Calibration Issues

[TiagoTiago]

Then rinse and repeat till both parts work right at the same time?

[Danny]

We did apply a "heading correction" for the hydra in the Driver. I've removed the correction in THIS VERSION OF THE SDK. It would be interesting to see what your results look like with this version.

[MrMormon]

Hmm... Could you explain more? Is the heading correction meant to fix a firmware bug or overcome a hardware limitation?

[Danny]

It was a simplified attempt to do what you guys are talking about. I don't believe it was a bug in the firmware but because of how magnetic fields work. The Hydra has great precision and if you put the controller in a specific position, the reported position and orientation are very repeatable, though they may not be "accurate". Doing what you guys are talking about here would increase that accuracy, but due to the fact that the environment is different per user it'd likely need to be calibrated by each user to get good results.

We noticed that the orientation would toe out as you moved further from the base along the x-axis so we adjusted the the yaw by 15 degrees by the time the controller had reached a foot away form the base along the x-axis in either direction.

[TiagoTiago]

It was a simplified attempt to do what you guys are talking about. I don't believe it was a bug in the firmware but because of how magnetic fields work. The Hydra has great precision and if you put the controller in a specific position, the reported position and orientation are very repeatable, though they may not be "accurate". Doing what you guys are talking about here would increase that accuracy, but due to the fact that the environment is different per user it'd likely need to be calibrated by each user to get good results.

We noticed that the orientation would toe out as you moved further from the base along the x-axis so we adjusted the the yaw by 15 degrees by the time the controller had reached a foot away form the base along the x-axis in either direction.

Was that deviation consistent across units from different batches?

[MrMormon]

I don't believe it was a bug in the firmware but because of how magnetic fields work. The Hydra has great precision and if you put the controller in a specific position, the reported position and orientation are very repeatable, though they may not be "accurate"..

So how raw is the data in the modified SDK? I mean, if we want to get the most accuracy possible, we should know what is technically being measured, have SDK access to that data, and attempt the theoretically correct correction, right? Are the three coils in this picture measuring the strength of perpendicular components of a magnetic field at their respective distances from the base? And what type of field does the base generate? A point field wouldn't give anything more than radial distance. For more accuracy, don't assume that the coils measure the field's strength exactly at their centers. Etc. Also, I have no idea what the large cylinder (magnet?) (removed from the remote in the picture) does.

due to the fact that the environment is different per user it'd likely need to be calibrated by each user to get good results.

Shouldn't that be done anyway?

[Danny]

The deviation was consistent.

The data in the SDK I linked to should be raw data.

The base station produces an AC electromagnetic field.

The large cylinder is a 25g weight.

[MrMormon]

Any progress? Because basically, right now, the Hydra doesn't work right. You realize that, right, Sixense?

[MrMormon]

Still waiting. What is(are) Sixense's opinion(s) on the fact that the data is distorted? Because this makes the Hydra's rotation information rather useless for my musical application, at least.

[Danny]

I'm actually waiting to hear of your results with the updated SDK I provided. The data should be more reliable with that SDK. You will still see bending but the optimization we made for the Hydra was removed. Someone could attempt to "calibrate" the Hydra. My understanding is that is difficult to do as both position and rotation affect each within the field.