Apple's count of "iPad apps" actually includes both iPad-specific and iPhone 4 "HD" apps. Given that a good percentage of Android apps are already "HD", comparisons aren't really as skewed as Apple likes to claim. In fact, given how well things like Firefox, Angry Birds, everything Adobe's released, etc. run on tablets, I'm not really sure how you get any more "tablet specific". I'm sure they'll come up with something...
Right. Buying direct from Apple follows the old Dell plan: if you sell direct, you can sell at a competitive retail price, rather than deal with distributers and retail stores and those various markups. But selling via the Apple store, they lose most of their advantage over places like Best Buy. Maybe all of it... Best Buy has 10,000 other products to help share that retail overhead.
The big remaining advantage of the Apple Store: only Apple products. You're not going to walk out of the Apple store with a Xoom or an HTC Thunderbolt.
Does anyone really claim that the iPad is so much better in the tablet market than the iPhone was in the smartphone market that its unassailable? Nonsense.
Don't forget the first Android phones... the G1 didn't even really try. And while the Droid was a pretty big boost to the "ecosystem", it didn't defeat the iPhone. The only case I can even recall of a single model of Android outselling the iPhone is recently, with the HTC Thunderbolt at Verizon outselling the iPhone 4... a brand new 4G product outselling last year's model.
And yet, even in the USA, Apple's stronghold, Android phones outsold the iPhone last year, and already beat RIM (the only other smartphone vendor to outsell Apple last year in the USA) this year. A market with multiple vendors always wins, in the end. That's why the IBM PC won, why DVD defeated DivX, why Blu-ray defeated HD-DVD, why Compact Flash and SD are the successful memory cards, etc. The only place this doesn't happen is in very limited niches markets only served by a few proprietary customers, or vertically integrated markets with unusual dynamics. The video game console market is an example of that; every vendor takes a loss on the hardware for a new console, only one hardware vendor per platform, etc. So sure, if Applet thinks they can make enough cash on the iTunes store to subsidize the market with super-cheap iPads, they win. But they'll never do it.
Price is an issue, though. Anyone who thinks the iPad is cheap is not paying attention -- the cost to make of any ARM tablet is lower than most netbooks. Apple made nearly as much money on the iPad last year as on the Macintosh. Just the hardware. Why ever change that model.
Some are starting to get the message. As much as Apple sells, they don't actually MAKE any component in the iPad. Samsung makes practically every component for a tablet themselves, including many found in the actual iPad. Is there any possible reason a Samsung tablet needs to cost as much as an iPad? None... they were simply setting the price based on Apple as the only competition. Another datapoint: Archos. They're on their fourth generation tablets now.... evolved out of the PMP world, never directly targeted at Apple. You can buy a 10" Archos with specs better than the iPad 1 for $300. The price, right now, should be keeping people away: ARM tablets at $500 or more look expensive when netbooks start at $200 and, in theory, offer more.
Yes, the iTunes store is a big magnet. It was with the iPhone too... but that didn't keep the iPhone on top. All of those Android users are getting the same kind of things from the Android Market, the Amazon Appstore, and maybe others. If there's really a market for tablets, this growing collection of users will be looking for Android tablets. Today's Xoom is really just the G1, in terms of the way Google's looking at the tablet market (despite the fact Android 2.x is just fine with tablets, I have found only one out of 50 I've tried on my "Adam" tablet didn't support full resolution -- vastly different than the iPad vs. iPhone story).
And that's even assuming something big doesn't happen. But we already know that today's B&N "nook" is getting an official update that will include apps as well as books -- the device is already proving a fine tablet among hackers. And Amazon's certain to release future Kindles with the full Android treatment as well. Neither device may be as open to other appstores as the average smartphone today, we'll see. But the same apps run everywhere.
Take one photo sensor site of 5um x 5um. Ok, good. Now take four photo sensor sites of 2.5um x 2.5um each. All other things being equal, we will have the same light hitting the sensor chip, and thus, four times as much light hitting that larger sensor. Let's assuming a 100% fill factor, to not make the smaller sensor any worse than it already will be.
Since our goal is to operate each photodiode in its linear range as much as possible, we can assume it usually does. Thus, the signal from that large sensor is four times larger than that of each smaller sensor... so I need four times the amplification to deliver the same signal level to my ADCs.
That means that, again, all things being equal, I will start to see noise in the image two f-stops (four times the light) sooner with the higher density sensor than with the lower density sensor. Chip-level noise is still significant.. once you hit the thermal noise floor, things may well change, but we're far from there today.
