DIY Head Tracker For A Tenner

[brumster]

Not that I'm aware of, no, it just tells the DMP to adjust the orientation. Those dots are the raw gyro/accel values, so by twisting it onto the side, the gyro dot will indeed go on the squiff - but provided the head moves correctly, it doesn't matter.

Rob (pocketmoon) mounts his on the side I think? I never have myself.

 

[pocketmoon]

Are there any disadvantages to side mounting the tracker ?

I used to play with a side mounted tracker and it was fine.

The important thing for low drift is for the gyros to be reading zero so what I do is:

With tracker flat/horizontal load calibration sketch. wait for warm up. do a few calibrations until gryos are zero and accell bias are close to zero.

Then rotate the tracker into it's desired position (usb down) and then ONLY USING the +/- buttons adjust the Gyro bias values so they are still zero.

The load the main sketch and set orientation to usb down etc and off I go

 

[Tar Stone]

I used to play with a side mounted tracker and it was fine.

The important thing for low drift is for the gyros to be reading zero so what I do is:

With tracker flat/horizontal load calibration sketch. wait for warm up. do a few calibrations until gryos are zero and accell bias are close to zero.

Then rotate the tracker into it's desired position (usb down) and then ONLY USING the +/- buttons adjust the Gyro bias values so they are still zero.

The load the main sketch and set orientation to usb down etc and off I go

THANK YOU
Can we put this in the FAQ/manual

 

[JupiterJoe]

Combatting Drift.

Possible method #1: Programmed so the drift compensation is variable to match real-time temp readings.

Possible method #2: Automatic reset of the device every 30 sec or so (Currently I have to manually press reset every couple of mins, which ruins immersion quite frankly). Programming it for 30 sec intervals may occasionally slightly interrupt targetting , but 99% of the time we're just flying around.

 

[brumster]

Possible method #2: Automatic reset of the device every 30 sec or so (Currently I have to manually press reset every couple of mins, which ruins immersion quite frankly). Programming it for 30 sec intervals may occasionally slightly interrupt targetting , but 99% of the time we're just flying around.

You couldn't do this; you just can't assume that people's heads will be facing perfectly straight every 30 seconds

I'm really surprised that you're getting drift that bad, the temperature must be fairly variable then, assuming the calibration is correct. Maybe try insulating it in some way - you could go the enclosure route but maybe even a simple bit of 'wrapping' with some non-conductive material (some fabric, some sponge foam, etc) could get the MPU on the top up to a stable temperature quicker.... worth a try maybe?

 

[JupiterJoe]

You couldn't do this; you just can't assume that people's heads will be facing perfectly straight every 30 seconds

Of course you're right. That was silly of me.

However, drift is really that bad for me and I am asking for help. During calibration (with the device sitting level on a table) it takes several clicks of 'Calculating Bias Values' to get the orange dot central. It always waivers slightly. I'm wondering if this is the root cause of my problems.

If not, I do live in a old house, that's a bit drafty. A modern, well insulated home, with double-glazing etc, is not going to have the temp fluctuations that I'm dealing with. It seems that this, combined with placing the device on top of my head (my head radiating heat of course), combined with the device warming up as it used is enough to throw it off.

 

[brumster]

Andy Edgson enclosure

Can I just say a big thankyou to Andy Edgson who has designed an enclosure for the EDTracker, and was kind enough to send me one - looks awesome in transparent plastic...! I believe he is open to the idea of printing them for people, for a small fee (which is only fair)....

 

[brumster]

However, drift is really that bad for me and I am asking for help. During calibration (with the device sitting level on a table) it takes several clicks of 'Calculating Bias Values' to get the orange dot central. It always waivers slightly. I'm wondering if this is the root cause of my problems.

Drop me a PM with some email details and if you're open to a desktop sharing session via Google Hangouts, I'll take a live look with you and see what's happening. Some times there's nothing that beats "seeing it live".

Did you buy it from one of us or DIY build? Other option is to post it off and I can check it.

Drop me a PM, or email dan@edtracker.org.uk.

 

[OillyRag]

I'm sure you guy's already have this stuff but just in case I've had a hunt around and i think I've found the motiondriver 6.1 library from invensense. this may help to get data from the compass via the DMP:

https://mbed.org/users/oprospero/code/MotionDriver_6_1/

also I've found a Library by a now defunkt company called 'Pansenti' that claims to allow us to obtain 9 axis fused data from the 9150 using an arduino:

https://github.com/richards-tech/MPU9150Lib

It might be of some use it might not

 

[AndyEdgson]

Can I just say a big thankyou to Andy Edgson who has designed an enclosure for the EDTracker, and was kind enough to send me one - looks awesome in transparent plastic...! I believe he is open to the idea of printing them for people, for a small fee (which is only fair)....

Glad you like it!
I can email out STL files if anyone wants them - or I can print them out - takes about an hour of my time (and the missus has to listen to the 3d printer )


£5 a pop plus postage at £3.50.

