Very interesting video from VectorNav showing what high frequency noise and magnetic fields actually do to an imu.
Nice visual representation!
Also this vectornav unit seems to be quite the imu!
G:)
You need to be a member of diydrones to add comments!
Comments
I see your point :)
Why don't they say exactly which sensor solution is used on the left?
From my experiences with IMU filtering, a slight change in Kalman filter parameters can make the system on the right to behave like a system on the left. Therefore it is not hard to make a good system look superb by producing a really under-performing one and compare them. The system on the left apparently uses just compass and accelerometers (it is referred as the system in use in many modern mobile phones and tablets), while any real IMU unit is based mainly on a very good gyroscope with the compass and accelerometer used only for initialization, drift estimation and cancellation.
To sum the things up, the video above has more of a negative effect than positive as they are comparing their 'novel' solution to a really under-performing one to make it look better.
I remember Bill Nesbitt referencing these sensors too - in addition to his own calibrated IMU - and he had one of the most stable quads I've seen. But since the Invensnse MPU-6000 is available now - does the Vectornav still offer any advantages? Is the MPU-6000 also calibrated? I remember Bill stating that IMU calibration was one of the most critical and difficult components to a stable platform.
It would be nice to see a comparison between vectornav ad mpu-6000
I visited an office building (225 Broadway St, New York) for a week and every time the Subway went through the basement, every CRT display in the place went crazy. So I guess flying your quad over subway trains is out as well. :-)
The last time I MRI'ed my quad, it didnt fly so well... ;)
Also, not a good idea to use magnets to keep your dome on...
One quick comment: the vectornav doesn't have a built in gps for correcting heading drift, so its built in kalman filter needs the magnetometer to avoid drifting. Even if you had a gps, these things could be used indoors or stationary where the gps isn't able to help with heading drift anyway. So I think it's an absolutely fair representation, as long as you understand their product and what they are showing. They also offer a temp calibrated version of their IMU. I know that a good filter can estimate bias, but I've also see that a good filter will converge *much* faster when fed non-biased sensor data.
For the product I am involved with, we just grab the raw sensor data from the vectornav + gps and feed it into our own kalman filter -- we aren't using their built in filtering, but their temp calibration really helps our filter lock in on true heading much much faster than the non-temp calibrated version.
Yuan Gao: also don't fly your quadcopter near electric motors. :-)
MRI machines: long ago I used to work in the MRI building of GE Medical systems -- this was back in CRT days and all our monitors got misaligned and the color guns each by a different amount. We had one conference room where you could throw paper clips at the white board and they would dance around on the order of feet before coming to rest somewhere random stuck on the wall. Pretty powerful crazy stuff.
The vectornav is not cheap, but you get really nice clean raw sensor data, you get a built in kalman filter if you choose to use it (that is very tunable), and temperature calibration -- which is more important than most people realize if you want fast IMU convergence and solid stable results. That said, you can get around the sky for much less money, but if you need accurate attitude estimation for camera pointing or georeferencing type work, it's worth the extra money to get some really nice sensors and feed that into a high end kalman filter.
So the moral of the story is: don't fly quadcopters near MRI scanners, maglev trains, particle accelerators, and rail guns.
But as always, "marketing puke" prevails!
-
1
-
2
of 2 Next