hi!

im trying to build a lcd mt similar to ms thinsight.

this is as far as i am:

http://www.vimeo.com/1413489

i used kingbright L-932p3c
and l-53f3bt and an arduino.

its all just a test to fiddle with the sensors, thus only crappy processing code to visualize.

todo:
- need to find a diffuser layer, i had to remove the original one
- need to find correct distance between the sensors, emitters and also the lcd
- maybe find better components
- try to do do it with a ftir setup
- maybe use daylight filters

does anyone has good ideas
regarding different componets, filters and the diffuser?

only requirement is, that it has to be low-cost…

cheers

Inspiring, brilliant, I was just wondering how Dell and HP are planning to do it as they have some multitouch tablets on the way in a year or so from what I’ve read.  While I was thinking about it I’d thought about this as a possibility.  I’m not sure about materials though.  If I ever have money again, I’d love to try this.  Keep posting, sorry I couldn’t be more help.

How much emitter/receptor can you handle with an arduino ?

i will put in some multiplexers and shift registers.
so thoretically more than enough. im just not sure if the speed is ok

as i want to have only one LED emitting at time and then capture the sourrounding phototransistors.

im still testing and now ordering new components as well as some ics (analog multiplexers and shift registers)

then there will be a test matrix of maybe 8x8 or so.

Keep up good work !

I’d like an idea to make thin MT installation.
I think it is possible to use some sort of combined emmitter\receiver lasers like in CD players. It can be more compact, than those from video. Or may be emitters can be separated from receivers.
How do you think, what count of those pairs do you need to handle touches accurate and not stop backlight? It’s the main problem for me in this setup.
Those receiver/emitter from CD players are 4.4mm in diameter (or separate SMD components, that are 2x3 mm each in sum 4x3 - about the same). So if you put them back to back you can get phisycal resolution ~120x100 on 24” display with little holes for backlight. It than cn be interpolated to a little bit suitable, but still not high reslution.

Great work , can’t wait to see how this project evolves.As a difussion layer use tracing paper , and also the tracking resolution is gonna be kinda low , tough someone on these forums (and also if you search on youtube for multitouch capacitive) talked/used an alghoritm for interpolating the blobs.

When you get your next test up with 8x8 matrix or whatever you are gonna use please post the results , I’m shure the community will be glad to help and take note from your work.

This might be a new method for multitouch input , and it’d be great for portable solutions.

I’ve read an article on ms thinsight, and they used a combined emmitter/ receiver. It’s called a retro-reflective optosensor . They used Avago HSDL-9100 for prototyping. Its seems quite cheap (2$/piece).
This looks like an interesting new technique. I will probably buy some and do some tests myself.

Here’s the link to the article, its quite interesting, and explains a lot about how its made:
thinsight

yeah I am considering getting some from http://www.newark.com/jsp/search/productdetail.jsp?SKU=72J6987&CMP=AFC-OP

It is pretty cheap, if someone could get it working well hardware wise, I would imagine the software would eventually come. They only have 895. It would cost around 150 dollars to have as many as they do for the thinsight.

I already have lasers and camera though.  But having thin multi touch is obviously the future. I wonder how many years until every laptop screen you find is multi touch.

I’m wondering why the thinsight doesn’t use FTIR over the top instead of their current setup.  Wouldn’t it save space and allow more opto sensors to be crammed onto a circuit board?  Certainly it would add thickness to the setup, but it seems possible and might add the resolution they’re looking for.  It also sounds like they’re using software to process the resulting input.  I wonder how hard it would be to use a microprocessor.  Anyone tried either of these things?

Very interesting keep up the good fight!

If you want I can post this on front page and try to direct some traffic towards this problem. I think it will take many of us to unite if we actually want to see our own thinsight developed.

Please nemster private message me and let me know where your research stands.

nuiman - 23 November 2008 12:42 AM

Very interesting keep up the good fight!

If you want I can post this on front page and try to direct some traffic towards this problem. I think it will take many of us to unite if we actually want to see our own thinsight developed.

Please nemster private message me and let me know where your research stands.

Sweet!!!! I would love to see more people try to work this issue out, and nemster, I would love to see your results if you don’t mind posting.  I’m planning on doing this soon, hopefully with my tax rebate.  However, my background in electronics is entirely self-taught so I don’t know how hard this will be.  But if we all share, we can get there.

Shouldn’t this be possible with an FTIR setup and some photoresistor that only see infrared? So when you touch your finger to the acrylic/compliant surface it reflects the light down, which is picked up by the photoresistors. This could be interpreted and interpolated with an Arduino setup and processing. It would cheap and if the soldering of the photoresistors was done well, you wouldn’t have too much occlusion in the backlight.

fuz3 - 23 November 2008 11:20 PM

Shouldn’t this be possible with an FTIR setup and some photoresistor that only see infrared? So when you touch your finger to the acrylic/compliant surface it reflects the light down, which is picked up by the photoresistors. This could be interpreted and interpolated with an Arduino setup and processing. It would cheap and if the soldering of the photoresistors was done well, you wouldn’t have too much occlusion in the backlight.

That’s about how I imagined it.  It might be a little complex, but I wondered if it were possible to pre-process the video input using a separate arduino or other microprocessor as well, thus relieving the computer’s CPU and hopefully speeding things up a bit.  That may require quite a bit of processing power though.

Also, while on the topic of trying new things....without having searched the forum on this, I understand that fiduciaries are not possible with FTIR. However, what if one were to light the fiducial patterns using an IR LED, sort of like an electronic puck?  Better still, light an array of IR LEDs to make a pattern.  Anyone have anything on that?  Not to take away from this topic, since I am most interested in the possibilities provided by this approach.

just wanted to let you know that im still thinking and working on this.

but im pretty busy at university, so it will have to wait until end of jan.

what i started this weekend was trying to read out pixeldata of an adns-2610 (18x18 pixel sensor of an optical mouse)
same idea but less data and more information. would need an lens tough.

also i have a setup with 6 ir fotodiodes > multiplexer > op amp running.
fotodiodes need an op amp and are way less responsive than fototransistors. this is actually a huge drawback. but they have a daylight filter integrated.

measurements where not that satisfying but i havent completed the research.

what i would need and where everyone could contribute is the area of algorithms to

- calibrate a matrix (nullyfy all sensor) (should be trivial)
- calculate a position from 4 calibrated sensor OR
- having a full matrix

my further agenda:
- experiment with the mouse sensor and see what could be done with that. main reason would be speed and shifting position computing into an uc.
- make 2 setups:  one with fototransistors and one with fotodiodes and compare them.

dunno if we succeed fast enough, i guess we will have some mt screens to buy soon.
also the ir-matrix setup does not really scale of course.

cheers
n