Group Actions
Engagement
  • Author Author: bwelsby
  • Date Created: Date Created
  • Views 1415 views
  • Likes 5 likes
  • Comments 6 comments
Related
Recommended

Raspberry Pi 3 temperature and cooling testing Part 2 with a HAT tests.

bwelsby

Raspberry Pi 3 temperature and cooling testing :

 

Part 1 initial tests.

Part 2 with a HAT tests.

Part 3 with Fan Speed Control. (Still to come)

 

 

 

The BeagleBoard  has capes, the The specified item was not found.  has Shields but the Raspberry Pi has HATS (Hardware Attached on Top) - an add-on board to provide additional functionality but which can closely cover the SOC chip and restrict the cooling airflow. There are many of these available some of which you can find here  The specified item was not found.

 

Following on from my initial temperature tests I then decided to see what effect adding a HAT would have on the running temperature of the RPi 3. However as I did not have any of these I decided to 3D print a dummy one to  RPF specifications here a OpenSCAD source for this is attached to Part 1.

 

One thing I should mention here though is that real HATs also contain electronic circuits that may generate some additional  heat.

 

I started out with one that sits 10mm above theRPi 3 PCB and repeated the tests as described in part 1.

 

10mm dummy HAT

imageimageimage

 

Not surprising there was very little difference in the results to the initial tests with enclosure cover installed, so close it was impossible to distinguish when plotted.

 

I therefore decided to try one that sits just 5mm above the RPi 3 pcb  which is about as close as I would expect to get, for this I needed to make a cutout to clear the HDMI connector. and a slot for the GPIO pins.  (after printing and fitting I realized I should have added a cutout  for the Audio connector too but as this is further away from the fan I didn't feel that the slight offset would make much difference.)

 

5mm dummy hat

image

 

The results for these tests were again very similar to the original "with cover" tests but this time there was a noticeable increase in the maximum temperature reached on those tests that maintained a temperature below the 80'C threshold where the processor reduces its operating frequency to help keep the temperature down. This was approximately 10'C as you can see on the graph of the results below. I tried to make things easier to see with this chart (hopefully).  The tests are T1 to T8 no HAT and T1H to T8H with a HAT installed.

The original test plan is:

Test#  Fan   Cover    Heatsink   Graph Colour

1         0        0           0               Brown

2         x        0           0               Red

3         0        x           0               Orange

4         x        x           0               Yellow

5         0        0           x               Green

6         x        0           x               Blue

7         0        x           x               Violet

8         x        x           x               Grey

 

And the results, if you click on the chart you should get a larger image.

image

 

Without a fan the 80'C threshold is reached much sooner but as you can see the heatsink still has an effect in delaying this.

 

Still to come:  Part 3  Adding temperature control to the cooling fan to reduce noise.

  • in reply to cstanton

      wrote:

     

      did you ever add speed control/monitoring for the fan?

     

    Sorry for the delay in responding, life's a bit chaotic at the moment.

    Yes I did get a prototype working using an ATiny85 but then I was side tracked and never got back to it. It was probably overkill really anyway. One thing I did learn was that the ideal PWM frequency for the fans is 20KHz  so is above audio range to prevent the fan squealing.

    The other thing I found was that just a small amount of air flow was sufficient so a slower, quieter fan would be ok too.  If you have seen the Pimoroni FanSHIM for the RPi 4, this works really well without any heatsinks and is very quiet.

     

    Brian

  • did you ever add speed control/monitoring for the fan?

  • in reply to shabaz

    Yes I agree, there wasn't a real problem with the first Pi  and possibly Pi 2 is ok but I will get to checking that on full  load too image. 

  • in reply to bwelsby

    That's a real concern. I've been looking at the HAT requirements here https://github.com/raspberrypi/hats

    and I couldn't find a height specified, so in theory people could produce very low profile HATs that would be incompatible with heat sinks.

    Pimoroni's two "Raspberry Pi 3 Heatsink" offerings on their website are 6mm and 7.5mm high for example, so added to the 1mm or so of the Broadcom chip, that is 7mm or 8.5mm high. I think it is a weakness, they should specify a minimum and maximum height. Maybe they already

    have, but I cannot see it.

    Looks like the thinner ceramic ones are the only option, and forced air cooling, at high performance levels.

    Since the HAT is directly above the processor, it has become a fundamental part that will affect cooling considerations, so needs to be

    specified better than it is today.

  • in reply to shabaz

    shabaz wrote:

     

    Are there any HATs that are 5mm? I'm guessing if there are then they would use the types of DIL headers that can allow the mating pins to pass through the PCB.

    I have not looked at the HAT spec yet though..

    Yes the Raspberry Pi Sense HAT is one with such headers, and the Foundation web site link details these at the bottom. They recomend mounting with standoffs so could go down to 5mm or 6mm if there's nothing under the HAT to short or interfere.  The SHIM RTC Realtime Clock Accessory Board for Raspberry Pi can be pushed down even closer but thats something else.