GPS not working

Eric Katzfey

@andremoura Yes, the errors related to px4io detection are expected since you do not have a px4io unit attached to the drone. And yes, you have to interact with things on the DSP (where the GPS driver runs) with qshell as you already figured out. Do you have any other Sentinel drones that you can compare against? Or other types of drones with GPS to compare against? When you are able to get 9 to 12 satellites what does qshell gps status show? It does appear to be wired up correctly as sometimes it is obviously communicating with the GPS unit.

Eric Katzfey

@andremoura It looks like there are at least 2 different issues here. One is that when you do see it working it has poor performance and doesn't get a lock. That is why I wanted to compare against other drones or other GPS units. If other units are working fine and this one isn't then maybe it's a faulty unit. If they all have poor performance then maybe it's something about the test location.

Eric Katzfey

@andremoura The other issue appears to be that you don't always get the driver to communicate with the GPS unit on startup. The start line in /etc/modalai/voxl-px4.config is qshell gps start -d 7 -b 115200. When you did qshell gps stop and then qshell gps start was it started with qshell gps start -d 7 -b 115200 or just qshell gps start?

Andre Moura

@Eric-Katzfey We don't have any other Sentinel drones. We have other drones with GPS, that detect at least the same or higher number of satellites but they all are able to GPS lock.

When we had that number of satellites, the qshell gps status seemed ok with number of satellites, latitude and longitude, but we aren't able to replicate that test right now because of bad weather conditions.

Andre Moura

@Eric-Katzfey I will test the Sentinel drone on a new better location next week, if weather gets better, and I will get back to you.

Andre Moura

@Eric-Katzfey GPS mostly hasn't beeen working on startup again.

We started it with qshell gps start, but we also tested qshell gps start -d 7 -b 115200. Both commands only worked a few times, we are not able to consistently start the GPS, having to stop and start again the service a few times before it works.

Eric Katzfey

@andremoura Is the LED on top of the unit (multicolor LED) coming on?

Andre Moura

@Eric-Katzfey said in GPS not working:

@andremoura Is the LED on top of the unit (multicolor LED) coming on?

Yes, every time voxl-px4 is running.

Eric Katzfey

@andremoura Okay, good. Holybro changed some of the components in their unit at one point and if that LED wasn't coming on it could indicate that it was one of the older units. But since it is on that indicates to me that it is the correct version of the Holybro unit.

Andre Moura

@Eric-Katzfey Ok, new find! It seems that when voxl-px4 runs, the GPS service isn't able to start immediately, but if we wait a for a bit and start it manually, it works.

Seems like something isn't ready to start immediately after voxl-px4.

Eric Katzfey

@andremoura Interesting! Okay, how do you accomplish that? Do you mean you are stopping the service and then starting voxl-px4 later manually or you are leaving voxl-px4 service enabled but commenting out the gps start line in /etc/modalai/voxl-px4.config?

Andre Moura

@Eric-Katzfey We are stopping the GPS service manually (qshell gps stop) and then starting it later manually (qshell gps start), not the voxl-px4.

Do you want us to do the same test with commenting the gps start line in /etc/modalai/voxl-px4.config?

Eric Katzfey

@andremoura Actually, can you try removing the -d 7 -b 115200 from the gps start line in /etc/modalai/voxl-px4.config and see if that changes anything? Thanks for trying all this stuff out!

Andre Moura

@Eric-Katzfey Hello, today I tested removing -d 7 -b 115200 from /etc/modalai/voxl-px4.config and it seems like the GPS service is working correctly on startup. I'm still going to test if I can get more fixed satellites and GPS lock on a new location with more open space and then get back to this post with more information.

Andre Moura

@Eric-Katzfey On the new location, we were able to get GPS lock after fixing 12 satellites, which allowed for take-off. During flight we reached 19 satellites at one point. Getting GPS lock with this sensor is a bit challenging, our previous location didn't have that many obstructions, but everything seems to be working.

Eric Katzfey

@Andre-Moura Thanks for all of the testing! We really appreciate it, this is all great feedback! It seems that Holybro may have changed the default configuration of the units and that could be why it wasn't coming up properly.

daniel

@Andre-Moura We have same issue here but we try to remove -d 7 -b 115200 from /etc/modalai/voxl-px4.config is still not working. Did you modify thr code like this?

Removing the 115200 baud rate is still not see GPS log.

