Hacker News Reader

Comment by tdicola | Hacker News Reader

Oh neat, I just reproed it with a Pi 2 and a Canon Speedlight flash. I'll put my scope on the power lines and see what's happening when you flash the board. Sounds like from the thread one of the power ICs is photo sensitive.

edit: Wow yeah, here's a look at the 3.3V power line when you flash the board, it drops almost down to 0V and then wildly fluctuates for about 100 nanoseconds: http://imgur.com/hG86pRy

edit 2: Another interesting measurement, with the board _totally unplugged_ and flashing it you can see a big voltage spike on the 3.3V rail. Up to 6-7 volts or so for a few nanoseconds: http://imgur.com/td262QK

I guess not only can you learn about electronics but also Einstein's photoelectric effect with the Pi 2!

View on HN

Are you sure that's not induced current from the power spike required to operate the flash or the spark-gap inside the xenon tube?

You've got the gear set up... would you mind terribly repeating the experiment but blocking the light with cardboard or something?

Alright I put the Pi 2 in a cardboard box and held the flash at the same spot (but the box obscures the light). It keeps running just fine and doesn't have the nasty spike. Definitely is something light related.

Thanks! Well done. Photoelectric it is. Now I wonder what color its most susceptible to...

The responsivity of silicon (and other semiconductors) by wavelength is pretty well characterized[1]. Any data sheet for a silicon photodiode will include a similar plot, and as photodiodes are just diodes (granted, operated in reverse bias), regular diodes will have a similar response if light can get past the part package. Photodiode junction doping and geometry are optimized to make them useful as optical devices, but they're still VERY close to regular silicon diodes.

This property of diodes can be used in a fun way: LEDs emit light, but they can also be used as photodiodes[2]. It's easy with a microcontroller pin that can act as both a digital output and analog input. LEDs are of course not optimized for this use (small sensor area, the package may absorb useful light particularly in the UV, phosphor if present may get in the way), but they can be used as photodetectors. There's even an Arduino sketch[3].

1. http://image.slidesharecdn.com/lect12-photodiodedetectors-13...

2. http://www.merl.com/publications/docs/TR2003-35.pdf

3. http://playground.arduino.cc/Learning/LEDSensor

Since the effect seems to be caused by a silicon component that isn't properly shielded from light, anything with a photon energy higher than the bandgap should do. The bandgap of silicon is 1.1eV, which means the effect should start in near-infrared.

What about glass, then that should work, right?

Not sure I understand you correctly.

The problem should happen with near-infrared, visible, and UV-light. Glass is transparent in the visible range, and depending on type, also in IR / UV, so glass won't work as a shield.

The easiest, reliable way to shield would be wrapping in aluminum foil (also helps against RF-interference).

Yeah you are right. I was assuming that the chips would block visible light but I forgot about UV.

So this would be easily rectified by using a case for the pi?

A case would fix it, but a cheaper fix would be to just put a dab of nail polish or other lacquer on chip U16, which seems to be the culprit.

Handy tip! Which - just tested - seems to work for me, at least with a couple of coats of black nail polish, and double-checked with a high-end Canon flash and a random green laser pointer. (Of course your mileage, and nail polish, may vary.)

I should change the hostname to nemi. ;)

There's another 'shiny' chip on the board too, next to the HDMI port, U8 I think, but it doesn't seem to give a damn what you do to that.

Compared to having to resolder all the Ethernet ports by hand, or USB power issues, this is a pretty minor errata! I presume the factory will pop a blob of epoxy on it for future runs, if they can't get this chip with a package on it.

Don't you get on the NSA's subversive person list for buying black nail polish?

If it were that easy, all the cool kids would be doing it. <g>

I'm not familiar with the photoelectric effect - would this be susceptible to a laser (and if so, how weak a laser? A laser pointer?)?

