EEVblog #716 - Raspberry Pi 2 Xenon Flash Problem Explained

[EEVmarty]

So.....What is the *FIX* for this problem? Just throw a blob of black resin (or nail polish) over/around the chip???

It's a bit like leaving the sticker off an Eprom window and waiting for it to go flakey lol

Marty.

[Howardlong]

What a terrible situation, amazing that they didn't consider such a use case during testing. I am sure none of us professional electronic engineers will have ever made such a fundamental life threatening school boy error as this in their lives.

I mean, the original Pi couldn't even deal with 12V shoved up it, what shocking design. http://youtu.be/lYf9HK-rI1s

In other news, 99% of eevblog viewers learned something new new today, while the other 1% pretended they didn't. 

/sarcasm

[EEVmarty]

99% of us don't bother with those 1 percenters  ;-)

What a terrible situation, amazing that they didn't consider such a use case during testing. I am sure none of us professional electronic engineers will have ever made such a fundamental life threatening school boy error as this in their lives.

I mean, the original Pi couldn't even deal with 12V shoved up it, what shocking design. http://youtu.be/lYf9HK-rI1s

In other news, 99% of eevblog viewers learned something new new today, while the other 1% pretended they didn't. 

/sarcasm

[SydB]

Dave, I think you give the engineering of that voltage regulator too much of a free pass.  Just because it settles back to regulating properly after the flash and doesn't go into some sort of oscillation or something is no reason to be complacent.  The fact is that it drops out, way outside of its presumed specification, in response to plain, ordinary light.  It is 100% common practice that packages are light-proofed and the manufacturer of this part has saved a quintillionth of a penny by ignoring this most sound best practice.  Shame on them.

This sort of package does not tend to be exposed to people firing xenon flashes during operation. These things are made to go inside phones, tablets, media players, etc. They're not actually supposed to be sat out in the open.

We have exactly an application where we must use a CSP package exposed. We need a suitable underfill product. Anybody know of one easily available in the UK? Have to be careful of cure shrinkage, need low viscosity etc.

By the way, WE had to tell the distributor that this could be a problem. Some datasheets note this, many do not!

[Fungus]

So.....What is the *FIX* for this problem? Just throw a blob of black resin (or nail polish) over/around the chip???

Ummm.... how about just not taking photos of it with Xenon flash tubes? Stick to the LED flash in your cellphone and you'll be fine.

[EEVmarty]

So.....What is the *FIX* for this problem? Just throw a blob of black resin (or nail polish) over/around the chip???

Ummm.... how about just not taking photos of it with Xenon flash tubes? Stick to the LED flash in your cellphone and you'll be fine.

No, it goes beyond that, i don't like having open can'o'worms on my boards if i use such devices.
While the Xenon tube scenario is in intense demonstration, i'm sure other sources of IR can be equally disruptive.

From now on, nobody should house their 'pi' in a clear box, or bare board in a cupboard etc where "at some stage" IR could flood the project and kill it.

Prevention..................better than cure!!!!!!!!!!

I don''t have a cellphone either, i use an odd thing called a camera....it might catch on ;-)

[mikeselectricstuff]

Has anyone tried a TV remote close-up? - the peak IR output is pretty high

[EEVmarty]

Has anyone tried a TV remote close-up? - the peak IR output is pretty high

or, that daylight stuff....it could stop a 'pi' in an outdoor project.

I think the next step will be to test various types of resins or potting compounds, maybe even paint or an acrylic nail polish as mentioned earlier, they might all shield intense IR light, they might not....they need to be proven though.

[Jimmyh]

Has anyone tried a TV remote close-up? - the peak IR output is pretty high

I tried a remote right on top of the chip, but it didn't cause a reset.  So far apart from a xenon flash, the only thing that causes a reset for me is the green laser pointer, with or without an IR diode to block the green light.

[JBaczuk]

Another gripe about those bare silicon chips is it's very difficult to see any marking in order to get the package assembled on the board in the correct orientation.  I had this problem with an EEPROM chip and it set us back a couple weeks.   

[Fungus]

Has anyone tried a TV remote close-up? - the peak IR output is pretty high

or, that daylight stuff....it could stop a 'pi' in an outdoor project.

Wouldn't it normally be in a box?

[kayvee]

Another gripe about those bare silicon chips is it's very difficult to see any marking in order to get the package assembled on the board in the correct orientation.  I had this problem with an EEPROM chip and it set us back a couple weeks.   

Ah... the virtues of First Article Inspection.  Must have been a painful experience

[tarheels100]

I wanted to clear up some of the science behind this effect.

First off, perhaps a better term (than the photoelectric effect) for the physical origin of this light sensitivity is the photovoltaic effect.  This is what is responsible for the light sensitivity of PN junctions in photodiodes, LEDs, transistors, etc.  The photoelectric effect, wherein an electron is ejected from the material into vacuum, isn't really going to be very relevant under the circumstances tested here.  The absorption of light by the silicon at the PN junctions in the device is the dominant source of photocurrent when exposing semiconductors to near-UV, visible, and IR light (like a xenon flash).  The bandgap of silicon is at a wavelength of about 1100nm and so any light with wavelengths shorter than that will be absorbed (up until about 400nm), inducing a photocurrent in the device.

