Safety concerns

Drop in for a chat and a cup of tea

18 posts
by recantha » Sat Jul 21, 2012 7:59 am
Hi.
With the release of the new Debian image which runs faster, have there been any studies with respect to safety and temperatures on the RPi? Before the Raspian release, my chips and components were getting near the temperature I'd start to worry about when students are using the boards. Since Raspbian, the temperatures are rising.

By design, the components are able to be touched while the Pi is running and I need to know how much health and safety procedure I need to go through to safeguard against burns etc.

--
Michael
My Raspberry Pi blog with all my latest projects and links to articles
http://raspberrypipod.blogspot.com. +++ Current project: PiPodTricorder - lots of sensors, lots of mini-displays, breadboarding, bit of programming.
Posts: 209
Joined: Mon Jun 25, 2012 10:41 am
by jamesh » Sat Jul 21, 2012 8:35 am
I don't believe there is an issue here. The chips should not get hot enough to burn, if they do I suspect there is a failure on the board somewhere.

That said, there are other issues with regard to the 'school/teaching' version, which means they will have to have cases anyway, which has always been the foundations plan.
Volunteer at the Raspberry Pi Foundation, helper at Picademy September and October 2014.
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 12120
Joined: Sat Jul 30, 2011 7:41 pm
by rurwin » Sat Jul 21, 2012 8:43 am
In my experience when you put a finger on the hot chips, your finger draws heat away faster than the chip can replenish it, so I don't believe there is a burn hazard. I'd be more concerned about all the trailing wires.
User avatar
Forum Moderator
Forum Moderator
Posts: 2933
Joined: Mon Jan 09, 2012 3:16 pm
by KenT » Sat Jul 21, 2012 11:38 am
Out of interest I measured the surface temperature of the chips after running at 100% load for tens of seconds. (very un-scientific just held the thermocouple on with my finger)

CPU - 32 deg C
Ethernet - 35 deg C

My hot water tank which became uncomfortable, but in no way burnt, after many seconds of holding the thermocouple - 40 deg C

I would be much more concerned about protecting the Pi from the students and I would have thought some sort of case was essential for primary and secondary education.
Pi Presents - A toolkit to produce multi-media interactive displays for museums, visitor centres, and more
Download from http://pipresents.wordpress.com
Posts: 606
Joined: Tue Jan 24, 2012 9:30 am
Location: Hertfordshire, UK
by Grumpy Mike » Tue Jul 24, 2012 9:38 pm
Before the Raspian release, my chips and components were getting near the temperature

Just because Raspian gives you faster results it does not mean the processor actually runs any faster, it is just that the instructions it does perform are more efficiently targeted towards the task.

The processor runs at a fixed speed and produces the same amount of heat no matter what instructions are being executed.
User avatar
Posts: 791
Joined: Sat Sep 10, 2011 7:49 pm
Location: Manchester (England England)
by mahjongg » Tue Jul 24, 2012 9:50 pm
completely agree with grumpy mike, with one small addendum, if you start using 3D graphics and such the GPU turns itself on (yes it does have a kind of sleep mode), and the PI might start using extra power. That should still not generate any heat issues through. The 700mA mentioned everywhere as needed for a PI is made up of 500mA maximum for the PI and two "single current units" (100mA) for each USB port. So at most the PI is using 2.5 Watt (0.5 Amp x 5.0 Volt).
User avatar
Forum Moderator
Forum Moderator
Posts: 5865
Joined: Sun Mar 11, 2012 12:19 am
by recantha » Tue Jul 24, 2012 9:57 pm
We should just all over-volt and overclock to 2Ghz, see that baby FRY ;-)
My Raspberry Pi blog with all my latest projects and links to articles
http://raspberrypipod.blogspot.com. +++ Current project: PiPodTricorder - lots of sensors, lots of mini-displays, breadboarding, bit of programming.
Posts: 209
Joined: Mon Jun 25, 2012 10:41 am
by Grumpy Mike » Wed Jul 25, 2012 9:07 pm
I never understood the obsession some people have with over clocking. At best you end up with a system that runs slightly faster than other peoples. This means that running other software could be too fast like in games and such. Or other people running your code would see it running slower and that might throw up problems that you didn't see and you could have fixed if you did see.
Also you never know if your processor is going to choke on an instruction at a certain temperature, so odd behavior might be left uninvestigated where as it might be just a programming error.

