News

All news

Xenon Death Flash: a free physics lesson

If you own a Raspberry Pi 2, congratulations: you’re also the proud owner of an elegant demonstration of the photoelectric effect!

At the weekend, Peter Onion, a veteran of our forums and of Raspberry Jams in Cambridge, Bletchley and surrounding areas (visible, costumed, in the background of this photo at the Christmas CamJam), discovered what we think might be the most adorable bug we’ve ever come across.

The Raspberry Pi 2 is camera-shy.

Peter’s bug report came via our forums. He’d been proudly photographing his new Raspberry Pi 2, and had discovered something peculiar: every time the flash on his camera went off, his Pi powered down.

The blip you’re seeing here is what happens when you point a xenon flash at a Pi 2.

Jonathan has spent much of the morning emitting flashes and poking an oscilloscope. We’ve found out what’s going on, and the good news is that it’s completely benign: your Pi will not suffer any permanent effects from being flashed at.

More good news: the effect only happens under VERY specific circumstances. Flashes of high-intensity, long-wave light – so laser pointers or xenon flashes in cameras – cause the device that is responsible for regulating the processor core power (it’s the chip marked U16 in the silkscreening on your Pi 2, between the USB power supply and the HDMI port – you can recognise it because it’s a bit shinier than the components around it) to get confused and make the core voltage drop. Importantly, it’s ONLY really high-intensity bursts like xenon flashes and laser pointers that will cause the issue. Other bright lights – even camera flashes using other technologies – won’t set it off. You can take your naked Pi 2 in the sunshine for a picnic or take it to a rave, and it’ll be perfectly solid. Just don’t take it on the red carpet at the Oscars. Jon is currently shining an 1800-lumen led light at a Pi 2 on his desk: not a wobble.

This component that’s causing the issue is in a WL-CSP package: a bare silicon die which has solder balls attached. This is a picture of the underside of a similar package (enormously magnified) – each circle is a minuscule ball of solder:

underside balls

WL-CSP packaging is a common technique for more high-tech electronics parts, as it means no further packaging of the device is required. It is also the smallest physical package possible, which designers of mobile things (and people making very tiny computers) really care about.

What’s causing the component to behave so oddly? It’s the photoelectric effect, where metals emit electrons when hit by light. The video below is a really good tutorial on how that works.

What you’re seeing with Pi 2 and xenon flashes is the same effect, but in semiconductor material, not metal. Semiconductors, like metals, have free electrons which can be ‘knocked off’ by photons. Photodiodes, solar cells and phototransistors all use this effect to function. If you’d like to learn more about how a solar cell works, there’s a nice explanation here at Physics.org.

Silicon junctions (the types that are responsible for making diodes and transistors and other such electronic miracles function) can be ‘upset’ by this photoelectric effect if it is large enough (i.e. if enough light of the right energy [i.e. colour] is fired at them). This seems to be what is happening to our power supply chip – somewhere in the complex silicon chip circuitry there are some transistors or diodes that malfunction when hit by high energy bursts of light, causing the power supply to ‘drop out’, so the Pi reboots.

Jonathan is actively investigating exactly what happens when U16 is flashed with a high energy pulse from a xenon flash tube, and we are also looking at possible ways to make future production Pis immune to this issue if we can – we know you like to take pictures of them.

We have found no evidence that ‘flashing’ your Pi2 with a xenon flash can cause any real damage, but we still don’t recommend doing it (it will crash or reboot, and this means you may corrupt your SD card). I’ve said it above, but it bears repeating, because I’ve seen some of you mention this in the forums and in comments sections elsewhere: common everyday light sources – e.g. bright sunlight, indoor lighting, angry cyclists* – don’t cause this to happen, so please don’t worry!

If you need to use your Pi 2 in a situation where it might be flashed at, our advice is to cover U16 (make sure you get the sides too) – the current easy fix is to use a small blob of Sugru or Blu-Tak covering the whole component (someone in the forums used a pellet of bread: the first yeasted bug fix we have encountered), or simply put the Pi in an opaque case.

Secretly, I’m kind of hoping for another (similarly benign) bug this abstruse. I love writing this sort of post.

*Note to the paranoid on Twitter: I wasn’t clear enough here. We are referring to ourselves, not dissing you and your bike. We’re in Cambridge, and a lot of us cycle (aggressively) to and from work: we’ve been shining our bike lights at Pis for much of this morning, because they’re the brightest lights any of us own.

194 comments

Great band name though! :)

I’ve already bagsied it.

I don’t know. Leaving bread on your Raspberry Pi sounds like it’ll just attract more bugs eventually.

Goddamnit xD

Instead cover your raspberry with a nice thick layer of cream or custard. Yummy!

So, I’m not the only person who thinks it’s odd to have a Pi and not be able to put whipped topping on it?

So, the Pi 2 doesn’t support Flash then?

Badoom tish!

The next time someone asks about Flash on the Pi, we can all note that Flash makes the Pi crash, and do it with a straight face.

I could do that BEFORE this bug was found…

I have a bright daughter.
Will this crash the Pi2 when she uses it?

Haaaaaaaaaaa~!!!!!!!!!!!!!!!!!!!!! I see what you did thereeeeee!!!! XD

Try some Loctite 237113 Opaque Epoxy potting compound.

Thx.
Will try on her.
Tom

One opaque epoxy is on way, thanks

I can imagine seeing this on future test plans: “Flash really big light at the board to see if it breaks.” Obscurity abounds!

If this gives us the opportunity to add a suite of FRICKIN’ LASERS to the office, I’m happy with that.

Liz, FRICKIN’ LASERS are only *really* cool when attached to sharks heads.

Hire frickin’ sharks with frickin’ laser beams on their frickin’ heads for the Quality Frickin’ Control Dept.

Hire the sharks from the superbowl halftime show!

