Menu
Forums
New posts
Search forums
What's new
New posts
New media
New media comments
New profile posts
Latest activity
Media
New media
New comments
Search media
Members
Current visitors
New profile posts
Search profile posts
Log in
Register
Search
Search titles only
By:
Search titles only
By:
New posts
Search forums
Search
Search titles only
By:
Search titles only
By:
Menu
Welcome DJI Spark Pilot!
Jump in and join our free Spark community today!
Sign up
Forums
General Forums
Regional Meet-Ups & Local Flyers
Canada
Dji Spark took off and now is lost
JavaScript is disabled. For a better experience, please enable JavaScript in your browser before proceeding.
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an
alternative browser
.
Reply to thread
Message
<blockquote data-quote="Virtual1" data-source="post: 95958" data-attributes="member: 4084"><p>FWIW, pitch, yaw, and roll combine to describe the direction the craft is facing and in what orientation. It has <em>nothing</em> to do with its elevation or position on the map. GPS signals are all about position and elevation. pitch/yaw/roll are all detected locally by the compass in the craft.</p><p></p><p>So a "yaw error" probably has nothing to do with the GPS. This sounds like either a compass calibration problem, magnetic interference, or a hardware failure. (could be a software failure too, unless there's fusion hardware in the compass, the software has to take the 9 axis information from the x/y/z compasses and combine them to create an actual compass bearing)</p><p></p><p>I've read a few people's accounts of being a bit manic (in my opinion) about re-calibrating their compass before every flight, or every time they go to a different location. My personal take on it is you should recalibrate your compass once a year as well as every time you make a significant change in your location. (flying in another city or state for example) But knowing what I do about compass fusion (and it's a bit of a geometry headache to do) I really don't see how it could be getting serious errors from being out of calibration. My calibrations usually only adjust by a few degrees in the x, y, and z axis. The only significant change problem is range errors, and any halfway intelligent fusion code should be able to detect when a sensor has gone out of range and simply adjust the scaling for that sensr to compensate. (all that turning over you do when calibrating the compass is ONLY being done to find the + and - max ranges for the three compasses, the software assumes the exact middle between + and - is the center - which is <em>usually</em> zero) You may lose a dozen degrees of accuracy, but you can still fly on that, and it's infinitely better than just throwing up your hands and saying "I give up! no more readings for you!"</p><p></p><p>At any rate, you SHOULD have been able to just let go of the sticks and it should have stopped moving. Compass says NOTHING about velocity or elevation - it doesn't matter if the compass readings are totally off the wall <em>if you're not moving.</em> But if RTH has been activated, a bad compass will make it very challenging for your spark to return to you. The compass and GPS will basically get in a fight, and you will probably see the spark start flying in a spiral.</p><p></p><p>It would be nice if the spark would have a tolerance level for bad compass readings, and at some point during RTH say "ok clearly the compass is not agreeing with the GPS, and the GPS is MUCH more trustworthy, so we're just going to stop listening to the compass for now." It should then be able to, by GPS alone, find its way to within a dozen or so feet of the home point, and set down. Once it's OVER the home point, the compass doesn't matter anymore because nobody cares what direction the quad is facing when it lands. I've read two account of people getting into a location that had magnetic interference and both of them reported the same problem, RTH was driving their spark in circles/spirals, and this is totally what I would expect to see happen. It's like a donkey walking toward a carrot being held out in front of them on a stick. If the rider holds the carrot off a bit to the right, the donkey will continue to turn right a little as they walk, and the path traveled will be a circle. Compass calibration errors <em>cannot</em> generate a straight flight unless the compass has been physically reoriented. AFAIK it's soldered down on the motherboard (or otherwise fixed in position), so that's not possible.</p><p></p><p>At the end of the day, my first suspicion on fly-away is always going to be "you didn't wait for a new home point to be set before taking off, and after RTH somehow got triggered, it tried to go to the last known home point, which was where you last flew, which may have been miles away from your current location, and to it just took off in that direction" DJI could patch for this - it's not too hard to calculate distance with GPS (yes it varies but the variance is KNOWN and can be adjusted for) so the software really ought to look and say "ok we want to RTH but home is 14 miles from here, there's no way in hell we're going to make it" and take some other course of action. WHAT else to do is of course the pickle, and I suspect is the reason they haven't made that change - I can't imagine what would be a good alternative.... you can't just land.</p><p></p><p>Inertial guidance maybe? That would be cool to see on a drone. Fighter jets use that for example, but they're using accelerometers and compasses that are SEVERAL orders of magnitude more accurate, and the calibrations they go through are absolutely insane. I don't know if "consumer grade" hardware could pull off inertial guidance good enough to RTH a spark without GPS. The limited flight time and fairly relaxed flights might be enough to make it practical? discuss?</p><p></p><p>(seguay - the military calibrates its accelerometers by taking them deep underground in a big cave where they BOLT the units onto a chunk of exposed bedrock then everyone leaves the cave and they start calibrations remotely, as the gyroscopes are calibrated <em>by sensing the rotation of the earth </em> - that's pretty hard-core!)</p><p></p><p>Background on that: the compass is NOT just a compass, and it's NOT just single axis. A proper compass is 9 DOF - 3-axis compass, 3-axis gyroscope, and 3-axis accelerometer. The compass readings are actually taken from the gyroscopes, which are corrected by readings from recently known orientation, which is updated from the acceperometers and being slowly tweaked by the compasses. It's quite a production. Inertial guidance works in a similar way, but lacks the self-correcting input, leading to drift over time. It's perfect when you start, but as you fly it slowly gets less accurate, and the more you flop around the faster it can lose accuracy. Inertial guidance is great because it's always got a position for you, unlike a GPS that updates once per second usually, so a GPS is usually used to keep the inertial guidance accurate the same way the compass keeps the gyro accurate. 15 minutes of mostly-level and slow-turning flight is actually pretty easy to stay on top of, even for consumer-grade sensors. (probably not workable on a racing quad though!)</p><p></p><p>I wonder if anyone at DJI is reading this?</p></blockquote><p></p>
