GamesCom, Shoulderbuttons and Heating

[EvilDragon]

Well, there weren't posts for a while, since we've been busy preparing flyers, etc. for the GamesCom.

And that took a lot of our time.

The shoulderbuttons

In the meantime, a green case arrived - the latest revision with the new shoulderbuttons.

Do they work?

Well... yes, basically, they do.

However, there's one thing I can tell you after having assembled the case:

While we've got four shoulder buttons and all of them can be pressed, it won't be comfortable to use the upper ones when gaming.

It's not possible with the extremely small and cramped space we got up there.

The lower ones need a bit of tweaking but will work comfortably then, but for the upper ones, you will have to bend your fingers a bit, which is not nice to do for extended periods of time.

They will be fine for functions you don't need that often in games and they will work fine as modifier keys when typing (as your thumbs are lower when you're typing than when you're playing).

If you're coming to the GamesCom, you can test them yourself and let us know if you got anymore good ideas, but we already discussed that with multiple users here, and no one got an idea how to solve that in a comfortable way.

The Heating

There's been a lot of discussion about heatsinks, special materials and even tiny fans on the other thread, since it seem the OMAP5 will get a lot hotter than the OMAP3.

Well, I finally found the time to properly measure it with a professional IR thermometer (without heatsink).

The result:

The OMAP5 internal measurement seems to be off quite a bit.

When it shows 80°C, it's actually 55°C on the surface of the SoC itself.

With 112°C (the maximum I could reach with maximum stress on both CPUs yet), it's 84°C.

The OMAP3 in the Pandora has a maximum of about 70°C, so it seems the OMAP5 (under full load) is about 15°C higher.

That's not that much. The Pandora doesn't use any heatsink, so a few copper layers as heatsink should be enough.

aTc also did some tests with his devboard.

The heatsink on the OMAP5 doesn't do much cooling actually - unless you use a fan.

So the only real way to lower the temperature a lot would be using a fan.

We probably don't want that, so the best thing we can do is distribute the heat inside of the case to prevent hot spots.

Shouldn't be much of an issue though.

The GamesCom

We're leaving for GamesCom tomorrow morning and hope to meet some of you there.

We'll have a Pyra devboard there as well as the prototype case and some nice PCBs to look at (and see how the CPU board will work and look like).

We'll take pictures as well, and I'll try to write small reports for you.

 

[Null]

We'll take pictures as well, and I'll try to write small reports for you.

Could we have a picture of the flyers as well (please)?

 

[TrashyMG]

When it shows 80°C, it's actually 55°C on the surface of the SoC itself. With 112°C (the maximum I could reach with maximum stress on both CPUs yet), it's 84°C.

I figured it was something like that..the numbers were way too hot to be real..  I just didn't have a good way of measuring it myself outside of sneaking the devboard into work.. 

Trashy's todo list.

   - Purchase DMM Thermocouple Adapter. 

   - A whole lot of other things..

   - Write Pandoralive article.

 

[Silent-Hunter]

So why the heck does the sensor do that?

 

[Honk5000]

Would the shoulderbutton layout from the Wii Classic Controller be an option?

 

[EvilDragon]

Would the shoulderbutton layout from the Wii Classic Controller be an option?

Only if we sacrifice all ports :/

 

[Creature XL]

So why the heck does the sensor do that?

The sensor is probably inside the chips housing. ED only messured the outside.

 

[Silent-Hunter]

Yes, but what do the specs reference? Core temp, or surface temp?

 

[TrashyMG]

So why the heck does the sensor do that?

These sensors are fairly accurate.. however without calibration, we don't know what offset is needed to reflect a true temperature reading. I'm not sure Ti has figured it all out yet or haven't shared a typical offset. 


At  work, I've seen some chips have over 20C degrees of variance... Sometimes chip production variances can sway the readings too.. I've seen parts that draw a tad more/less current or parts that run a bit faster/slower change the readings of the internal temperature sensors.

 

[CFWhitman]

Have we absolutely ruled out buttons placed in-line (somewhat similar to NeoGeoX, I thought) due to space considerations?  Or would it be worth testing that arrangement as well?

 

[EvilDragon]

Have we absolutely ruled out buttons placed in-line (somewhat similar to NeoGeoX, I thought) due to space considerations?  Or would it be worth testing that arrangement as well?

There's no space.

We've got less space than with the Pandora shoulderbuttons.

 

[edgex004]

Have we absolutely ruled out buttons placed in-line (somewhat similar to NeoGeoX, I thought) due to space considerations? Or would it be worth testing that arrangement as well?

There's no space.


We've got less space than with the Pandora shoulderbuttons.

What about two buttons on the underside of the case?

 

[TrashyMG]

What about two buttons on the underside of the case?

Eww.. 

 

[Grench]

The shoulder buttons - is there any chance of a side by side instead of over-under layout?  Or has that already been tried?

The Heating - I fear there is confusion between spot temperature and heat.  Those core temps AND your laser surface readings may both be accurate.

Many CPUs and SoC's have a small built in heat spreader (sink).

The spreader/sink will have some natural dissipation to the air around it when run lid-off.

Running them hard until they self-throttle shows where the max temperature is at - good to know - but not how much heat is being dissipated during either max or normal load.

To truly know 'how much heat' there is to dissipate, in THEORY that would be measured in calories per minute (heat/time).

One calorie is the amount of heat that it takes to increase the temperature of 1cc of pure liquid water 1*C at STP.

That said, calorimeters for silicone are rare, stupid expensive and rarely very accurate.