Liz, have you thought of writing up a lesson plan around this? As you say, a good example of the photoelectric effect.

The bug then becomes a feature :)

Difficult – it’s a nice idea, but the risk of corrupting SD cards means it’s not something we’ll be doing!

Don’t know, Liz. Perhaps if one were to anticipate using the Pi 2 in this way, just reserve one Pi 2 just for this and make sure you have a handful of SDs standing by. It is just too good to ignore. You get the plank relation (E = hf, in that it is wavelength, or, inversely, frequency dependent) and everything. Not quantitative but qualitative.

maybe with no SD card inserted? At the rainbow screen?

Well maybe we can write a small bare metal piece of code that does not access SD card after boot for this demo? An animating rainbow screen maybe?

Bare PC boards, in general, need to be treated with “kid gloves”. In particular, electrostatic discharge can kill modern electronics quite easily if precautions aren’t taken. My Pi is currently encased in a self-built LEGO case to keep it safe from fingers that might accidentally zap it with static electricity. I’d think that same LEGO case would also protect it from camera flashes too.

Electrostatic discharge (ESD) events are a well-understood phenomenon. This first popped up in the 1980s when CMOS logic came to the fore – here’s a (now extremely dated) technician training video from Apple on the matter:

https://www.youtube.com/watch?v=4WHbcd4NGJQ

Modern electronic components incorporate robust ESD protection on all I/O pins – in fact we type-test the Pi to 4kV HBM (human-body-model) which involves zapping a powered board repeatedly with an ESD testing probe on the exposed metal shells around the edges. Immunity to ESD and fast transient events are part of the requirements for CE compliance.

FYI if you have your hand in the wrong place while testing the board, 4kV *hurts*.

One of the early public incidents involved a MOSFET proximity detector at trade shows. Every time someone touched the unit (instead of merely moving their hand near it), the sales critter had to replace the MOSFET. IIRC, the MOSFET in question, which had no bleed resistor for protection, had an input impedance of something like 10**14 ohms.

**? That’s FORTRAN’s notation for exponent right?

Fun facts: FORTRAN, Python, Ruby and a bunch of others. Some lesser-practiced use caret (like Lua). Some silly ones, like C# don’t have an operator for it.

I think I see an alternate version of the “Will It Blend” craze in the making. Or computers, games consoles, and especially Hi-Fi equipment being subjected to flash strobes to see if any negative effects can be found…

I should send to a framed copy of the first picture I took that started all this :-)

s/to/you/

I’m not sure if you’re joking – but we’d TOTALLY display that at Pi Towers. (It’d need a little plaque.) :D

I’ll bring you a copy to the Birthday Party :-)

Caption “One small flash, one giant step for Vcore”. Or similar.

Windows 10 was only announced and already you got a Flash vulnerability :D

So I can’t turn on my UV temperature gun and point it at it? Or turn on my UV desk light?

Your UV temperature gun will be fine – it’s not actually firing any emissions, just reading them (Cool gadget by the way, I must get one!)

Your UV lamp sounds interesting – try it and see – my guess is it won’t cause a problem unless it has an extremely rapid turn on/turn off time…

What about when I activate the red laser pointer on the UV gun? That fires out a red light for improved accuracy.

Oh, if it’s got a red laser pointer it might have an effect. (Other people are saying this elsewhere.) So the only way to know is to try.

Tried with my laser pointer and no effect at all.

Depends how bright your pointer is – they come in different strengths. Like chilli sauce.

The UV light is just a UV lightbulb in a base.

Hey, I got a mention. I’m a very angry cyclist. I don’t take prisoners on Winchester Rd, Basingstoke.

NEWS FLASH! DATELINE CAMBRIDGE NOT APRIL FOOLS DAY STOP

THE ONION REPORTS RASPBERRY PI BEING FLASHED CAUSES IT TO FALL OVER LIKE THOSE SILLY GOATS WITH THE FRIGHT GENE STOP

FOUNDATION RECOMMENDS MAKING PI WEAR TIN-FOIL HAT TO WARD OFF LASER BEAMS AND XENON STROBES STOP

SALES OF LASER POINTERS AND OLD-TIMEY XENON STROBES OFF THE CHARTS ON EPREY AND AMAZON STOP

FILM AT 11 SINCE KLIEGLIGHTS DON’T CAUSE FAINTING OF STARLET STOP

STOP THIS SILLINESS STOP

NO REALLY STOP IT STOP

… and scene!

Brought to you courtesy of the Not Ready for Pi Time Players :D

Has anyone tried the obvious even-older-timey flash bulbs/cubes, or do they exhibit an intense broad-spectrum emission?

This begs for a case with an aluminized Mylar layer that reflects external light sources, requiring an internal light source like the electroluminescent tubes that were all the rage years ago in gamers’ hopped-up cases with transparent sides, updated to use addressable LED strips often driven by Arduinos. No LED lasers though – Feinting Pi Syndrome (FPS) would ensue!

This is just hilarious – more “features” like this, please!!!

Feinting Pi Syndrome should not be confused with Fainting Pi Syndrome, where in the former a Pi makes you think it’s going to zig when it really is going to zag! I hereby invoke the Bart Simpson Defense: “I didn’t do it, nobody saw me do it, you can’t prove a thing! Don’t have a cow, man.” ;)

“Hey mouse, say cheese!”
From the episode where they went to Itchy and Scratchy World. The robots went beserk, and they used camera flashes to stop them!!

Brilliant! (Literally, brilliant…)

“RASPBERRY PI BEING FLASHED CAUSES IT TO FALL OVER LIKE THOSE SILLY GOATS WITH THE FRIGHT”

That was my big Laugh Of The Day!

Well Einstein got a Nobel for his description of the photoelectric effect. So you’re in excellent company for your educational demonstration…

Wouldn’t covering the components that with something like an opaque nail polish work? Just take some black polish and paint the little bugger and be done?

