DIY Head Tracker For A Tenner

[brumster]

@brumster
Is this a known problem, and is there likely to be any fix for it?

I thought I remembered someone else reporting this as well...

It's not something reported with any regularity; there very well may have been someone else who reported it (it's been a few years now!) but I'm not aware of it as a "known problem"... first step would be to try and re-create with the latest firmware. I vaguely remember some issues where people set their EDTracker down for an extended period of time and, for one reason or another, the device came close to something magnetic (one user finally realised that his folding headset was closing up and putting the tracker right over the earpieces!). This "tugs" the magnetometer off over a period of time (fnaar fnaar); it thinks "Whoa, I'm drifting big time here, I best ramp up the correct a bit!"... and it gets worse the longer you leave it. Then, when you come back to it and sit it on your head, the magnetic field around the device reverts back to normal and the compensation (that's ramped up over the period) is no longer right, it's way overcorrecting.... and it takes a while to restore back to normal levels. In reality, it's quicker to just re-center once or twice rather than wait.

I'm not sure what we could really do about that. I think we contemplated only applying magnetometer correction while we were sensing a certain degree of movement, but we never tried it/implemented it long term. Rob may have more comments. It's maybe something to look into longer-term but I'm not sure it's that critical; I don't believe it affects a massive number of people.

I do the same as NeilF, keep it upright.
I flashed my 9150 with EDTracker2_9150 4.0.4 , no problems, shouldn't Lordshadowfax try that and why is the temperature in the screencaps showing zero ?

@Brumster.. If I wanted to build a second tracker would a 6150 be a bad idea ( so cheap on Ebay right now ) ??

I'd steer clear of 6050's now; not many people bother with them because of the lack of a magnetometer. As a cheap learning exercise, fair enough, but given the price difference I don't imagine many people really bother

 

[Zplintz]

Aww thanks buddy, it's feedback like that that makes it all worthwhile!

Np at all.... Now about that rares route you promised

On a down side my daughter tried it out while driving an srv around and now she wants one too :facepalm:

 

[wstephenson]

Head tracking VR with Google Cardboard, EDTracker, NVidia and Android

May I draw your attention to this reddit article?

https://www.reddit.com/r/EliteDange..._vr_with_your_phone_how_to_and_some_thoughts/

The cool part is in the middle.

It suggests using an Android app called Moonlight Game Streaming (https://play.google.com/store/apps/details?id=com.limelight&hl=en) to emulate an NVidia Shield tablet.

Together with that, a Google Cardboard, your trusty EDTracker and OpenTrack (for roll) you have 1080p60 HMD VR over wifi.

Run, don't walk!

Caveats: FOV is more limited than with a real HMD. The aspect ratio usable for streaming does not give 1:1 together with ED's side by side graphics setting, so things stretch as you roll the view.

 

[brumster]

Ah, fantastic! Thank you sir I'm sure it will save me some agony, and hopefully others that come late to the project but still want to try a DIY build with the latest components.

Arrgh, the video might have to wait, I came to cut it together and the audio is all screwed up I need to re-record it some time. Sorry!

 

[Junky Juke]

Hello folks. I bought EDtracker and received it acouple of days ago. I have to say it works fine except for a very disturbing slow drift along the yaw axis. I discovered it's a common issue.
May I ask if there is a known solution for this problem? it is preventing me to read correcly the left and right panels in game because it slowly gets out of position.

 

[brumster]

What have you bought? EDTracker Pro? It should not drift - if it does, it's more than likely down to incomplete/incorrect calibration; have you done the mag calibration? Have you gone through the checklist on the website?

http://www.edtracker.co.uk/support/troubleshooting

 

[Junky Juke]

I bought EDtracker Pro and I went through the checklist. I have a 2.0 data/power usb cable. I did 5 minutes calibration and Auto Bias. I did it multiple times and everytime it dirifts on yaw axis.

Ok I guess I found a solution. I just set smoothing to 100 and now there's no drifting at all. It's an odd fact and it's a partial solution.

 

[wstephenson]

You might need to repeat calibration a few times to get a good one. I only got a calibration that didn't yaw when decentred when I calibrated with the device held at head height - the magnetic environment when calibrating in front of myself whilst seated would appear to be different enough from that near the top of my head to influence stability.

 

[sewerratuk]

I bought EDtracker Pro and I went through the checklist. I have a 2.0 data/power usb cable. I did 5 minutes calibration and Auto Bias. I did it multiple times and everytime it dirifts on yaw axis.