Andre Moura

@daniel I think that's all I did, but here is my /etc/modalai/voxl-px4.config, in case you want to look for any differences:

#!/bin/sh
# PX4 commands need the 'px4-' prefix in bash.
# (px4-alias.sh is expected to be in the PATH)
. px4-alias.sh

# Figure out what platform we are running on.

# Eventually there will be a utility called voxl-platform that will
# return the platform tag or an error code. This utility is not yet
# ubiquitous so it may be that it isn't available. Trying to call a
# non existent program will generate an error code by the OS. If the
# program exists and doesn't return an error code then use the results.
PLATFORM=`/usr/bin/voxl-platform 2> /dev/null`
RETURNCODE=$?
if [ $RETURNCODE -ne 0 ]; then
    # If we couldn't get the platform from the voxl-platform utility then check
    # /etc/version to see if there is an M0052 substring in the version string. If so,
    # then we assume that we are on M0052. Otherwise assume M0054.
    VERSIONSTRING=$(</etc/version)
    M0052SUBSTRING="M0052"
    if [[ "$VERSIONSTRING" == *"$M0052SUBSTRING"* ]]; then
        PLATFORM="M0052"
    else
        PLATFORM="M0054"
    fi
fi

# We can only run on M0052 or M0054 so exit with error if that is not the case
if [ $PLATFORM = "M0052" ]; then
    /bin/echo "Running on M0052"
    if [ "$RC" == "CRSF" ]; then
        /bin/echo "Error, crossfire not supported on M0052!"
        exit 0
    fi
elif [ $PLATFORM = "M0054" ]; then
    /bin/echo "Running on M0054"
else
    /bin/echo "Error, cannot determine platform!"
    exit 0
fi

uorb start
muorb start

# In order to just exit after starting the uorb / muorb modules define
# the environment variable MINIMAL_PX4. (e.g. export MINIMAL_PX4=1)
# This is useful for testing / debug where you may want to start drivers
# and modules manually from the px4 command shell
if [ ! -z $MINIMAL_PX4 ]; then
    /bin/echo "Running minimal script"
    exit 0
fi

# Sleep a little here. A lot happens when the uorb and muorb start
# and we need to make sure that it all completes successfully to avoid
# any possible race conditions.
/bin/sleep 1

# IMU (accelerometer / gyroscope)
# Start this first because it gets the high rate interrupts coming in to the
# DSP. Without this the DSP will oversleep and miss critical timeouts.
# TODO: Why is that the case?
if [ "$IMU" == "ROTATE_IMU_YAW_180" ]; then
    /bin/echo "Starting IMU driver with yaw 180 rotation"
    qshell icm42688p start -s -R 4
else
    qshell icm42688p start -s
fi

/bin/sleep 1

# Load in all of the parameters that have been saved in the file
param load

# Start logging and use timestamps for log files when possible.
# Add the "-e" option to start logging immediately. Default is
# to log only when armed
logger start -t

# We do not change the value of SYS_AUTOCONFIG but if it does not
# show up as used then it is not reported to QGC and we get a
# missing parameter error.
param touch SYS_AUTOCONFIG

# Start all of the device drivers on DSP

# Magnetometer
if [ "$GPS" == "MATEK" ]; then
    # Use this line for the magnetometer in the Matek Systems M8Q-5883 module
    /bin/echo "Starting Mateksys M8Q-5883 magnetometer"
    qshell qmc5883l start -R 10 -X -b 1
elif [ "$GPS" == "HERE3" ]; then
    # Use this line for the magnetometer in the Here3 GPS module
    /bin/echo "Starting Here3 ak09916 magnetometer"
    qshell ak09916 start -R 2 -X
else
    # Use this line for the magnetometer in the Holybro GPS module
    /bin/echo "Starting Holybro magnetometer"
    qshell ist8310 start -R 10 -X -b 1
fi

# LED driver for the Pixhawk 4 GPS module
# Older units have i2c address 0x39 (57) and newer ones 0x38 (56)
# M0054 only supports the newer one. M0052 can support either.
if [ $PLATFORM = "M0052" ]; then
    qshell rgbled_ncp5623c start -X -b 1 -f 400 -a 57
fi
if [ "$GPS" == "HOLYBRO" ]; then
    /bin/echo "Starting Holybro LED driver"
    qshell rgbled_ncp5623c start -X -b 1 -f 400 -a 56
elif [ "$GPS" == "HERE3" ]; then
    # Use this line for the Here3 GPS module LED controller
    /bin/echo "Starting Here3 LED"
    qshell rgbled start -X -f 400
fi

# Barometer
qshell icp10100 start -I -b 5

# ESC driver
# We start this even if there is a PX4IO module. If there is
# a PX4IO (M0065) module it will be plugged into the RC port
# and so this driver will output commands on an empty port.
qshell modalai_esc start
qshell mixer load /dev/uart_esc quad_x.main.mix

# APM power monitor
qshell voxlpm start -X -b 2

# On M0052 the GPS and RC drivers run on the apps processor
if [ $PLATFORM = "M0052" ]; then
    # Pixhawk 4 GPS module
    gps start -d /dev/ttyHS2