Nail polish has solvents in it which might cause corrosion – we haven’t tested the various bits of metal on there with it. Use at your own risk.

You can get nail polish that’s free from formaldehyde, toluene and other baddies. Not suggest anyone buys it for the purpose (not cheap) but if it’s, well, lying around e.g. on a dressing table or summat…

Nail polish can contain toluene??? Good golly. I wish I had a daughter so I could forbid her to wear it.

Xenon (Pi) 2: Megablast

https://www.youtube.com/watch?v=y0cACVRFguk

(Remember, middle-aged spods, it’s not “retro” if it’s your original copy that’s just been sitting in a disk box for 30 years. It’s just *old*.)

My Fav game … which I’ve been thinking about constantly since I read about this glitch on the register … I must be retro or old to remember this one with such affection! … All computers need a weekness in case they try and rise up and take over the world.

Sure it’s not retro, it’s original. Like my “Never Mind the Bollocks” and “London Calling” albums. Not retro, like “have you heard this band, Queen, they’re like amazing”.

Get off my lawn.

Did you know you can play the Acorn Archimedes version of Xenon 2 on the Raspberry Pi NATIVELY if you use Risc OS on it?

Take a look at this AWESOME project:

http://forums.jaspp.org.uk:9000/forum

The bad news is we only have 14 hours to save the Earth!

It’s common to reboot after flashing the memory of a device, but this takes it to another level. Excellent explanation!

Obviously I’m going to try this when I get home, but might the Kapton tape from the HDMI socket solve the problem? Is it sufficiently opaque?

I notice that my two Pi 2s from RS Components last week didn’t have any Kapton tape on their HDMI ports. Thankfully I have (goodness knows why) kept some Kapton tape from the older 1B/B+s. My youngest five year-old goblin… er, daughter, was almost tearful as she realised that her job as Kapton-tape remover was now obsolete. She previously considered it as her key contribution to the field of computing, and a vital stage in setting up the numerous Pis we’ve raised donations for at Cotswold Raspberry Jam. I have temporarily assigned her the role as third Pibow hexnut tightener, which has assuaged any domestic factory-line revolt for now.

(Out of interest, anyone know why some Pi 2s are being pictured with Kapton tape, and others, like mine, have a more closed/solid design of HDMI port which I’m guessing the sucker component placement robots don’t need assistance with? Is this an indication of UK-made vs. foreign-made Pi 2s? Or a change in the Newport production line?)

I think RS and Farnell are using different HDMI adaptors, and different CSI and DSI ports. (It’s to do with where they could each get the best deals on components.) There are no mechanical differences, but there is a cosmetic difference. They’re still being made on the same line: on RS days, different reels are used from the reels used on Farnell days.

A-ha! Thanks. That would account for the black vs. notblack (white? cream? me=colourblind) CSI/DSI ports too.

Different component reels for different wholesalers, wow. That Newport factory really is the absolute business in electronics fabrication flexibility, isn’t it?

Let’s call it off-white for safety. And yes: the Pencoed plant is an amazing, amazing thing.

You have to remember the scale of this operation.

I’m not sure exactly what connector they use on the pi but picking a similar connector from molex http://www.molex.com/molex/products/datasheet.jsp?part=active/0526101572_FFC_FPC_CONNECTORS.xml they quote 1000 connectors per reel.

I’ve heard that they are now making something like 20000 pis per day. Each pi has 2 flatcable connectors so that would mean 40 reels of connectors PER DAY.

Hi,
while everything you write in this article is true, you are still missing one important point.

Any exposed silicon die is affected by light if you shine it on the side where the active parts are. The photons will hit the junctions and generate currents which will confuse the logic and cause malfunction.
This chip is mounted upside down on the PCB, which means none of the active parts are visible, so how does the light get through.

Silicon has another interesting feature. It is quite obvious that it is opaque to visible light. When you look at it, it has a silvery grey color, and it is impossible to see through unless it is lapped down to nanometer thickness. It is however transparent to infrared light. This is a feature that is utilised in failure analysis using microscopes with infrared slight source and infrared camera.

If you just shine some visible light on the backside of the die, nothing really happens, but if you have a light source with high IR component, you will get glitches. I’m guessing the Xenon flash emits a lot of IR.

Even a thin layer of paint would fix the problem. Nail polish is mentioned. Nail polish is mostly dissolved in acetone, which should work just fine.

Cheers,
Kjell
(chip designer and failure analysis engineer)

Ladyscience here: acetone will dissolve nail polish, but it’s not the solvent *in* nail polish, which can be anything from toluene to butyl acetate. (The acetone in our office doubles as flux remover and nail polish remover.)

Obviously I don’t use nail polish that much. Didn’t know that they used toluene. I thought that stuff was even less healthy than acetone.

Anyway, I would just paint over that die with whatever I have at hand, but as Liz says, it is at your own risk (which is probably pretty low).

Kjell

Besides solvents, consider that nail polish might contain unexpected conductive materials (carbon black for example). I’m going to go the clean Blu-Tack route – it’s easy, not as permanent as Sugru or epoxy/paint and is mentioned in the post here so it ought to be warranty-friendly too.

I think you’re right to be careful about the constituents of paint-like coatings on active electric connections. I went looking for “dielectric black paint”, and found this: http://www.plastidip.com/home_solutions/Liquid_Tape

I would think that “Liquid Electrical Tape” perfectly fits the bill here, as it will coat the back and sides of the chip, right down to the board, and is designed for use on active electronics.

On some ham radio boards, they discuss painting antennas to protect them without interfering with them, and the PlastiKote paints are mentioned by name.

The depth of light penetration in silicon increases rapidly as you fall off the red end of the visible light spectrum – additionally chip package in question is only 0.32mm thick.

