T-Mobile G1 multi-touch demonstrated in video

[Ryan Gardner]

Originally Posted by Seraphic

I'm still not sold by the video. It doesn't look that different to capacative averaging that works on non-multitouch displays. To be honest, if you see how unresponsive certains parts are, I'd say it shows that its not properly supported at hardware level.

The reason it is not smooth is because of how this is hacked together. The most recent touchscreen event buffer is being written to /dev/tsout on my device. I have a thread in the java program constantly polling that looking for data - but the likelihood of the thread always polling it at the right time to get the touch data is pretty low - and it probably drops most of the touch events.

Regarding the hardware level... That was already proven unequivocally by the debug logs which were posted last week - the hardware without any doubts supports and does track multiple fingers.

If anything, the fact that such a crude and rough hack works at all shows a lot of potential for what can be done once the proper support is put in place.

Also, please refrain from making up terms like "capacitive averaging" in an attempt to justify your skepticism.

[Ryan Gardner]

Originally Posted by inphu510n

Yep. Either the program isn't optimized or the hardware isn't quite fast enough to track both fingers accurately. Looking at how it traced, it is possible to implement something fairly reliable for pinch/expand actions. Though, I wouldn't count on this making it into any functionality released by Google for this device.

It's most certainly that the program isn't optimized. Not only is the program not optimized, the method of grabbing the data itself is a complete hack. I'm really busy with my day job this week, but later on I will post the source code and document everything - I will put at least one basic optimization in place (currently, it is regexing every time it polls that device - which is unnecessary since I can spit the data out in a format that i can just use .split(" ") and get the same result - but I was lazy at 2:00AM and didn't want to recompile my kernel for my phone then)

It's unlikely that Google / T-mobile will ever push out an update to this device that includes official multitouch support, but it's entirely reasonable that multitouch support can be built into the android platform and that such support can be enabled very easily by developers for the G1 with a custom firmware.

The question of how to best implement support for multitouch in the android platform must be addressed - if Google doesn't have a multitouch device in the pipepline it might not be a top priority for them, but as a community we can work with them to get support properly put into the platform so that any future official device will be easy to implement, and any G1 owners who want to run custom builds could get multitouch support as soon as the stuff is in place.

As far as patents and such are concerned, let the lawyers deal with that.

[Jorsher]

Well, I was going to back you up Ryan, even though the information you posted was clearly stated in the complete article, but you handled it with grace

[EclipsedExistence]

From what I've heard, Apple has patents on multi-touch, which makes Google hesitant to release multi-touch. I don't recall where I read this, but it was somewhere on the net, on one of the tech sites.

[DesignDawg]

Originally Posted by Ryan Gardner

Also, please refrain from making up terms like "capacitive averaging" in an attempt to justify your skepticism.

Literally LOL at that line. Hilarious. I like you, Ryan. Welcome to AC.

[Seraphic]

Originally Posted by Ryan Gardner

The reason it is not smooth is because of how this is hacked together. The most recent touchscreen event buffer is being written to /dev/tsout on my device. I have a thread in the java program constantly polling that looking for data - but the likelihood of the thread always polling it at the right time to get the touch data is pretty low - and it probably drops most of the touch events.

Regarding the hardware level... That was already proven unequivocally by the debug logs which were posted last week - the hardware without any doubts supports and does track multiple fingers.

If anything, the fact that such a crude and rough hack works at all shows a lot of potential for what can be done once the proper support is put in place.

Also, please refrain from making up terms like "capacitive averaging" in an attempt to justify your skepticism.

I say what I see. I said I'm not sold on the idea; I didn't say it proved either way. Yes, the program was hacked together and it's a good example of what could be done, and I hope the hardware supports true multitouch - I'd love my G1 to support it but until I see a working, practical example I'll remain skeptical.

Terms such as "capacitive averaging" aren't made up at all, anyone who works with touchscreen technology will tell you this - there's no exact name for it, it's what we call it at Panasonic. Almost all touchscreens use a form of capacitive averaging to determine positional input data. Take a look here if you must for a basic example. There are projected non-capacitive averaging screens which only take the very first contact with the screen into account. These are commonly used in touch-keypads and anything that doesn't require 'movement' of the finger/stylus.

[Seraphic]

Originally Posted by DesignDawg

Literally LOL at that line. Hilarious. I like you, Ryan. Welcome to AC.