Some personal stuff - and prototypes building!


I didn't like very much how the Pandora was using the USB as a dissipator... I know there wasn't much choice, and it actually always worked well when the port was not in use, but when something was plugged in the temperature could reach very high levels.
The Pandora wasn't using the USB ports as a dissipater was it? I remember you could tell if your CPU was overworking by feeling for a point between the nubs where the CPU was. The Pyra on the other hand has this heatsink connected to the CPU and resting against the USB port, as well as having a number of vias acting as thermal conduits through to the internal copper layers which can also buffer a little heat. I never noticed any of those on my Pandora PCB. The Pandoras 1GHz OMAP3 could only do 2.0DMIPS for every MHz ticked, while the OMAP5 scores up to 4 on each core, therefore it can be hitting the RAM faster and heating that up, but the weedy CPU in a Pandora would rarely benefit from any kind of cooling, therefore it was never implemented there.
 
Not necessarily, it depends on the amount of heat produced and whether you have a fan or not.
The fan would be used to blow air through the fins and basically blow the heat away - in this case, you'd need to have an area where air can be sucked in and blown out.

Without a fan, the air wouldn't move around a lot - not even if there's an opening. The cooling is (of course) a lot less than when using a fan.
The fins are simply there to increase the overall area where heat is in contact with air.
The air WILL still cool it, even inside the plastic case.

And as long as the air is cooler than the CPU, then it will work :)

I fully agree, but as there is so little air inside the case and no ventilation holes I don't think it will be cooler than the cpu for very long :S.

The USB port doesn't have any fins, so the area where it's in cotact with air is VERY little.
Also, as the USB port only has a thin metal layer (compared to the thick heatsink), it would heat up a lot faster.

Combine a faster heating with less contact to air to cool off and you can understand why this would probably make things worse.

That is why i suggested replacing the heatsink with a copper block (basically a heatsink without fins) so it has more thermal capacity than the little amount of air we have inside of the case.

To actually get the heat out of the case, the USB ports seam like the ideal option (by connecting the heatsink or copper block to them) as we do not have a lot more options, there are no air vents and plastic is a bad thermal conductor.

Although, all this needs to be tested before we can be certain of what is actually the better option.

Th best option would probably be to have a metal IO shield on the back of the case and use that to dissipate the heat, but that is something for the version 2 of Pyra.

TLDR: As we do not have much air circulation inside the case I think thermal capacity is a lot more important than air contact area for the heatsink.
 
I think there might be more air circulation that you expect to be honest. The heatsink fits between the USB port and the bottom of the case. There's a whole other part of the case to be attached from the photo I copied currently at the top of the page, and air can move between the right hand side and the left hand side of the case (I don't believe there's a route for air to the air channel on the other side of the battery)

I do observe however that in that photo, the vanes of the heatsink appear to be oriented the wrong way to encourage the movement of air; it pretty much fills that valley of plastic, so each of the air masses between the interval fins are pretty much sealed, only the outside of the two edge pins will likely warm up surrounding air. It might be worth testing runs with the heatsink round each way, which would help to show whether the convection of air within the case is having any measurable effect.

I also note the close proximity of that hole where the battery clip goes so if significantly warmed air does start circulating within the case, it's going to enter the battery cage with ease there. All attempts to stop heat conducting through the plastic to the battery cage might be lost there. Perhaps there are benefits to having most of the vanes from the heatsink blocked.

I don't know how much of this is a real concern however. Under normal use, the CPU and RAM may not be as stressed all of the time as it's impossible to be. But I can rest assured the likes of ptitseb and notaz are likely to be trying to stress it as much as possible when porting some of the most demanding software to the system.

Either way, this needs to be tested. You can normally get thermal reports from linux systems by looking in /sys/class/thermal; on my laptop currently, there's a file called cooling_zone0/temp in there that currenly reports 58000 which translates I believe to 58 degrees.
 
I have the idea/dream/fantasy of, using a rotary tool, cutting a hole through the back of a Pyra case then creating a wax mold of the whole space+hole, then texturing to increase the surface area that pokes through the hole as an external radiator and using that to create a solid silver fitted one-off heat sink & radiator. Note - I won't be doing that to my prototype - and if the Pyra 'as is' has adequate thermals, I won't even bother to think about it again. Just a thought of, "how can I max this out." Might just be a complete waste of time & money (literally) though.
 