Now of course, if you were to average all four pixels together, that would tend to average out the random noise.. but then you're back to your 1/4 resolution.
Now in practice, things are not always equal. Vast improvements in chip-level noise have been achieved over the years, but it's really improved since everyone switched to CMOS sensors (despite the other issues, like rolling shutter effects). The fact is, this year's 18Mpixel APC sensor is better than the 6 Mpixel sensor from some years back: better fill percentage (better microlenses, maybe it's a backlit sensor), more sensitive photosites, far lower noise CMOS process. And going forward, there are very definite walls on pixel shrinking we're already hitting (the 1/4" sensor used on the iPhone is diffraction limited at about f2.7). Other features will be selling camera sensors, not megapixels.. though I'm sure plenty will have "marketing" pixels, which may actually produce a lower quality image.
First is the fact they mention Sony. If they were not already using a back illuminated sensor, they want one. That's Sony's big thing... they're using smaller back illuminated sensors in their consumer camcorders and competing with larger conventional sensor pretty favorably. Not crazy differences... a Sony with a 1/2.88" sensor is outperforming a JVC with a 1/2.33" sensor, and doing well against a Pansonic with three 1/4" sensors, that kind of thing. But in short, this is exactly why you'd go to Sony.
Then there's the fact that the iPhone has to compete. Since the beginning of 2010, the standard for nearly every high-end smart phone not from Apple has been 8Mpixel: HTC EVO, Desire, Droid Incredible; Sony's own Xperia X10, U5i; Samsung i970, Galaxy S II, Memior, Motorola XT720, Droid X; Nokia N86, E7, C7; LG GC900, Renoir, etc. Actually, this started in 2008, though back then (six 8 Mpixel models were introduced in 2008), some were "CAMERAphone" models. Today, 8Mpixel is mid-to-high, 5Mpixel is entry level.
And so Apple must put in 8Mpixel... it's expected and required. That's marketing, and it has nothing to do with the physics of the actual camera. They can write at length in detail, as many here, why the 5Mpixel senor is as good or even better than the 8Mpixel in every competitor's camera. They can push through articles in magazines, ads on TVs, etc. making this claim. And still, 80% of the audience doesn't understand them, 10% of those remaining doesn't believe it. Or they put the 8Mpixel camera in there as expected, and do the best things possible to make it a good camera. Then hand this over to the marketing people, who will come up with some press that claims this is now Fleonard's gift to cameras in phones, and you probably can't tell it from my expensive Canon DSLR most of the time.
Until that 12Mpixel camera ships in iPhone 7, and renders every form of photographic expression before it little more than a long, long nightmare.
The earlier iPhones had pretty bad cameras.. basically glorified webcams from Micron, mostly. The iPhone 4 camera was the first one Apple seems to get a little serious about.
Apple is still subject to physics, at least outside of the reality distortion field. The only way they can get a substantially better image from an 8Mpixel sensor at the same 1/4" sensor size will be to go to a lower aperture, like f1.8 or so. Which means a substantially larger lens. If they stay at f3, there's not going to be much difference in actual resolution. And that's even assuming the lens is good enough to make the sensor the limiting factor.. that, also, is rather questionable.
Of course, they could also cheat. Drop a DVD to HD style upscaling algorithm in there, and most people will believe the camera is better. They have enough CPU to do that today. Of course, if they already are doing that, then they'll need better-still upscaling, which is kind of a trick.
Now, you might ask, how can upscaling help when the actual output is 5Mpixel, same as the input. The key here is that the sensor is interpolated... each sensor site is masked by either R, G, or B filter (which, incidently, cuts out 2/3 of the light), and they use Dr. Bayer's pattern and, more or less, the same interpolation algorithms. The end result is that a 5Mpixel Bayer image is much softer, due to this interpolation, than a 5Mpixel RGB image would be.
Supposedly about a 1/2" sensor (you have to know about the history of videcon tubes and other ancient video stuff to know about sensor measurements... anyone who just gets out a rules and measures will get the wrong answer). That's not bad. Higher end P&S cameras like the Canon G12 or Nikon P7000 usually rock something like a 1/1/7" sensor.
Moving upward, the Panasonic and Olympus DSLR, EVIL, and ILCC's sport 4/3" sensor... APS-C, APS-H, and full frame 35mm move up from there. Looking in the other direction, cheaper P&S cameras from reputable companies go down to about 1/2.5", which is also close the largest sensor you'll find in a consumer camcorder (I think JVC have a few at 1/2.3").
