IMU Data Collection

[Baron Johnson]

Hello,

I was looking for some feedback as far as what a good IMU & onboard recorder setup would be with relatively clean signals (I know this depends in large part on installation, especially with accels) for aircraft characterization. Looking for something to put on ~3ft and larger aircraft.

I am looking at using the MEMsense nIMU and Eagle Tree Systems recorder. If anyone has experience using that particular IMU I would be very interested in hearing your experience. Or, if there is a better IMU choice I would be interested. The Eagle Tree FDR requires I2C input, but I would be open to other recorders as well if it's not too time consuming or complicated. The nIMU/Eagle Tree system is almost plug&play, which makes it appealing. But, obviously, data quality is more important.

Thanks for any input.

-Baron

[Unterhausen]

what do those units cost?

Here is a link:
http://www.memsense.com/products/nimu.asp

[Baron Johnson]

I think the nIMU runs about $2400.

This would be used for research at the University of Florida, so it wouldn't be out of my own pocket ;)

[clolson]

I've done something similar with the Xbow MNAV. And in addition I can take the collected data and replay it in FlightGear to get a 3d visualization of the flight (and control surface deflections.)

The downside to this is that the MNAV open source kalman filter estimate of pitch and yaw are not the greatest. Roll angle generally seems pretty reasonable.

But I'm not trying to pick on Xbow here, more accurate solutions are probably going to cost you money. The Microbotics MIDG is really good as far as these MEMS sensor based IMU's go, and they are up in the $5k - $6k range.

Curt.

[Draganfly]

Baron, what sensors does your ideal IMU have? How much processing should the IMU do for you (do you want orientation estimations based on well-designed algorithms, or raw sensor outputs, or filtered sensor outputs?)? What data recording / logging features do you want, and at what sample rate?

-Adam

[dmgoedde]

I think the nIMU runs about $2400.

This would be used for research at the University of Florida, so it wouldn't be out of my own pocket ;)

WHOA! $2400? I could send you something to test free, and if you like to keep after short trial period the price would be around $500. This would be 2 axis gyro with temperature compensation data, and 3 axis accel +/- 3g range. Data from IMU 12 bit resolution, at 120 Hz over 8 channels (each an being the mean of 30 measurements). Data would be logged to micro-SD card. This is a stand-alone unit, only requiring GND and +4to +13 volt input. Tunr it on and data starts recording to SD.

Dean

[Draganfly]

Also, do you need the data in real time inside the airframe or on the ground, or is recording on a SD card fine?

[Baron Johnson]

Baron, what sensors does your ideal IMU have? How much processing should the IMU do for you (do you want orientation estimations based on well-designed algorithms, or raw sensor outputs, or filtered sensor outputs?)? What data recording / logging features do you want, and at what sample rate?

-Adam

Adam,

3-axis gyro, accel, & magnetometer, and 4 control servo positions. Pressure sensors for a pitot-static system could be used, but not essential. Temperature corrected output would be nice, but not absolutely necessary. Orientation estimation would be great... definitely would save some time, but also would want the raw/filtered data for potential uses. For sample rate, probably something like 40Hz or so. Recording onboard would be perfectly acceptable (in fact, preferable), no need for real time data on the ground and no need for extremely long flights

-Baron

[Baron Johnson]

WHOA! $2400? I could send you something to test free, and if you like to keep after short trial period the price would be around $500. This would be 2 axis gyro with temperature compensation data, and 3 axis accel +/- 3g range. Data from IMU 12 bit resolution, at 120 Hz over 8 channels (each an being the mean of 30 measurements). Data would be logged to micro-SD card. This is a stand-alone unit, only requiring GND and +4to +13 volt input. Tunr it on and data starts recording to SD.

Dean

Dean,

I'd be very interested in something of this sort! Have you found +/-3g accels work well for small aircraft? I know some people here at UF have had issues with vibration saturating 3g accels, but it could've been installation issues or something else. With a 2-axis gyro, which axis do you typically ignore? Maybe roll, since you don't want to ignore yaw without a magnetometer? Is there any capability to store servo positions? How much does a unit like that weigh?

-Baron

[dmgoedde]

Dean,

I'd be very interested in something of this sort! Have you found +/-3g accels work well for small aircraft? I know some people here at UF have had issues with vibration saturating 3g accels, but it could've been installation issues or something else. With a 2-axis gyro, which axis do you typically ignore? Maybe roll, since you don't want to ignore yaw without a magnetometer? Is there any capability to store servo positions? How much does a unit like that weigh?

-Baron

I won't exaggerate - but weight is about 15 grams if I make you a unit that does just what you're asking, and 30 grams if GPS module is included. This is home-made all-SMD stuff, attached are photos of some of the components.

In my testing I had the IMU oriented to capture roll and pitch rates, plus of course all three Accel axes XYZ. You could re-position to capture any two rate axes. I'm not sure if you could use 5 DOF to get full state estimate, however in special circumstance with appropriate assumptions this may work to a degree.

I have flown this device on a 9 ounce 26" span straight-wing aileron V-tail plane. I captured onto SD the 7-channel A/D data (5 DOF IMU + temperature +Vref (of gyro))data at 120 Hz, 4 servo positions at 25 Hz (though I can do full 50Hz), and happened to capture 1 Hz GPS as well, the GPGGA and GPVTG nmea sentences. There is a line-formatter on each row so you can later parse and split-out data types (IMU, servo, GPS). 20 minutes of flight results in a 6 MB .txt file (about 2,500 pages if opened in Word!!!). This is just a FIREHOSE of data, and you need something besides Excel to analyze (I uses SAS JMP, which can open files up to 1,000,000 rows). Start/Stop of data logging is controlled by placing left Tx stick (throttle/Rudder) to top left, and STOP is bottom right. This aligns OK with arming function of speed control at the beginning of a flight.

Vibration was observable in the XYZ accelerometer data despite taking into account bandwidth via capacitor choice between the XYZ accel and 8 channel A/D. My setup was rigidly mounted to airframe. I think wrapping in foam and possibly suspending in a box with rubber bands (stiff though) would cut down on the higher frequenct noise. Nonehtless, the XYZ accel data I got is usable. This might be where a Kalman filter would shine in its ability to filter noisy data into something more usable.

I can provide snippet of the data output (it's not on my work computer)