Now, the thing that everyone seems unable to explain is why a xenon flash can do this while a flashlight or other light sources can't.  The explanation for that is all in the peak intensity of the light.  Most light sources are continuous, meaning they put out essentially constant light intensity.  A xenon flash, however, is a pulsed source, delivering a huge amount of light in a very short period of time (on the order of microseconds).  The peak intensity of a xenon flash is likely several orders of magnitude higher than even very bright flashlights.  This means that the flash is able to produce a large amount of photocurrent all at once, destabilizing the IC.  A flashlight on the other hand, even though very bright, only produces a trickle of photocurrent by comparison, which the IC is able to handle (dissipate) easily.  You can imagine that PN junctions form a sort of RC high pass filter that passes the xenon flash but not the flashlight.  Likewise, an LED flash doesn't cause the effect since it produces a longer pulse of light and with much less intensity.  So in short, the xenon flash is unique in the shortness of its burst of light.

Overall, I would predict that light in the wavelength range of 1100nm to ~600nm is the main culprit and that there is a peak intensity threshold necessary for causing the RPi to crash. 

[artag]

I tried a remote right on top of the chip, but it didn't cause a reset.  So far apart from a xenon flash, the only thing that causes a reset for me is the green laser pointer, with or without an IR diode to block the green light.

I've tried a red laser pointer. That caused a reset. With some masking around the edge of the chip it didn't (but a xenon flash still did).

[Fungus]

So in short, the xenon flash is unique in the shortness of its burst of light.

What if it's in daylight and you walk in front of it?

Do the sudden changes between light/shadow crash it?

[BarsMonster]

Dave, silicon is transparent to IR light, right where PN-junctions starts to loose sensitivity.

So photons to blame are between 1050nm and 1100nm wavelength: silicon is somewhat transparent there (so light can reach bottom of the die, where transistors are) and PN-junctions are still slightly sensitive.

Light with other wavelength (i.e. red laser pointer) can cause reset only if you shine it from the side, so that light does not need to go though the die.

[Howardlong]

If I shine a decorator's lamp with a 400W R7s halogen bulb at a Pi 2, it goes la la too.

The tube itself needs to be about 6" to 8" away to reboot it (mine reboots, it doesn't seem to hang). You can simply bring it closer and closer in, and then it goes bye bye.

Anecdotally, coming in from an exposed side of U16 seems to be slightly more susceptible, but it resets from the top and even the back of the PCB.

[Paul Price]

For some unreasonable reason, you have all fixated  your thoughts that light is causing the problem. You haven't proved anything if you fail to investigate electromagnetic EMI effect.

How do you know it is not due to an electromagnetic effect?  It could easily be that a voltage is induced in some wire on this pcb by the very high electromagnet energy,  the high voltage delivered to trigger or the xenon lamp or the power dump itself into the lamp creates a high power EMI pulse  that initiates a reset

[ruffy91]

For some unreasonable reason, you have all fixated  your thoughts that light is causing the problem. You haven't proved anything if you fail to investigate electromagnetic EMI effect.

How do you know it is not due to an electromagnetic effect?  It could easily be that a voltage is induced in some wire on this pcb by the very high electromagnet energy,  the high voltage delivered to trigger or the xenon lamp or the power dump itself into the lamp creates a high power EMI pulse  that initiates a reset

This has been discussed numerous times and has been disproved as it is possibe to reset the Pi with a laserpointer.

[tecman]

We had done some work on similar issues some years ago and found that the major culprit was not light, but the EM pulse.  As stated, there is a cap charged to several hundred volts and discharged in milliseconds.  There is a high current of many 10's of amperes in the form of a fast pulse.  This generates a significant high frequency EM radiation which can infiltrate and disrupt any number of devices.

I suggest that Dave use his fancy EM probes and look at the levels produced by the xenon flash.

paul

[Monkeh]

We had done some work on similar issues some years ago and found that the major culprit was not light, but the EM pulse.  As stated, there is a cap charged to several hundred volts and discharged in milliseconds.  There is a high current of many 10's of amperes in the form of a fast pulse.  This generates a significant high frequency EM radiation which can infiltrate and disrupt any number of devices.

I suggest that Dave use his fancy EM probes and look at the levels produced by the xenon flash.

paul

Bluetac or a plastic box should not be effective at stopping this effect. And yet..

[German_EE]

We had a kitchen timer that has a similar problem. The outside case is white plastic and when the kitchen light is turned on (a ceiling fluorescent) the startup pulses reset the timer. The cure was some aluminum foil on the inside of the case.

The chip in this case is a COB under a layer of epoxy.

[SeanB]

EM pulse will be a common stopper for electronics, but here it is the visible light penetrating the substrate.

I wonder if a drop of Loctite black superglue applied to the chip will act as an effective protection. It is opaque and sets into a flexible coating. Otherwise black non magnetic toner mixed with some thin laquer might be a cheap fix as well, otherwise you will need some underfill material.

[Howardlong]

For some unreasonable reason, you have all fixated  your thoughts that light is causing the problem. You haven't proved anything if you fail to investigate electromagnetic EMI effect.

How do you know it is not due to an electromagnetic effect?  It could easily be that a voltage is induced in some wire on this pcb by the very high electromagnet energy,  the high voltage delivered to trigger or the xenon lamp or the power dump itself into the lamp creates a high power EMI pulse  that initiates a reset

Er, no. Look at my post immediately before yours.

I deliberately tested for a transient by bringing the halogen lamp in gradually with it switched on. As the light is gradually brought in closer, it resets.

[TheEPROM9]

Dave one thing occurred to me while watching the video. It would be interesting to repeat the test but with the camera at different distances so you could separate the EMI noise from the light and see how it effects the circuit. Also it might be worth repeating the test but attaching different wavelength filters to the flash and seeing how that effects the results and what wavelengths are the ones responsible for causing the reg to drop out.