Some of the "Flip" style webcam devices have used the same 1/4" sensor you usually find in a cellphone, and a lens to match. But they better of these have slightly better sensors today, the reasonably-good-for-a-Flip-clone Kodak ZI8 has a 1/2.5" sensor. Though the N8 still bests those.
To bad Nokia's dumping all their interesting stuff for a cozy spot in bed with Microsoft on Window 7 Phone.
We're already at the point in APS sensors where diffraction limits are kicking in at f8... that's APS-C at 18Mpixels. We don't see it yet, though, because they're all doing Bayer interpolation, which increases the effective circle of confusion by about a factor of 9. So there's still some practical resolution to be had in APS sensors, but it's not going on forever.
The ironic thing here is that, assuming Sigma/Foveon ever get their act together on an RGB (three sensors per pixel site) senor chip with a modern spatical resolution, they'll be just in time to still suck, due to diffraction.
It's going to be the same size as the current iPhone, 1/4".
As for 5-6Mpixel... the Mpixel matters not... it's the sensor site pitch and fill that makes the most difference. Today's DSLRs go well into the 20MPixel range and can continue to improve into the next few years at least. Medium Format DSLRs have been shooting 60Mpixel images, and can keep going.
But with smaller sensors, there are two problems. Each sensor site getting smaller means less sensitivity, so the only improvements are making the sensor more sensitive (which is what the "backlit sensor" is all about) or lowering the noise levels (this happens with each new generation of sensor).
Physics, as usual, puts limits on all of these. Diffraction, as I mentioned before, limits the value of continually adding pixels to a fixed-size sensor. Apple will absolutely need an f1.8 lens or so to actually get a resolution increase on the 8Mpixel sensor versus the 5Mpixel sensor of today, both being 1/4". They can keep loweriing the noise floor for awhile, but there is the thermal noise floor, a limit of physics. After that, they'll have to figure out how to shoot a bunch of quick exposures and average out the noise, without too much subject blur, or some other trick... physics gets the final say.
Sony's a big proponent of backside illumination photo sensors. That's probably why Apple's interested -- the 1/4" CMOS sensor is just too frickin' tiny to be useful at 8Mpixel without employing nearly all of the magic tricks these days. And back illumination is one of the best ones.
They used Micron sensors for the earlier phones... don't know offhand who made the iPhone 4 sensor.
They're going to 8Mpixel because every other high-end smartphone has had an 8Mpixel sensor since last summer or so. That doesn't make it a good idea.
Take the current iPhone... 5Mpixel 1/4" sensor, f3 fixed aperture lens. If you do the math, you find the diffraction limit on that sensor is just above f3. The resolution limit is the point at which the circle of confusion of the lens matches the size of the Airy disc (the disc image formed due to diffraction). Basically, once you're there, the image gets fuzzier as you stop the lens down, and adding more megapixels just lowers the f-number (eg, same sized Airy disc, smaller sensor site). So unless Apple spends a little more, risking esthetics and thickness, the move to 8Mpixel may do nothing, because the lens they're using now will, if anything, look a bit fuzzier on an 8Mpixel sensor. And they're sticking to 1/4", of course, to keep things otherwise the same.
Not that Apple's above selling you "marketing pixels" anyway. And of course, being a single sensor, the image is already softened due to Bayer interpolation. So I'm sure they're not expecting anyone to notice, and expecting all Apple Fanbois to praise this new sensor endlessly. And they're probably honing their image enhancement algorithms (already well at work on camera phones, "Flip" cameras, P&S, and any other imager used by consumers as a replacement for real camera or camcorder), so everyone will agree this is a better camera.
I switched from Sony to Panasonic camcorders a year and a half back... but primarily because Panasonic simply made a better product. And year, part of that reason is that Sony, more than anyone, was pushing some of their corporate vision on the features of different models, rather than just doing what they could do with the technology. That ultimately kills market share... it has with Sony on many fronts, and it will with Apple too.
And those who leave probably don't come back, even when if they fix those problems.