I have found (as other have, e.g. with red laser pointers) that it is indeed longer wavelengths that cause the most pronounced bounces in voltage, which tallies with that hypothesis.

Xenon lamps are particularly bad offenders for near-IR emission – they have some very strong peaks in the regions of interest.

http://zeiss-campus.magnet.fsu.edu/articles/lightsources/xenonarc.html

> Kjell says:
> 9th Feb 2015 at 3:29 pm

> Hi,
> while everything you write in this article is true, you > are still missing one important point.
………………………………………

Yes, and the manufacturer of the (I presume switching regulator) warns about shining light on the package (chip). Maxim, Analog devices, TI, all do. These parts are almost exclusively aimed at mobile sector, hence no plastic package. And it’s closed behind some non-opaque chassis, so no need to worry there. But the Pi is open and therefore vulnerable to this (CMOS) effect – the same as on a CMOS chip in a digital camera.

My explanation for this behaviour is that the induced charge increases voltage on some feedback loop (more than on other parts of the regulator), and the switching regulator pulses less power to the output, to lower the voltage – and to reach the voltage goal. I guess it’s a 1.2V or 1.8V one for the memory or the core voltage vreg.

Solution – either use non WL-CSP package (but that would be bigger) or paint the exposed chip at the factory. There are resins that can be used to cover it including sides and underside of the chip. It must be insulating, and black. And better yet, it insulates the delicate solder balls from water and oxidation ;)

Splendid! Congratulations on the whimsical engineering!

As one who finds that imperfections contribute to beauty, may I humbly suggest you consider leaving the design of the PI 2 as it is.

I think that STEM learning can also benefit understanding the roles of imperfection and humour in the technical world.

Liquid tape.

Let’s officially call it the #FlashGate

Ok, you guys shined bright flashlights on them with no effect. I wonder how an IR flashlight would effect it. I just ordered my Pi2 yesterday and now I am REALLY itching to receive it. Though, I already feel sorry for all my other Pis that wont be played with as much. I guess I am going to have to give them permanent tasks (water the garden or something) and take them out of the test farm.

Wow, I am impressed!

How many commercial companies would investigate and then publish such a detailed analysis of a rather obscure bug so quickly?

You sure set an example for other companies with this one.

How about a dab of black silicone?
It would offer the flexibility not granted by epoxy.

The article is missing the most important info – will you keep it, or change the IC for next hardware revision?

We’re looking at changing it, but it’ll be later this year; lead times on alternative parts are some months long in the volumes we’ll require.

Either use the non WL-CSP package, or just cover the chip with Dow Corning’s Sylgard ### resin/epoxy. Just select the right color and ###number for the application. a drop or two might suffice :)

Sylgard 170 or Sylgard 170 FAST CURE might do the job

So glad it doesn’t happen with LED lights (like the PiGlow). I can imagine a big problem with my ISSAbove system if it suddenly developed a “feature” where it would self-reboot every time the International Space Station was nearby.

However a lot of customers do send me photos of themselves in front of their ISS-Above setup. I’m going to have to warn them of the possible reboot when I start shipping a Pi2 compatible version of the code.

Anyhow… looking forward to meeting ya’ll at the Birthday bash in Cambridge on Feb 28th. I’m travelling all the way from California to join in the fun.

All the way from CA! Wow – thank you SO much. Come and grab me and say hello at the Bash.

I’ll be sure to do that Liz. Of course I missed the boat as far as getting an official ticket to the party (via eventbrite) but PiMoroni were looking to squeeze me in via their allocation :-)

I will have a bunch of free ISS-Above SD cards to give-away… so I will come bearing gifts of an out of this world kind.

How would I go about finding your ISSAbove program?

well.. you click on my name and it takes you to issabove.com. I will be giving away a bunch of ISSAbove SD cards at the Pi Party Bash in Cambridge on Feb 28th.

Contact me for the link to get the download version if you are not going to the party. .

Very interesting effect! I look forward even more to receiving my Pi2 to play with. I gather the light can come down through the backside and bulk silicon die thickness, or sideways from the edges, reflect off the PCB, and onto the chip topside. I guess these bare silicon die devices are not intended to be used on boards without an external case.

Interesant

So, Xenon is Pi’s Kryptonite then? Super cool!!!

My bottle of black nail polish says: Ethyl Acetate, Butyl Acetate, Nitrocellulose, Acetyl Tributyl Citrate, Isopropyl Alcohol, Adipic Acid/Neopentyl Glycol/Trimellitic Anhydride Copolymer, Acrylates Copolymer, Stearalkonium Bentonite, Styrene/Acrylates Copolymer, n-Butyl Alcohol, Benzophenone-1, Silica and MAY CONTAIN Bismuth Oxychloride/CI 77163, Black 2/CI 77266, Ferric Ammonium Ferrocyanide/CI 77510 and a long list of various other coloring agents. Wow. I’ve used it for anti-reflection coating various optoelectronic prototypes without ever seeing conductivity or corrosion issues, but I never tried it on a bare die. I have also used black acrylic paint but found it slightly conductive; maybe it uses carbon black as the coloring agent.

This is great for a raspberry pi cluster. You can literally shoot the other node in the head (STONITH)

Try gluing some aluminium foil on top of the chip and see if the problem persists.

Cheers.

Don’t do that. You don’t want to use anything conductive. It has the potential to make electrical currents go places they shouldn’t.

Having placed a piece of bread over my Pi 2, and then tested its efficacy using a flash of a really, really bright light, I found that it is now toast.

Any other bright ideas?

Yours, Angry of somewhere near Wales

I am surprised you haven’t suggested trying to further dope the Si-die with BaCoN Dave :)

What a great article, Liz!
Not sure if this was mentioned but Albert Einstein was awarded the Nobel Prize for …..?
His 1905 paper explaining the photoelectric effect. What a great opportunity to use the Pi 2 as an prop for demonstrating the photoelectric effect in the classroom! Don’t fix it! Just supply a little cap or something for the offending component.