[QUOTE="Virtual1, post: 95958, member: 4084"] FWIW, pitch, yaw, and roll combine to describe the direction the craft is facing and in what orientation. It has [I]nothing[/I] to do with its elevation or position on the map. GPS signals are all about position and elevation. pitch/yaw/roll are all detected locally by the compass in the craft. So a "yaw error" probably has nothing to do with the GPS. This sounds like either a compass calibration problem, magnetic interference, or a hardware failure. (could be a software failure too, unless there's fusion hardware in the compass, the software has to take the 9 axis information from the x/y/z compasses and combine them to create an actual compass bearing) I've read a few people's accounts of being a bit manic (in my opinion) about re-calibrating their compass before every flight, or every time they go to a different location. My personal take on it is you should recalibrate your compass once a year as well as every time you make a significant change in your location. (flying in another city or state for example) But knowing what I do about compass fusion (and it's a bit of a geometry headache to do) I really don't see how it could be getting serious errors from being out of calibration. My calibrations usually only adjust by a few degrees in the x, y, and z axis. The only significant change problem is range errors, and any halfway intelligent fusion code should be able to detect when a sensor has gone out of range and simply adjust the scaling for that sensr to compensate. (all that turning over you do when calibrating the compass is ONLY being done to find the + and - max ranges for the three compasses, the software assumes the exact middle between + and - is the center - which is [I]usually[/I] zero) You may lose a dozen degrees of accuracy, but you can still fly on that, and it's infinitely better than just throwing up your hands and saying "I give up! no more readings for you!" At any rate, you SHOULD have been able to just let go of the sticks and it should have stopped moving. Compass says NOTHING about velocity or elevation - it doesn't matter if the compass readings are totally off the wall [I]if you're not moving.[/I] But if RTH has been activated, a bad compass will make it very challenging for your spark to return to you. The compass and GPS will basically get in a fight, and you will probably see the spark start flying in a spiral. It would be nice if the spark would have a tolerance level for bad compass readings, and at some point during RTH say "ok clearly the compass is not agreeing with the GPS, and the GPS is MUCH more trustworthy, so we're just going to stop listening to the compass for now." It should then be able to, by GPS alone, find its way to within a dozen or so feet of the home point, and set down. Once it's OVER the home point, the compass doesn't matter anymore because nobody cares what direction the quad is facing when it lands. I've read two account of people getting into a location that had magnetic interference and both of them reported the same problem, RTH was driving their spark in circles/spirals, and this is totally what I would expect to see happen. It's like a donkey walking toward a carrot being held out in front of them on a stick. If the rider holds the carrot off a bit to the right, the donkey will continue to turn right a little as they walk, and the path traveled will be a circle. Compass calibration errors [I]cannot[/I] generate a straight flight unless the compass has been physically reoriented. AFAIK it's soldered down on the motherboard (or otherwise fixed in position), so that's not possible. At the end of the day, my first suspicion on fly-away is always going to be "you didn't wait for a new home point to be set before taking off, and after RTH somehow got triggered, it tried to go to the last known home point, which was where you last flew, which may have been miles away from your current location, and to it just took off in that direction" DJI could patch for this - it's not too hard to calculate distance with GPS (yes it varies but the variance is KNOWN and can be adjusted for) so the software really ought to look and say "ok we want to RTH but home is 14 miles from here, there's no way in hell we're going to make it" and take some other course of action. WHAT else to do is of course the pickle, and I suspect is the reason they haven't made that change - I can't imagine what would be a good alternative.... you can't just land. Inertial guidance maybe? That would be cool to see on a drone. Fighter jets use that for example, but they're using accelerometers and compasses that are SEVERAL orders of magnitude more accurate, and the calibrations they go through are absolutely insane. I don't know if "consumer grade" hardware could pull off inertial guidance good enough to RTH a spark without GPS. The limited flight time and fairly relaxed flights might be enough to make it practical? discuss? (seguay - the military calibrates its accelerometers by taking them deep underground in a big cave where they BOLT the units onto a chunk of exposed bedrock then everyone leaves the cave and they start calibrations remotely, as the gyroscopes are calibrated [I]by sensing the rotation of the earth [/I] - that's pretty hard-core!) Background on that: the compass is NOT just a compass, and it's NOT just single axis. A proper compass is 9 DOF - 3-axis compass, 3-axis gyroscope, and 3-axis accelerometer. The compass readings are actually taken from the gyroscopes, which are corrected by readings from recently known orientation, which is updated from the acceperometers and being slowly tweaked by the compasses. It's quite a production. Inertial guidance works in a similar way, but lacks the self-correcting input, leading to drift over time. It's perfect when you start, but as you fly it slowly gets less accurate, and the more you flop around the faster it can lose accuracy. Inertial guidance is great because it's always got a position for you, unlike a GPS that updates once per second usually, so a GPS is usually used to keep the inertial guidance accurate the same way the compass keeps the gyro accurate. 15 minutes of mostly-level and slow-turning flight is actually pretty easy to stay on top of, even for consumer-grade sensors. (probably not workable on a racing quad though!) I wonder if anyone at DJI is reading this? [/QUOTE]
Verification
Post reply
Forums
General Forums
Regional Meet-Ups & Local Flyers
Canada
Dji Spark took off and now is lost