Still like my PS3, and I still use Sony Media Software. You can't actually blame every division of a company for the evils of some. But I certainly do look elsewhere before even considering a Sony product. And I will never own another Apple product (I did, once, it was a gift). It's hard to say for certain if, across the board, Apple or Sony has been more evil. I was more directly affected by Sony... they felt it necessary to protect the copyrights of my own live MiniDisc recordings by not allowing direct digital transfers... even after they built a device that allowed direct digital transfers TO MiniDisc. But over all, Apple has been far more successful, lately, at locking people in, at controlling their use of their own hardware, pushing completely proprietary formats as "standard" (like the still-mysteriously-growing use of Apple's totally proprietary ProRes video CODEC in professional video -- you can't even create a ProRes file on a non-MacOS PC), putting DRM on nearly everything, etc. That doesn't mean Sony isn't as evil in their collective heart, but if they are, Apple's still much better at it. Particularly in areas that actually matter.
Yup. And Apple is a product integrator, not a component manufacturer. They don't make a single thing on their own. But they sell so many, this isn't a problem. And in fact, much of their strategy is to maximize volumes by selling only a very limited number of models of a product, and then reusing these parts as much as possible. Thus the single version of the iPhone, iPod Touch, and iPad each year, using many of the same parts in each.
And in fact, Apple's increasingly buying parts from direct competitors. Samsung made mode of the expensive ICs in the iPhone 4 and iPad 1, not sure if they've found a different SOC foundry for the iPad 2's A5 chip. Of course, Samsung is the second largest IC company in the world, and a primary supplies of memories of all kinds... like Sony with imaging products, they're hard to avoid. Only a few companies make LCDs... Samsung (again), Sharp, Toshiba, and some relatively unknown companies in China make up the bulk of them. Samsung also makes nearly all of the AM-OLED displays used in volume production
And you can't get away from any one company so easily, anyway. Sony has invested heavily in joint projects with Samsung, Toshiba, and Sharp in displays. Apple's done similar deals with LG, Sharp, and Toshiba. Panasonic outright bought Sanyo in 2009, primarly to make themselves the king of all rechargeable batteries, at least outside of China. These things happen.
Sony had been supplying components and subsystems to Apple for decades. Back when I worked at Commodore, I took various Macs and Mac IIs apart, and they usually had large modules, even power supplies, from Sony. They used Sony picture tubes in their monitors in the old days (as did anyone selling a "Trinitron" tube.
This is not such a shock... the two companies have been very similar in many ways, again for decades. Their idea of product design, shooting for the high end based on marketing and reputation as much as actual features, good and interesting mechanical, design, etc. And yes, their ideas of product lock-in and the ability of the consumer to control only so much of the hardware, these ideas were invented by Sony and perfected by Apple. Apple has been doing it better, lately... Sony's actually conceeded on the device lock-in (Sony cameras now offer standard memory card slots along with the proprietary slot) and DRM lock-in (Sony still does DRM, but it's cross platform... Apple's still locking you into Apple-only products).
If you have seen any real different between Apple and Sony on their policies, you've been looking only at Sony Music, and ignoring Apple entirely. They're cut from the same cloth. And it's not an accident; Sony, and in particular the Sony of the 70s and 80s was one of Steve Jobs' primary models for business, to this day.
I just replaced an 8Mpixel camera ith an 18Mpixel camera, and the image quality is dramatically better. That's because these are DSLRs with large (APS-C) sensors. On a much smaller sensor, you're already diffraction limited... more pixels will do nothing to deliver a sharper image. Before you factor in the blurring due to Bayer interpolation, a 1/4" sensor is already diffraction limited at f2.8 and up... at 2Mpixels. Increasingly, sensor resolution boosts are just "marketing bits".
Sure, you would. But when was the last time you recall Apple making even one concession to function over form? They'll put the 8Mpixel sensor in because that's the largest any reputable company makes in 1/4", same size as the iPhone 4 camera.
The camera in a smart phone is a useful tool for scanning, augmented reality, videophone, and "better than nothing" snapshots. Apple needs 8Mpixel because other top tier smartphones have had them since last summer, and more than ever, Apple actuqlly has to compete with features,not just hype. But if you want a camera, buy a camera.
The problem with this arrangement is that, now with the Nokia deal, they're in a special position. Thus, every other hardware vendor is second tier in the Windows 7 Phone world. Sure, Google picks favorites, but only for a single new software release (Xoom with Android 3.0, Nexus One for 2.1, Droid for 2.0, etc).
This is similar to the "strategy" Microsoft and Toshiba used to rule the market with HD-DVD. Oh, wait, Toshiba's special position there kept all other hardware people away. Maybe MS was slight smarter, waiting until they had a few other companies signed up for W7P before the Nokia deal was announced. But I really question whether those guys will stick around in the long term. Even if W7P amounts to little more than a software port and name change for the Android phone the companies were going to build anyway.