Colours available so far are opaque (as modelled by brumster) white, black, pink, blue, red , silver (well greyish), luminous (sort of), buttercream and lime.

 

[brumster]

Just for the record....

I'm sure you guy's already have this stuff but just in case I've had a hunt around and i think I've found the motiondriver 6.1 library from invensense. this may help to get data from the compass via the DMP:

https://mbed.org/users/oprospero/code/MotionDriver_6_1/

The hex firmware for the MPU in that is exactly the same as what we're using today. I did a byte compare; it's identical. As far as I'm aware, it's 5.1, not 6.1 - I think the author is maybe incorrect in his labelling (or at least the DMP firmware image is no different).

also I've found a Library by a now defunkt company called 'Pansenti' that claims to allow us to obtain 9 axis fused data from the 9150 using an arduino:

https://github.com/richards-tech/MPU9150Lib

As above - it's the same DMP firmware. The DMP does nothing clever for us on the mag side - they are just fusing the data manually outside of the DMP, much as we will have to in the 9150 version. Could be useful to examine their code though, maybe borrow some of it's principles!

There's a LOT of sensor fusion approaches out there, so should be fairly easy to set up something.

Unfortunately, expecting the Invensense MPU to take some of the load off and do it for us like it can with the accel/gyro sensors - pffft, seems it's not going to happen

 

[OillyRag]

Unfortunately, expecting the Invensense MPU to take some of the load off and do it for us like it can with the accel/gyro sensors - pffft, seems it's not going to happen

Ah well, I guess I best take off my rose coloured glasses and go back to scratching my head.

 

[brumster]

Ah well, I guess I best take off my rose coloured glasses and go back to scratching my head.

<laughs> I suspect one day in the future, when sales have dropped and the buzz of the 9150 has calmed down, Invensense might be a little more open with the capabilities of the DMP and how to program it. One big problem, of course, is having to load the firmware image/config into it every time you power it up, which means it's 3k of data space we're losing in the AVR by having to store it in there

I sometimes wonder whether dumping the DMP approach completely and moving to simple AVR-based fusion is the better way to go :-/ but for head tracking, the DMP seems perfectly adequate....

Keep the ideas coming sir, they're all valid and good to hear!

 

[Gimi]

Glad you like it!
I can email out STL files if anyone wants them - or I can print them out - takes about an hour of my time (and the missus has to listen to the 3d printer )


£5 a pop plus postage at £3.50.

Colours available so far are opaque (as modelled by brumster) white, black, pink, blue, red , silver (well greyish), luminous (sort of), buttercream and lime.

Would very much like the STL file. See PM.
Would you consider uploading it to a print on demand site and let people order it from there. Some have a setup where you get a small cut when people order it.

Would also be an idea to make a version without the rubber band hooks. I Velcro mine to the brim of a cap, so I don't need the hooks.

Finally, would be great if the EDTracker site had a page with pictures of the different case solutions. Might inspire people.

 

[n0ahg]

Spotted this http://oddsheepgames.com/?page_id=134 which is similar head tracking to what I was doing with FaceTrackNoIR and an Android phone with the stereoscopic vision added. The streaming works perfectly on my Nexus 5 and the head tracking worked on a quick test with mouse head look but I don't have any lenses to make a headset (like Google cardboard).

 

[nowhere.elysium]

I've just built one of these, and I wanted to say thanks to the OP for their hard work in not only getting the firmware nailed down, but also building a really good flashing/calibration tool, too.

Of course, nothing will stop the mockery that my wife is delivering for having something with flashing lights strapped to the top of my headphones, but there we go.

 

[sthalik]

That fellow's cooperative so far, let's see how it goes.

https://github.com/opentrack/opentrack/issues/48

We're all discussing Android support in this issue.

 

[Munshine]

I know that the goal behind EdTracker was to provide a tracking device at the cheapest cost as possible. But unlike other 6 dof tracking hardwares, Edtracker is only 3 dof.

It was said previously that a Stem technology combined with a gyro would allow true 6 dof.
Any idea when it will be available and what would be its cost compared to a trackir-like device ?

 

[n0ahg]

I know that the goal behind EdTracker was to provide a tracking device at the cheapest cost as possible. But unlike other 6 dof tracking hardwares, Edtracker is only 3 dof.

It was said previously that a Stem technology combined with a gyro would allow true 6 dof.
Any idea when it will be available and what would be its cost compared to a trackir-like device ?

STEM uses A/C electromagnetic field to determine the position and orientation of up to five tracking points so doesn't need a gyroscope or suffer from drift! See http://sixense.com/wireless

It is available for pre-order and hopefully I'll have mine next month.

 

[Munshine]

STEM uses A/C electromagnetic field to determine the position and orientation of up to five tracking points so doesn't need a gyroscope or suffer from drift! See url]http://sixense.com/wireless[/url]

It is available for pre-order and hopefully I'll have mine next month.

Thank you for the link. Heck that's expensive even for head tracking only!