How about applying a snip of black vinyl electrical tape? Or gaffer tape? Insulating, removable, opaque.

It’s good that you are publicizing this. There are pranksters that delight in building and using things like TV-B-Gone, so anyone exhibiting a Pi2 based design publicly should consider taking precautions if the PCB is exposed. It might be easy for an attacker to use a laser pointer to mess up someone’s demo from a distance.

Let’s pretend the Pi2 is not boxed and is square on to you and there are no obstructions. Hitting that chip from even 5m, you’ve got an arc of about 0.2 degrees to play with. Basically, if you are near enough to disrupt the chip with a hand held laser pointer, you may as well reach over and turn the power off. With your nose. ;)

Class!

Yeayh, yeah…. I´ve heard this kind of non-sensical confidence before…. “Any attack made by the Rebels against this station would be a useless gesture, no matter what technical data they have obtained…..” and look how they ended.

I am guessing this may be an issue for sending the Raspberry Pi 2 to the ISS.

I know nothing of the required testing they do, but lots of nasty photons and other “stuff” floating around up there. Of course it wouldn’t be running life-support quite yet.

Glad it is only specific flashes that caused this issue.

There is a new market for some very tiny baseball hats!

The Pi2 isn’t going up, it’s the B+. There simply wasn’t time to do the testing — it’s VERY rigorous. A tiny mission patch would have solved the problem though :)

After minutes of flashing and lasering, I found out that it doesn’t work with the B versions.

Too bad. I was just about to set up my flyback transformer to produce some really nasty UV radiation. :D

So how long till someone starts building pi powered autonomous vehicles tasked to seek out and “flash” other pi powered machines?

Has anyone come up yet with a game of laser tag using a few B2 Pis – all you’d need is a little program that blinks a LED when “live” and stops blinking when hit…

Any idea of when this will be fixed in future version 2 Pis? (If it will, that is.)

“We’re looking at changing it, but it’ll be later this year; lead times on alternative parts are some months long in the volumes we’ll require.” – liz

As a teacher I’d deliberately hunt down V1 boards even when this got fixed, it’s quite brilliant :)

German resellers Watterot have posted a video of this on Facebook and titled it “Geblitzdingst”, which is my favourite thing so far in the whole of today.

Surly a quick blob of araldite (epoxy) would cure that?

SONY cybershot – 5 year old camera took it out.. must have xenon flash .. really neat lesson .. thanks for the explanation .. I wonder at what distance it will be effective? I was only a foot away …

@ 3 feet it had no effect.. so you really have to be doing close up work I guess.

Hey. Some folks have been whinging for an off-switch for years. No one thought you’d raise that to remote control.

First thought.

Great to see Blu tack being a useful fix for a British computer.

http://en.m.wikipedia.org/wiki/ZX81

Clive Sinclair recommended Blu tack to keep the 16k RAM pack in place on the ZX-81.

Even better, you can now get Raspberry coloured blu-tak (well ok, pink) !!

http://www.amazon.co.uk/Blu-Tack-801608-Bostik-Pink/dp/B0016Z4AYK

As a purely theortical question:
Has any testing been done with the Pi2 in apparently transparent cases?
I’m wondering if any currently on the market either filter or diffuse the light sufficently that this issue doesn’t arise, while still leaving the Pi on display.
Also, is there any essentially transparent material that could filter or diffuse the light sufficently to build such a case from.
Just random thoughts…

Fairchild have an application note that alludes to this problem (see the last paragraph of page 3) https://www.fairchildsemi.com/application-notes/AN/AN-5075.pdf . They suggest that it’s not something to be concerned about because it is only a problem with long wavelength light (red/IR).

Amazing that it only took a couple of days to be discovered on the Pi2.

I must confess to not having read the above 100 plus comments, but surely a simple solution is to just house your lovely new Raspberry Pi 2 in one of the many attractive and functional cases currently available!

I have a work light like the one at the link. It is quite strong and it resets Raspberry Pi 2 at close range.

http://www.lowes.ca/work-shop-lights/utilitech-500-watt-halogen-stand-work-light_8783751.html?af=3632&cse=3632&gclid=CNKXj_-R1sMCFXRo7Aod5QwAHA

I have one of those 500W halogen worklights too. It is worth mentioning that anything at close range to that light will eventually melt, smoke, and/or catch fire. So a cpu lockup might be the lesser of the various problems caused by that situation.

Sticky lumps of chewing gum fixed the ZX81 RAM pack wobble problem, and so should solve the RPi Xenon Death Flash issue.

Just wondering… Well, was it really confirmed that the problem is with the WL-CSP chip? I would rather put my bet on the board’s LEDs, because LEDs have another property besides emitting light: they can also act as a mini photocell, GENERATING a small voltage when illuminated!

Maybe the intense light pulses are making them generate pulses sent back to Raspberry Pi 2 and ruining some logic state inside the ports. As I do not own a Pi 2, would someone care to test this theory?

[minor edit by mod as per commenter request]

Based on the forum thread, it has been confirmed to be that WLCSP voltage regulator chip. The problem can be triggered when a laser pointer shines on that one chip and nothing else, and the problem can be fixed by an opaque mask over just that chip. Theory and experience agree that bare silicon die like that are light sensitive.

I translated this article in Japanese, do you I can be posted on my blog?

Yes, but please link back to this site and make it clear that the post was translated from something written here!

thanks! I posted article this blog
http://bokuweb.hatenablog.com/entry/death_flash

It’s “Pi2electric effect” :)

All semiconductors are light sensitive – not just silicon.
I discovered this the hard way – more than 50 years ago when I scraped the paint off my 1st transistor to see what was inside – in those days transistors were in little glass tubes.

