Please select Into the mobile phone version | Continue to access the computer ver.
ATTI - Flyaways - NYUT - MEMS mysteries ....
461 1 2018-10-27
Uploading and Loding Picture ...(0/1)
First Officer
Flight distance : 4698533 ft
United States

still relevant thread.. that seems DJI has yet to fully address and continues to show its ugly head..

there are several misconceptions here and perhaps a lot of ignorance .. perhaps on both sides as to the gravity of the issue..

keeping a quad in the air... is a monkey standing on the head of a pin balancing on a ball bearing.. IE its not simple..

people think that the flight control boards and MEMS packaged 'gyro' boards are just plug and go... that they produce acceleration compensated position, heading, and attitude reverence and that is inaccurate ...

nothing there could be further from reality.

one of my favorite quotes from 'The Right Stuff' ..

you know what makes these birds go up...?

hell the science alone would take so long to explain....  

and that perfectly summerizes MEMS attitude and positional systems and AHRS/INS

The integrated MEMS gyros used in just about anything are raw sensor boards.. IE they don't have the processing algorithms built in to take the raw measurements and filter and process them for angular and accelerated motion to calculate AHARS...

to try and dumb it down into layman's terms..

when a MEMS package is at rest.. a 5 year old could process the data coming from the on package gyro, accelerometer and compass ... its easy math

now add into the mix movement and acceleration in all axis and now you need a combination of kalman and complimentary filtering (advanced mathematical calculations) in order to keep a horizon level.. let alone navigate ...

gravity aint gravity when something is moving or spinning... its centripital force and for heading its called motion induced precesion

ie.. without these mathematical equations.. if you held a package level with the horizon and started spinning around in a circle, say in a barbers chair,  the horizon would tilt due the the centripetal forces.. i.e. .. the artificial gravity..and the compass would not be accurate,  the filters have take the inputs from the accelerometers and compass .. sense its spinning and take those rates to null the gyro to show the horizon as level.. as it still is.. not where the centripetal (gravity) is pointing and correct the compass heading for acclerative precession if present.. etc.

This 'MATH' is generally applied 'off package' ... i.e. the MEMS board provides raw data and its up to the engineer to code 'flight software' to make the magic happen. in the FC package

in an attempt to find a paper that further highlights this...


When DJI release a new firmware, we don't know what 'tweaking' they are doing to this mathematical soup.

keeping further in mind that there is no redundancy in the spark.. any gross change or incoherent (incorrect) data coming from a sensor can cause catastrophic issues if not filtered properly (i.e. mathematical BS flag ) and in some cases CANT fully be correct for. (oh **** now what response.. ie ATTI)

without seeing the raw sensor data, and the algorithms that process it.. its IMPOSSIBLE to say for certain what is occurring. ..or how DJI are processing the data stream and handling 'exceptions'

what is definitive I believe here is this..

the compass is a very (the) weak link in MEMS AHARS solutions.. and there are MANY MANY videos out there showing DJI or any drone product really, where the compass heading is moving all over the place, when the drone is not turning ..

GPS can't produce or even estimate aircraft heading since quads can hover and fly in any direction unlike an airplane.. GPS can only calculate heading made good (course) but has no idea where the aircraft is pointing

there are only 2 sensors on the aircraft that relate to aircraft heading.. the compass and the accelerometer in the yaw axis.. if the rotation of the aircraft compass and the yaw angles felt by the accelerometer don't match up.. the software has to HANDLE THE EXCEPTION ..

similarly .. if GPS data begins to deviate from MEMS calculations/estimations of motion, an algorithm has to try and deal with that.. am I moving or is the GPS position drifing becase of satilite issues

depending on the calculations being used in the complimentary filters to handle the euler angle and Quaternions .. if not dealt with properly can effectively hit mathematical gimbal lock (deadman) where the math falls apart...

add to this filters and tables to deal with vibration, drift, precession, thermal instability of MEMS, MEMS package degradation, GPS positional integration, etc..

its not simple...

I 'believe' that the failure mode that we witness where the events seem to appear in a distinct order where  the drop to atti  starts with yaw errors, then compass warnings, then the GPS working error message.. followed by a period of apparent 'unusual behavior and sometime lack of input control' is in fact the system RESETTING itself  it reached a level of mathematical confusion to the point where the engineers coded a reset.. i.e. it gives up and restarts.

if the GPS is degraded to the point where navigation is not capable... or the GPS position moves drasically.. . it will stay in ATTI or perhaps overcorrect to that GPS data in ATTI .. until it regains enough sats... before returning to P-Mode

now factor in that some have made the observation that these issues are more prevalent in 'REFURB' products and perhaps with latter produced sparks..  a different manufacturer.. different tolerances, installed improperly? more/less vibration due to tightness of mounting hardware... etc..

there are literally hundreds or possible reasons that could cause additional issues

also keep in mind that its 'easy' for the engineers to hide these warnings from us.. just adjust the thresholds that trigger the warning.. it doesn't mean they are not happening ... etc

there is a lot of secret sauce that goes on here and it doesn't take much to screw up the recipe...

this is the principle reason that real aircraft still RARELY rely on MEMS based AHARS for anything except battery powered backup systems like a standby attitude indicator.. primary navigation position INU systems still use mechanical INS gyros like ring laser.. or very very expensive and highly redundant hybrid MEMS units.. every military plane I flew had a mechanical gyro for its primary position and  were so sensitive that during aircraft spin-up would take the rotation of the earth into account and its initial position had to be very accurate to get a good alignment.

I noticed with my 2 sparks that were calibrated identically, at the same time, on the same leveled engineers marble plate, at the same temperature etc... had 2 very different flight characteristics..

one would sense lateral motion caused by wind blowing off course and correctly tilting into the crosswind while the other didn't

one would point as much as 45 degrees off during RTH while the other was maybe 10
one would always autoland facing way off its takeoff heading

one spark would rotate the video (in camera yaw correction) while at rest while the other rarely did..

ie.. the IMUs where behaving markedly different ... and it seemed it was compass/yaw stability issues..

the spark that had greater errors had more flight time on it.. neither had experience a crash or hard landing... both sparks were less than a month old

any who.. food for thought...

Use props


I also think something is wrong with the code. Either sync of the gps vs compass data or different sampling rates or something third.... we will never know, tho... Low quality, cheap compass chips are prone to mechanical stuck as well as magnetic inclination/declination issues. Degaussing helps sometimes. That's why serious drones have redundant sensors.
Use props
You need to log in before you can reply Login | Register now

Credit Rules