New Compass Calibration Recommendations (from the v2.0 manual)

[tcope]

Calibrate as often as you want, it doesn't hurt, it's not like there are little cogs inside that can wear out. Moreover, the more you do it, the smoother you do it, you learn little tricks how to better hold it and rotate smoother.

It could hurt as each time it is done you are opening the door for some problem with that setting. If you are flying and it's set correctly, it is better not to take a chance that something could go wrong with a resetting.

 

[sar104]

I think during calibration it goes into "I'm accepting reality as is" mode, i.e. it's not comparing data to previous calibration to say "move away from ferromagnetic objects". So, if someone calibrates in a room next to powerful speaker, or in the car, it can be screwed up.

If the external distortion is too strong it will swamp the calibration, or if it is local enough that it varies at the magnetometers as the aircraft is rotated during calibration then yes - it could incorrectly identify those varying components as being on the aircraft. Likely the calibration will simply fail in the former case, but could, conceivably, lead to an accepted bad calibration in the latter.

 

[Xtreme Drone Pilot]

It could hurt as each time it is done you are opening the door for some problem with that setting.

What problem? If you're doing calibration by the book, why would you ever have a problem.

Here's the deal. Compass has certain thresholds within which it's considered to be operating normally. Let's say for the sake of example, that some value X must be close to 0, but up to 10 is acceptable. By calibrating, you make it close to 0. If you don't calibrate for long time or in vastly different locations, the value X can be 9.9... still "good", but then in flight in flies over some small magnetic anomaly and X goes over 10, and BOOM! Mavic goes berserk.

 

[tcope]

What problem? If you're doing calibration by the book, why would you ever have a problem.

"If". Take a few thousand compass calibrations and someone is bound to make a mistake or the software or hardware might screw up. When they are working correctly and not changed there is a much greater possibility that something won't go wrong. Once you enter numerous variables into the situation, the chance of an error are greatly increased.

If you don't calibrate for long time or in vastly different locations, the value X can be 9.9... still "good", but then in flight in flies over some small magnetic anomaly and X goes over 10, and BOOM! Mavic goes berserk.

It was already established that doing the calibration when changing locations and when not flown in sometime was correct. You then added that it could not hurt to do every time. That is different then you mentioned here. I never disagreed with calibration when making a substantial change in location or over an extended period of time, as recommended. Only that calibrating every time you fly was not a good practice.

 

[msinger]

So, if someone calibrates in a room next to powerful speaker, or in the car, it can be screwed up.

Unless the powerful speaker or car is attached to the Mavic while it's rotating, the calibration attempt should fail.

 

[sar104]

Unless the powerful speaker or car is attached to the Mavic while it's rotating, the calibration attempt should fail.

Obviously I don't know the calibration algorithm being used - just the principle - but in that context we can probably assume that the FC is expecting a fixed magnetic field in the earth frame of reference (which should be just the earth's magnetic field) and a fixed magnetic field in the IMU (aircraft) frame of reference that will be different but constant at each of the magnetometer locations.

If that is the case (which is the ideal condition for a calibration) then taking the measured magnetic field vector as a function of orientation at each magnetometer, the FC should be able to decompose that to a linear combination of the two fields in the two frames of reference and store the field due to the aircraft for subtraction from the measured field in flight. That would be a good calibration.

If there are other magnetic field components that are neither fixed in the aircraft or the earth frame of reference, such as something magnetized being worn by the person calibrating that moves during the process, or a locally-generated field (such as a nearby vehicle or speaker magnet) that varies enough with position that the vector at the magnetometers changes as the aircraft is rotated, then there should be no unique linear combination of two fields in different FORs that account for the measured data, so that should be detectable as a bad calibration.

In other words I think that your example, above, is likely a correct interpretation of the situation, and that is also consistent with @BudWalker's hypothesis that, in general, a calibration in a locally non-uniform distorted field will fail. The corollary of that is that a calibration in a locally uniform distorted field should succeed, but that would be fine because the FC is not interested in characterizing the field in the earth FOR - only the one in the aircraft FOR.