Learning the language is not the same as learning to program. But take my case... I taught myself BASIC when I was 12, then FORTRAN and Assembler. In college, I learned Pascal and LISP. I didn't get into C until I was 20, working the summer at Bell Labs. That took a day to learn. My C code was and still is more formal than most of what you found in the K&R days, thanks to the Pascal that came before it. And I'm sure I've improved over the years -- but that had nothing to do with learning the language -- your ability to program is independent of the language.
So back to this kid... if he's got some supercharged brain for mathematics, and already basically had a calculus-sized hole in this mathematical understanding, maybe he took this in very fast. Calculus itself is conceptually very easy -- the hard part is the algebra, and understanding how to apply it.
I taught myself C, C++, Python, and Java each in a day... not the same week. LISP and assembler took longer, but they were only my 5th and 3rd languages, respectively. And I was 14 for assembler (Z-80), 18 for LISP, 19 for C.... I only taught myself BASIC when I was 12. That took about a month, but I didn't have an actual manual on the BASIC I was using, and no real book on BASIC either, only old copies of "Kilobaud" microcomputing. And I was learning how to program at the same time.. but that was more of an on-going process. And my IQ tested at 161, peak.. and that was under the older systems that allowed scores of 200 or more.
Of course, once you know how to program in several languages, you're just learning the language when you do this. It's very different to claim you leaned to program complex systems in a week. And while I certainly leaned Java in a day... and was actually producing code for a project before the end of that day, I don't claim to have learned, much less mastered, all of the standard library functions. Much less LISP... I'm not sure anyone (ok, maybe Guy Steele Jr.) knew most of MacLisp or InterLisp back in the day, much less Common Lisp. I'm not sure anyone tried -- every LISP programmer carried around books of functions. When you needed something, you looked it up, outside of the general purpose stuff.
There's a big difference between learning the language and learning everything that goes into that ecosystem as a standard resource.
As for leaning calculus, I learned some over the course of ten months in High School... and more in four more advanced college courses. And a good six or more in Mathematics and Electrical Engineering (I double majored at CMU) that applied those methods. So when did I actually "learn calculus"? I do find it hard to image he took in five or six college level courses in a week.
The Microsoft thing is BS.. not the idea in general.
The whole FCC idea of "Whitespace" is that we have a huge chunk of the best overall spectrum put aside for OTA television. But in most areas, most of that spectrum isn't used.. even given the losses due to original cellular (channels up to 83) and the more recent 700MHz auction for 4G (channels in the 60's on UHF).
So the idea of whitespace radio is simple: treat it as ISM radio (like 900MHz and 2.4GHz in the USA) once you acertain that the channel (in 6MHz chunks, just like TV, in the USA) is not used.
The problem is, just using sensing, you can't know if the channel you pick is clear. Your receiver can go into spectrum analyzer mode and not see a thing, but it's still very possible your transmitter is going to interfere with the guy down the street. who for whatever reason (rooftop antenna with 40dB LNA) can actually get that OTA channel.
Thus, the current plan for whitespace radiio... radios need to be location aware, and only use channels legal for that specific location. This is trivial to do, and it pretty much just works. Nothing MS is doing here improves this, far as I can tell. You can't be correct about the usability of a channel from a single monitoring point, whether you spend $100 or $100,000 on that spectrum analyzer. And so, given the need for one node in the network to have a separate internet connection, nothing MS does online is an improvement over the basic idea -- we absolutely know where the licensed radio is, because it's LICENSED! That license is for a certain areas, and no army of MS spectrum analyzers can be certain that your neighbor can't receive that channel, within the licensed area. Beyond that area, it just doesn't matter -- you get to use that channel anyway.
Years back anyway, CMU taught one "Introduction to Programming" course for Freshmen, anyone in Engineering or Science took this. The gateway to computer science was a largely theoretical course, with very little actual programming (after I took this, 15-211 back in 1980, they split it into two courses a year or two later). The goal was to teach the foundations of computer science, not lock you into any given methodology.
I think the key is to teach different methodologies, and not to rely too heavily on any specific language.
When I was at CMU, they didn't really get into parallel programming. Later, working at Commodore on hardware and learning AmigaOS programming, I caught on that "thread" is just another regular programming construct, just like using "class", "function", "structure", or "method".
And yet, the fact that mainstream software is still not up to that level is a shame. In programming Windows some years back, I found that the Windows API was actually working hard to serialize thing (via the single application message queue) that I had imagined were happening in parallel, as they would have been on any reasonable OS. Linux doesn't get a pass either... it's a royal pain in the butt to do asynchronous I/O in Linux, something that, as an Amiga programmer, I believe fundamental enough it should be a literal no-brainer.
