There's a Weird Problem with the New Raspberry Pi Computer

CommanderFrank

CommanderFrank

Cat Can't Scratch It
Joined
May 9, 2000
Messages
75,399
Just about all computers have some embedded quirks that can’t be easily explained away. Take the new model Raspberry Pi 2; early testers have found that when you try to take a picture of the diminutive computer with a Xenon flash, the whole kit and caboodle crashes. Nothing worse than a camera shy computer. :D
 
Very neat...

although he doesn't look to have an oscilloscope, he has a multimeter which does not have the resolution to detect drops in voltage over that short of a time frame.
 
So when the robot apocalypse comes, all we need is a camera with xenon flash.
 
Gonna have to go with some sort of EM pulse since the camera recording the whole thing also glitched when the camera flashed.
 
the Pi is a development board. that's why finished/retail electronic products have Shielding to block emi.
 
Reminds me of light sensitive eeproms. (The version I heard goes like this) some guy was pulling apart a camera while it was still powered only to find it was still working without the optics. They tried it again but it didn't work until someone peeled the eeprom sticker off. Very interesting stuff.

I always wonder why we don't have a light based computer yet.
 
Well known phenomena with poorly shielded solid state electronics. Because of RFI/EMI issues, there are places I cannot take my digital camera at the Nukes I work at. I would hate to be the cause of the China Syndrome.:eek:
 
Retroactive said:
Well known phenomena with poorly shielded solid state electronics. Because of RFI/EMI issues, there are places I cannot take my digital camera at the Nukes I work at. I would hate to be the cause of the China Syndrome.:eek:
Click to expand...

Last i heard those stockpiles are getting dismantled from the START treaties.Of course, that doesn't mean all of them. Minuteman's right?
 
viscountalpha said:
Last i heard those stockpiles are getting dismantled from the START treaties.Of course, that doesn't mean all of them. Minuteman's right?
Click to expand...

Those treaties are only going to reduce the stockpiles somewhat but I have no doubt that there are secret stockpiles in all participating countries. Its not like its hard to build a deeply buried base and store whatever you want in the middle of nowhere.

That treaty will be used as a way to close down openly known silos and allow programs to "relocate".
 
viscountalpha said:
Last i heard those stockpiles are getting dismantled from the START treaties.Of course, that doesn't mean all of them. Minuteman's right?
Click to expand...

I would think most of our nuclear capability is in the depths of the oceans where it is less detectable and less window of opportunity (for them) and probably not fully acknowledged.
 
Here's the thread where the problem is diagnosed:

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

It's definitely a U16 chip that is part of the power supply circuit, and that chip may be a substitute or even a counterfeit chip.

A really interesting bit is: "Blu-tak around the edges [of the U16 chip] was sufficient to stop the laser pointer affecting it but not the flash."
 
evilsofa said:
Here's the thread where the problem is diagnosed:

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

It's definitely a U16 chip that is part of the power supply circuit, and that chip may be a substitute or even a counterfeit chip.

A really interesting bit is: "Blu-tak around the edges [of the U16 chip] was sufficient to stop the laser pointer affecting it but not the flash."
Click to expand...

I doubt the chip is counterfeit. CMOS circuits are naturally light sensitive. For most chips, the solution to this problem is to use black opaque packaging. The issue is due to fully encapsulating parts, the actual die is much smaller than the package. This tends to be a problem for small portable devices since it leads to wasted board space. The U16 chip is a WLCSP (wafer-level chip scale package) part.

WLCSP are a new type of package that removes the need for extra packaging by directly placing the solder bumps on the bottom of the CMOS wafer and not on a separate substrate in the package. A good example is the Nordic nRF51822 BLE chip. The QFN part is 6x6mm, but in a WLCSP package, the chip size is only 3.2x3.2mm making it take just over a quarter of the board space.