Ok I guess I found a solution. I just set smoothing to 100 and now there's no drifting at all. It's an odd fact and it's a partial solution.

I had this problem with pitch.
I just kept persevering with re-calibration and now have no problems with it

 

[Junky Juke]

You might need to repeat calibration a few times to get a good one. I only got a calibration that didn't yaw when decentred when I calibrated with the device held at head height - the magnetic environment when calibrating in front of myself whilst seated would appear to be different enough from that near the top of my head to influence stability.

Yes I tried also that way. No luck.
And tried without putting it on the headset. Still no luck.
My post-calibration graph is now a nearly perfect sphere with no spikes/anomalies and still I have the drift (again it disappears when I set smoothing at 99).
I think I'll have to stick with my compromise.

 

[Szyslak22]

Arrgh, the video might have to wait, I came to cut it together and the audio is all screwed up I need to re-record it some time. Sorry!

No worries brumster! Thanks for the attempt.

 

[brumster]

Yes I tried also that way. No luck.
And tried without putting it on the headset. Still no luck.
My post-calibration graph is now a nearly perfect sphere with no spikes/anomalies and still I have the drift (again it disappears when I set smoothing at 99).
I think I'll have to stick with my compromise.

That is very odd. In cases like this, what we normally ask people via email is to put the GUI into DEV mode (About->Mode->DEV) so that all the additional technical data is shown on the right hand side of the GUI. Then, open the magnetometer tab and re-do the mag calibration just so we can visualise the green/red spheres on-screen. Then put the device down flat on the desk (away from headset/speakers, etc), leave it still, click the auto-bias if you haven't done it once already and when complete screenshot the entire window and send it in.

The motion and gyro numbers should all be zero, with maybe the odd "blip" to +/- 1 or 2 if things move a bit.

Accel Z should be about 1G which should be ~-16387, the other accel numbers should hover around zero but there will be quite a lot of fluctuation because the device is sensitive - so +/- 40 is quite normal. If the device is definitely flat and still, and the first 2 accel numbers are way off (say +/- 400 or more) then you can check the "Force Accel Bias" checkbox and re-do the auto bias button, and it should fix it. This is a very rare occurrence where some of the MPU-6050 chips from the factory come with ridiculously-set accelerometer bias values. We can't understand why and presume it's an Invensense calibration issue, but I hasten to add this is very rare... like 1 in 200 MPUs maybe?

The Mag numbers are somewhat arbitrary and dependent on your environment, but they should at least be fairly stable (again, they will fluctuate by +/-10 or less).

The Y/P/R values should be pretty much rock solid at 180/180/180, they are the resultant output from the previous values so any change here is most likely related to odd readings on the accelerometer, gyro or magnetometer.

 

[Junky Juke]

That is very odd. In cases like this, what we normally ask people via email is to put the GUI into DEV mode (About->Mode->DEV) so that all the additional technical data is shown on the right hand side of the GUI. Then, open the magnetometer tab and re-do the mag calibration just so we can visualise the green/red spheres on-screen. Then put the device down flat on the desk (away from headset/speakers, etc), leave it still, click the auto-bias if you haven't done it once already and when complete screenshot the entire window and send it in.

The motion and gyro numbers should all be zero, with maybe the odd "blip" to +/- 1 or 2 if things move a bit.

Accel Z should be about 1G which should be ~-16387, the other accel numbers should hover around zero but there will be quite a lot of fluctuation because the device is sensitive - so +/- 40 is quite normal. If the device is definitely flat and still, and the first 2 accel numbers are way off (say +/- 400 or more) then you can check the "Force Accel Bias" checkbox and re-do the auto bias button, and it should fix it. This is a very rare occurrence where some of the MPU-6050 chips from the factory come with ridiculously-set accelerometer bias values. We can't understand why and presume it's an Invensense calibration issue, but I hasten to add this is very rare... like 1 in 200 MPUs maybe?

The Mag numbers are somewhat arbitrary and dependent on your environment, but they should at least be fairly stable (again, they will fluctuate by +/-10 or less).

The Y/P/R values should be pretty much rock solid at 180/180/180, they are the resultant output from the previous values so any change here is most likely related to odd readings on the accelerometer, gyro or magnetometer.

I will check all the data you kindly posted here and let you know if there are anomalies. Thank you.

EDIT: ok, I did a new calibration and this is the graph:


Then I did the Auto Gyro Bias and this is a screen of right window's parameters:


Accel first two parameters were higher than 400 so I applyed the force accel bias and did auto bias again:


Drifting still there.

