Touchscreen Accuracy: iPhone Compares Favorably to Other Smartphones

[Minimoose 360]

After reading this story, I took my friends Droid and crushed it with my hands for being too inaccurate.

Not serious, but this serious thread is very serious. None of you people designed the devices, so none of you 100% know what you are talking about.

 

[cmaier]

None of you people designed the devices, so none of you 100% know what you are talking about.

Because observation and the scientific method are incapable of being used to figure out whether a device can detect straight lines?

Well, we might as well shut down all institutions of scientific learning and live in huts, I guess.

 

[kdarling]

I would not be surprised if kdarling is somehow connected with one of the other handset makers and view the iPhone as a threat to their livelihood.

Instead of acting so paranoid and attacking posters when they say something you disagree with, why don't you try responding to their points?

Don't take my word for it. Do a search for kdarling displaying posts by that user. I have often had to correct them on CDMA versus W-CDMA by posting various wikipedia and industry articles ...

Yep, please search. I and others have attempted to correct your misconceptions about W-CDMA and LTE being compatible protocols. Your enthusiasm is commendable, even if your technical knowledge is lacking. If you wish, please start a separate LTE thread where we can figure it all out.

I agree. kdarling is crazy.

Yep, as crazy as any engineer with thirty years' experience would be around some who prefer myths over facts and adult observations

All that some of us are trying to do is figure out how they got those results and what could cause them. I don't understand why that throws some of you into a fit. That's just weird.

.

 

[paul4339]

This study has no more scientific legitimacy than any fourth grade science project. The variables are not well enough controlled, leaving the independent variable too reliant on oversights in the testing protocol.

For example, how do we know the subject's finger is moving in a straight line? While not intentional, the ideomotor effect would easily lead to the outcome desired or expected by the proctor. Why weren't multiple subjects used? Why weren't the tests blinded?

Virtually all criteria for ensuring scientific accuracy are missing in this experiment. Something as simple (and obvious!) as a straight-edge ruler placed along the subject's finger could have greatly improved these results.

Don't get me wrong: this test is excellent on the preliminary level, and gives one a hypothesis for further testing; however, the current results are not worthy of publication nor our attention.

We have some folks reproduce and post the iPhone tests.
It'll be nice if someone with a Droid (or other phones) can simple reproduce the results and post.
Good tests should be repeatable.

 

[CFreymarc]

After reading this story, I took my friends Droid and crushed it with my hands for being too inaccurate.

Not serious, but this serious thread is very serious. None of you people designed the devices, so none of you 100% know what you are talking about.

Speak for yourself. There is a very interesting distribution of people here including those that have designed mobile systems from the board layout, to the embedded OS to the apps on.

Then there are those that post from their parents basements living off the second mortgage taken on the place after getting their folks to sign a living will ... and you know who I am specifically talking about here!

 

[Journojulz]

We have some folks reproduce and post the iPhone tests.
It'll be nice if someone with a Droid (or other phones) can simple reproduce the results and post.

Easy to do. All you need is a straight edge and a sketching app.

My Palm Tungsten had dead spots so i laid a ruler diagonally across the screen and drew lines across the whole screen.

It showed huge blank spots and that one jpeg shortened a call to tech support.

Go. Rule. Draw.

And for the record - i love the iphone touch interface, it feels 'right'.

 

[VulchR]

Apple doesn't have the touch interface right.

I like this article because it is one of the few attempts to systematically measure mobile phone performance. It also highlights a systematic error in the routines/hardware by which Apple determines where you finger is contacting the touchscreen. As cmaier posted above, Apple are obviously using the centroid (2-D centre) of the points activated by your finger. As you finger moves off the screen, the geometry changes so the centroid no longer matches the centre of your finger. Put another way, the algorithm Apple acts as though your finger shrinks and elongates as it moves to the edge of the screen. This is one of the few instances I have seen in which Apple has been caught with a fundamentally flaw in the user interface. Imagine if a mouse were that systematically inaccurate. Perhaps in the next revision of the iPhone software they'll fix this obvious bug....

 

[Habakuk]

This study has no more scientific legitimacy than any fourth grade science project.

+1

I tried some of the "iPhone killers" even new Android phones (with—tataaa!—multitouch, but not in all parts of the world, and not spanning the whole system): The touchscreen technique (capacitive, multitouch) works best on iPhone and iPod touch by far as well as the software behind. The best way to test is not a primitive draw pattern but challenging draw line games where you need to be fast and accurate. Not to mention some of the car navigation systems or—quite new—some cameras with touchscreens ==> all crap comparatively.

 

[TheMacBookPro]

That test is obviously Apple-biased. Read the comments and see 'real-life' tests.

Oh yeah, look at all that stair-stepping

 

[Fiveos22]

So Apple got the touch technology right. Over two years ago.

Hardly surprising.

Or that could mean that they got it right with technology that they learned from the days of the Newton (dev 1989-1998).

