Injecting 12v on to the GPIO 5 volt pin to power RPi - don't

[ReadiesCards]

To save having to have the USB micro cable sticking out I thought I'd power the Pi by connecting the +5v and 0v pins (Header P1 pin2 and Header P1 pin6) to my 5v PSU - the same technique worked lovely with the Arduino.

However on the Pi, the red LED came on, there was a little tiny 'phut' then the red light faded away.... leaving me with a dead Pi.

So don't try this at home

Paul

[fruitloaf]

I'm pretty sure you've done it wrong as I'm powering my Pi that way for the last two weeks and its still rumbling along ok.

[Burngate]

It would be interesting to know where the 'phut' came from - is there any evidence, such as a hole in RG2, the regulator that provides the 3v3 that drives the red LED?

Also of interest would be what supply you were using, and whether you connected it correctly.

[jbeale]

Many people including me have powered their Pi from the GPIO pins with success. My guess is you connected to the wrong pins, it's easy to do. For example 3V3 on P1-01 is right next to 5V on P1-02 and connecting 5V to the 3V3 rail could probably do what you describe. The red LED is powered from the 3V3 rail through a 1k resistor R34. It might have been the BCM2835 SoC which uses the 3V3 rail, that went "phut".

[ReadiesCards]

Ah I just found the issue - brain fade!

I had assumed a project that I was converting from Ardunio to RPi had a 5v PSU .....it did not instead it was 12v . Doh!

Glad to hear people powering pi through gpio pins - makes installation in a small box easier

[shortcircuit]

So you probably killed RG2, but the other parts may still be intact. A good candidate for conversion to using a switched regulator if all the other parts look ok.

[ReadiesCards]

So you probably killed RG2,

RG2 visually looks ok. Can I cut it off and solder another one in place?

A good candidate for conversion to using a switched regulator if all the other parts look ok.

Could you kindly expand? I had assumed it was unrescuable

[kaos]

RG2 visually looks ok. Can I cut it off and solder another one in place?

Yes. Well, unsolder rather than literally cut, to avoid damage to the circuit board, but it is certainly replaceable. You can get one from many sources; here is a link to Farnell http://export.farnell.com/on-semiconduc ... dp/1652365, only because their website is reasonably easy to use Sound effects not withstanding, it can be blown without any visible damage, and it is one of the components I would expect to go first under the circumstances. Mind you, it is possible that something else is blown (I think the SOC, while primarily operating from 3.3V, does have some 5V feeds also), but maybe worth risking a 60p spare part to find out?

A good candidate for conversion to using a switched regulator if all the other parts look ok.

Could you kindly expand? I had assumed it was unrescuable

A switching regulator wastes less power than the linear type which is used in the Raspi. The switching regulator requires a few extra components and is therefore a slightly more expensive option, which is why the foundation went with linear type. This kind of conversion is of particular interest to those that want to run their Raspi from batteries, solar cells or suchlike.

--
Best regards,
Kári.

[geoffr]

On the topic of powering the Pi via the 5V pin on P1:

I have noticed that I am getting only 4.75V on the 5V line (pin 2) of P1 when I feed power the "normal" way via the micro-USB. By comparison, I am seeing 5.05V when feeding power directly to the 5V line on P1, using the same USB power supply.
Can anyone clarify the source of that voltage drop?

The voltage drop is interesting to me, as I want to provide for sufficient margin above the pickup voltage of the relays I am using. The relays are being driven by a ULN2803A, which is connected to a MCP23017 on the I2C bus.

[kaos]

That voltage drop is across the polyfuse, which you bypass when powering the Raspi through the GPIO connector. That again means that you should set up your own protection when doing so, whether polyfuse, regular fuse or other.

--
Best regards,
Kári.

[shortcircuit]

Mind you, it is possible that something else is blown (I think the SOC, while primarily operating from 3.3V, does have some 5V feeds also), but maybe worth risking a 60p spare part to find out?

