Is your Rift's view tilted / slanted? Here's how you fix it

[Articate]

I am writing this thread in response to my experience with Oculus support. Oculus support has an extremely quick first line support, however they did not have the tools to deal with this issue, and the second tier seem to go through this dance of "please provide pictures of the issue", which seems like a waste of time. I am writing this thread both to let Oculus know their routines on this issue should be improved, and to help those who anxiously await the support they want.

Fear not! Most of the time, the Rift isn't faulty!

There is a calibration tool for the IMU (Inertia Measuring Unit - it's the part that measures how the Rift is rotated, tilted etc) that I used with great success. I want to link to the guide I used, and point out some important parts about this:

1) YOU'RE DOING THIS AT YOUR OWN RISK
This is important. It seems like Oculus do want some sort of control of who uses the tool since it's not that easy to get. It would be pretty easy for the Oculus team to pack this into a nice GUI with steps of how to perform a recalibration, so since they haven't done that, it could indicate that in some cases, this might not be the right thing to do. Without the support team saying "yes, you can run this tool", you should not assume that this will fix your problem, nor be certain that this won't make things worse. At worst, it can be harmful to use this tool, so be cautious when using it, and do this at your own risk

2) This is specifically for issues where the horizon line is slanted or tilted. I think the best way to notice this is just to put on the HMD, pull up a grid view (like the universal menu). If you notice it here, it most likely is there. This rules out the surface you put it down on being slanted in itself, etc. This is never due to a slanted sensor, as the sensor knows what is up and down.

It is very important that the calibration is run when the Rift has been completely disconnected for hours and hours. Leave it alone overnight. This is because the tool calibrates the IMU at all temperatures, since temperature affects the IMU. It runs for ten minutes so it can go from cold to hot and calibrate all the steps along the way. It is also important that it is placed on the most stable surface you can find.

Follow all the steps and use this guide, you can download the tool there:
http://www.vrheads.com/how-fix-oculus-rift-calibrations-problems

This fixed my problem. I hope this can help others. Remember that this might be the right cause of action. Contact Oculus support if you are experiencing this issue, perhaps even say that you do intend to use the calibration tool, and ask if it's the right thing to do. I hope this post can make Oculus' response to this issue be more efficient, as it is such a well-known issue that there should be a standardized response.

[Scope666]

cybereality said:

Yeah, anything is possible. At this point the best thing would be to create a support ticket. This will help us track the reports coming in better, and maybe see a connection if we can get multiple people with the same issue. 
https://support.oculus.com/

I opened a support ticket when I first noticed the issue (the day the Rift arrived, ticket number 338629)

I've commented in this thread and the others to try and get a sense of how many others have this problem, and give it some additional visibility.     

[flexy123]

I just had it bad while checking out the (otherwise fantastic!) "Mission ISS".
I just came up with an interesting thought:

Do people who experience the slanted floor have a non-symmetric sensor setup, with one sensor possible higher than the other? And how many sensors do you have?

My setup is two sensors with one sensor approx 0.5-0.6ft higher up than the other. Just a shot in the dark here, but maybe there is a connection. (Darn I am dizzy now after this and the slanted floor didn't help with that either...)

I am trying to reproduce this, maybe the ISS experience is a good way to do it.

[kojack]

There's only two ways in current VR to determine floor position: steam has you put the hmd on the floor, oculus has you enter a player height then stand still while the sensors do that animated sweep thingy. The steam way is probably less error prone, but oculus sensors have only 70 degrees verticle fov compared to lighthouse 120 degrees, so seeing the floor as well as reaching up like in The Climb is less likely.

Scope666 said:

I have the tilted horizon issue with 1.12, and re-calibrating the IMU does not help it at all.  From my understanding the IMU is more about acceleration and tracking movements and not so much determining what level is.

An IMU has three parts: accelerometer, gyroscope and magnetometer. (and temperature sensor, so really 4).
The accelerometer tracks acceleration. This could be due to movement (moving the controller from one place to another involves accelerating and decelerating). But gravity is also a constant acceleration of 1G (actually varies around the world due to the composition and thickness of the earth below you). When the hmd is left on a stable surface, it doesn't read 0 acceleration. It will read 1G in some direction. The direction of acceleration is used to determine the level direction.

The gyroscope measures angular acceleration, which is acceleration and deceleration of turning. But it has no reference point, it doesn't understand how the hmd is facing in world space. It only understands changes in facing. You still need the accelerometer to determine gravity direction.