This are screens of post calibration without headset.

I'll post a screen of post calibration/autobias (with force accel bias checked) graph with the device mounted on my headset:

 

[stephen_usher]

I've had a report on my blog post that EDTracker + Opentrack 2.2 (stable) has stopped working after yesterday's Elite: Dangerous update.

Now, my system seems fine, but has anyone else found their system stop working?

 

[brumster]

EDIT: ok, I did a new calibration and this is the graph:

snip

Those honestly look fine, so I can't figure out how it's drifting. What I think is maybe best is you drop us a line via your order email and arrange return and replacement. We can send you another one out, and I'll get that one back and test it to see if I can figure out what's going on with it. Naturally we will cover postage/etc.

With it lying still, if you watch the numbers over a short time do they "jump about" at all - the accel/gyro values - maybe they are generally stable (+/-40 or so, let's say) but then every now and then there's a spike and it throws it off?

 

[Junky Juke]

Those honestly look fine, so I can't figure out how it's drifting. What I think is maybe best is you drop us a line via your order email and arrange return and replacement. We can send you another one out, and I'll get that one back and test it to see if I can figure out what's going on with it. Naturally we will cover postage/etc.

With it lying still, if you watch the numbers over a short time do they "jump about" at all - the accel/gyro values - maybe they are generally stable (+/-40 or so, let's say) but then every now and then there's a spike and it throws it off?

Yes there are spikes of say +/-50 up to 80 when I keep it still on my desk, but only if I rotate the device out of zero (center) position (and keep it still). The wider is the angle the higher are the spikes. This happens only on the accel values.

 

[brumster]

Cool, ok. Well, the accelerometer should not be spiking like that, irrespective of the orientation of the device, if it's definitely still. It's sensitive - so it will even pick up vibrations in a desk for example - but it sounds to me like something is wrong with the accelerometer aspect of the MPU. It would certainly explain why it's yawing off. Drop Becky an email

 

[Noam]

Hmm

I bought the kit from hobby components and it came with the MPU 9250

It has a 10 pin as opposed to an 8 pin, so if I follow the guide at edtracker.org (specifically Bartybees_Bare_EDTracker_BuildGuide.pdf) I get a bit lost.

If I use the 10 pin, I can't use solder the button in. I read that the button isn't really needed, but how do I wire it without the button? Do I just leave out the button wires?

EDIT: Screw it ordered the pre-etched PCB, want to support the makers

 

[Junky Juke]

Cool, ok. Well, the accelerometer should not be spiking like that, irrespective of the orientation of the device, if it's definitely still. It's sensitive - so it will even pick up vibrations in a desk for example - but it sounds to me like something is wrong with the accelerometer aspect of the MPU. It would certainly explain why it's yawing off. Drop Becky an email

you are very kind to offer me a replacement of the device. But I do not trust italian mail enough to risk to lose it.
As I said some post ago, I found a compromise. Infact setting up "smoothing" at 99 the drift effect becomes very very low for some reason and it works perfectly with Elite Dangerous in direct mode (I mean without OpenTrack). So I'm a bit puzzled if to take the risk the device will never arrive at destination or keep it as it is. I will need some time to take a decision.

 

[brumster]

Hmm

I bought the kit from hobby components and it came with the MPU 9250

It has a 10 pin as opposed to an 8 pin, so if I follow the guide at edtracker.org (specifically Bartybees_Bare_EDTracker_BuildGuide.pdf) I get a bit lost.

If I use the 10 pin, I can't use solder the button in. I read that the button isn't really needed, but how do I wire it without the button? Do I just leave out the button wires?

EDIT: Screw it ordered the pre-etched PCB, want to support the makers

You just leave it off, don't wire up the pins

There is a note here about it :
http://www.edtracker.org.uk/index.php/11-news/72-using-the-mpu-9250

We have a revised PCB in the pipeline just to fit the 9250 easier but it'll be a month or so before it's out of the factory yet.

you are very kind to offer me a replacement of the device. But I do not trust italian mail enough to risk to lose it.
As I said some post ago, I found a compromise. Infact setting up "smoothing" at 99 the drift effect becomes very very low for some reason and it works perfectly with Elite Dangerous in direct mode (I mean without OpenTrack). So I'm a bit puzzled if to take the risk the device will never arrive at destination or keep it as it is. I will need some time to take a decision.

That's fine, honestly. Well, just let us know if there's anything we can do .