12

Is it possible to power a Raspberry Pi 3 directly from another Raspberry Pi 3 that is powered from a Micro USB socket.

I understand there is a 5v rail and I am thinking that linking the 5v rails together might work.

I think the 5v rail bypasses the protection circuitry but that shouldn't be needed since the first Pi will be plugged in via USB.

Is this a good idea or am I going to blow stuff up?

Robert3452
  • 237
  • 2
  • 4
  • 3
    Why do you want to do this? What advantage is to be gained? – Steve Robillard Jan 04 '17 at 19:53
  • 3
    @SteveRobillard If there were only one power supply available and two raspberries needed to be powered? – Hydraxan14 Jan 04 '17 at 23:12
  • 2
    @Hydraxan14 That is a kludge of a fix. What are the Pi's doing that could not wait until a second power supply could be found/procured? This may be true on a macgyver episode or at some remote antarctic outpost but seems contrived for most scenarios. Like I have said below the first time you experience a problem (power related) you would need a second power supply just to diagnose/fix the problem and that would negate the whole premise of the OP's question. – Steve Robillard Jan 04 '17 at 23:16
  • 1
    @SteveRobillard Yeah, it's not ideal, but maybe Robert is in a classroom setting and some of the power bricks have been lost, and they're just doing some low-demand assignments. A potentially working kludge is sometimes better than nothing (or an exploded pi). – Hydraxan14 Jan 04 '17 at 23:25
  • @Hydraxan14 But you are neglecting time here. The fact that he had time to ask and presumably (why ask if not) wait for an answer kind of negates that type of rationale. And kludges are not a replacement for proper planning. – Steve Robillard Jan 04 '17 at 23:27
  • 1
    Maybe parallel is the way to go here :D – goldilocks Jan 04 '17 at 23:41
  • Remember: When the red light flickers, bad things are happening. What you propose is going to make the red light flicker. – Michael Hampton Jan 05 '17 at 05:13

7 Answers7

13

You may not blow anything up, but it is still not a good idea. The Pi-3 has a 2.5 amp polyfuse which means that those 2.5 amps would need to be shared between the 2 Pi's. A quick look around this site will turn up dozens if not hundreds of questions related to under powering the Pi and the hard to troubleshoot issues that result (e.g WiFi not working, SD card corruption etc.).

Steve Robillard
  • 34,687
  • 17
  • 103
  • 109
9

I think as per the other answers this is feasible in theory, but may not work out well in practice.

I'm reproducing this graph from an article in our blog, Exploring the 3.3V Rail (thanks PM). Note that this graph shouldn't be considered definitive (keep reading):

enter image description here

Simple enough -- the green line is the 5V rail, tested via the breakout. The load is additional, i.e., sunk via a pin, and the Pi itself (actually a B+) is headless and idle. Part of what this illustrates is the 5V rail isn't as well regulated as the 3.3V rail (it isn't regulated at all, in fact, beyond some overcurrent/overvoltage protection on the microUSB jack).

If you look around, I think you will find a Pi 3 needs nominally ~450 mA to operate properly, because it will draw that much when the processor revs up high enough. Based on the graph, if the rail were providing an even 5V, at 450 mA it is going to drop to 4.6V -- not so good. At that point you run the risk of Pi #2 cutting out (or both, since as per Steve's comment the voltages will probably match up).

Why that drop occurs is debatable (see comments below) and maybe somewhat inevitable (due to "some 5V components [being] subject to changes in resistance and conductivity under load"), or may be in large part because a linear supply is used (for which "the voltage provided...will vary with changes in load impedance", which has been frequently observed by users here particularly WRT the greedy Pi 3).

I.e., YMMV -- but it should be easy enough to check your own scenario with a cheap multimeter. Just busy loop (x4) Pi #2 and see what happens.

Is this a good idea or am I going to blow stuff up?

You won't blow anything up -- if you try hard, you could possibly trip the polyfuse on the first Pi. How wise that is or how often you should do it I don't know, but I would stop after round 1 and think about where all the draw is.

However, done sanely that is unlikely, and the worst thing that could happen to the second Pi is it intermittently reboots due to voltage drop. This could cause filesystem corruption, but not physical damage.

