Best low-power method to detect Orientation-change?

I’m looking for the best way to detect the change of orientation of a device I’m working on (i.e. rotation on any of the XYZ-axes). It is absolutely critical that it is a low-power method (< 100uA) as I’m planning on powering this device with a single D cell and achieve a lifespan of 2+ years.

So far, I’ve been using an ADXL-362 hooked up to a Particle Electron (in Deep Sleep) and send an interrupt signal whenever sufficient movement/vibration is detected (after which code runs to monitor the orientation) The issue however with this is twofold:

(1) Often, the device picks up on some minor vibrations, runs through its code only to realize it was a false positive
(2) When the device is in fact rotated along one of its axes, if not sufficient vibration is produced it doesn’t wake up and run its code

As far as I know, when the MCU is in Deep Sleep the ADXL-362 can only be configured to send the INT signal once an acceleration threshold is exceeded; it can’t continuously measure XYZ-values (please correct me if I’m wrong)

So, I’m looking for another type of sensor that I could use alongside the ADXL to improve the performance of my device. I’ve been considering a tilt sensor, but the issue is that I don’t know exactly in advance what orientation my devices will be in (I’m planning to get a few 100 of them made). So I want to be able to have devices, let’s say, upside-down, upside-up or any other orientation and not have it impact its ability to detect on orientation change.

Three mechanical tilt switches should be able to detect any orientation.

You can probably find an accelerometer that will detect orientation change without needing any vibration to wake it up...

Tough one. I think you have one of the best sensors for this task. Maybe you can find one with "tap detection" to improve/modify the sensitivity.

Is it something that normally sits on a flat surface like a desk? Then it is possible to spin it on the desk and that will make no detectable accelerations which that sensor could use. I would try to constrain the way that it is moved, for example, put a handle on it in a way that you automatically rotate it when you pick it up.

A D cell is a very big battery. You probably don't need to get your power consumption down to nano-amps. Maybe a MPU6050 or MPU9250 with gyros will help.

westfw:
Three mechanical tilt switches should be able to detect any orientation.

You can probably find an accelerometer that will detect orientation change without needing any vibration to wake it up...

I will look more into this.

MorganS:
Tough one. I think you have one of the best sensors for this task. Maybe you can find one with "tap detection" to improve/modify the sensitivity.

Is it something that normally sits on a flat surface like a desk? Then it is possible to spin it on the desk and that will make no detectable accelerations which that sensor could use. I would try to constrain the way that it is moved, for example, put a handle on it in a way that you automatically rotate it when you pick it up.

A D cell is a very big battery. You probably don't need to get your power consumption down to nano-amps. Maybe a MPU6050 or MPU9250 with gyros will help.

No, the devices are to be mounted onto a waste container (either side, the back, or on the outside somewhere). The way it is programmed now is that it can be mounted in any possible position, as the device collects historical data on its orientation so it "knows" what orientation its something to be in while in resting state. Maybe I can find a way to add something to the device that amplifies the vibrations or something...

Thanks for your suggestion, I will have a look...nano-amps indeed are not needed because the Particle Electron itself already has a 130uA consumption rate in Deep Sleep.

So your device has one flat side that's going to be magnetically stuck or glued to a flat surface which is usually vertical? Then you have some chance of setting up the orientation of the sensor to make sure no one axis is parallel to the rotation you want to detect.

I find it hard to believe that a sensor originally intended for a phone can't detect a waste bin being disturbed. There must be more tuning or options available. The ADXL345 is very similar but it may have a different algorithm for detecting 'wake up' events.

Unlike tilt sensors, gyros are active sensors which create vibrations and translate changes in the vibrations into acceleration or angle. Such sensors don't come at zero energy consumption.

For zero energy I could imagine a sensor floating on a gel. In rest the gel will form a horizontal surface, on which a floating tilt sensor is inactive. If the entire thing is tilted or moved, the sensor is activated until the gel returns to a horizontal surface.

And you think it is a good idea to build 50 of those gel baths and mount them to waste bins outdoors? What if the gel freezes?

Just a few tilt switches in different orientations should work. Maybe as few as 2, if you can put "this way up" on it.

MorganS:
And you think it is a good idea to build 50 of those gel baths and mount them to waste bins outdoors? What if the gel freezes?

Just a few tilt switches in different orientations should work. Maybe as few as 2, if you can put "this way up" on it.

I forgot to mention it, but some of the devices will actually be used in up to -40C (Northern Alberta)

Vitesze:
I forgot to mention it, but some of the devices will actually be used in up to -40C (Northern Alberta)

That's outside the rated temperature range for many components and sensors.

Simple tilt switches are the lowest power solution but not accurate (only an on/off output), and as long as they stay dry the temperature should be no problem.