Somebody else said a laser pointer works, but I can't confirm myself yet as I don't yet have a pi 2

The most recent post [1] seems to indicate that covering one of the chips with blu-tac prevents the freeze

[1] http://www.raspberrypi.org/forums/viewtopic.php?f=28&t=99042...

I'm thinking of trying a more permanent fix with some Sugru[1] when my Pi 2 arrives. Their about says it's insulating and okay for electric repairs under 24 volts.

[1] http://sugru.com/about

Maybe not such a good idea according to: http://sugru.com/questions/can-i-use-sugru-to-secure-shock-m...

Not knowing what the IC is, one thought is that the blu-tac could be blocking light, or providing a wee bit of extra capacitance to tame a wicked electrical signal.

If I understand correctly the Speedlite is battery powered, so as such shouldn't cause any voltage spike on mains.

It's not a mains spike, but an EM discharge caused by the rapid voltage and current changes in the flash, that the parent comment is considering (IIUC).

Yes thankyou. This is what I'm getting at. The momentary power flow in a xenon flash is very large. A great big capacitor charges up and then discharges very rapidly to create the bright flash of light. There is also a high voltage exciter coil that fires to kick the xenon tube into conduction at the beginning of the flash cycle. Both of these could be very disruptive to nearby digital electronics.

I thought the original poster would be in a very good spot to help us determine if the effect if photoelectric or the result of EM because he's got everything set up. All he has to do is repeat the experiment while blocking the light from hitting the Pi.

From the original thread it does appear to be a photoelectric effect, since blocking a specific chip on the board causes it to survive the flash.

I'll put it in a box and try in a little bit, sorry have to do a few things first. I'm 99% sure it's light dependent though, I had to aim the flash at the board and keep it somewhat close to get it to lock up (like 6" away or so).

The historical importance of the photoelectric effect is to demonstrate the relevance of the work function Ω. It works like this:

V ~ E_0 - Ω

where hf is the energy per photon and Ω is the work function of the target. If Ω > E_0, no voltage is produced. This provided the basis for Einstein's demonstration that energy is proportional to frequency (E_0 = hf), and won him the Nobel prize.

What does this have to do with xenon? Xenon lamps produce not only visible light, but also ultraviolet light, in fact, xenon lamps extend further into the UV than natural sunlight:

http://en.wikipedia.org/wiki/Xenon_arc_lamp#mediaviewer/File...

LEDs by contrast produce almost no UV light:

http://en.wikipedia.org/wiki/Light-emitting_diode#mediaviewe...

With this data we can determine that the work function of silicon (in the Raspberry Pi 2) is somewhere in the range of hc/(400 nm) < Ω < hc/(250 nm) or 3.1 eV < Ω < 5 eV. Using data from

http://journals.aps.org/pr/abstract/10.1103/PhysRev.127.150

http://www.sciencedirect.com/science/article/pii/00223697599...

we see that the work function of silicon is around 4.7-4.9 eV, which agrees well with our observations.

draws a little box

The problem with your theory is that, as others have noted, the effect seems to happen with a red laser pointer as well.

To interfere with the electronics, the wavelength only has to as large as the band gap in silicon (as someone else has noted, around 1.1 eV). This means that IR light (and all visible light) can cause the interference too.

A strong enough light source will lift a lot of electrons from the valence band across the band gap into the conduction band, effectively turning a transistor into a conductor for the duration of the flash. Since the sensitive part seems to be related to the power supply, it's quite plausible that this leads to the voltage fluctuations on the power rail.

My money is still on an electromagnetically coupled pulse onto either the 3.3V or ground rail.

I have seen that happen plenty of times, but I have never seen a light flash disrupt a power rail like that.

I'm not saying it couldn't happen, just playing the odds...

It can and did happen, as the afflicted chip was the SMPS controller.

At last, a use for all those write-protect stickers we collected for floppies...

I want to see this with a bare regulator. I might have some time Monday.

Reply on news.ycombinator.com