I can't see the advantage of stepping out of the group experience.
User avatar
Posts: 791
Joined: Sat Sep 10, 2011 7:49 pm
Location: Manchester (England England)
by recantha » Wed Jul 25, 2012 9:17 pm
I think the group experience is tending towards over-clocking. It's just another way of testing out the limits of technology, or revving the engine if you like.
My Raspberry Pi blog with all my latest projects and links to articles
http://raspberrypipod.blogspot.com. +++ Current project: PiPodTricorder - lots of sensors, lots of mini-displays, breadboarding, bit of programming.
Posts: 209
Joined: Mon Jun 25, 2012 10:41 am
by mahjongg » Wed Jul 25, 2012 11:39 pm
Really? I only think that less than 1% will be interested in such a concept of overclocking, unless it really makes a difference in the user experience.
User avatar
Forum Moderator
Forum Moderator
Posts: 5865
Joined: Sun Mar 11, 2012 12:19 am
by jamesh » Thu Jul 26, 2012 8:12 am
Well, it's completely safe to overclock to about 850, and 1Ghz with mild overvolt. And those are pretty high percentage improvements which make a big difference to usability.
Volunteer at the Raspberry Pi Foundation, helper at Picademy September and October 2014.
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 12120
Joined: Sat Jul 30, 2011 7:41 pm
by mahjongg » Thu Jul 26, 2012 12:37 pm
"mild overvolt" will make for "mild warranty-void", or rather full warranty-void.
Maybe because the PI will be used more by "hacker-types" (that is, curious kids) that overclocking will be more common, I don't know, all I know is that on most other platforms overclocking is very rarely done, but perhaps you are right.

I do think though that grumpy mike made some good points, although many kids wouldn't be deterred by them, I rather think they will make a sport out of it.

Tip for those interested, the hotter the device the faster it gets (which is a bit of counter intuitive, but true.). Obviously it also -ages- faster when it does.
User avatar
Forum Moderator
Forum Moderator
Posts: 5865
Joined: Sun Mar 11, 2012 12:19 am
by meatballs » Sat Jul 28, 2012 8:03 am
Worry about a warranty on £25? :lol:
Posts: 14
Joined: Sat Mar 03, 2012 3:24 pm
by Colly » Sun Jul 29, 2012 7:56 pm
mahjongg wrote:Tip for those interested, the hotter the device the faster it gets (which is a bit of counter intuitive, but true.). Obviously it also -ages- faster when it does.


Counter intuitive? As conductors heat, the molecules vibrate faster and electrons are more easily knocked loose and electrical resistance is thereby reduced. A heated conductor has lower resistance than the same conductor when cold. As the electrons in heated wire can move around more easily, it seems logical, rather than counter-intuitive, that the chip can perform faster when hot. Of course, if it heats to the point where microscopic connections come unsoldered, or insulation barriers melt and break down, then it should be expected to fail.

Over-clocking will cause switching bits to change state faster, which will cause voltages to drop, so that's why overclockers boost voltage a bit to avoid that drop. However, package designs are created to dissipate certain amounts of generated heat and increasing the voltage, increases the watts of heat being generated. At some point, one will exceed the rate at which the chip can dissipate heat and the interal connections will fail.

Want a long lived Raspberry Pi? Run it at design levels. Want a fast Pi? Overclock it - after you buy a spare to have handy in case you burn it out.

Next up: Who has connected tiny Peltier chips to cool the hot spots on their Pi for some extreme over-clocking? Yes, they would likely cost more than the Pi, but if you're out to set over-clocking records, that seems the way to go. Peltier chips with, perhaps, water cooling added to the hot side of the Peltier chip. You'd be sucking the heat right out of those chips. Over kill? Of course, but isn't that what over-clocking is about?

