800 Gas Helicopter Governor help

Ferrosan · December 11, 2020, 12:57pm

here another user with the futaba:

Might worth to try using a logic analyzer, that will provide info about voltage level and signal.

ChrisOlson · December 11, 2020, 3:26pm

Actually, as far as that goes, I used the GV-1 back in the day on my nitro heli and it didn’t work all that great, period. The only way I could get it to work at all was to use the RSC_SETPOINT feature and the rpm was all over the place. The old three-point throttle curve worked better than the GV-1.

Adi_Fly · December 20, 2020, 7:32am

I used the align Governor sensor but no luck. No rpm being shown… either the sensor is faulty or I have installed the magnets wrong.
Is there a correct way to install the magnets?
The distance between the sensor and magnet is less than 2mm.
The align had a different connector, all I did was cut that and solder a servo connector to its end to connect it to aux5.
Lost as to what’s happening!!!

Thanks and Regards,
Adarsh

Implicit · December 20, 2020, 9:06am

Take the sensor off, hook it up to either a oscilloscope or even just a multimeter and measure what whappens when you wave a magnet in front of the sensor.

jinchengde · December 20, 2020, 9:46am

I got the internalError 0x2000000 too with futaba hall sensor
but apm doesn’t disarm

Is that some error for GPIO?

github.com

ArduPilot/ardupilot/blob/master/libraries/AP_InternalError/AP_InternalError.h#L65


    main_loop_stuck             = (1U << 15),  // 0x08000  32768
    gcs_bad_missionprotocol_link= (1U << 16),  // 0x10000  65536
    bitmask_range               = (1U << 17),  // 0x20000  131072
    gcs_offset                  = (1U << 18),  // 0x40000  262144
    i2c_isr                     = (1U << 19),  // 0x80000  524288
    flow_of_control             = (1U << 20),  //0x100000  1048576 for generic we-should-never-get-here situations
    switch_full_sector_recursion= (1U << 21),  //0x200000  2097152
    bad_rotation                = (1U << 22),  //0x400000  4194304
    stack_overflow              = (1U << 23),  //0x800000  8388608
    imu_reset                   = (1U << 24),  //0x1000000 16777216
    gpio_isr                    = (1U << 25),  //0x2000000 33554432
    mem_guard                   = (1U << 26),  //0x4000000 67108864
    __LAST__                    = (1U << 27),  // used only for sanity check
};


// if you've changed __LAST__ to be 32, then you will want to
// rejig the way we do sanity checks as we don't want to move to a
// 64-bit type for error_t:
static_assert(sizeof(error_t) == 4, "error_t should be 32-bit type");


uint16_t last_error_line() const { return last_line; }

jinchengde · December 26, 2020, 6:14am

I fall back firmware to 3.6.12 and it work

bnsgeyer · December 26, 2020, 6:16pm

@jinchengde and @Adi_Fly I think we need to move this to the Copter 4.0 thread. Since this sensor works in 3.6 and doesn’t in 4.0, then there is something that changed in the code that is causing a problem. I will start a post in Copter 4.0 and reference this post thread. Thanks to @jinchengde and @Ferrosan for doing the investigative work.

tridge · December 26, 2020, 8:41pm

@jinchengde that GPIO_ISR internal error is triggered when we see more than 10k pulses in 100ms, so a rate of 100k pulses per second, which would be 360000 RPM. The reason we check for this is that interrupts take priority over all threads, so if we get too many interrupts per second then we get starved of CPU for flying the heli. I chose 10k per 100ms as at that level we are using just one or two percent of the CPU for interrupt handling, and it is a level beyond any reasonable RPM value. If we didn’t have this check and the number of pulses per second went over about 700k/second then you would fall out of the sky as there would be no CPU left to fly with.

There are a few possible reasons this is happening:

it could be a short lived transient from this sensor
it could be that the sensor is missing a pullup or pulldown to prevent the pin from “floating”. A floating pin can generate a lot of interrupts from electrical noise
it could be that we were getting this issue in 3.6 and the errors were being suppressed using the simple filter we have in the RPM code (which rejects pulses that are too fast).
maybe our software pullup/pulldown settings are different in 4.0 than 3.6? If you tell me what flight controller this is and what pin you are connected to then I can check

Ideally we’d get a trace (eg. oscilloscope trace) of this pin when the problem happens.

