We managed to get our hands into a Raspberry Pi 3 and decided to give APM a try with it using the PXFmini. Here’s a walkthrough over some of the tests that our team conducted: 

Benchmarking

So, down to the benchmarks, we performed 3 types of tests using sysbench and the default Raspbian images (no APM running for now). SysBench is a modular, cross-platform and multi-threaded benchmark tool for evaluating OS parameters that are important for a system running a database under intensive load. The output of sysbench looks like this:

sysbench --test=cpu --cpu-max-prime=20000 run

sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:

Number of threads: 1

Doing CPU performance benchmark

Threads started!

Done.

Maximum prime number checked in CPU test: 20000

Test execution summary:

   total time:                          477.3324s

   total number of events:              10000

   total time taken by event execution: 477.3236

   per-request statistics:

        min:                                 47.69ms

        avg:                                 47.73ms

        max:                                 85.54ms

        approx.  95 percentile:              47.72ms

Threads fairness:

   events (avg/stddev):           10000.0000/0.00

   execution time (avg/stddev):   477.3236/0.00

 

The output is graphed against the Raspberry Raspberry Pi 2 (note that smaller bars indicates better results):

sysbench --test=cpu --num-threads=1 run



sysbench --test=memory run --memory-total-size=2G

sysbench --test=memory run --memory-total-size=2G --num-threads=1 --memory-oper=read

The 64 bits CPU of the Pi 3 dramatically improves the results of this particular tests over the Pi 2. It’s 40% - 60% faster.

Mounting the autopilot:

We mounted the shield as follows which proved to be robust enough. Vibrations were kept at a reasonable level as will be shown later in the log analysis:

Flying with APM and log analysis

After a short flight we started some checks. The official documentation of APM will tell us:

“Check the scale on the left and ensure that your vibration levels for the AccX and AccY are between -3 and +3.  For AccZ the acceptable range is -15 to -5.“

Fine on this. We also felt that the drone was reacting really good so made a few plots that convinced us about the autopilot responses with the Raspberry Pi 3:

 

Conclusion

Raspberry Pi 3 is a great candidate to make APM Linux autopilots using the PXFmini!. Having Bluetooth and WiFi is indeed a great asset removing the need of using additional USB dongles unfortunately, WiFi is solely b/g/n and does not support the 5 GHz frequency band (which comes handy when flying with RC controllers that work in the 2.4 GHz band).

Views: 26318


MR60
Comment by Hugues on March 2, 2016 at 8:00am

Nice. What's the hardware specs of RPi3 : CPU, RAM, etc ?

Comment by Víctor Mayoral on March 2, 2016 at 8:35am

Hi Hugues,

According to raspberrypi.org:

  • A 1.2GHz 64-bit quad-core ARM Cortex-A53 CPU (~10x the performance of Raspberry Pi 1)
  • Integrated 802.11n wireless LAN and Bluetooth 4.1
  • Complete compatibility with Raspberry Pi 1 and 2

It does indeed feel like a big improvement.


MR60
Comment by Hugues on March 2, 2016 at 8:39am

And in terms of connectivity and powering the board, is there any improvement vs Rpi2 ? (where a powered Hub was required which is a pain)

Comment by Víctor Mayoral on March 2, 2016 at 8:43am

None that i've been able to identify. For now, we are powering the Pi3 through the autopilot shield and using several dongles on it. Raspberry pi org mentions that it may be more power hungry though:

This time round, we’re recommending a 2.5A adapter if you want to connect power-hungry USB devices to the Raspberry Pi.

Comment by JB on March 2, 2016 at 8:45am

I like forward to the new A53 Pi compute model based PXF mini....when will you make it? The new compute module will be here in a few months. ;-)


Developer
Comment by Kabir on March 2, 2016 at 8:59am
Nice!
Does the single-sided mounting make plugging in connectors difficult? Does the board not flex or bend?
Comment by Víctor Mayoral on March 2, 2016 at 10:24am

JB,

I have no experience with the pi compute modules but sounds like an interesting idea. Will give it a though.

Kabir,

No it does not unless you apply a lot of force. The header and the spacers seem to do a pretty good job.

Comment by Jerry Giant on March 3, 2016 at 12:43am

Great report for the timing and quality.

I will begin tests on nVidia TX1, hope we can work together.

Comment by Jerry Giant on March 3, 2016 at 12:44am

Pi computing Module has been obsolete now, RS-online now dropped it.

Comment by Víctor Mayoral on March 3, 2016 at 3:01am

Hi Jerry,

The TX1 is indeed interesting. I'll PM you on this matter.

Regards,

Comment

You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2017   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service