SparkFun Forums

[ala42]

Do I read your graphs correctly, all axes reached their min/max values at exactly the same time at point B, so there is no time lag between the max temperature reading and the min/max values of all axis ?

[sebmadgwick]

There are no “time lags”. The rate of heating is many orders of magnitude below any relevant time constants. Each measurement is the mean of 1000 samples taken over 2 seconds, the rate of heating is less than 0.5˚C per minute and the thermometer is built into the MPU-6050’s 5mm x 5mm package.

[olliesa]

I'm attempting to do the same thing with a batch of MPU-6050's and I have observed a similar "hysteresis" effect on the gyro bias with respect to temperature. The bias exhibits a different relationship to temperature when the device is being heated compared to when it is begin cooled. I'm curious to know what, if anything, you ended up doing to improve your error model.

[sebmadgwick]

My current thinking is that the behaviour is accounted for by the random walk in some way being excited by the dynamics. I have found that the temperature curve seems to be repeatable but that it is ‘shifted’ up or down by a limited amount giving the impression of a hysteresis if only a limited number of cycles are observed. I have also found that the shift is reduced (or perhaps just less likely?) if the temperature change is monotonic.

I am hoping that effective calibration can be achieved by first modelling the temperature curve based on a monotonically changing temperature dataset and then tracking the low-bandwidth random walk at a higher level (e.g. AHRS sensor fusion algorithm).

[olliesa]

Thanks for following up.

I believe that at least part of the 'shift / hysteresis' I observed was actually caused by the high frequency vibrations in my thermal chamber. These vibrations were different when I was heating the devices compared to when I was cooling them.

[adrianj]

Hi,

I notice the images for the temp graphs discussed here have been deleted. Are they still available somewhere ? Would be very useful, since we just started using this device.

[victoraa]

Hello,

How to get the gyro correction with the temperature? Invensense says that if you use the MPU library you have temperature compensation, but they don't say how to do it. If we known how the temperature act, we could include in your program and correct the gyro values.

Ploting Gyro (Z) with the temperature I can see a graph, different if I plot the X or Y.