Adi_Fly · December 27, 2020, 4:55am

I finally managed to get the align sensor working for the rpm… now the rpm is being displayed in the gcs… next is to switch on the Governor mode.
Thanks guys… will keep you posted… but ya the futaba sensor is def a no no till it’s rectified…

Thanks & Regards,
Adarsh

jinchengde · December 27, 2020, 1:36pm

@tridge
thanks for you reply

I used cuav v5 nano with futaba gv-1 hall sensor in AUX3(pin 60)

rotems · January 8, 2021, 11:38pm

Hello Adarsh,

Can you please advise how did you manage to get reading from the Align RPM sensor?
I’m trying but with no success.

Thank you
Rotem

Adi_Fly · January 16, 2021, 2:41am

Sorry for the late reply.
I think I had a faulty sensor which I removed from my old helicopter. I ordered a new one and it’s working.
I installed a aerospire gov sensor on another Heli, and that’s working too…

Thanks again for all the help everyone.

Kind regards,
Adarsh

rotems · January 19, 2021, 12:17pm

Hello Adarsh,

Thank you for your replay.

I think mine is faulty as well, I’ve ordered a new one.

Rotem

bnsgeyer · January 25, 2021, 2:13pm

@ChrisOlson see post 28 in this thread from @tridge. He had thrown out some ideas on what could be causing the problem but needed some one with an oscilloscope that could help troubleshoot. If you could support this trouble shooting, I’d really appreciate it. I don’t have an oscilloscope and I am not seeing this issue on my controllers.

@tridge what exactly do you need to see on the oscilloscope. What would be the setup in order to do this test? Here is the post on what Chris found.

@tridge this is serious bug that will impact tradheli users that want to use the existing governor in 4.0 and the improved governor that Chris Olson has been work on for 4.1.

Ferrosan · January 27, 2021, 4:27pm

Hi @tridge and @bnsgeyer,
as discussed with Bill I report here what we have observed on our setup:
-Flight control: CUAV Nora, powered by CAN HV power module;
-Servo Rail: connected to external BEC set at 8.4V;
-RPM sensor connected to servo rail and RPM_PIN set to 63: https://www.sm-modellbau.de/magnetischer-Drehzahlsensor-fuer-UniTest-2-UniLog-1-2

RPM sensor is working (led changes status upon each magnet passage) but reading on the flight control is not.
If rpm sensor is powered instead by external 5V power supply with ground and signal connected to flight control, rpm reading in the flight control works.
Sensor itself outputs 3.3V level signal.
Additionally, we noticed on this flight control that the servo rail voltage is not logged:

ChrisOlson · January 27, 2021, 5:23pm

I assume you mean pin 53?

Since all modern digital tach’s are pulse train, what is the thoughts on implementing a pulse train frequency counter in the Heli RSC specifically for the governor and rpm functions in heli, and not use the RPM library with an interrupt? The RPM library was originally written only for logging and was called from the main code at 10Hz. I increased that to 40Hz for the governor, but it is still not ideal and really needs to run at 400Hz (or main loop rate) with the rest of the governor code.

Ferrosan · January 27, 2021, 7:46pm

no Chris, it’s connected to pin 63, I followed this:

github.com

ArduPilot/ardupilot/blob/master/libraries/AP_HAL_ChibiOS/hwdef/CUAV-Nora/hwdef.dat

# hw definition file for processing by chibios_hwdef.py for CUAV-Nora

# MCU class and specific type
MCU STM32H7xx STM32H743xx

# crystal frequency
OSCILLATOR_HZ 16000000

# board ID for firmware load
APJ_BOARD_ID 1009

FLASH_SIZE_KB 2048

# with 2M flash we can afford to optimize for speed
env OPTIMIZE -O2

# ChibiOS system timer
STM32_ST_USE_TIMER 2

# bootloader takes first sector

This file has been truncated. show original

Also, we have a fuel flow sensor connected to pin 62. They both work fine if powered by a 5V source, otherwise if powered by the servo rail (HV) straight the autopilot doesn’t read anything.
Both sensors are rated for HV power supply, as side note.

ChrisOlson · January 27, 2021, 10:30pm

I will try stepping my sensor drive voltage down to 5V and see if that works.

ChrisOlson · January 28, 2021, 1:15am

That did not work to reduce the drive voltage to 5V on the sensor. Further, I do not get an rpm signal from the RPM library with the Pixhawk 3 Pro on ChibiOS with Copter 3.6 either. It only works with NuttX.

Ferrosan · January 28, 2021, 8:48am

Chris, sorry to hear that. On our Drotek Pix3Pro it works fine with 4.0.5 (using the same sensor as above linked). Furthermore, on the pix3pro it is powered directly by servo rail which runs on HV.
In our scenario, only the CUAV Nora shows no readings if sensors are powered directly from servo rail (but sensors are working, we checked with a logic level analyzer).