage protection.

[SN]

Many many years ago I worked on the final testing line for one of ICL's mainframe ranges and one day we left one of these on a jig (PCB's bare and exposed but cabled together much like a raspi) running some soak tests overnight.

The following morning we came back on to the line to find the closer we got to that jig, the ("what`s all this black stuff?") unusual fine layer of black flakes of dust got thicker and thicker as we approached the offending jig, a completely burnt out chip on one of the PCB's - given that this was the SINGLE line that this particular range was produced on, if this had took hold (and provoked a full sprinkler incident), I suspect Government IT Departments across the UK would have some significant delays in sourcing these new mainframes for the next few months...

So never mind the magic smoke, if you see magic black flakes, then you know its really 'b**gered'

[larsth]

SN said:


So never mind the magic smoke, if you see magic black flakes, then you know its really 'b**gered'


Yes that is true, if the magic smoke trapped inside an electronic component has been freed, then that component does not work any longer'

[thexman]

just a case i had where it says 24v and theres a small o near to it dont think it means 240v. lol i once made that mistake only to fry the PCB and all that was connected to it but i was right there was a 2 4 small-o then V on all the boards apparently its where the board should have been drilled for the mounting screw but wasent by the factory...lol

[jdobmeier]

rew said:


But then again, it will be difficult to break the whole thing. It is likely you'll blow up a single output driver, but the rest will remain working.


Unfortunately I seem to have the perhaps dubious distinction of being to first to actually admit bricking a pi. I did it exactly this way: grabbed the wrong jumper when trying out the GPIO for motor control and ended up first sending 7.5V then ground to a pin configured as output when it was in the high state. Long story short the power light comes on but gets really dim after the RG2 3.3V regulator starts to overheat.

I might try to replace the SoC with a hot-air soldering gun, otherwise they are back ordered until when, sometime in July-August?

Moral of the story: optical isolators on GPIO pins! (or at least during prototyping)

[Joe Schmoe]

I don't understand what good having a hot air soldering gun on the board would do.

[hippy]

jdobmeier said:

Long story short the power light comes on but gets really dim after the RG2 3.3V regulator starts to overheat. 
It could be worth removing RG2, providing 3.32V to the GPIO power pins from a Bench Supply and seeing if anything good happens. It really depends on if you've totalled the regulator, the SoC, or everything.

I'm always cautious of plugging something I've blown up into something else in case that leads to cascade failure and knocks out something else but monitoring the UART pins for the boot messages should be fairly safe to try.

[jdobmeier]

I think I can dig up a sacrificial linux box to read the console just on the off chance. Thanks, I'll try it.

Edit: I should mention that the power light is the only lit led.

[johnbeetem]

jdobmeier said:


I think I can dig up a sacrificial linux box to read the console just on the off chance. Thanks, I'll try it.


If the messages are long enough and the Baud rate not too fast, an LED and a resistor are a quick way to see if there's any transmit data at all.  UART data is normally high, so have it pull LED low to show presence of data.

[tufty]

Joe Schmoe said:


I don't understand what good having a hot air soldering gun on the board would do.


You mount it on the board, and any attempt by the user to connect random stuff to the GPIO results in burned fingers.  Admittedly, it increases the power consumption a little, but it's a small price to pay for protecting the board.  I can't imagine why it's not fitted as standard, to be honest.

jdobmeier said:


I might try to replace the SoC with a hot-air soldering gun, otherwise they are back ordered until when, sometime in July-August?


Nah-ah.  The SoC isn't available for purchase (unless, I suppose, you work for either Broadcom or one of the companies manufacturing stuff using it), it has a RAM chip piggy-backed on it, it's a super-fine pitch BGA, and that's only some of the reasons why you're not going to be able to do that.

Simon

[Phil Spiegel]

But he'd be the first with an 'Exclusive Insider' Photograph to accompany an article on 'How it (doesn't) work (anymore)  ... or spreading the Raspberry Pi..  'Raspberry Exposed'  .. what a balls up... etc etc ... it'll save any of us takingone apart to see what it looks like

[jdobmeier]

tufty said:


Joe Schmoe said:


I don't understand what good having a hot air soldering gun on the board would do.


You mount it on the board, and any attempt by the user to connect random stuff to the GPIO results in burned fingers.  Admittedly, it increases the power consumption a little, but it's a small price to pay for protecting the board.  I can't imagine why it's not fitted as standard, to be honest.


You must be an apple fan, I heard a similar system will be standard on the next iPhone.

[IT_Sean]

Chris Rowland said:


Slight correction – no one has admitted to letting the magic smoke out…


DanielDaniel, a member of the IRC channel, has admitted to killing a Pi.  It's the first dead pi i've seen reported.  It was a couple of weeks ago.  Apparently, he dropped a solder blob onto it, and then powered it up.

[ledow]

I have such a cable + PSU combo on my workbench which I use to power up ALIX boards that I use. Now a different colour from the others!

One of the reasons that I insist that all PoE cables and PoE capable switches are coloured/labelled separately from everything else where I work. I use red cables throughout the school I work just for PoE. Nobody touches the red cables, and there's no "extending" them or "replacing" them or "moving" them or anything else. Outside the other end of a suitable PoE-sucking device, on it's unpowered daisychained port, sure we just chop and change and crimp Ethernet like normal.

Although you're talking about a "homebrew" style of PoE, and the standardised versions are much safer, the only thing stopping that voltage running down the cable is the sensing of a certain resistance between two of the pins, and then when it's "live" you can have dozens of watts pumping down some cable.

Otherwise, it's only a matter of time before some smart-alec workman thinks "I'll just cut that and extend it" and gets 25W up his tool or, worse, manages to push it into other unpowered data wiring within the same cut and take things out.

[jdobmeier]

It could be worth removing RG2, providing 3.32V to the GPIO power pins from a Bench Supply and seeing if anything good happens. It really depends on if you've totalled the regulator, the SoC, or everything.

I tried this with an led/resistor on the tx pin but to no avail. The tx pin never even went high. In addition the SoC became extremely hot, which makes sense considering that the bench supply indicated it was sinking more than two amps.