 

[MorphingDragon]

That test is obviously Apple-biased. Read the comments and see 'real-life' tests.

Oh yeah, look at all that stair-stepping

The test is only fair when you have the same back-end code.

So comparing to a program optimised for the Nexus against something whipped up for the iPhone or Droid in 5 mins is hardly a fair test.

Unfortunately when it comes to testing things like touchscreen you cant get the same pressure over and over. Theresfore these test you can ultimately be chucked out the window.

 

[GnillGnoll]

It's strange they tested only diagonal lines. While I believe the devices used all have a grid of horizontal and vertical wires (you can see them under the right lighting conditions) and thus should usually do well with horizontal/vertical lines, there are other devices, like the HTC HD2, which use diagonal wires (on the HD2 they are particularly visible).

I like this article because it is one of the few attempts to systematically measure mobile phone performance. It also highlights a systematic error in the routines/hardware by which Apple determines where you finger is contacting the touchscreen. As cmaier posted above, Apple are obviously using the centroid (2-D centre) of the points activated by your finger. As you finger moves off the screen, the geometry changes so the centroid no longer matches the centre of your finger. Put another way, the algorithm Apple acts as though your finger shrinks and elongates as it moves to the edge of the screen. This is one of the few instances I have seen in which Apple has been caught with a fundamentally flaw in the user interface. Imagine if a mouse were that systematically inaccurate. Perhaps in the next revision of the iPhone software they'll fix this obvious bug....

The proper way to fix this would be to extend the touch sensing surface beyond the edges of the screen, preferably to the entire front surface. That would also prevent annoying "sliding off a virtual D-pad" mistakes and allow Palm Pre-like off-screen gestures.

 

[distortedloop]

Because observation and the scientific method are incapable of being used to figure out whether a device can detect straight lines?

Except this test followed the scientific method in only the most fleeting of ways. No controls for the shakiness or variances in speed or pressure in the tester's hand.

Good tests should be repeatable.

Exactly. That's a basic of the scientific method; your tests must be repeatable and produce similar results.

Easy to do. All you need is a straight edge and a sketching app.

The fact that MOTO didn't use a ruler, and the clearly different amount of shakiness in the Droid test take away most of the credibility of their results to me.

The test is only fair when you have the same back-end code.

So comparing to a program optimised for the Nexus against something whipped up for the iPhone or Droid in 5 mins is hardly a fair test.

I call BS on that one! IF it's all about the experience, then you want code that is exactly optimized for the device you're using. It's not like the typical user is going to run around using software whipped up in 5 minutes every day. Isn't it a common battle cry for the Apple devote around here to point out that even though the iPhone doesn't have the best hardware that it's optimized OS and software still gives it the edge? You can't have it both ways.

 

[MorphingDragon]

I call BS on that one! IF it's all about the experience, then you want code that is exactly optimized for the device you're using. It's not like the typical user is going to run around using software whipped up in 5 minutes every day.

So how many fart Apps have been downloaded over the world?

---

Re-Read my post, you're missing something.

Isn't it a common battle cry for the Apple devote around here to point out that even though the iPhone doesn't have the best hardware that it's optimized OS and software still gives it the edge? You can't have it both ways.

You can actually. But you still aren't grasping the point.

 

[cmaier]

It's strange they tested only diagonal lines. While I believe the devices used all have a grid of horizontal and vertical wires (you can see them under the right lighting conditions) and thus should usually do well with horizontal/vertical lines, there are other devices, like the HTC HD2, which use diagonal wires (on the HD2 they are particularly visible).


The proper way to fix this would be to extend the touch sensing surface beyond the edges of the screen, preferably to the entire front surface. That would also prevent annoying "sliding off a virtual D-pad" mistakes and allow Palm Pre-like off-screen gestures.

Agreed. It would be almost impossible to fix with software since all software could do is guess as to where the restnof the finger is likely to be, perhaps base on remaining finger shape.

 

[distortedloop]

So how many fart Apps have been downloaded over the world?

---

Re-Read my post, you're missing something.

You can actually. But you still aren't grasping the point.

Either that or you're not making it very well.

I got a chuckle out of the fart app response, however, that hardly proves whatever point you're trying to make. We're not discussing using fart apps as a way to prove the touchscreen performance superiority of one platform over another...

 

[WeegieMac]

28 people rated these news negative on an apple forums website, with test clearly showing an apple products superiority in one respect...

Don't you people have anything better to do than frequent apple forums and bitch and vote negative because you own non apple products?

This site attracts so much of this type of crowd, it's mind boggling...

It's one thing, even as a semi-regular poster, that I notice a lot and do sit and wonder, "Why".

I actually think MacRumors could run a story saying, "Apple find cure for Cancer" and there would be people who lurk here that would rate it as negative for the sake of doing so.

 

[Habakuk]

I actually think MacRumors could run a story saying, "Apple find cure for Cancer" and there would be people who lurk here that would rate it as negative for the sake of doing so.