The real critical thing here... what gets taught, or not taught, at the college level today ultimately sets the scale of things like APIs we're all going to be working with in the future. As much as you'd kind of hope that actually innovations in computer science make it into new OSs, it's ultimately going to be those two 2nd year programmers writing that one module you're interacting with. If they don't understand the need for multithreading, much less the fact it should be ubiquitous, they will screw things up in ways that will at least give you a headache (as a user or programmer), at worse have negative impact on the quality and evolution of software.
When I was at CMU, Math/CS majors (they didn't have specialized CS/CE degrees at the time... I did "computer engineering" by double-majoring in EE and Math/CS) took "Introduction to Pascal" as a Freshman course (then, as now, you declared your major when you applied). At least the course I took was extremely well done, It taught everything from a modular programming perspective -- at least in as much as that's possible, given late 70s/early 80s Pascal running on large computers (DEC-20s). Having known only BASIC, FORTRAN, and Assembly as an in-coming Freshman, this was a very good course.
The second level course, 15-211, was hardcore computer science. It started with FSMs (Finite State Machines), went though lambda calculus, all sorts of things that might be possible in programming languages... and all book and paper work, until near the end. The point was not to teach any programming language, but to give you the fundamentals for any language that might follow.
And since those days, I've done at least one major project in over 40 different languages. The idea that OOPL was required is a bit odd... beyond the 15-211 course (which had a 50% attrition rate in 1980, and got splt into to course covering the same material a year or two later... I got an "A", and yeah, it was hard enough to make me brag about that "A", even now), there weren't many specific requirements. You have to have certain number of courses at certain levels... I took CAD and Comparative Languages (Pascal, APL, LISP, and SNOBOL4 were the main languages... we spent a day on FORTAN and a day in InterCal), I could have taken Computer Engineering (the large project course... but we did that in CAD).
Languages themselves are often fads. There's no reason a university CS course should get too language specific, but you do need to program in order to learn how to program, and to really get the higher level ideas of CS. The best thing all that extra money I spent at CMU (and paid back over the next ten years) bought me was improved tools for taking on new things. So I picked up Java in a day when I needed it, same with Python and Ruby... RF Hardware Engineering took me a bit longer. But you really don't know where you're going to end up after school... 5, 10, 25 years later.
Slashdot Mirror
Apple's count of "iPad apps" actually includes both iPad-specific and iPhone 4 "HD" apps. Given that a good percentage of Android apps are already "HD", comparisons aren't really as skewed as Apple likes to claim. In fact, given how well things like Firefox, Angry Birds, everything Adobe's released, etc. run on tablets, I'm not really sure how you get any more "tablet specific". I'm sure they'll come up with something...
Right. Buying direct from Apple follows the old Dell plan: if you sell direct, you can sell at a competitive retail price, rather than deal with distributers and retail stores and those various markups. But selling via the Apple store, they lose most of their advantage over places like Best Buy. Maybe all of it... Best Buy has 10,000 other products to help share that retail overhead.
The big remaining advantage of the Apple Store: only Apple products. You're not going to walk out of the Apple store with a Xoom or an HTC Thunderbolt.
Does anyone really claim that the iPad is so much better in the tablet market than the iPhone was in the smartphone market that its unassailable? Nonsense.
Don't forget the first Android phones... the G1 didn't even really try. And while the Droid was a pretty big boost to the "ecosystem", it didn't defeat the iPhone. The only case I can even recall of a single model of Android outselling the iPhone is recently, with the HTC Thunderbolt at Verizon outselling the iPhone 4... a brand new 4G product outselling last year's model.
And yet, even in the USA, Apple's stronghold, Android phones outsold the iPhone last year, and already beat RIM (the only other smartphone vendor to outsell Apple last year in the USA) this year. A market with multiple vendors always wins, in the end. That's why the IBM PC won, why DVD defeated DivX, why Blu-ray defeated HD-DVD, why Compact Flash and SD are the successful memory cards, etc. The only place this doesn't happen is in very limited niches markets only served by a few proprietary customers, or vertically integrated markets with unusual dynamics. The video game console market is an example of that; every vendor takes a loss on the hardware for a new console, only one hardware vendor per platform, etc. So sure, if Applet thinks they can make enough cash on the iTunes store to subsidize the market with super-cheap iPads, they win. But they'll never do it.
