2

Problem

I have a 5 inch Waveshare Touch Screen. I was testing something today and had power going to the screen (not the pi) which is okay because they share power when connected correctly. I was paying attention to something else and aligned the pins wrong. I noticed almost instantaneously but it was too late, I could smell the damage. I had baked my pi.

Scenario

I made an image to show you exactly what I did wrong. Most of this is self-explanatory but the last part may not be so here is what is going on underneath "INCORRECT OVERLAP":

  • I've labeled only the pins that touched.
  • GPIO(15) on pi made contact with 3V3 on screen, pi 3V3(17) to SDA etc etc

    click on the image for a larger/clearer version enter image description here

Question

What I'm curious to know is why did this fry my board?

  • 2
    I'd presume the screen's power is intended to come from the pi, so by that mistaken arrangement it hasn't got any. It's normal to use to use 3.3V as active in a button circuit with a pull-down, but the resistor in that case is to protect the 3.3V from the ground used as a pull down on the input. If it could really short directly through the input, then it would -- there's no resistor there when the button is down. An input itself isn't grounded (or else it would not work as an input), so you can't short to it. Of course as an output you could short from it. – goldilocks Apr 4 '16 at 2:57
  • 2
    The SPI CLK in this would connect to a ground and I think that would short. – goldilocks Apr 4 '16 at 3:00
  • 2
    "I'm less concerned with the signals connections (SCL, SDA, MOSI, etc.) ... since they are always data lines and never need to draw/sink large currents ... 5v/3v3, though, are meant to supply large currents." -> In that case it would be the connection to GPIO 3 if the driver is using it as an output and initially drives it high, since in the wrong version that is connected to the screen's 3.3V sink. – goldilocks Apr 4 '16 at 13:51
  • 2
    @goldilocks Yep, I agree. But I think everything marked 3v3/5v on the screen is connected to a Pi GPIO and might have caused an excessive current draw. It could have been any of these four connections or the combined load, if you ask me. The data lines might be an issue too, e.g. you mentioned that a clock line was connected to ground, but there's often an internal resistor on data lines because they never need much current (thus, the manufacturer can safely install internal resistors without risking limiting functionality. Not sure if it's the case here though.) – jDo Apr 4 '16 at 16:16
  • 2
    Yeah, there's also "4" and "5" there now connected to 5V sinks (the BCM numbers on those would be 22, 23, 24, but I guess that doesn't matter). Sounds like you have a valid enough answer... – goldilocks Apr 4 '16 at 16:25
5

Sorry about your Pi - hope you got a new one :)

You've probably seen the discussion between goldilocks♦ and I in the comments and have a pretty good idea about what my "post mortem" will say. However, we both missed the fact that the Pi wasn't powered which changes the scenario a bit. How was the screen powered anyway if it's meant to get juice via the Pi pin header?

Since we're dealing with 26 * 2 pins that might also be inter-connected via the screen's internal circuitry, there's an awful lot of complexity here. However, important factors always are:

  • pin direction (in/out)
  • internal resistors (enabled/disabled)
  • pin state (high/low)
  • pin voltage (3v3/5v)
  • impedance of the Pi pins

The first three bullet points mainly apply to a running Pi (they're software-configurable) but the last two, the voltage level and the pin impedance, are always important since the Pi's GPIOs are not 5v tolerant and since the impedance will determine how much current can flow into the Pi.

When the Pi is running, a pin that's configured as an output and set to a high state will attempt to supply all the current it possibly can to whatever you connect. If you're lucky, you've connected it to a high impedance input (or you've added a resistor) and nothing will happen; if you're unlucky, you've connected it to a low impedance/resistance input or a ground connection and it'll fry. The only connections that can be ruled out upfront are the ground to ground connections; providing that the ground potentials of the Pi and the screen are the same, these should be safe.

Anyway, since your Pi wasn't powered, it's more likely that it acted as a current sink - allowing the screen to find alternative routes to ground. I don't have a Pi and multimeter at hand right now but if an un-powered Pi pin has a low impedance/resistance, connecting a power source directly to it could easily create a path to ground that would allow hundreds of mAs to go through the pin - add to that the over-voltage and you have two serious issues. The data lines (SCL, SDA, etc.) might also be problematic but I think the 5v screen connection is the most likely candidate. I've highlighted some of these connections below and added brief explanations.

Connections that could have fried the Pi

Disclaimer: I'm no electrical engineer; just a nerd who's seen plenty of components go up in smoke - including a Pi!

  • 1
    You know when we had that conversation I had missed the point that there was power to the screen not the pi. In that case, since the (three) overlapping grounds would have supplied the completion of a circuit, those live 5V connections to the pi's 3.3V rail and a GPIO would definitely be a cause. I don't think there is any protection between the those and, e.g., the actual (3.3V) processor. – goldilocks Apr 5 '16 at 11:06
  • 1
    @goldilocks I completely missed that as well and had to re-think and edit my answer. The Pi wasn't supplying anything with power but, rather, it acted as a ground connection for the screen. Simply connecting the grounds shouldn't be an issue but screen power to Pi GPIO or screen power to Pi ground definitely are. "I don't think there is any protection between the those and, e.g., the actual (3.3V) processor." I didn't know that but that would certainly be a problem as well. The Pi burning out might actually have saved the screen from doing the same then. The Pi was an expensive fuse :D – jDo Apr 5 '16 at 11:19
  • 1
    It looks like your #1 and #2 are flip-flopped a bit as far as the red lines go but since they both go to power it should matter much. That being said, thank you for putting so much thought / effort into your answer. I very much appreciate that knowing that it's part of what makes the SE communities great. – Jacksonkr Apr 6 '16 at 14:42
  • 1
    @Jacksonkr What a nice comment; you're welcome :) ...and you're right! goldilocks♦ and I had a long discussion, I uploaded one picture, we then realized we were wrong and I then uploaded a second picture - a bit too fast, it seems. I'll correct it. – jDo Apr 6 '16 at 14:52
  • 1
    Also, FYI for you who may be interested - the waveshare screen still works. I just tested it on the rpi 3 I ordered to replace this burnt rpi 2. – Jacksonkr Apr 6 '16 at 15:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.