 

[BudWalker]

Obviously I don't know the calibration algorithm being used - just the principle - but in that context we can probably assume that the FC is expecting a fixed magnetic field in the earth frame of reference (which should be just the earth's magnetic field) and a fixed magnetic field in the IMU (aircraft) frame of reference that will be different but constant at each of the magnetometer locations.

If that is the case (which is the ideal condition for a calibration) then taking the measured magnetic field vector as a function of orientation at each magnetometer, the FC should be able to decompose that to a linear combination of the two fields in the two frames of reference and store the field due to the aircraft for subtraction from the measured field in flight. That would be a good calibration.

If there are other magnetic field components that are neither fixed in the aircraft or the earth frame of reference, such as something magnetized being worn by the person calibrating that moves during the process, or a locally-generated field (such as a nearby vehicle or speaker magnet) that varies enough with position that the vector at the magnetometers changes as the aircraft is rotated, then there should be no unique linear combination of two fields in different FORs that account for the measured data, so that should be detectable as a bad calibration.

In other words I think that your example, above, is likely a correct interpretation of the situation, and that is also consistent with @BudWalker's hypothesis that, in general, a calibration in a locally non-uniform distorted field will fail. The corollary of that is that a calibration in a locally uniform distorted field should succeed, but that would be fine because the FC is not interested in characterizing the field in the earth FOR - only the one in the aircraft FOR.

Just a small clarification. The term distorted may be ambiguous. A successful calibration requires that the external field be uniform and straight within the confines of the compass dance. The external field can be oriented different from would be considered normal but it has to have the same orientation for all parts of the dance.

The calibration requires an external field in order to detect the magnetic effects of objects on the AC. There isn't anything special about the geomagnetic field, it's just convenient.

 

[BudWalker]

IMHO the risk of obtaining a bad calibration is way, way overstated. It may be possible but, AFAIK, there has not been a documented case where this has happened.

 

[Robert Mitchell]

IMHO the risk of obtaining a bad calibration is way, way overstated. It may be possible but, AFAIK, there has not been a documented case where this has happened.

^^^This. Is what I wanted to hear.

So... the precautions presented by DJI for the calibration process, e.g. no metal keys etc, aren't to prevent BAD calibrations, they are to prevent FAILED ones, correct?

 

[sar104]

Just a small clarification. The term distorted may be ambiguous. A successful calibration requires that the external field be uniform and straight within the confines of the compass dance. The external field can be oriented different from would be considered normal but it has to have the same orientation for all parts of the dance.

The calibration requires an external field in order to detect the magnetic effects of objects on the AC. There isn't anything special about the geomagnetic field, it's just convenient.

Right - I should have clarified "distorted" by which I meant, as you realized, not the unmodified field due to the earth, even though it could be uniform - or straight, as you mentioned.

 

[BudWalker]

^^^This. Is what I wanted to hear.

So... the precautions presented by DJI for the calibration process, e.g. no metal keys etc, aren't to prevent BAD calibrations, they are to prevent FAILED ones, correct?

Well, I don't presume to speak for DJI. They have been know to put stuff in their manuals that many people don't believe. But, yes, the intent of taking the precautions you mentioned is to prevent a failed calibration which results in the "interference move to another location" message.

 

[Xtreme Drone Pilot]

Agree with sar104 and BudWalker: magnetic field doesn't have "flavor" or "smell", once constant magnetic field from some source (such as speaker) is mixed with Earth's MF, Mavic thinks it's the true Earth's MF and passes calibration with flying colors. When you take it to fly, it says goodbye and flies to Zimbabwe.

While there's no need to calibrate before each flight, if you notice *any* deviation from straight flight, calibrate. It happened twice to me, about 2 months ago in the Sierras (Mavic drifiting at about 1m/s to the right during full forward flight) and couple of days ago (when I accidentally put a speaker on top of Mavic's case), it had slight drift to the left. Both cases I performed calibration by the book (removed both smartwatch and dumbwatch, phone from pocket, clear area, far away from any metal) and both times the problems were solved 100%.

