This is not good enough because you can't detect when the glasses are upsidedown. Euler angles require 2 of the angles to be -180 to 180, and 1 of the angles to be -90 to 90.

Also, the SDK documentation is wrong. It says pitch and roll only go up to +/-70, which is false. They go up to +/-90. Which is the correct behaviour for pitch, but the incorrect behaviour for roll.

The hardware is perfectly capable of supplying the full range of roll values without modification to the firmware. It just needs to use the atan2 function on the horizontal and vertical accelerometer values.

GlovePIE can read the full roll with VR920.StrictRoll or VR920.SmoothRoll if you use the menu to calibrate the VR920 accelerometers.

The accelerometers can only detect tilt. They can't detect rotation around the axis of gravity (yaw). On the other hand, the magnetometers can't detect rotation around the axis of magnetic north.

By the way, the magnetic north axis is not horizontal. Where I live it is almost straight up, but tilted slightly to the north. But it is enough to allow it to track yaw.

If it did contain gyros, they would not be the same as the gyroscopes in an aircraft. They would be minature, solid state, rate gyros. The SIXAXIS contains a gyro, but I can't hear it.

Gyros can only measure rotation velocity. Which means that by themselves, there is no way of knowing which way you are facing other than by keeping track of how much you have turned. It is unreliable, so it drifts. Turning 90 degrees then turning 90 degrees back won't leave you facing exactly the same way you were before. The Trimersion only uses rate gyros, but since it only emulates mouse-look it doesn't matter. The Z800 combines rate gyros with accelerometers and magnetometers, so it can get more accurate tracking.

Anyway, accelerometers are perfectly capable of detecting when the glasses are upsidedown (because the y accelerometer will go negative), but that information is not returned by the SDK anywhere.