But gravity can only tell you pitch (looking up and down) and roll (tilting side to side). You can't tell yaw (turning around while looking level) from an accelerometer. If you know the initial direction (like the direction you face when starting a game), then the gyroscope can track changes from that. But these values will drift over time (so your front will slowly become wrong).

The magnetometer tries to correct that. It measures magnetic fields and used those to determine which way you are facing. A bit like a compass. It's not accurate enough to use for much, but it can help counteract drift.

The IMU is read hundreds of times per second, far more often than the frame rate of the rift. This lets us do things like timewarp, which uses multiple IMU reads per frame.


That's the hmd (and how the DK1 worked). Now for sensors (DK2 and CV1).
The sensors are webcams. They look at the pattern of IR lights on the HMD and calculate where the HMD is relative to the camera, and which way it's facing (also relative to the camera). The sensors have no ability to track their own position or tilt. Instead they look at the HMD and compare how they think the HMD is tilted to how the HMD's accelerometer says it's really tilted:
sensor: "I see the HMD is tilted 20 degrees to the left"
hmd: "I'm tilted 5 degrees to the left"
sensor: "oh, that means I must be tilted 15 degrees to the right!"


Then sensor fusion happens. This combines the high speed IMU values (like 500 times per second, but can drift) with the slow speed (60 times per second) values from the sensor. The sensor values can't drift, but are too slow to use on their own. Mixing both, we get stable tracking without drift.

The touch controllers also have IMUs and do sensor tracking with sensor fusion. If you cover up a touch controller (hide it from the sensors), it will continue to work fine for a second or so, the IMU can track movement for about that long before it becomes too inaccurate (accumulated errors). Then the touch will stop moving in the game, but it will continue to rotate, because rotation is easier to track with IMUs than position. You'll get drift again, like a DK1 hmd. It will return to normal when the sensors can see the touch again.

[Scope666]

Perhaps there are bugs in 1.12's "fusion" code as a LOT of people have this issue, and it's almost always the left side lower.  I've done the IMU calibration numerous times and the slant always remains exactly the same.  The really odd one is before my Rift arrived, I noticed the slant in the menus of Robo Recall using Revive, and that's a completely different headset.  That leads me to believe it's a code problem and not the HMD itself.

[kojack]

Hmm, if revive is doing it too (which means different hmd and tracking system), then yeah, that sounds like a software issue.

[Anonymous]

I think this tool won't work anymore with current version of Oculus software installed.

Support also won't send it anymore:

We are no longer distributing the IMU calibration tool as it is no longer an effective solution since we released the display optimization a while back.

[Adiel79]

Hi all,

I have a question for people in the know. 

I bought my CV1 from Amazon. It shipped from the US to my country in the Mediterranean.

I've had many issues with my system, including this one, and this one -

https://forums.oculus.com/community/discussion/51600/sudden-tracking-problem-all-games-unplayable-lo...

Could it be that my headset is calibrated for the US and is having issues due to the temps in my region or something of the sort?

[cybereality]

Yes, the way the internal IMU sensors are calibrated can be affected by the location, for example weather conditions like the temperature, etc. We have a recalibration tool that can help resolve this in some cases, but it's not a silver bullet. But it's worth contacting support to see if it will help.

[Anonymous]

Owned a Rift for a few weeks, and just got this issue yesterday. I tried the IMU tool multiple times to no avail, and like Scope 666 noticed that the IPD adjustment screen with the "+" is perfectly level every time I try and recalibrate the device, suggesting it is a software issue.

One thing I did notice is that it's always reported that the tilt is to the left. I looked at my Rift, laid upside down, and noticed it tilting right (i.e. to the left of the device if it were right-way up). I then realised that this is because the HMD's cable is on the left side.

Before yesterday, whenever my Rift was not in use, I would pack it away. Yesterday was the first time I left it out, and I think it's at least partially something to do with the weight of the cable tilting the device left, and that affecting it when you subsequently turn it on.

I tried weighing down the left side of the Rift with a Touch controller while re-running the IMU with this in mind, but that didn't resolve the issue either. Still, I don't think it's a coincidence that all these reports have the tilt being to the left to various degrees (the extent of the tilt changed for me from time to time in the past 24 hours).

Does anyone else feel like I'm on to something here? 

[phillabust]

just got my rift this week. everything worked fine the first night, didn't play the second night. the third night went to play again and the whole world was tilted to the left.. to the east. when i turn around it is relatively tilted to my right. (still to the east) i was really hoping to have some VR experiences, but with the world tilted to the side... i'd rather not risk VR sickness. it's a noticable amount of tilt, feels close to 10 degrees or more.