Hi team!

Now during the IMU calibration you lean the Phantom 4 in
all directions as well as calibrate it upside down.

Could this mean that the surface you perform the IMU
calibration on no longer has to be exactly horizontal?

Regards
Georg

Do you wish to have a completely straight horizon?



RedHotPoker

Yes, of course. I was thinking that the new way of doing the calibration would make the need for a 100% horizontal surface superfluous. Maybe that's why they changed they way the calibration is done.

george_007 Posted at 2017-7-16 13:44
Yes, of course. I was thinking that the new way of doing the calibration would make the need for a 100% horizontal surface superfluous. Maybe that's why they changed they way the calibration is done.

Admittedly, I'm inexperienced with this new P4 drone, but it sounds exciting  ;-)


I only know the IMU calibration done on a 90* flat surface has always been the suggestion.


RedHotPoker

I'm sure it should remove some of the precision from the equation, but the more I think about it, the more there still has to be some form of true north. Assuming the surface is relatively level and craft stays pointed in the same direction through each of the steps, and doesn't move in any direction, it probably takes the mean average of each measurement and gives a more accurate result on uneven durfaces. But if the surface is way off, I don't see how it can auto correct just because there are numerous measurements.

Like a laser spirit level... It will correct itself up to a point.

I thought the same initially, then not, but then back again :-)
At some point during calibration it needs to know what 'level' is, otherwise how can it judge when it's not level (I know). So lets supposing you've had a bad night and you think that the desk is level, it is in fact a 10 degree slope. At two points during the calibration it would expect to be level, one when on its feet (taking a reading at +10 degrees), and two when sitting on the motors(taking a reading at -10 degrees). So to compensate for any error between the two readings it would 'split the difference' and end up with a dead level setting ? And the four side calibrations give it an indication of how far it can tilt in each direction.

Aardvark Posted at 2017-7-16 14:41
I thought the same initially, then not, but then back again :-)
At some point during calibration it needs to know what 'level' is, otherwise how can it judge when it's not level (I know). So lets supposing you've had a bad night and you think that the desk is level, it is in fact a 10 degree slope. At two points during the calibration it would expect to be level, one when on its feet (taking a reading at +10 degrees), and two when sitting on the motors(taking a reading at -10 degrees). So to compensate for any error between the two readings it would 'split the difference' and end up with a dead level setting ? And the four side calibrations give it an indication of how far it can tilt in each direction.

Just my $0.02 - I have to agree that the surface still needs to be level. It's logical that it relies on that to be able to assume a zero baseline to set any required IMU off-sets (as determined from the angles). Again, using logic, the 'dance' may simply be a way for the dual sensors to self-check the settings from multiple angles - which is what you just said Aardvark.

But did I use 'logical' twice then? (hehe)

1+1=1

You could very well be correct.

My thinking was that, as an example, if the surface is off by 10 degrees "sideways" ( let's call it "+10 degrees") the drone
would lean to one side at the first measurement but assume that it is level. In the last measurement you turn the drone
upside down so now the drone will notice a -20 degree change compared to standing up. It would then calculate the
average of the two (difference between assumed horizontal level and -20 degees off) and correctly set 0 degrees as level
(ergo adjusting by -10 degrees from first measurement).

Now, in theory that would work for a surface leaning "sideways" but there are no two measurements that handle the same
in forward/backward direction.

Keep it square... ;-)



RedHotPoker

Although the procedure has been updated, the first step is with the bird setting normally. This is still what the IMUs are using a a reference horizon...make the the surface is level in both axises. The additional steps are probably done to insure linearity to the first level reference.

fester Posted at 2017-7-17 10:02
Although the procedure has been updated, the first step is with the bird setting normally. This is still what the IMUs are using a a reference horizon...make the the surface is level in both axises. The additional steps are probably done to insure linearity to the first level reference.

"This is still what the IMUs are using a a reference horizon."

You state this as a fact. Do you know this to be the case?
(not trying to be an asshat, just want to understand what the statement is based on)

george_007 Posted at 2017-7-17 15:32
"This is still what the IMUs are using a a reference horizon."

You state this as a fact. Do you know this to be the case?

     Based on 25 plus years of engineering experience in various fields, I'm 99% certain that this is indeed the case. IMU's are used anywhere from our Phantoms to most modern aircraft and Multimegaton Nuke carry ing ICBMs. Each one of these needs a calibrated reference to a gravity referenced horizontal horizon. An aircrafts Autopilot maintains straight and level flight by keeping its wings level to that known calibrated Horizon. There is a thread over at Phantom Pilots on IMU Calibration that might be helpful as well.

fester Posted at 2017-7-19 13:08
Based on 25 plus years of engineering experience in various fields, I'm 99% certain that this is indeed the case. IMU's are used anywhere from our Phantoms to most modern aircraft and Multimegaton Nuke carry ing ICBMs. Each one of these needs a calibrated reference to a gravity referenced horizontal horizon. An aircrafts Autopilot maintains straight and level flight by keeping its wings level to that known calibrated Horizon. There is a thread over at Phantom Pilots on IMU Calibration that might be helpful as well.

Good answer. Respect! :-)

For the calibration to be able to work out what is level it would be necessary to turn the aircraft through 180 degrees at some point so that it could take an average, just tilting it to each side is not enough.

So I think it is still important to do the IMU calibration on a level surface, and then after the IMU calibration do a gimbal calibration.

Nigel_ Posted at 2017-7-22 12:42
For the calibration to be able to work out what is level it would be necessary to turn the aircraft through 180 degrees at some point so that it could take an average, just tilting it to each side is not enough.

So I think it is still important to do the IMU calibration on a level surface, and then after the IMU calibration do a gimbal calibration.

It is rotated 180. On its feet and sitting on motors. A simple test would be to calibrate on an obviously non level surface, say a ten degree slope, and see what the sensors report when sitting on a level surface.

Aardvark Posted at 2017-7-22 13:40
It is rotated 180. On its feet and sitting on motors. A simple test would be to calibrate on an obviously non level surface, say a ten degree slope, and see what the sensors report when sitting on a level surface.

That would depend on which way you rotate it, now I am wondering if all those people who are having problems rotated it the wrong way!

When it is upside down during the calibration, is the camera still facing the same direction or is it now rotated and facing the opposite direction?   To correct the horizon for an inaccurately mounted gyro sensor would require the camera facing the opposite direction...