goldilocks
  • 58,859
  • 17
  • 112
  • 227
  • 1
    At 4.6V the undervoltage indicator (lightning bolt) in the corner of the screen would already have tripped, likely on both Pi's. which means correcting the problem would involve a second power supply - the exact thing the OP was trying to avoid. – Steve Robillard Jan 04 '17 at 22:38
  • The drop on the 5V rail is almost certainly the power supply; nothing to do with the Pi, because the the ONLY things which can affect the 5V are the polyfuse and connecting cable. – Milliways Jan 04 '17 at 22:47
  • @Milliways Hmmm...what about "This is expected as some 5V components are subject to changes in resistance and conductivity under load and there is no regulation to compensate for this"? – goldilocks Jan 04 '17 at 22:49
  • Yes. just wanted to say the same about the power supply. If you go and get a switching power supply you will not see the same issue due to the more efficient switching circuity. The 5V "rail" is nothing more than decoupling capicitors.. there is no actual voltage regulator.. the 3v3 is regulated and will always be 3v3 because the GPU runs on this voltage. The only issue with 5v0 dropping so low are USB peripherals like KB/Mouse, WiFI and HDMI issues dropping out. I suspect headless you wont even notice any issue. – Piotr Kula Jan 04 '17 at 22:50
  • Well, the one used in the test is described as "a bench linear power supply" and the Pi is headless, no accessories. I guess it would not be hard to do your own test here with a multimeter. – goldilocks Jan 04 '17 at 22:55
  • Linear is the keyword which means its just a coil rated for a specific ameperage. The more load you give it the less efficient it becomes (it voltage drop at the coil) So all it has is a rectifier, coil and some other cheap stuff like some decoupling capacitors and not much more. Switching power supply has several coils, transistors and sensing mechanisms to always provide the most efficient output (so prevent voltage drop) There are also other benefits like, Linear usually uses full Amp at plug even in no load, while switching uses close to load amps. Switching also provides clean DC waveform – Piotr Kula Jan 04 '17 at 23:05
  • 1
    @ppumkin Cheers! Will edit. – goldilocks Jan 04 '17 at 23:07
  • 1
    @ppumkin sorry, but a non-switching power supply with a linear regulator can by no means generally regarded to have poor load regulation or not a clean DC waveform. Linear power supply does not imply it's unregulated whereas switchung PS does not necessarily imply being regulated. – Ghanima Jan 04 '17 at 23:16
  • 1
    Voltage drop is inevitable, that is a simple consequence of Ohm's law. However the magnitude of the drop varies. I have built 10A lab supplies which only drop a few 10s of mV on load. Of course, you don't get that kind of performance on a $10 supply that most people use. – Milliways Jan 05 '17 at 00:38
4

This is not how electronics works. Just think about it. Look at your light on your ceiling.

Is that light daisy chained to your bedroom light or lounge light?? Nope!Why? Because there is something called like a Common Rail that provided 220v/110v to the a single outlet / fitting which is also fused up and marked well. And that fuse is powered by another common rail that comes out of the street.. and that common rail is powered by another common rail.. yea you get it now... (also each house has it own common rail.. we are not daisy chained with our neighbors, not even in Brazil)

Common Rail in your application is to have a nice power supply, like a 10A Switching Power supply (50Watt) - Get a good decent cable of 1mm core and then use that to power all your Pi's, USB Hubs, Touch Screens ...

You can run about 25 - 50 metres of 1mm core cable @5v before you start to notice voltage drops at the end. I'm talking about 4.9v~4.8v

Daisy chaining as you wish to do it never ends well.. or in any electronics application.

Piotr Kula
  • 17,307
  • 6
  • 65
  • 104
3

You will have to make sure the power supply of the first Pi provides enough current for both Pis and peripherals, without exceeding the maximum intake of the first Pi, this is not a huge range.

See the documentation page and FAQ for typical current draw.

Also, see related answer here

Glorfindel
  • 620
  • 1
  • 8
  • 15
osadezu
  • 31
  • 1
3

No, this will not go well. The Pi has a limit on the power it can pass, as well as a requirement on how much power is needed to function. This means that if one Pi, powered via USB, needs even just 1.5A, you won't have enough power left for a second Pi.

John Keates
  • 349
  • 2
  • 5
  • 1
    If one pi needed 2.5 A all to itself, there would be no point in having a 5V (and 3.3V) rail on the breakout -- or 4 USB host ports. – goldilocks Jan 04 '17 at 22:24
  • 1
    You may find this interesting: https://raspberrypise.tumblr.com/post/144555785379/exploring-the-33v-power-rail – goldilocks Jan 04 '17 at 22:25
  • I have edited my answer to reflect the case more correctly. Surely it doesn't eat 2.5A just for the Soc and/or USB/Ethernet SMSC chip etc., but even if it just needs 51% of all available power it won't work out. – John Keates Jan 04 '17 at 22:30
0

The short answer: Don't.

I can say with absolute certainty, that you should never do that, because I have tried it. It completely bricked the Pi that got power from a Pi. When I hooked it up, it seemed to turn on and off several times due to being under powered. After a short while, the green diode that signifies activity turned off for the last time.

A friend and I have made several attempts to get the Pi back to life, but without success. It is quite dead. Only the red power-light is on when it is powered and it is impossible to install anything on it.

Martin Vinter
  • 1
0

When I ran a Pi 3 on 2.0 amps, it ran, but I got the lightning bold on the upper right corner telling me it was under powered. When I used a a 2.5 amp adapter, that lightning bolt went away and everything worked. When I tried to run the Pi 3 from a 1.0 amp adapter, it booted up, but stopped after 3 seconds. I don't think the Pi 3 could power another Pi 3.

Russell Hankins
  • 201
  • 2
  • 7