search

gabriel-milan / TinyMPU6050

Tiny implementation for MPU6050 focusing on performance and accuracy
MIT License
22 stars 10 forks source link

Measurement slow when when upside down #24

Closed mcer12 closed 3 years ago

mcer12 commented 4 years ago

Hi @gabriel-milan first off, thank you so much for this, it works the best with ESP8266 from libraries I tried :)

I don't know if this is a bug or it's a "feature" of the sensor but when I orient it upside down (MPU6050 chip facing down), the measurements are slow / delayed. It's around one second delayed. Once I reorient it back, it's fast again. This is very unfortunate as I plan on using it oriented like that.

EDIT: Z-axis seems to stay fast, only X and Y are slow. Tested on Angles example. The difference can be easily observed when making sudden fast moves

gabriel-milan commented 4 years ago

Hi @mcer12, thank you for appreciation!

Honestly, I probably haven't tested the library in such conditions, as I wrote it for different purposes. Can you assure it's a library issue? Have you tried it with another MPU6050/ESP8266 pair? Unfortunately, I have no access to mine, so I can't tell whether it's reproducible.

I don't think this should be regular behavior, though. There may be a bug I couldn't find yet.

mcer12 commented 4 years ago

@gabriel-milan quite honestly I wasn't able to make other libraries work (esp8266) :) MPU6050_tockn example returns the same number and doesn't react to changes. MPU6050_light example crashes during calibration

I suspect this delay might be caused by the calculation somehow when the gyro and acceleration data are merged. I will take a look if this delay is also happening with acceleration / gyro angle only.

gabriel-milan commented 4 years ago

I wasn't able to make other libraries work

I can't tell for sure, but maybe it's a sensor issue. I'd be pleased if you would investigate on that, all collaboration is very welcome.

mcer12 commented 4 years ago

The sensor works perfectly OK with your library, I think it's the compatibility with ESP issue with other libs.

I viewed the values in serial plotter side by side mpu.GetAngAccX() and mpu.GetAngX() and here are the results:

GetAngX() is instant and perfectly copies and smoothens GetAngAccX() when positioned normally, on point. When positioned upside down, GetAngAccX() is instant but GetAngX() takes time to catch up. So it's absolutely a calculation issue. The graph looks like it tries to average to the wrong side... it overshoots and then slowly catches up. I can share the plotter screenshot if that helps.

mcer12 commented 4 years ago

I played a bit with the plotter and the mpu.GetAngAccX() is fast but is also a bit less precise or responsive. So I will probably redesign the project to use it in normal position. The sensor probably doesn't like to be positioned like that.

gabriel-milan commented 4 years ago

I can share the plotter screenshot if that helps.

That would be nice to see, even if you're going to redesign your project.

The sensor probably doesn't like to be positioned like that.

I don't think that should be an issue. It's likely a library issue. Such a pity I'm not able to help on debugging/fixing these right now (and I'm not sure when I'll be able again).

github-actions[bot] commented 3 years ago

Stale issue message