InvenSense MPU-6000


I think it's perfect!



- Tri-axis gyroscope and tri-axis accelerometer in one package

- Programmable scale ranges:

    - Gyro: ±250, ±500, ±1000, and ±2000°/s

    - Accelerometer: ±2g, ±4g, ±8g and ±16g

- Output: rotation matrix, quaternion, Euler Angle, or raw data format

- Digital Motion Procssing (DMP)

- 16 bits DACs

- Sample rate at least 1 kHz

- Digital temperature sensor

- 400kHz FastMode I2C or SPI

- Support for external I2C magnetometer

- No need for calibration

- Support for programmable interrupts

- Price: $15 (1-99)   Farnell: 43,33 € (1-9)

- Size: 4 x 4 x 0.9 mm

E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones


  • I have just received my nice new APM boards from udrones (Mexico) which arrived on fri, afternoon (Cyprus), which was great as I didn't order them until Tues. afternoon, Ditto a small cable from DIY drones. Thanks guys that is what I call service. I see that I have a IDG500 gyro installed which has a dual output of 500 deg/sec and also a 4x output equivalent of 110 Deg/sec. That's great, can someone tell me how those outputs are integrated into the system i.e. does the flight control use both outputs or is the 110 deg/sec reserved for aux. camera stab?


  • I.S. 

    on 3. Is it reducible to a neat equation? This I suspect is a matter of debate. For 20 years, I've seen only two schools of thought on these kinds of problems: neatly reduced curve fitting, and lookup tables based on testing. All the designs which start with neat theories, end up adding lookup tables in the end anyway (to adjust for x & y). In this example, simple rotation on a single axis doesn't really exist, rather real rotation is - or could be - occurring in ~19 dimensions simultaneously - while imaginary rotation can be occurring in all 3 axis (vibration induced error).  So pure math will only get one so close to a solution - in the end, the serious solutions are lookup tables and weighted averages - anyone else?


    on 5. GPS are not very accurate in 3d space, they can jump 200 feet in a single reading; however, the GPS signals allow some units to report relative movement with much greater precision than their resolved location.


    (GPS resolvement is another example of throwing loads of piss poor data on the wall to see what sticks. Modules are rated on the millions of guesses per second.)


    Finally - how good is the integrated IMU than the discrete?, the answer starts with groking the limitations of the contained sensors - for games, antiskid traction control, antishake, ipad gaming, etc... it is a fine system; however, once the 3rd dimension is no longer an assumed norm, things can go wrong; and by go wrong is meant the flaming spiral of death.


    I've been thinking that a better understanding of airspeed (multiple pressure sensors, angle of attack sensors) etc,,, might could offer an alternative sense of self, but the only three thus far demonstrated are infrared, IMU, and the self-stabilized airframe.


  • 3D Robotics
    bGatti: good job on the explanation! It's perhaps the most misunderstood part of IMUs. The only thing I'd add is the GPS only gives meaningful heading information when in motion, so for things that hover or can remain stationary, such as quads, helis and rovers, it's better to use a magnetometer for yaw drift correction.
  • @bGatti

    Thanks for your explanation.

    at point 3 you mean:

    ac= v^2/R = Rxw^2 right? or some other equivalent equation?

    at point 5 you mean use the GPS to obtain bearing?

    Anyway I knew about the "large" MEMS drifting, I just asked myself how good this gyro&accels of MPU-6000 are comparing to other MEMS used on other cost-effective IMUs (eg Oilpan)

  • @I.S

    Perhaps no topic has been more discussed to less effect here and elsewhere :)

    I'll have a go - but it's tricky:


    1. Silicon gyros are crap. Compared to aircraft grade gyros which allow the Space Shuttle to orbit the earth a hundred times and land straight and level days later, your best Silicon Gyro will go wheels up about 3 minutes after you turn it on - which is a small price to pay for a 100-1000? times improvement in size, power, and cost.

    2. Because of 1, it is necessary to "reset" the gyro constantly. To do this, one can use a static accelerometer which will point "down"; however accelerometers are affected by, well acceleration, one special variant of which we call centripetal acceleration - caused by changing direction.

    3. One can use a compass to determine change of direction; however turn rate is only half-a-term as centripetal force is a product of both speed and turning rate (Delta V).

    4. needs a means to repair the gyro which doesn't assume we are standing still, or at least attached to a wheelbase which is right-side up 99.99999% of the time.

    5. ... Enter the GPS - which often gives meaningful movement information (sometimes jumps 200 feet).


  • Could you please explain further (or redirect somewhere) about this chip not having centripetal correction?


    On the other hand how good is this chip compared to stand alone gyros and accels? (in terms of vibration rejection, accuracy,...)

  • @bcr,
    One reason is to offload the engineering. If Ivensense can output rational values - that's be great.

    Chris's point is interesting; I guess that solid state gyros are so unstable compared to the models used in current airplanes that one cannot build a "solid state" facsimile of their spinning counterpart?
  • It's nice to combine the gyro + accel, but with modern MCUs I hardly see a need to offload the math. You can run a super-naive C floating point of DCM with like 5% of a $3 STM32.

  • 3D Robotics
    Yes, we discussed this last year when it was announced. It is not available now and won't be for a while. We've been working closely with Invensense and will be the first autopilot to use this chip. But using the built-in sensor fusion is not easy, since it wasn't designed for moving vehicles and centrifugal force is not corrected for. We're in discussion with Invensense about ways to input a GPS direction vector, but in the first iteration I suspect the chip will be used mostly just for the raw sensor data, not the built-in Kalman filter.
  • Developer
    We are just waiting for the SPI version to release the APM v2.0 around Q3-4
This reply was deleted.