Apple iPhone "cured" my diabetes mellitus Type 2. When I've got the phone I was (and still am) so occupied with that gadget that I almost forgot to eat. Lost 30 Kilograms of weight in five months and don't need any medication any more (glucose in blood—normal). No joke!

http://en.wikipedia.org/wiki/Diabetes_mellitus
Be careful with that!

Want to lose weight easily without missing anything?

1. Don't eat carbohydrates = starch (such as cereals, bread, grain products, rice, potatoes and pasta, banana, beans, corn) or sugar (such as candy, jams and desserts). Or try to eat as few of it as possible (if you can't avoid it then absolute natural, biological and most expensive products, NO WHITE SUGAR).
   
2. Get some Apple gadget (instead of wonder-diet plans, books, DVDs, receipts...) and play/work with it for distraction of food themes. Don't watch any food/cooking/commercials on TV.

http://en.wikipedia.org/wiki/Carbohydrate

Try it! It works for most people! No yo-yo effect. I am keeping my normal weight now almost one year.

Sorry for off-topic but those touchscreen tests are ridiculous.

 

[SkyBell]

Texting? Texting seems fine. I can go right to the very outer edge on any letter or number sitting on the edge, like, Q and P and 1 and 0, etc. Or did you mean something else?

Eh, maybe it's just me, but there seems to be a large loss of accuracy on the letters near the edge.

 

[miketcool]

That test is obviously Apple-biased. Read the comments and see 'real-life' tests.

Oh yeah, look at all that stair-stepping

See, I knew I was foolish for question those tests! Now there is evidence that my skepticism might be founded!

 

[kasplash]

One more data point

We have some folks reproduce and post the iPhone tests.
It'll be nice if someone with a Droid (or other phones) can simple reproduce the results and post.
Good tests should be repeatable.

I tried this on my Moto Droid, and my lines came out clean and beautiful, just like the iPhone in the video. The app I used is Paint Note.

 

[tbrinkma]

Except this test followed the scientific method in only the most fleeting of ways. No controls for the shakiness or variances in speed or pressure in the tester's hand.

The video seems to show that visibly similar (if not actually identical) testing was done on all the phones involved in the test. No, it doesn't meet the strictest scientific repeatability standards, but somehow I doubt the guy has access to a multi-million dollar lab with the equipment necessary to reach that standard. Then again, if having that equipment were actually necessary to perform reasonably good science, we'd still be living in caves. The man demonstrated his methodology. If you disagree with it, run your own tests and show where his methodology went wrong *and* poisoned the results. Until then, your complaining about his methodology is *less* scientific than his methodology is.

I call BS on that one! IF it's all about the experience, then you want code that is exactly optimized for the device you're using. It's not like the typical user is going to run around using software whipped up in 5 minutes every day. Isn't it a common battle cry for the Apple devote around here to point out that even though the iPhone doesn't have the best hardware that it's optimized OS and software still gives it the edge? You can't have it both ways.

Actually, since it seems to have been a test intended to reveal finger-position plotting accuracy of the various phones, completely unoptimized software that simply plots out the contact point reported through the phones' APIs would be the best way to go. Specifically, you *don't* want to use drawing software with algorithms designed to smooth out such defects. (I don't know enough about the software he used on the various phones to know whether some or all of the software meets this criteria.)

Even better might be software that randomly shows a dot on the screen, and then waits for a user to touch it and measures the relative position of the dot and the measured touch. That gives a pin-point target and the variance around different areas of the screen can be compared. Tracing a diagonal lines is a good way to easily cover large areas of screen real-estate quickly to gain similar results.

 

[VulchR]

Agreed. It would be almost impossible to fix with software since all software could do is guess as to where the restnof the finger is likely to be, perhaps base on remaining finger shape.

You have an excellent suggestion, for guessing finger position wouldn't be too hard:

(1) Determine the average size (height and width) of the finger spot on the touchscreen while the swipe is in the safe part of the screen.

(2) As the finger moves toward the edge (say, within the radius of the average size of the touchscreen spot created by fingers), check to see if spot width>>height or width<<height. If so, recompute finger location as though width=height.

Simples.

 

[DougB541]

Wow...really?

All these modern OS's (mobile OSX, Android and WebOS) have very accurate screens for what the phone needs you to do.

What it neesd you to do is point and hit an icon the size of the pad of your finger.



I used to hate benchmarking computers for games....but now we have touchscreen benchmarks?

 

[cmaier]

You have an excellent suggestion, for guessing finger position wouldn't be too hard:

(1) Determine the average size (height and width) of the finger spot on the touchscreen while the swipe is in the safe part of the screen.

(2) As the finger moves toward the edge (say, within the radius of the average size of the touchscreen spot created by fingers), check to see if spot width>>height or width<<height. If so, recompute finger location as though width=height.

Simples.

Far better would be to simply extend the touch radius one-half finger around the edge of the screen (approximately 20 pixels, assuming the current dot pitch). This would also allow for new gestures involving the bezel.