Angular/orientation sensing - Gyro, accelerometer, or IMU?

[Mayo2017]

Working on a project right now where I need to measure orientation of a device for processing with a PIC16 micro. Trouble I'm having is that there's a large amount of solution out there for angular measurement, and it's a bit overwhelming when it's not a group of sensor types I've ever worked with. I get the basic principles behind gyros, accelerometers, and combined IMUs, but I just need a nudge in the right direction as to what sort of thing I should be looking for.

I'm looking to measure the device's angular orientation to at minimum a precision of 0,25° - 0,1° or better being ideal. I don't anticipate needing a high-G-capable sensor, expecting a max of 2g on a really bad day, and the update rate doesn't really need to be any better than 1Hz, thought better is of course appreciated. The actual rate of change it'll be experiencing is well under 1°/second, so it doesn't need to be able to do quick updates. I2C-capable is a good thing since it would keep things simple with other devices.

Anyone have anything they've used down at sub-degree resolution without too much trouble? Not looking for someone to go out and find a product for me, just wondering if anyone already had something they knew of, or even what kind of device I should be looking for. 

[richard.cs]

Orientation on which axes and over what kinds of timescales?  An accelerometer can tell you which way is down, gyros can give you relative orientation (conpared to their starting position) over short timescales,  if you need absolute orientation over hours then expect to pay several thousand for the gyros and have some way of setting the initial orientation. At this point an external reference becomes more practical.

I think we need more details of your budget,  application,  etc.

[hli]

Reference: slaa518a.

Link: http://www.ti.com/lit/an/slaa518a/slaa518a.pdf (Nine-Axis Sensor Fusion Using the Direction Cosine Matrix Algorithm on the MSP430F5xx Family)

[AndyC_772]

Is this device subject to significant acceleration besides gravity?

If it just needs to know which way is 'down', you can use a 3 axis accelerometer. Several manufacturers make them with I2C interfaces, and they're cheap.

Accuracy might be an issue, though... your best bet might be to pick out the most accurate one you can find, and test it. If you only need 1 Hz update rate, you can at least average many samples to get the necessary resolution, then test it for linearity and drift.

[Mechatrommer]

gyro is to measure angular speed not orientation. acceleratometer is to measure linear acceleration, not orientation. what you are looking for is magnetic sensor... but i'm not sure what accuracy a cheapo one can provide. maybe its propositional to the price.

[AndyC_772]

A magnetic sensor will tell you which way is north, not which way is up.

[miguelvp]

light pulse and light sensors in the jig.
Sensors hooked to interrupts and a processor that is fast enough and has a clock accurate enough to determine the times.
Once you collect the times then you can calculate the position, orientation etc.

Higher frequency chip for more precision and of course a precise clock to match that precision.

[mikeselectricstuff]

If you need to know which way is up on something that isn't moving fast or vibrating,  an accelerometer will do it.
For that accuracy  you will probably need to do multi-point calibration.
I used a Freescale accelerometer, I think 14 bit - whatever was the best one they did at the time - to sense the position of the London Eye (big wheel) & dynamically remap the lighting to produce stationary displays regardless of wheel position.
ISTR that with a fixed rotational axis, without calibration it was accurate to a degree or so, but seemed stable enough that you;d probably get quite a lot better with some calibration.
There may be issues with things like offet and gain drift over temperature.

If you're trying to detect level with movement in multiple axes it gets more involved but should still be doable.

[f5r5e5d]

0.1 degree sounds challenging when all noise, tempco and drift terms are added up

search did lead to a accelerometer tech new to me: http://www.memsic.com/accelerometers  maybe worth looking at