Hey everyone, long time lurker first time poster. I’ve been so thankful over the past few years at everyone on these forums who give up great information, and now it is my turn to return the favor.
A few years ago I ordered a GY-521 breakout board for the MPU 6050 from Amazon:
If you look at the product pictures carefully, you can see they are actually showing two different boards. The second picture is the one I got in January 2014, and the first picture is the one I got about a month ago when I ordered a few more. It may be easiest to see the difference if you look at the voltage regulator - one of them has a capacitor between the voltage regulator and the jumper pins, and one doesn’t.
The newest version I got a month ago was giving me absolutely HORRIBLE data. When I had the digital low pass filter as high as it could go, and the accelerometer range set to ±2g, I was still getting ±0.5g of noise. The value would jump from -0.5g to 0.5g while the sensor was sitting still. The sensors I got in 2014 only had about 0.01g of noise.
So after a few weeks of investigation including oscilloscopes, pulling another MPU6050 out of an old Galaxy S3, and a bunch of other stuff I finally figured out the problem. One of the capacitors is the wrong value.
According to the datasheet, the Regulator Filter Capacitor should be 0.1 uF, but the one on the board is around 2 nF. Replacing this capacitor brought the deviation of the accelerometer down to 0.01g without any low pass filter.
Check out the attached picture to see which capacitor needs to be replaced.