Epoxy encapsulation keeps light away, but does not totally exclude it. Some of the white ceramic packages (used in the early days of microprocessors) were translucent and shining strong lights on circuitry could disrupt it.

In the old days one of the most popular transistors was the OC71 germanium PNP transistor, sold in the UK by Mullard, who also sold a photosensitive version called the OCP71 phototransistor. It was well-known that you could convert an OC71 to an OCP71 by removing the paint – worthwhile as there was a premium for the phototransistor!

One use I remember popped up in electronic mags to construct an “optical telephone” using a paintless OC71 to recieve the output of a modulated lightbeam.

Mullard eventually put the OC71 in a black epoxy case, stopping such pastimes, and soon after the OC71 itself fell out of favour in the face of smaller and cheaper NPN devices.

I know it’s somewhat off the main topic but I thought “the first yeasted bug fix” was penicillin.

I can’t hold a candle to that story. Ewwww …

Here’s watcha gotta do. Mount a light sensor between the Pi and any possible flash source (maybe keep the Pi in a dark corner). When a pulse from the sensor (caused by a flash) causes an interrupt, you put the Pi to sleep for a microsecond or two, then after it wakes up (unharmed) and before it resumes normal service, you get the audio to say “Bang! Whollop!” People will think it’s meant to be like that.

Just have posted this blog translated in Japanese to Japanese forum.if you are Japanese please check out the link.

http://www.raspberrypi.org/forums/viewtopic.php?f=82&t=99466&p=690358#p690358

Two (three?) questions:

U8 also looks like it uses a WL-CSP package and might be similarly affected. What does that IC do? (Will Pi 2 schematics be available sometime soon so we can answer these questions ourselves?)

Also, some people have claimed that flashing the Pi causes a brief glitch on one of the voltages (3.3v, IIRC), bringing it up to over 5 volts. Could this potentially damage the Pi, or is it too brief to cause any real problems? SD card corruption, while annoying, probably won’t cause any permanent damage, and you could probably work around that by mounting the filesystem read-only.

U8 is an HDMI auxiliary line ESD clamping device. I²C/CEC/HPD lines are protected by this chip (note that the HDMI differential signals themselves have integral protection built in to the BCM2835/2836 silicon to comply with HDMI specs).

U8 *would* be affected by the photoelectric effect, but it has no real impact. The reason for this is that ESD clamps are reverse-biased avalanche diodes. The photoelectric effect only increases reverse leakage through hole-electron creation – and you’d need a very high amount of incident flux to start to affect the I2C/CEC lines – they have very strong pull-up resistors on them.

https://en.wikipedia.org/wiki/Avalanche_diode

Under the photoelectric effect, the I-V curve for the diode shifts down and to the right. In the lower right quadrant, this effect is used for photovoltaic cells.

There is no effect of the flash on the 3.3V or 1.8V lines. Xenon discharge tubes also generate large electric fields – using poorly-shielded scope probes causes a large false reading.

So I can use laser pointer to reboot it. Like a remote :)

hahaha,
we should rename RPi2 to
“Myotonic Fainting Goat”
https://www.youtube.com/watch?v=AnVv0RkiG4U

I have to admire how a design flaw has been turned into an educational opportunity to teach about the photoelectric effect. However, I hope that something else is going on in the background too.

The Pi is marketed as a do-it-all computer with amazing embedded potential. Imagine what might happen with a Pi controlled drone if it’s hit by some excited photons.

The spectrum of responses from the institute ranges from product recall to sending a 1 cm strip of black sticky tape to everyone.

Seriously though, I’d probably look to the institute sending everyone who has one of these devices an email informing them of the defect and a link to this article. Now that the flaw has been discovered, they have a duty of care.

Not really a bug at all, but perhaps a specification update. It would at least make an interesting addition to the otherwise boringly thick and redundant booklet of “safety information” that presumably some lawyers have added to the Pi’s element14 shipping box.

Most printed circuit boards remain inside an opaque chasis for their entire life and hardly ever see any light, so mounting naked dies makes perfect sense. What makes the RasPi unusual is just the fashion of putting it inside transparent cases. Some of those could easily be modified very slightly to shadow the photosensitive components (e.g., small edits to layers 3 and 4 of the Pibow).

So rather than frantically searching for a board fix, I’d just update the safety-information data sheet with a warning of this effect, and leave it to users or case manufacturers to decide what to do about it.

Perhaps the next Pi could even have a naked CPU, such that I can update the registers with a microscope and a laser pointer? Smartcard hackers have long called this “semi-invasive attack”.

I am typing this on an elegant demonstration of the photoelectric effect!

Saw this on Sunday and had to try it out.

Wheeee!

I suppose it all depends on how powerful the Xenon flash is, my Canon G11 failed to upset my Pi2 beyond 50cm. Provided you keep the silly at a safe distance, you should be safe. ish.

Of course, a certain misinformed gossip site had their own interpretation on the issue: “C**p hardware, not surprising”. Ho-hum!

technically, taking the raspi in a rave may be a problem, considering lots of lasers and very bright lights are used for lighting effects, including xenon flashing stroboscopes !

“the current easy fix is to use a small blob of Sugru or Blu-Tak…”

How very ZX81

I have a vague recollection of reading that germanium transistors (a long time ago!) which came in painted glass packages could be converted into photo-transistors by the simple expedient of scraping off some of the paint.