Still air is a shitty heat conductor. Convection moves warmed air upwards, iff there is an unhindered passway and cool air can freely flow in from underneath or from the sides. Convection does not cause a lot of pressure to force circulation through bottlenecks, let alone a series of those.
 
The Pandora wasn't using the USB ports as a dissipater was it?
I actually don't remember who told this, but was Craig or DaveC or someone involved in the project in the early years, I wasn't even registered at the time, just lurking around collecting every information...
 

Sorry for off topic, and don't want to derail thread, but where is he now?

Remember him from GP2X days. He was very passionate about screens and sound problems in emulated classic
arcade games. Didn't he design Pandora dpad as well?
 
DaveC made the Case Design for the Pandora, Craig dit some testing and some Videos and pr and selling in his English Shop
Mweston made the PCB Design (the electronic)
 
A thermal block of copper would be the best choice. It will take approx. 30 sec to warm it up from 20°C to 75°C. (6W dissipation, 0,864 cm³ of copper) under optimal conditions, such as very low thermal resistance of the copper tape. The time for booting the device, i guess. The additional copper mass is also not significantly higher in comparrison of the copper mass of the cpu-board. The problem with the thermal mass is, once it is on temperature, it will take a long time to cool down. Without convection, the heat cant go anywhere. The thermal radiation is also minimal. This thermal solution isnt worth the effort and also not worth further delays.
 
What one can do, when controlled (by design and thermal principals or with moving parts) convection ain't in the cards, is to move the heat to your outermost parts (the case) and if that's a shit conductor, widen the cross section you use. Like - spread the heat to the case, but to every point of it you can reach.

Or build in a heat exchange chamber with a door or drawer, where you can regularly change fluid water for solid water.
 
A thermal block of copper would be the best choice. It will take approx. 30 sec to warm it up from 20°C to 75°C. (6W dissipation, 0,864 cm³ of copper) under optimal conditions, such as very low thermal resistance of the copper tape. The time for booting the device, i guess. The additional copper mass is also not significantly higher in comparrison of the copper mass of the cpu-board. The problem with the thermal mass is, once it is on temperature, it will take a long time to cool down. Without convection, the heat cant go anywhere. The thermal radiation is also minimal. This thermal solution isnt worth the effort and also not worth further delays.

The current solution (as can be seen here) has already been proven years ago that it improves the situation vastly.
While a Pyra without that heatsink warms up running at 1GHz, the one with the heatsink applied stays cool.

ptitSeb has a Pyra with the heatsink and also did some compiling at 1.5GHz, it stays A LOT cooler than without.
Without the heatsink, the CPU throttles after 5 seconds to 300MHz when maxing out at 1.5GHz,

Pictures with thermal cameras also show that the hotspot is moved away into the heatsink.
So yes, it definitely isn't the perfect solution (which would be a lot more complex), but tests have shown that it is still vastly improved, so why not do this?
 
Pictures with thermal cameras also show that the hotspot is moved away into the heatsink.
Just to avoid confusion in readers, that aren't aware. The heatsink cannot be warmer than the heat source - not without a Wärmepumpe. (What's Wärmepumpe in English? UPDATE: It's heat pump. Could have guessed that.)
[doublepost=1555434140,1555433678][/doublepost]
The current solution (as can be seen here) has already been proven years ago that it improves the situation vastly.
While a Pyra without that heatsink warms up running at 1GHz, the one with the heatsink applied stays cool.

ptitSeb has a Pyra with the heatsink and also did some compiling at 1.5GHz, it stays A LOT cooler than without.
Without the heatsink, the CPU throttles after 5 seconds to 300MHz when maxing out at 1.5GHz,
Just to avoid confusion in me. @EvilDragon Were those "open-air" events?
 
Just to avoid confusion in readers, that aren't aware. The heatsink cannot be warmer than the heat source - not without a Wärmepumpe. (What's Wärmepumpe in English? UPDATE: It's heat pump. Could have guessed that.)

Yes, that's for sure :)

[doublepost=1555434140,1555433678][/doublepost]
Just to avoid confusion in me. @EvilDragon Were those "open-air" events?

A fully assembled Pyra, inside the GamesCom halls (and in my home)
 
The USB port doesn't have any fins, so the area where it's in cotact with air is VERY little.
Also, as the USB port only has a thin metal layer (compared to the thick heatsink), it would heat up a lot faster.

