EEVblog #901 - Raspberry Pi 3 Photoflash Problem

[SimonR]

Hi SeanB

You're pretty much spot on with that explanation. I used to work for a programmer manufacturer and there were literally hundreds of variations of that algorithm across the many manufacturers. They were mostly just twists on the same theme as everyone patented their own particular version. The ultimate one was called Intel quick pulse. 12 VPP, 6V VCC 10uS pulses up to 25 pulses. I'm not sure if this one had an over prograrm. pulse or not. some needed an extra pulse of number of short pulses required time some other pulse length. My brain is hurting trying to remember it all.

[EEVblog]

You're pretty much spot on with that explanation. I used to work for a programmer manufacturer and there were literally hundreds of variations of that algorithm across the many manufacturers.

As someone who wrote and sold EPROM programmer software 20+ years ago I can testify to that!

[Tek_TDS220]

I'm sure the cause is UV light, but keep in mind that the black tape is probably filled with carbon black.  Can black tape block the electrical pulse from the flash electronics, even though the conductivity of the tape is not high?  I would have used a piece of PET (which absorbs UV) cut from a water bottle, or even a piece of 'soft' glass.

[Monkeh]

I'm sure the cause is UV light, but keep in mind that the black tape is probably filled with carbon black.  Can black tape block the electrical pulse from the flash electronics, even though the conductivity of the tape is not high?

Err, carbon black in insulating tape meant to have a breakdown voltage of 40kV/mm? This seems unlikely.

[firewalker]

Could the problem occur due to rapid temperature rise of the flash?

Alexander.

[sibeen]

I must be old. I remember having to black texta all over OC81 transistors (if memory serves) to ensure that they weren't acting as inadvertent photo transistors.

This *may* have been a few years ago.

[mikerj]

Why are people attempting to come up with alternative theories?  Flip chip devices are well known to be more sensitive to light, and Dave has done plenty enough to confirm this is the cause in this case.

[SNGLinks]

I must be old. I remember having to black texta all over OC81 transistors (if memory serves) to ensure that they weren't acting as inadvertent photo transistors.

This *may* have been a few years ago.

I remember scraping the paint off an OC71 to make an OCP71.

[eugenenine]

So what if it has a minor known issue.  Known issues are easier to workaround.  I'll take $29.99 with a known issue over a $500-$1500 system with a close source OS that has random unpublished issues.

[MK14]

So what if it has a minor known issue.  Known issues are easier to workaround.  I'll take $29.99 with a known issue over a $500-$1500 system with a close source OS that has random unpublished issues.

There was a time, when we would often get 3 or 4 blue screens of death, per day. I think (famously) Bill Gates himself was doing a demonstration in front of lots of people. Then suddenly got the blue screen of death. It was/is hilarious.

I think with a number of embedded micros, e.g. Pic. There is potentially a list of erratas anyway. I.e. the MCU can be slightly faulty, even before you design it into a system or computer.

[eugenenine]

My work provided laptop did a BSOD last weekend, its a 6 month old laptop and cost $1200.  Back when I worked for a reseller we had one customer with an $Apple$ costing $2000 and he must have sent that thing back to Apple 6 times in the first year.  I upgraded my web server from an RP2 to an RP3 when the 3's were first available, the 2 ran for over a year, maybe once or twice I rebooted when a kernel update was released but no problems.

[EEVblog]

Why are people attempting to come up with alternative theories?  Flip chip devices are well known to be more sensitive to light, and Dave has done plenty enough to confirm this is the cause in this case.

Yes, it's obviously the light, do the experiment yourself. This is a well known problem with bare die, no mystery at all.

[Monkeh]

My work provided laptop did a BSOD last weekend, its a 6 month old laptop and cost $1200.  Back when I worked for a reseller we had one customer with an $Apple$ costing $2000 and he must have sent that thing back to Apple 6 times in the first year.  I upgraded my web server from an RP2 to an RP3 when the 3's were first available, the 2 ran for over a year, maybe once or twice I rebooted when a kernel update was released but no problems.

Comparing a high performance laptop running unknown userspace applications on an extremely large and complex codebase with an unstressed webserver.

Apples and peanuts.

[Brumby]

So what if it has a minor known issue.  Known issues are easier to workaround.  I'll take $29.99 with a known issue over a $500-$1500 system with a close source OS that has random unpublished issues.

You pragmatist, you.   

[Fungus]

So what if it has a minor known issue.  Known issues are easier to workaround.  I'll take $29.99 with a known issue over a $500-$1500 system with a close source OS that has random unpublished issues.

Is it really an issue?

It's really just an experiment to confirm a known theory.

An issue would be if Pis randomly crashed if you took them outdoors at noon or something like that.