As I booted my Raspberry Pi B+ yesterday [9 February 2015] for the first time, I took some pictures with my tiny Rolleiflex 550 full HD camera. With Flash. No problem.
Rollei seems to utilize some wavelengths that do not disturb the Raspberry so far…

Speaking as someone intending to buy one when the excitement of the ‘new’ dies down, I can say the rapid blog post about the non-problem and the entertainment of the posts from all the enthusiasts convinces me that this is the ideal bare bones computer for me at least

This could be good in a Pi Robot war, you could disable the enemies from a meter or so :)

I was surprised by the Sugru recommendation as their customer support site[1] notes it might cause corrosion if used on PCBs. Though in a rather vague fashion: “… With regards to PCBs, this depends on some factors, but it may react with copper, causing it to oxidate. …”

[1] http://sugru.com/questions/can-i-use-sugru-to-secure-shock-mount-a-sensitive-pcb

Hey Liz,

At Western Electric we used to environmentally coat our printed circuit cards with a spay on application called “glip-ticol” or a liquid simiar to “lock-tite” I’m curious if a simple spray on application could work here? Figure if we mask off the interconnections and chip heat sink locations and give it a good dose, then come up with waterproof interconnection covers and then we could create underwater PI’s… (Would this then be a Sub-Pi or maybe a Pi-Marine, yep, can see it now a Pi with a Green Beret. Or better yet create a high flying Pi, strap it to a solar RC plane and add a 10w segment for the Pi and create a Royal Pi Force. :) Just have to stay away from lightning until it’s sealed up… )

Cheers!
Dave Cooke

This issue is actually slightly ironic, my family owns an Event Photography company in Las Vegas and I ordered some Pi 2’s to use at some of our events. Events where there are normally multiple flash umbrella’s and various lighting and high end DSLR’s flashing constantly. Don’t think these will do too well in that environment. :P

I heartily recommend buying cases.

As amusing as this photosensitive part is, it apparently impacts the reliability of the new Pi in a way that could lead to unexpected crashes. Even if an immediate crash does not result, the resulting electrical noise could slowly lead to future data corruption and error.

Could the Raspberry Pi engineers please put together a safe, effective and permanent fix able to be performed by non-technical types in the field? A detailed how-to along with a cheap kit, if necessary, that could be purchased from the swag store with the correct paint or whatever would really help.

” …the resulting electrical noise could slowly lead to future data corruption and error.” Source / evidence?

As for a fix — Sugru or Blu-Tack is safe, effective and permanent. As is an opaque case.

Dear Clive,

Thanks for your quick reply recommending sugru or blu-tack. I was hoping to use a product whose design specifications included the required dielectric and optical properties. This would ensure some makeshift fix doesn’t lead to corrosion or damage a few months down the road. While Blue-Tack found in the U.K. may be fine, how do I know a similar looking generic will be suitable. Anyway, my Pi 2 hasn’t arrived yet and chances are this light sensitivity issue will be resolved at the factory by Fall semester.

The evidence that light sensitivity makes the current Pi 2 less reliable has already been presented. The scope readout indicates light can create noise with amplitude between -.25 to +.1 volts on the 1.2 volt processor core power supply. This 20% fluctuation is enough to shut the processor off. Smaller amounts of power supply noise in other processor designs can lead to bits flipping. While I have no proof, it seems plausible that the same could happen with the Pi 2.

Or put it in a case? Mind you, I’ve been using one bare for a few weeks, and not had any problems with it in an office environment, which is where most will spend their time.

@Kaldon–your RPi2s should be ok if you put them in opaque cases.

“If you need to use your Pi 2 in a situation where it might be flashed at, our advice is to cover U16 (make sure you get the sides too) – the current easy fix is to use a small blob of Sugru or Blu-Tak covering the whole component (someone in the forums used a pellet of bread: the first yeasted bug fix we have encountered), or simply put the Pi in an opaque case.”

I heard from a mate, who’s brother was in a Cambridge pub recently, that this is a secret design feature, put in by the Pi engineering team. Apparently some bloke called John nipped back from around 100 years in the future & said it would be a jolly good idea.

Light sensitive electronic is comon. I remember when i build photo-transistor with comon bipolar pnp or NPn like AC132. Those were germanium in aluminium case. If you remove the case the beta became dépendent on the ambient light. Older transistor from the 60’s, like OC71, were build in glass paint in black, if you scratch the paint they are good for light détection (sorry for my english). Anyway it’s the coolest bug ever !

The point is, it recovers. EMC requirements say that a device should be able to safely reset after conducted/coupled/inducted electromagnetic events.

This reminds me of a Simpsons episode way back when in ItchyLand the robots would die because of camera flashes :P

I don’t have a PI2 (yet) so I can’t check how warm U16 normally gets but with some of the fixes, like excessive use of blu-tack, are probably “thermally unhelpful”. Sure, small packages like this usually lose more heat conductively, through their connections to the PCB but still, encasing it in a thermal insulator might negatively affect the long term reliability. Maybe a small square of self-adhesive copper tape is better. Just don’t make it too large and short something! ;-)

U16 doesn’t seem to heat up, only the Processor and the USB/Lan chip get hot.
https://bmsdoug.files.wordpress.com/2015/02/ir001808.jpg

Rather than gumming up the chip with Sugru, why not cover it the same way an EPROM erasure window is covered? Take a small piece of black vinyl electrical tape, cut out a 1/4 inch (6 mm) square and stick it on top of the chip?

The black vinyl tape is almost transparent to INFRARED light, but it’s the ULTRAVIOLET light that a xenon flash tube generates that’s glitching the chip (the tape will stop all the UV no problem).

It would look a lot better than a “glob of glue” dribbled on the board.

The sides of the chip are vulnerable too, sadly. The chip itself is <1mm across, so you're going to have to be an accurate snipper!

You can always use black nail polish to cover the chip.

Nowadays i am sure you can find a teenage girl in school that has such artifact :)

It is cheap and available very easily and it does not affect the electronics.

Read the comments above – we can’t recommend nail polish because some of the solvents which are sometimes used in it may cause corrosion.

granted some do.

How about the plain coloured ones with no additive like mink oil, glitters or nail hardener. These uses plain enamel that should not effect the metal parts nor the plastic used in the chips casing. They come in small enough sizes and the paint brush is always included.