If the tape contacts both the USB shield and the heatsink nothing prevents adding an exterior "USB" radiator.

Just a dummy copper plug (that doesn't short out the USB 5V) with fins.
Could even make a larger version that powers a tiny fan from the USB port but I doubt it'd be necessary.

Want to do a build or play heavier games? Plop that radiator into the USB port.
Need to store the Pyra away or just surfing the web, remove the external radiator.

It'd be similar to water-cooled laptop docks.

...but something much smaller than the Asus GX800.
 
Without convection, the heat cant go anywhere.
Bro, plastic also conducts heat, albeit slower than metal: from there it'll dissipate into the air, desk, or hands. The CPU produces an amount of heat energy in a concentrated area, the heat spreader and heat sink take that energy and spread it around the case; lower concentration of heat means lower necessary transfer rate. Yes, the CPU is producing more heat than can be dissipated in its tiny window, but by spreading it around the amount of heat transfer required is much lower; assuming EvilDragon isn't lying then we can be assured that the amount of heat, spread over the whole surface, is lower than the heat transfer rate of the plastic, making it 100% sufficient, probably better than.
 
Bro, plastic also conducts heat, albeit slower than metal: from there it'll dissipate into the air, desk, or hands. The CPU produces an amount of heat energy in a concentrated area, the heat spreader and heat sink take that energy and spread it around the case; lower concentration of heat means lower necessary transfer rate. Yes, the CPU is producing more heat than can be dissipated in its tiny window, but by spreading it around the amount of heat transfer required is much lower; assuming EvilDragon isn't lying then we can be assured that the amount of heat, spread over the whole surface, is lower than the heat transfer rate of the plastic, making it 100% sufficient, probably better than.

Well, I am pretty sure it will be impossible to keep it running for an extended period of time with full CPU strength.
That doesn't work for ANY system out there. Heck, I don't even know a smartphone that doesn't get REALLY warm in your hand if you play some complex games for extended periods of time.

So yes, unless we use a fan, it will NEVER be possible to max out the CPU. More recent ARM SoCs (anything newer than A9, excluding the low-power ones) have not been designed for that.
They have been designed to be able to PEAK at full CPU power for a few seconds, so that webpages render faster, etc.

It won't work, unless you have the space for a REALLY huge heatsink.

So yes, I can assure it's not enough if you plan to max out the CPU usage with 1.5GHz.
But it's enough to not heat up in your hands when doing normal stuff (or playing games), which is better than what most smartphones manage.
 
I hope the USB socket does not get so warm that it desoldering itself. :oops:

Well, I am pretty sure it will be impossible to keep it running for an extended period of time with full CPU strength.
That doesn't work for ANY system out there. Heck, I don't even know a smartphone that doesn't get REALLY warm in your hand if you play some complex games for extended periods of time.

So yes, unless we use a fan, it will NEVER be possible to max out the CPU. More recent ARM SoCs (anything newer than A9, excluding the low-power ones) have not been designed for that.
They have been designed to be able to PEAK at full CPU power for a few seconds, so that webpages render faster, etc.

It won't work, unless you have the space for a REALLY huge heatsink.

So yes, I can assure it's not enough if you plan to max out the CPU usage with 1.5GHz.
But it's enough to not heat up in your hands when doing normal stuff (or playing games), which is better than what most smartphones manage.

They never built any Smartphone that does not throttle under full load? No crazy cooling solution or anything innovative? I mean "Gaming-Smartphones" exist I've heared. I just ask because I see some market possibilities if you can speed up everything under load without faster chipsets, just by keeping the peak load longer than other devices. This would have valuable usages, dont forget: they built folable Smartphones just because they can! (even if nobody ever asked for this xD )
 
I hope the USB socket does not get so warm that it desoldering itself. :oops:
The melting point of solder (including the lead free stuff used in a RoHS compliant Pyra) is north of 200°C. If you've got that sort of heat inside a polycarbonate case, I'd expect something else to go pop before the solder starts to melt. And in case that worries you, we already know that the CPU will throttle itself if it gets above a specified temperature which'll be well south of 100°C (my x86 machines tend to hover around 60 degrees before they start spinning up their fans and throttling back processors)
 
Back
Top