Lego vs Microinfinity Gyro Comparison

[lvoc]

I have been testing the new Lego EV3 gyroscope and run some tests comparing it against the Microinfinity XG1300L. I took a video showing these two sensors being tested side-by-side. The details of the experiment and the video can be found at:
http://www.robotnav.com/gyroscopes/

[mightor]

Hi Lauro,

Thanks for doing the experiment. It would be very useful if you could have some kind of conclusion on that page with regards to what your findings were.

Regards,
Xander

[HaWe]

Thank you for this test!

So finally the extremely bad additive error of the Lego EV3 Gyroscope which already has been reported in the very early days of the ev3 EDU set using the EV3 GPL could be verified.
I'm actually curious which programming language has been used in this experiment...?

But all over 5° error of the Lego EV3 Gyroscope by roughly 10 times moving the gyro around and back and forth - that's really awful, that speaks for itself.

I agree with Xander that the documentation of the experiment should be improved:
- which programming language
- source code of testing program, especially reading loop speed

wishful would be additionally for statistical reasons
- sensor sensitivity,
- sensor drift,
- error dependency from turning speed, from turning amount, and from reading loop speed (quantified and specified)
- error range over time and related tu turning angle

[mattallen37]

For such an inexpensive toy, 5 degrees error in the integral isn't all that bad after that much movement. Sure it isn't ideal, but it's also not an expensive solution.

If the values are strictly integrated without any other input (such as magnetometer or accelerometer), there will be an accumulation of error which cannot be avoided.

Doc, I'm not certain, but I think he used the accumulation mode of the EV3 Gyro (where the gyro does the integration automatically), not the rate mode (where it reports the current rate of rotation).

Based on my limited testing of the EV3 Gyro, it is not tilt compensated at all. Based on these tests it appears that it isn't helped by a magnetometer either (which I figured).

Does the XG1300L use an accelerometer or magnetometer? Is it tilt compensated?

[HaWe]

that's why I asked which software program has been used.
But 30 EUR == cheap ?? "much" movement ?? not "ideal" ?? the results currently are crap IMO - curious if it could be improved by software.
A IMU-6050 is 15-17 EUR.

I ordered a cruizcore XG1300L IMU today. Here you may get more information about how it works:
http://www.minfinity.com/eng/page.php?M ... ub=1&tab=5
http://hackedgadgets.com/2011/01/19/mic ... -lego-nxt/

[mattallen37]

In the USA, the Gyro sensor costs $29.95 (equivalent to 21.90 EUR according to Google). Compared to several hundred to several thousand dollars (or more) for a "real" gyro (e.g. the ones used in aircraft), $30 is a cheap toy that I wouldn't expect a whole lot from. We are all entitled to our own opinion about the performance of the EV3 Gyro.

Thanks for those links.

[HaWe]

Matthew, you can not seriously use products for commercial aircrafts as a comparison...
And compared to thousands of EUR for commercial products you mentioned (I don't which products exactly you might mean) even 85 EUR for a toy is cheap and inexpensive but it is worth the money.

But even the Hitechnic gyro is a lot better with respect to additive error.

AFAIK you have lot of experience with Mindsensors and Dexter products -
Compared with the Lego Gyro and the Cruizcore Gyro: how accurate and how powerful are they (Gaussian error, additive error, drift)?

[mightor]

Helmut,

The programming language used for the EV3 is largely irrelevant considering all of the math is done on the actual sensor. You merely retrieve the current heading, which is calculated internally. The same applies to the XG1300L.

I have the XG and I think it's an awesome sensor, it's very accurate and his practically no drift

= Xander

[lvoc]

The XG1300L is a single axis gyro (one degree of freedom). It does NOT have or use any magnetic reference (which would be useless indoors). It does include a 3-axis accelerometer but their outputs are not used to compensate the gyro rate/heading estimation. The accelerometers are extra sensors that can be used for other applications such as tilt estimation in (near) static experiments, collition detection, etc.

[HaWe]

AFAIK the cruizcore gyro values are internally compensated by sensor fusion with the accelerometer and a Kalman filter - pls CMIIW....
(don't know exactly where I read this, I may be wrong, I just remember this roughly...)

The HT Gyro gives just rotation speed as an output value which have to be integrated by software by one own, it's not done internally.
This might be done for the Lego gyro, too.
I don't know if the cruizcore sensor is integrating or also just providing rotation speed - which is actually enough to have for angle processing.