Why use a Due for the WEC?

[ryancoe]

From Steve's email:

• Due with its ARM processor partly for its significantly faster 84MHz clock (other Arduinos have AVR processor w. 16 MHz clock)
• has some special functions for quadrature decoding on certain timer pins

@DeepFriedDerp @nickross4444 - Is this all still relevant with our current set up?

[nickross4444]

@ryancoe @DeepFriedDerp The faster processor seems to be useful, as I'll explain further down. We aren't using the quadrature decoding functionality right now, but we are having a little trouble with the SPI to the external encoder buffer, so we might end up using. That being said, Gerrit and I have some confidence in being able to figure out the SPI.

I did some testing on the interrupt timing on the Due and mega. Both of these were done with all functions active(except maybe the jonswap), so I believe them to be accurate to what we will see with hardware runtime. Each arduino has 2 interrupts, the control loop and the serial loop. I kept the serial loop at it's standard 32hz and pushed the control loop down until it stopped working:

Arduino Due:

• Fastest stable control loop period: .002s(500hz)
• Interrupt variation: 0 microseconds
• Control loop exception time: 182 microseconds
• Significant control loop functions: Velocity calc, PD calc, value mapping, pwm writing.
• When pushed to fast, serial stopped all together.

Arduino Mega:

• Fastest stable control loop period: .003s(333hz)
• Interrupt variation: 0~100 microseconds
• Control loop exception time: 1130 microseconds
• Control loop without SPI: 850 microseconds
• Significant control loop functions: Velocity calc, PID calc, value mapping, SPI writing.
• When pushed to fast, The serial interrupt never had time to execute.

I was pretty surprised with the speed of the interrupts, I thought the Due would be much faster than the Mega. I think the most important factor here is the interrupt variation. The mega's variations were less than 1% but still much worse than the Due. I think for that and the significantly faster execution time it's still worth using the Due for the WEC controller. I also still think that the mega is fine for the Wavemaker, since interrupt tightness is less important for that control algorithm.

@gbacelli Could you look at these numbers and let us know if they seem acceptable? If not, we can talk about switching to the Zephr ROS, or to a teensy, which should operate up to a Ghz and might be easier to convert our code to.

[nickross4444]

Will close this issue once we have tested the control loop on the hardware.