But forget about removing it after as the solvent (nail polish remover) contains acetone which may affect the plastic on some connector.

Q: have you considered the possibility that a Xenon discharge is a large source of electromagnetic radiation? Have you placed a piece of black tape on top of the chip and fired the flash at the same distance to make sure that it is indeed a light effect and not the EM radiation?

Yes we have – and you would be *amazed* to see how much EM testing the Pi goes through before it can be given CE or FCC certification. It’s definitely the photoelectric effect, and definitely not EM.

lol even pimoroni got in on it… chack their description of the pi 2.

Beautiful Pibow case (select your style above)
900MHz ARMv7 quad core processor
1GB RAM @ 450MHz
40pin extended GPIO
HAT compatible
10/100 ethernet
Xenon-flash reset function <——
4 x USB port
Use a microSD card for your operating system and data
microB USB connector for power
4 pole stereo output and composite video port
Full size HDMI
CSI camera port
DSI display port

I really want to buy the Raspberry Pi 2 but this problem is really off putting, guess I could find a way to cover it up somehow.

I was reading the thread on the forums and noted the thread was closed and locked now. The one question I have now is not addressed there. It also leads to another depending on your answer.

What can we do to protect our oversensitive U16 that will not violate our warranty? If you have not come up with a supported solution should we expect one?

I for one would like to eliminate this as a potential source of any future issues.

This was answered about the fifth post into the forum thread (and in the subsequent BBC news article on the subject): Blu-Tack.

We could simply take the electrons from the photoelectric effect, and charge the Pi2 with it! Should be 5V.

Got here by chance (link from Lifehacker.
Had huge fun reading all comments.
The upshot is clear: Liz rocks!

Aw – thank you!

OMG, what if you buy a lot of Pis and put them on robot and start a laser tag war!!! when you hit the U16 the robot will turn off. :D

instead executing sudo reboot, just flash it

I wouldn’t – you risk corrupting your SD card. (Great name, though!)

Hi,
is there coming a update, to eliminate this?
cheers

All that training and mentoring I gave you, Jonathan, and you spend your time now flashing (your lights) and innocent computers.

At least I wont’ have to worry too much about this bug up here in Aberdeen (except, maybe, at 3:18pm on July 19th – when Summer Time is expected to arrive, and remain for then following 11 minutes).

Will a flash of that intensity and duration do any permanent damage?

Will Factor 50 sun-tan lotion work?

Cheers,
Niall
Aberdeen

This was an unusually-entertaining, cheerful-spirited thread! (^_^) The “chip shot” was quite interesting.

Anyone contemplating using curable silicone should be careful not to use the common type that smells of acetic acid (vinegar). That corrodes copper. A respected manufacturer didn’t know that, and coils in their then-new inductive proximity sensors failed in the field.

NB
Waltham, Mass.

I’m reminded of a mystery in our forums that was solved yesterday: someone’s month-old Pi had mysterious spots of corrosion; he was going to return it, because he assumed it was our fault.

Long story short, turned out that the cupboard he was keeping it in had mice in it, and they’d been weeing on it. Urine will corrode solder joints just as surely as vinegar will…

could it be possible for a sleeping pi to be activated by a powered off pi through the hdmi cec function by exploiting the photo electric issue that way a pi could tell pi what to run?

Just got a model B+ and a pi 2 for some projects. The instructions on the little manual that comes with it state that a “maximum” current of 1500 to 2000mA should be used.

This should be a “minimum” instead. Slight correction!

Is it an RS or Farnell Pi? Those companies are responsible for the booklets: let me know which it was, and I’ll let them know there’s an error.

Got it thru Amazon so I can’t really tell. There is a link to rs-online.com at the bottom, maybe that helps.

It’s RS. Thank you. I’ll let them know.

now there is the big bang theory, have to love electronics. its going to do what it wants and sometimes we can’t stop it. 73 peoples.

I don’t understad what you mean by a bug, I thought this was a feature. When the end of the world comes via our old friend the nuclear holocaust it is well known that the EMP will destroy all electronics currently running.

With this feature, the flash from the nuclear explosions will automatically power down the Pi so that the EMP does not affect it. Then, once the end of the world is over we can all get back to playing … I mean working on our Pi!

Guys this is such a minor problem that it doesn’t warrant a silicon change. If people think it is a problem then use a sticky paper dot as per old eprom windows.

ok… when can i get the raspbian update to make my pi 2 i’m saving up for brave (oh, and throw in the maynard desktop and adobe flash, too)

What is the function of U16? I don’t see it on any schematics.

Comments are closed

News

Latest posts

Dragonfly Spectral Line Mapper

Refurbished vending machine

DSO-500K oscilloscope | HackSpace #67

Antarctic Picam | The MagPi #130

Next Post

Prepare for Pi Day!

Previous Post

Emulation on Raspberry Pi 2

Share this post

194 comments

David Honess

Peter

Thomas

Ravenous

Shannon Spurling

W. H. Heydt

critter42

Tom

Tom

Jimmy Dansbo

Adam

Tracy

Tong G

Chris Combs

Jeff Findley

jdb

W. H. Heydt

Silent-Hunter

Trent Williams

Ravenous

Peter Onion

Peter Onion

Peter Onion

Ravenous

Misel

Aaron

Ravenous

Aaron

Ravenous

Aaron

Dougie

Jim Manley

Jim Manley

Andrew

Ravenous

Ravenous

Ken MacIver

Alex

Clive Beale

Tom Westq

Vanfanel

Gerard

Kendalf

Peter Green

Kjell

Kjell

Chris-S

jdb

Lada

Greg Johnston

Luis

Steven

Haxor

Lada

Lada

Brian

JBeale

JBeale

fgimenez

Codu

W. H. Heydt

Pierre

Tong G

Keith

Clive Beale

Clive Beale

Phosphenius

Tony