So, the real question is:  How much heat does the SoC need to dissipate (not just store in a sink) to prevent it from ever getting hot enough to need to throttle under normal circumstances?

Experimental method:

Find 10 copper coins - clean with silver polish.  (US pennies pre-1965 excl WWII tin variety work.)  Alternatively a 10g copper balance weight would be perfect.

Smear them with thermal grease and seat them on one SoC or the other.

Leave the top of the upper most coin 'clean' - that's the one you'll be firing your laser temp reader at.

Record the start temp, record temp every minute on the minute during the experiment.  This will yield the rate of change K/t.

Measure the weight of the coins (or copper weight) on an accurate balance.  This gives the mass of a specific substance which we can find in the heat capacity tables.

Reference this page:  http://en.wikipedia.org/wiki/Heat_capacity

Knowing those factors it should be possible to calculate a benchmark of how much heat any give surface (two SoCs) are shedding in relation to each other.

You'll also then be able to calculate pretty near-to-exactly how large the heat sink needs to be to handle SoC heat over X time period.

Then, if you choose to, you could also calculate how much surface area to air radiator interface is needed to hold the SoC at a steady state of <=Y*C in a room of Z*C.

Or you could just assemble one, let 'er rip and see how hot it gets.  If it maxes out then your thermal solution is inadequate.  

Personally I'd aim for no more than 80% of the temperature difference between, 'normal ambient' (30*C) and max (throttle).

For that it's pretty easy - run the proposed design flat-out until it reaches steady state and see if it ever crosses 80% of max SoC temp at room temp of 30*C.  30*C+((112*C - 30*C)*0.80) = 95*C.  If it crosses 95*C then the thermal design is inadequate and other solutions should be investigated.

Alternatively one could claim that the device is not designed to run flat-out continuously.  Then you may want to figure out what load is reasonable, apply it and take measurements to ensure that it can handle your expected heat load at your 'normal' processing load.  

It's not uncommon these days for device manufacturers to assume that their product will NOT be completely tasked for extended periods.

Just my humble opinion though - I'm not an engineer on this project.  I have every confidence in ED's abilities and thought processes.  I'm simply adding my thoughts because I find the project fascinating.

 

[Null]

What about two buttons on the underside of the case?

 

[EvilDragon]

Have we absolutely ruled out buttons placed in-line (somewhat similar to NeoGeoX, I thought) due to space considerations? Or would it be worth testing that arrangement as well?

There's no space.


We've got less space than with the Pandora shoulderbuttons.

What about two buttons on the underside of the case?

No space - and I don't want to push them each time I put my unit down

 

[EvilDragon]

So, the real question is:  How much heat does the SoC need to dissipate (not just store in a sink) to prevent it from ever getting hot enough to need to throttle under normal circumstances?

Well, I already made quite extensive experimental testings a while ago.

I cut an SD Card case in half (those small plastic cases), which I put completely over the OMAP5 (without heatsink).

So the OMAP5 was trapped inside the plastic (worse than what we will have with the sandwich PCB) with little air inside.

Even at 110°C, it was a LOT less hot than touching the surface of the OMAP5 directly.

Doesn't seem to dissipate that much heat.

 

[Grench]

Looking at the button pics here:  http://boards.openpandora.org/topic/16700-a-small-photo-story//URL]

It looks like the lower button is long - goes to case edge and the top button is short.

What if that were swapped?  I.e. have the upper button go to the case edge with a touch of wrap and the lower button be recessed from the corner?

I could see it working to be able to snag the edge of the top button at the case edge and the lower button down and 'inward' of that.

Or is there something else in the way to prevent that?

 

[edgex004]

What about two buttons on the underside of the case?

Eww.. 

What about two buttons on the underside of the case?

Have we absolutely ruled out buttons placed in-line (somewhat similar to NeoGeoX, I thought) due to space considerations? Or would it be worth testing that arrangement as well?

There's no space.


We've got less space than with the Pandora shoulderbuttons.

What about two buttons on the underside of the case?

No space - and I don't want to push them each time I put my unit down

Yeah not my first choice either. If they were flush with the case like the Pandora's, then they wouldn't get pushed.

Honestly even a slightly inelegant solution is a good solution for me in this situation. Utilizing shift and control on the Pandora's shoulders make keyboard usage so much more efficient. Any way to get fn and alt on the shoulders would be very much appreciated.

Could the shoulder button be made bi-directional? Push back towards your palms for R1/L1 but up towards the lid for R2/L2?

 

[Grench]

So, the real question is:  How much heat does the SoC need to dissipate (not just store in a sink) to prevent it from ever getting hot enough to need to throttle under normal circumstances?

Well, I already made quite extensive experimental testings a while ago.

I cut an SD Card case in half (those small plastic cases), which I put completely over the OMAP5 (without heatsink).

So the OMAP5 was trapped inside the plastic (worse than what we will have with the sandwich PCB) with little air inside.

Even at 110°C, it was a LOT less hot than touching the surface of the OMAP5 directly.

Doesn't seem to dissipate that much heat.

Sorry - I'm trying hard to avoid too much Physical Chemistry here.

Temperature <> Heat

Weight <> Mass.

Heat:  http://en.wikipedia.org/wiki/Heat

All that your experiment did was to find out at what temp the SoC throttles to self regulate by insulating it.  It appears to be somewhere around 110-112*C.

That you could touch the SD card case on the hot OMAP only means that plastic (the SD card case) is a poor thermal conductor (an insulator) - not that the SoC was 'good with that'.

You seem to have your mind set though.  I'm going to back off and let you work it through.