MEMS Gyroscope bias drift correction?

&I use such gyros. The zero drift depends on temperature mainly. It has a not always linear evolution. The best is to use rest periods to measure the "zero" and compensate following measurements. The complexity of the correction depends on the required uncertainty of results. What do you want to use it for?

If you have no limit on energy supply it is possible to build a stabilized temperature environment and reduce very much the potential errors.

Hi. Thanks for answering... Im trying to use a low-cost gyro with random drift in about 0.1 deg/sec for some sort of navigation system with 3-axis gyro. Gyro data is integrated to get relative angle information. System is temperature stabilized. Suppose that I use periodic rest-time zero measurements for compensating further gyro readings. Is it possible to get some reasonable gyro readings after an hour of straight line moving (0.2 m/s speed), with error less than 10 deg/sec ?

I do not understand your question : for a gyro a linear velocity should not have an influence the device being sensitive only to rotations.

Any way the errors depend on the sensitivity and please do not forget that amplifiers amplify also basic noise not only signal. If your sensor has a random sensitivity drift (i assume without stabilized temperature) of 0.1 °/s then for 1 hour you should assume that the random value is equal to no more than one standard deviation. If the distribution is normal then the maximal probable deviation is ± 3 std dev with a probability of >99.%.

I see that you use a 3 axis gyro. This has a reason to be used only when the displacement is in 3 dimensions i.e. diving or flying. For such applications the gyros should not be low quality since the errors at the end due to integration can be too important. It can become dangerous especially if you compute a displacement on the vertical.

If you give more informations the comments can be better matched to your problem.

sorry...forgot to login...:)

Hi,

are you talking about temperature dependant bias drift or (and?) random drift?

What about g-sensitivities, etc?

There are in good gyros around 5 to 20 drift components to be calculated or measured and compensated.

What level of performance for what application?

Simplest solution: buy a better one.

The good ones are not allowed to export!

Regards

RHABE

What kind of gyros did you analyse? Usually they are not sensitive to g (if you mean accelerations) of course it depends which principle they use thus my question. For the types i use the main drift of the "zero" (no move output) is dependent on yemperature and it can be compensated by a thermostatic control (maintain at constant teperature). With quite simple devices (for car applications or photo compensation) the uncertainty is quite good. Of course a calibration is required since the sensitivity is not the same. I did not use gyros for flighng systems so that i am pearhaps not familiar with the parameters you (5 to 20) you mention. Could you give more details?

Thanks

Dear Rahbe,

I would be very thankful if you could answer my last question since it is a problem which is for me of interest and importance. Thank you in advance.

Dear Nickname,

I was involved in some gyro development and similar high precision instrumentation devices mostly on the component level in the last 30 years.

This was related to high precision gyros of 0.01 degree per hour random drift rate.

As one of my todays specialties is precision measurement, precision bearings and precision flexures I am naturally interested in MEMS gyros but not really involved deeply.

So what are your questions?

Regards

RHABE

Sorry for the delay but for an unknown reason i do not get daily the post as before. I one of your previous comments you mentioned between 5 and 20 parameters. Since my experience resides in the use of simple low cost gyros for precisions at the industrial level and not at the navigation level as you have (at least i presume) i am interested to know more since i consider this as an opportunity to enlarge my horizon.

There is as well a trend in low cost to increase the stability and the sensitivity and it is good to know what problems are met in such situations.

Thank you for your readiness to answer. By the way in some mythologies the "Rhabe" is connected to knowledge. I assume this is the reason you made the choice for it.

Hi Nickname,

to relate drifts to physical origins it is necessary to know the construction of the sensor.

In principal these result in drifts that have a fixed component and uncertainties from the change of the contributing factors.

For example in rotating gyros the radial mass unbalance (rotating with synchronous speed) multiplies with the axial synchronous 1n vibration to result in a coordinate fixed DCtorque that will cause drift. If the radial mass unbalance is changing (may be multiple reasons for a change) or the axial synchronous vibration is changing we see a changing drift rate. Very often it is not possible to produce accurate enough items (impossible or too costly) to get rid of these drifts. So the total radial mass unbalance is equivalent to the balancing accuracy. The change in radial mass unbalance is originating in material instability (long term mostly) and temperature effects (residual stress is generating bend and bow by change of elastic modulus with temperature and by bimetallic distortion with temperature and ordinary distortion with tempco of expansion and temperature gradients. The axial vibration is originating from nonideal bearings and nonideal shaft and (worst) nonideal bearing seats in the housing.

As there is no rotation in a modern MEMS "gyro" there are similar but not identical mechnisms. These have to be minimised by the designer of the sensor.

If you think at the user level then it may be suitable to think more globally:

the drift is made up by an error budget: "constant" drifts of bias and scale factor, these are not constant but dependent on temperature and temperature gradients (3orientation dependent and one time dependent), acceleration-sensitivities, acceleration-squared-sensitivities (anisoelasticity), axis orientation, switch-on to switch-on instability, day-to-day-instability, mounting to remounting instability.

(These errors to be completed if construction is known and after a look into the literature, I am sure I forgot some error sources in this fast and not complete list.)

What literature do you know? I expect this discussion may be beyond the interest of other users, so please send your email adress and location.

To go deeply into this field would require some time and effort but is worth while.

Regards

RHABE

I thought that you think about mechanicle sensors but since the trend is since many years toward optical or mem type sensors i was not sure.

You have my mail address and the location as requested.

Expect to continue the contact.

syapnh@tiscali.fr

France

When you will have some time i would appreciate to have your ccordinates on my mail address

Hi Nickname,

I did write to you an email, this was on July 17th, and did send it once more on July 19th.

Did you receive this or is there some email cruncher on tiscali? Big Brother or else?

RHABE

Especially for low-cost inertial instruments, it's important to be aware of many error sources that are motion-sensitive. Tip-of-the-iceberg: mounting misalignment. See http://jameslfarrell.com/gyro-mounting-misalignment/