Price is an issue, though. Anyone who thinks the iPad is cheap is not paying attention -- the cost to make of any ARM tablet is lower than most netbooks. Apple made nearly as much money on the iPad last year as on the Macintosh. Just the hardware. Why ever change that model.
Some are starting to get the message. As much as Apple sells, they don't actually MAKE any component in the iPad. Samsung makes practically every component for a tablet themselves, including many found in the actual iPad. Is there any possible reason a Samsung tablet needs to cost as much as an iPad? None... they were simply setting the price based on Apple as the only competition. Another datapoint: Archos. They're on their fourth generation tablets now.... evolved out of the PMP world, never directly targeted at Apple. You can buy a 10" Archos with specs better than the iPad 1 for $300. The price, right now, should be keeping people away: ARM tablets at $500 or more look expensive when netbooks start at $200 and, in theory, offer more.
Yes, the iTunes store is a big magnet. It was with the iPhone too... but that didn't keep the iPhone on top. All of those Android users are getting the same kind of things from the Android Market, the Amazon Appstore, and maybe others. If there's really a market for tablets, this growing collection of users will be looking for Android tablets. Today's Xoom is really just the G1, in terms of the way Google's looking at the tablet market (despite the fact Android 2.x is just fine with tablets, I have found only one out of 50 I've tried on my "Adam" tablet didn't support full resolution -- vastly different than the iPad vs. iPhone story).
And that's even assuming something big doesn't happen. But we already know that today's B&N "nook" is getting an official update that will include apps as well as books -- the device is already proving a fine tablet among hackers. And Amazon's certain to release future Kindles with the full Android treatment as well. Neither device may be as open to other appstores as the average smartphone today, we'll see. But the same apps run everywhere.
Take one photo sensor site of 5um x 5um. Ok, good. Now take four photo sensor sites of 2.5um x 2.5um each. All other things being equal, we will have the same light hitting the sensor chip, and thus, four times as much light hitting that larger sensor. Let's assuming a 100% fill factor, to not make the smaller sensor any worse than it already will be.
Since our goal is to operate each photodiode in its linear range as much as possible, we can assume it usually does. Thus, the signal from that large sensor is four times larger than that of each smaller sensor... so I need four times the amplification to deliver the same signal level to my ADCs.
That means that, again, all things being equal, I will start to see noise in the image two f-stops (four times the light) sooner with the higher density sensor than with the lower density sensor. Chip-level noise is still significant.. once you hit the thermal noise floor, things may well change, but we're far from there today.
Now of course, if you were to average all four pixels together, that would tend to average out the random noise.. but then you're back to your 1/4 resolution.
Now in practice, things are not always equal. Vast improvements in chip-level noise have been achieved over the years, but it's really improved since everyone switched to CMOS sensors (despite the other issues, like rolling shutter effects). The fact is, this year's 18Mpixel APC sensor is better than the 6 Mpixel sensor from some years back: better fill percentage (better microlenses, maybe it's a backlit sensor), more sensitive photosites, far lower noise CMOS process. And going forward, there are very definite walls on pixel shrinking we're already hitting (the 1/4" sensor used on the iPhone is diffraction limited at about f2.7). Other features will be selling camera sensors, not megapixels.. though I'm sure plenty will have "marketing" pixels, which may actually produce a lower quality image.
The rumor is completely credible.
First is the fact they mention Sony. If they were not already using a back illuminated sensor, they want one. That's Sony's big thing... they're using smaller back illuminated sensors in their consumer camcorders and competing with larger conventional sensor pretty favorably. Not crazy differences... a Sony with a 1/2.88" sensor is outperforming a JVC with a 1/2.33" sensor, and doing well against a Pansonic with three 1/4" sensors, that kind of thing. But in short, this is exactly why you'd go to Sony.
Then there's the fact that the iPhone has to compete. Since the beginning of 2010, the standard for nearly every high-end smart phone not from Apple has been 8Mpixel: HTC EVO, Desire, Droid Incredible; Sony's own Xperia X10, U5i; Samsung i970, Galaxy S II, Memior, Motorola XT720, Droid X; Nokia N86, E7, C7; LG GC900, Renoir, etc. Actually, this started in 2008, though back then (six 8 Mpixel models were introduced in 2008), some were "CAMERAphone" models. Today, 8Mpixel is mid-to-high, 5Mpixel is entry level.