    # RC input may be coming via either an external M0065 (aka px4io) module
    # or a directly attached Spektrum receiver. First see if the M0065 is detected.
    px4io detect
    PX4IO_STATUS=$?
    /bin/echo "PX4IO detect returned value $PX4IO_STATUS"
    if [ $PX4IO_STATUS -eq 0 ]; then
        /bin/echo "Found M0065, starting px4io"
        # On M0052 px4io only supports RC input, not PWM output
        px4io start
    else
        /bin/echo "M0065 not detected, starting Spektrum RC driver"
        spektrum_rc start -d /dev/ttyHS1
    fi
# On M0054 the GPS and RC drivers run on SLPI DSP
else
    if [ "$GPS" == "MATEK" ]; then
        # Use this gps start line instead for Matek Systems M8Q-5883 module
        /bin/echo "Starting Mateksys M8Q-5883 GPS"
        qshell gps start
    elif [ "$GPS" == "HERE3" ]; then
        # Use this line for the Here3 GPS module
        /bin/echo "Starting Here3 GPS"
        qshell gps start
    else
        # Only the newer Holybro unit is supported on M0054
        /bin/echo "Starting Holybro GPS"
        qshell gps start
    fi

    # If RC is TBS Crossfire then start its driver.
    # Otherwise, it could be either a directly connected Spektrum receiver
    # or an external M0065 (aka px4io) module.
    if [ "$RC" == "CRSF_MAV" ]; then
        /bin/echo "Starting TBS crossfire RC - MAV Mode"
        qshell tbs_crossfire start -m -p 7 -b 115200
    elif [ "$RC" == "CRSF_RAW" ]; then
        /bin/echo "Starting TBS crossfire RC - CRSF Mode"
        qshell tbs_crossfire start -c -p 7 -b 250000
    else
        qshell px4io detect
        PX4IO_STATUS=$?
        if [ $PX4IO_STATUS -eq 0 ]; then
            /bin/echo "Found M0065, starting px4io"
            # On M0054 we also support PWM output from M0065 (aka PX4IO)
            # If there are no pwm escs then that connector will be empty.
            qshell px4io start
            qshell px4io recovery
            qshell mixer load /dev/px4io quad_x_io.main.mix

            sleep 1
            qshell pwm rate -c 1234 -r 400
			qshell pwm disarmed -c 1234 -p 900
			qshell pwm min -c 1234 -p 1060
			qshell pwm max -c 1234 -p 2000
            qshell pwm failsafe -c 1234 -p -1 -d /dev/px4io
        else
            /bin/echo "M0065 not detected, starting Spektrum RC driver"
            qshell spektrum_rc start -d 8
        fi
    fi
fi

/bin/sleep 1

# Start all of the processing modules on DSP

qshell sensors start
qshell ekf2 start
qshell mc_pos_control start
qshell mc_att_control start
qshell mc_rate_control start
qshell mc_hover_thrust_estimator start
qshell land_detector start multicopter

/bin/sleep 1

# Start all of the processing modules on the applications processor

rc_update start

dataman start

navigator start

commander start
commander mode manual

# This is needed for altitude and position hold modes
flight_mode_manager start

# Start our service to update system time from GPS
modalai_gps_timer start

# As of 1.4.11 mavlink configuration has been changed!!! All communications are
# now done via the mavlink server and voxl-vision-px4!!!
param set MAV_BROADCAST 0

if [ "$RC" == "CRSF_MAV" ]; then
    # Slow everything way down if we are using TBS Crossfire
    mavlink start -a -x -u 14556 -o 14557 -r 10000 -n lo -m minimal

    /bin/sleep 1
else
    mavlink start -x -u 14556 -o 14557 -r 100000 -n lo -m onboard

    /bin/sleep 1

    # slow down some of the fastest streams in onboard mode
    mavlink stream -u 14556 -s HIGHRES_IMU -r 10
    mavlink stream -u 14556 -s ATTITUDE -r 10
    mavlink stream -u 14556 -s ATTITUDE_QUATERNION -r 10
    # speed up rc_channels
    mavlink stream -u 14556 -s RC_CHANNELS -r 50

    /bin/sleep 1
fi

if [ $PLATFORM = "M0052" ]; then
    # Start up the IMU server to support VIO. This is done differently on M0054
    imu_server start
fi

mavlink boot_complete

# For now we use a mavlink bridge for TBS Crossfire support
if [ "$RC" == "CRSF_MAV" ]; then
    mavlink_bridge start
fi

# Optional MSP OSD driver for DJI goggles
# This is only supported on M0054 (with M0125 accessory board)
if [ $PLATFORM = "M0054" ]; then
    if [ "$OSD" == "OSD" ]; then
        /bin/echo "Starting OSD driver"
        msp_osd start -d /dev/ttyHS1
    fi
fi

If your config is like this and it still doesn't work, maybe @Eric-Katzfey can help you better.

daniel

@Andre-Moura Thank You! We fixed our GPS setting problem.

JustinDev23

@daniel what ended up being your fix? I was seeing the same issues yesterday with our Sentinel trying to do GPS only outdoor flights. It wouldn't ARM due to the GPS Speed Accuracy. Flashing blue for a minute or two and then turning green after what i assumed was VIO getting healthy.

@Eric-Katzfey I don't want to use the Sentinel VIO so was hoping there was a GPS Only No VIO params file. It seems like the one posted on the git is still looking at VIO.