Just for the heat dissipation method: Maybe someone might need this to help ruggedize the Pi for application in a hostile environment, say in a desert setting being run by solar power where any Pi enclosure might be getting excessive heat added by sunlight. Perhaps the Pi is controlling the tilt of some solar array panels in a desert setting? One might want to implement both Peltier and fan cooling, just to keep ambient heat extremes from toasting the Pi.

At only $35 each, no doubt many people will be testing Raspberry Pi in all sorts of extreme environments. Model rocketry to log altimeter readings? Landings could be rough when parachutes fail to deploy. Possibly even more risky than the preschool classroom I work in, with 20 four-year olds who have yet to find something they can't break. :)
Posts: 17
Joined: Mon Jul 16, 2012 4:27 am
by mahjongg » Mon Jul 30, 2012 1:22 am
Counter intuitive in the sense that overclockers desperately try to cool down the CPU of the PC they are trying to overclock, and it seems they are trying to make it much cooler than a normal system is.

But if you do know a bit more about it, it seems that these people do not exactly know what they are doing, because indeed the hotter the silicon the faster it becomes! You can even say quite exactly how much percentage ponts it will become quicker with each ten degrees increase in temperature.

Also increasing the switching voltage als increases the switching speed, as (parasitic) capacitors in the logic get filled up faster. Obviously increasing the voltage generates (much) more heat.
User avatar
Forum Moderator
Forum Moderator
Posts: 5865
Joined: Sun Mar 11, 2012 12:19 am
by Colly » Mon Jul 30, 2012 3:42 am
Doh! I got it completely backwards! Resistance INCREASES when conductors heat up! It's explained very well here: http://wiki.answers.com/Q/What_would_lo ... _of_a_wire

Also, it makes sense, considering somethings actually become superconductors at ultra-low temperatures. That'll teach me to check my facts next time before I post! :)
Posts: 17
Joined: Mon Jul 16, 2012 4:27 am
by Colly » Mon Jul 30, 2012 3:54 am
mahjongg wrote:Counter intuitive in the sense that overclockers desperately try to cool down the CPU of the PC they are trying to overclock, and it seems they are trying to make it much cooler than a normal system is.

But if you do know a bit more about it, it seems that these people do not exactly know what they are doing, because indeed the hotter the silicon the faster it becomes! You can even say quite exactly how much percentage points it will become quicker with each ten degrees increase in temperature.

Also increasing the switching voltage as increases the switching speed, as (parasitic) capacitors in the logic get filled up faster. Obviously increasing the voltage generates (much) more heat.


Perfectly explained! Now I get you. And yes, I bet there's a formula somewhere that would look nice in a graph showing how hot, gets how fast, right up to the point where the heat overcomes the chip and the connections or insulation fails and everthing goes poof. Just like formula race cars! How hard can we push the engine and get the car to go as fast as possible without (quite) blowing up the engine?

I think I'll plan on running mine at normal speeds and hope it lasts a long time, since I'd probably slip a decimal point somewhere and toast mine on the first attempt at overclocking. Also since my knowledge of the involved physics and electronics is sadly lacking.
Posts: 17
Joined: Mon Jul 16, 2012 4:27 am
by mahjongg » Mon Jul 30, 2012 12:08 pm
common wisdom is that silicon based electronics breaks down at about 160 degrees Celsius.
That doesn't mean that other stuff on board won't suffer earlier, for example electrolytic capacitors are normally rated 85 degrees Celsius, and are known to explode when they get a bit warmer than that, because the liquids inside them starts to boil!

It is also said that lifetime halves for every 10 degrees Celsius increment, so that is one reason to keep expensive electronics cool.
User avatar
Forum Moderator
Forum Moderator
Posts: 5865
Joined: Sun Mar 11, 2012 12:19 am