One disadvantage with WLCSP parts is since it is essentially just a bare CMOS die, is unlike packaged parts, they are light sensitive. To help with the light sensitivity and to improve the structural integrity during handling, most WLCSP have either an epoxy coating or LC tape applied to the backside of the wafer. Unfortunately, the back side coating only shields the backside and the edges of the chips where they are diced and the bottom are left exposed, so they are still effected by a bright light. Most of the time, this isn't an issue since the pcb is then put into an opaque sealed case.

For the Raspberry Pi, the device is shipped as a bare PCB, so it isn't surprising that the WLCSP parts would be light sensitive. The interesting thing is several pictures and videos in the thread show the top of U16 as shiny and not black. This makes it look like a manufacturing defect where somewhere in the assembly process, the backside coating was somehow removed from the component. This would make the chip very light sensitive.
 
Just use it inside your farady cage, and run cables. Problem solved.
 
rpeters1969 said:
Bloody fascinating! Reading th Pi forums, it seem that the SMPS IC is photosensitive:

http://www.raspberrypi.org/forums/viewtopic.php?f=28&t=99042
Click to expand...

This was exactly my first thought. When I worked in a research lab playing with DRAM wafers, we used to have to let the wafers sit in total dark for a while before probing down because if we accidentally probed across transistor pads, the charge built up from absorbing the fluorescent light in the lab could cause damage to the wafer. Even if we didn't damage the wafer, the measurements were all skewed for a time until the charge could be dissipated.

It's interesting though, usually IC packaging is designed to make them resistant to this sort of thing. The joys of cost cutting to keep the Pi in the market it's intended for, I'm guessing.
 
Not all that surprising, if you put an HID Xenon light near a working computer, the one-off errors go through the roof.
 
When I was in college, we used a cardboard box with all of the seams covered with duct tape to cover our probe station when we tested wafers from our microfab lab. It was cheaper than commercial alternatives. Just place the probes, then cover it up!

The probe lab had windows. Lots and lots of windows.
 
hardcase said:
When I was in college, we used a cardboard box with all of the seams covered with duct tape to cover our probe station when we tested wafers from our microfab lab. It was cheaper than commercial alternatives. Just place the probes, then cover it up!

The probe lab had windows. Lots and lots of windows.
Click to expand...

Heh, we made Faraday cages out of MDF and aluminum foil. There was a lot of equipment running in the lab and the signal noise was terrible without them. Cheers to novel solutions though! And people think because it's university research, you can just throw money at the problem. Pfft!
 
spugm1r3 said:
This was exactly my first thought. When I worked in a research lab playing with DRAM wafers, we used to have to let the wafers sit in total dark for a while before probing down because if we accidentally probed across transistor pads, the charge built up from absorbing the fluorescent light in the lab could cause damage to the wafer. Even if we didn't damage the wafer, the measurements were all skewed for a time until the charge could be dissipated.

It's interesting though, usually IC packaging is designed to make them resistant to this sort of thing. The joys of cost cutting to keep the Pi in the market it's intended for, I'm guessing.
Click to expand...

My guess is the component is shaved to fit strict real estate requirements. The stock version of the component may be fully protected.
 
Or the people sourcing the component didn't know they were ordering unshielded components while shielded ones exist. A simple manufacture engineering screw up.
 
lcpiper said:
Light isn't in the EM Spectrum is it?
Click to expand...

Specifically (from wikipedia): The visible spectrum is the portion of the electromagnetic spectrum that is visible to (can be detected by) the human eye. Electromagnetic radiation in this range of wavelengths is called visible light or simply light. A typical human eye will respond to wavelengths from about 390 to 700 nm.[1] In terms of frequency, this corresponds to a band in the vicinity of 430–790 THz.
 
Of course it's an EM pulse from the flash. Bad grounding or noise isolation, probably in the clocking circuit causes the CPU to crash.

Put it inside a leaded glass container, then photograph it ... bet it doesn't glitch then.

Welcome to the difference between great engineers/designers and the run of the mill shlubs out there in hardware design world.
 
You must log in or register to reply here.
Back
Top