[WIP] Dev dshot

[Igor-Misic]

Describe problem solved by the proposed pull request
This PR adds drivers for Dshot protocol. (This is not full support for Dshot protocol, only drivers.)

Test data / coverage
Dshot protocol was tested on Pixracer. This still needs to be integrated into the whole system and tested.

Describe your preferred solution
We have discussed on a slack channel to add this driver support.

Describe possible alternatives
No alternative, at the moment this is how other PWM drivers work.

[dagar]

How were you testing this?

[dagar]

Suggested change

	rPSC(timer) = ((int)(io_timers[timer].clock_freq / dshot_pwm_freq)/DSHOT_MOTOR_PWM_BIT_WIDTH) - 1;;
	rPSC(timer) = ((int)(io_timers[timer].clock_freq / dshot_pwm_freq)/DSHOT_MOTOR_PWM_BIT_WIDTH) - 1;

[Igor-Misic]

With calling init function with different DSHOT frequencies.

example:

up_dshot_init(channel_mask, DSHOT1200);
up_dshot_init(channel_mask, DSHOT600);
up_dshot_init(channel_mask, DSHOT300);
up_dshot_init(channel_mask, DSHOT150);

And then sending the arming command:
up_dshot_arm(true);

Setting throttle value:

    uint16_t throttle = 100;
    up_dshot_motor_data_prepare(0,throttle, true);

And triggering the message:
up_dshot_trigger();

I've connected logic analyzer on dshot (servo wire) to see frequencies.
ESC was working for my setup Pixracer at 168 MHz on all Dshots modes. It was able to arm and to spin the motors.

For each Dshot message, frequency needs to be near selected.
For example dshot 1200, the frequency will be around 1.2 MHz. It depends on the selected system frequency, and this formula will calculate the closest one by rounding it down because of the integer.

[dagar]

Realistically will a user need/want to configure the equivalent of PWM_{MIN, MAX, DISARMED} when using dshot?

[mhkabir]

@dagar Nope.

Edit: Maybe only for setting the threshold for when motors start spinning, etc. (varies)

[dagar]

At the moment I'm thinking a simple dshot "driver" that you optionally run in place of px4fmu is going to make the most sense. That should also work well as a place to handle telemetry.

[Igor-Misic]

How were you testing this?

If you mean by the whole driver, I added some extra code inside boards/px4/fmu-v4/init.c at the end of int board_app_initialize(uintptr_t arg) function. At that point, I'm sure Nuttx initialized all NVIC and DMA2 so you can start using stream5 and code for nuttx will handle interrupts.

This is the code which I use for testing drivers

At the beginning of the file:
#include <drivers/drv_pwm_output.h>

At the end of the board_app_initialize function

    //Start debugging for dshot
    uint32_t channel_mask = 0x3F; //b0111111 -> TIM2: b0110000+ TIM1: TIM2: b0001111
    up_dshot_init(channel_mask, DSHOT1200);
    up_dshot_arm(true);
    uint16_t throttle = 100;
    up_dshot_motor_data_set(0,throttle, true);
    up_dshot_motor_data_set(1,throttle, true);
    up_dshot_motor_data_set(2,throttle, true);
    up_dshot_motor_data_set(3,throttle, true);
    up_dshot_motor_data_set(4,throttle, true);
    up_dshot_motor_data_set(5,throttle, true);

    for(uint32_t i = 0; i < 1000; i++) {
        up_dshot_trigger();
        usleep(1000);
    }

    up_dshot_arm(false);

    // This should not do anything since is disarmed.
    for(uint32_t i = 0; i < 1000; i++) {
        up_dshot_trigger();
        usleep(1000);
    }

[Igor-Misic]

Reopening pull request because accidentally removed branch.

[dagar]

We'll need to go through each board and get the DMA map under control. This was a bit of a pain for SPI. #12436

[hamishwillee]

Thanks @Igor-Misic . I have created a docs page to track here PX4/PX4-user_guide#560

That sounds like a good plan. Sounds like we don't need work on this until "end user support" is in. Please ping me when you think we're ready for that.

[jkflying]

For ESC support, I believe BLHELI_S also supports DShot up to 600kHz, but not bidirectional.

[dagar]

Continued in #12854.