 

[wytnyt]

Agree with sar104 and BudWalker: magnetic field doesn't have "flavor" or "smell", once constant magnetic field from some source (such as speaker) is mixed with Earth's MF, Mavic thinks it's the true Earth's MF and passes calibration with flying colors. When you take it to fly, it says goodbye and flies to Zimbabwe.

While there's no need to calibrate before each flight, if you notice *any* deviation from straight flight, calibrate. It happened twice to me, about 2 months ago in the Sierras (Mavic drifiting at about 1m/s to the right during full forward flight) and couple of days ago (when I accidentally put a speaker on top of Mavic's case), it had slight drift to the left. Both cases I performed calibration by the book (removed both smartwatch and dumbwatch, phone from pocket, clear area, far away from any metal) and both times the problems were solved 100%.

how far from cars is safe?

 

[sar104]

Agree with sar104 and BudWalker: magnetic field doesn't have "flavor" or "smell", once constant magnetic field from some source (such as speaker) is mixed with Earth's MF, Mavic thinks it's the true Earth's MF and passes calibration with flying colors. When you take it to fly, it says goodbye and flies to Zimbabwe.

While there's no need to calibrate before each flight, if you notice *any* deviation from straight flight, calibrate. It happened twice to me, about 2 months ago in the Sierras (Mavic drifiting at about 1m/s to the right during full forward flight) and couple of days ago (when I accidentally put a speaker on top of Mavic's case), it had slight drift to the left. Both cases I performed calibration by the book (removed both smartwatch and dumbwatch, phone from pocket, clear area, far away from any metal) and both times the problems were solved 100%.

Just to clarify - if a constant external magnetic field is mixed with the earth's field then I'm actually arguing that it doesn't matter, because the calibration is not trying to measure the earth's magnetic field - it's trying to measure aircraft's magnetic field, and it can do that in any constant, locally spatially uniform, external field. In theory, at least, the calibration should pass and should be good.

Now it's also possible that the firmware has some acceptance criteria for the external field - i.e. if it does not plausibly look like the earth's field at that location (if the field vector were vertical, for example, or much too weak or strong) then it might throw a failed calibration flag.

 

[Xtreme Drone Pilot]

how far from cars is safe?

I was about 50ft away in both calibrations.

 

[Xtreme Drone Pilot]

Just to clarify - if a constant external magnetic field is mixed with the earth's field then I'm actually arguing that it doesn't matter, because the calibration is not trying to measure the earth's magnetic field - it's trying to measure aircraft's magnetic field, and it can do that in any constant, locally spatially uniform, external field. In theory, at least, the calibration should pass and should be good.

Now it's also possible that the firmware has some acceptance criteria for the external field - i.e. if it does not plausibly look like the earth's field at that location (if the field vector were vertical, for example, or much too weak or strong) then it might throw a failed calibration flag.

Both points are solid, agree.

 

[Robert Mitchell]

... But, yes, taking the precautions you mentioned will prevent a failed calibration, not a successful but flawed calibration.

Hmmm, sounds like you're saying exactly the opposite.
Sounds like you're saying that taking the recommended DJI precautions, can still result in "successful but flawed calibrations"

 

[BudWalker]

Hmmm, sounds like you're saying exactly the opposite.
Sounds like you're saying that taking the recommended DJI precautions, can still result in "successful but flawed calibrations"

Sorry 'bout that. It was a little confusing. I changed the wording in that post. Have a look.

 

[Robert Mitchell]

Sorry 'bout that. It was a little confusing. I changed the wording in that post. Have a look.

Thanks for the clarification.

 

[HiKen51]

When I was flying with an iPhone the app always want to calibrate the compass every time I started it up. now I'm flying with android and I can't find a way to recalibrate manually even when I have moved a long ways from my previous location. Hoe do you make it calibrate if it's not telling you to?