Ah yes, apparently it does. I'm a bit surprised since it shouldn't really need 5V for anything. That means there is a big risk the SOC is damaged rendering the whole board useless. I'd try to cut the leads of RG2 (carefully, the package is fragile) anyway and check if it works on it's own. If it doesn't there's still some hope a regulator swap could fix things.

Could you kindly expand? I had assumed it was unrescuable

Replacing the linear regulators with switching regulators is discussed in this thread:
http://www.raspberrypi.org/phpBB3/viewtopic.php?t=12387

[geoffr]

That voltage drop is across the polyfuse, which you bypass when powering the Raspi through the GPIO connector. That again means that you should set up your own protection when doing so, whether polyfuse, regular fuse or other.
.

Thanks - makes sense. Given that I probably need to provide my own 5V power to the board I am building, it is increasingly looking like it will make sense to feed 5V to the Pi via P1 (one less cable). - As you point out, I will just need to provide my own fuses.

[repton]

On the subject of powering your RasPi from the GPIO connector...

I'm currently designing some expansion boards for the RasPi and wish to be able to power the RasPi from them via the GPIO connector.

My question is this: is the RasPi PCB designed in such a way that the traces from all of the ground connections are capable of carrying the full 1000mA the RasPi could potentially use? I'm asking as I had originally planned to connect +5V to pins 2 and 4, and 0V to pins 6 and 9 in order to spread the load over more than one pin, however this is proving hard to route to my satisfaction and I was considering connecting 0V to pin 9 only. However I don't want to do this if the trace on the RasPi PCB that connects to pin 9 is too thin to safely carry 1000mA. I don't have access to any of the other ground pins as I'm only using a 2x5way header as my boards are I2C related and so I don't need to be able to connect to the higher numbered pins.

Any ideas?

TIA,
Paul

[Grumpy Mike]

That voltage drop is across the polyfuse,

No that is not the case.

The voltage drop comes from the resistance of the USB connector. I have found that this varies considerably with the exact cable you have. After measuring several cables I have found a spread between 0.5R per contact to 2.5R. As this is per contact it is in effect doubled. So the worst case cable will have in effect a 5R series resistor. Assuming the Pi is taking half an amp that can be a drop of 2.5 volts. Even with the best cable you loose in the order of 0.5V.

[repton]

My question is this: is the RasPi PCB designed in such a way that the traces from all of the ground connections are capable of carrying the full 1000mA the RasPi could potentially use? I'm asking as I had originally planned to connect +5V to pins 2 and 4, and 0V to pins 6 and 9 in order to spread the load over more than one pin, however this is proving hard to route to my satisfaction and I was considering connecting 0V to pin 9 only.

Answering my own post, I've now managed to route the traces so that the 0V is supplied on both pins 6 and 9 so there shouldn't be an issue.

Paul

[Grumpy Mike]

Answering my own post, I've now managed to route the traces so that the 0V is supplied on both pins 6 and 9 so there shouldn't be an issue.
Paul

It is not just the pins, it is the tracks on the PCB to the source. Paralleling up the pins is good but if at the end of the day the PCB tracks won't take it then it makes no odds, Chain / weakest link and all that.
As the PCB layout has not been published it is imposable to say. Mind you I haven't seen my Pi drawing any more than 300mA no matter what it is doing so maybe 1A is a bit pessimistic.

[repton]

Answering my own post, I've now managed to route the traces so that the 0V is supplied on both pins 6 and 9 so there shouldn't be an issue.
Paul

It is not just the pins, it is the tracks on the PCB to the source. Paralleling up the pins is good but if at the end of the day the PCB tracks won't take it then it makes no odds.

Indeed and the question in my original post was relating to the PCB tracks rather than the pins. A datasheet I found for a very similar PCB header on Farnell's website suggests the pins themselves are probably good for 3A, so one pin would be plenty on that basis, but I was just wondering if anyone had any inside info as to the tracks hidden away in the inside layers of the PCB.

Paul