And so Apple must put in 8Mpixel... it's expected and required. That's marketing, and it has nothing to do with the physics of the actual camera. They can write at length in detail, as many here, why the 5Mpixel senor is as good or even better than the 8Mpixel in every competitor's camera. They can push through articles in magazines, ads on TVs, etc. making this claim. And still, 80% of the audience doesn't understand them, 10% of those remaining doesn't believe it. Or they put the 8Mpixel camera in there as expected, and do the best things possible to make it a good camera. Then hand this over to the marketing people, who will come up with some press that claims this is now Fleonard's gift to cameras in phones, and you probably can't tell it from my expensive Canon DSLR most of the time.
Until that 12Mpixel camera ships in iPhone 7, and renders every form of photographic expression before it little more than a long, long nightmare.
The earlier iPhones had pretty bad cameras.. basically glorified webcams from Micron, mostly. The iPhone 4 camera was the first one Apple seems to get a little serious about.
Apple is still subject to physics, at least outside of the reality distortion field. The only way they can get a substantially better image from an 8Mpixel sensor at the same 1/4" sensor size will be to go to a lower aperture, like f1.8 or so. Which means a substantially larger lens. If they stay at f3, there's not going to be much difference in actual resolution. And that's even assuming the lens is good enough to make the sensor the limiting factor.. that, also, is rather questionable.
Of course, they could also cheat. Drop a DVD to HD style upscaling algorithm in there, and most people will believe the camera is better. They have enough CPU to do that today. Of course, if they already are doing that, then they'll need better-still upscaling, which is kind of a trick.
Now, you might ask, how can upscaling help when the actual output is 5Mpixel, same as the input. The key here is that the sensor is interpolated... each sensor site is masked by either R, G, or B filter (which, incidently, cuts out 2/3 of the light), and they use Dr. Bayer's pattern and, more or less, the same interpolation algorithms. The end result is that a 5Mpixel Bayer image is much softer, due to this interpolation, than a 5Mpixel RGB image would be.
8's been kind of the high-end smartphone minimum since last summer, unless you're Apple.
Supposedly about a 1/2" sensor (you have to know about the history of videcon tubes and other ancient video stuff to know about sensor measurements... anyone who just gets out a rules and measures will get the wrong answer). That's not bad. Higher end P&S cameras like the Canon G12 or Nikon P7000 usually rock something like a 1/1/7" sensor.
Moving upward, the Panasonic and Olympus DSLR, EVIL, and ILCC's sport 4/3" sensor... APS-C, APS-H, and full frame 35mm move up from there. Looking in the other direction, cheaper P&S cameras from reputable companies go down to about 1/2.5", which is also close the largest sensor you'll find in a consumer camcorder (I think JVC have a few at 1/2.3").
Some of the "Flip" style webcam devices have used the same 1/4" sensor you usually find in a cellphone, and a lens to match. But they better of these have slightly better sensors today, the reasonably-good-for-a-Flip-clone Kodak ZI8 has a 1/2.5" sensor. Though the N8 still bests those.
To bad Nokia's dumping all their interesting stuff for a cozy spot in bed with Microsoft on Window 7 Phone.
We're already at the point in APS sensors where diffraction limits are kicking in at f8... that's APS-C at 18Mpixels. We don't see it yet, though, because they're all doing Bayer interpolation, which increases the effective circle of confusion by about a factor of 9. So there's still some practical resolution to be had in APS sensors, but it's not going on forever.
The ironic thing here is that, assuming Sigma/Foveon ever get their act together on an RGB (three sensors per pixel site) senor chip with a modern spatical resolution, they'll be just in time to still suck, due to diffraction.
It's going to be the same size as the current iPhone, 1/4".
As for 5-6Mpixel... the Mpixel matters not... it's the sensor site pitch and fill that makes the most difference. Today's DSLRs go well into the 20MPixel range and can continue to improve into the next few years at least. Medium Format DSLRs have been shooting 60Mpixel images, and can keep going.
But with smaller sensors, there are two problems. Each sensor site getting smaller means less sensitivity, so the only improvements are making the sensor more sensitive (which is what the "backlit sensor" is all about) or lowering the noise levels (this happens with each new generation of sensor).
Physics, as usual, puts limits on all of these. Diffraction, as I mentioned before, limits the value of continually adding pixels to a fixed-size sensor. Apple will absolutely need an f1.8 lens or so to actually get a resolution increase on the 8Mpixel sensor versus the 5Mpixel sensor of today, both being 1/4". They can keep loweriing the noise floor for awhile, but there is the thermal noise floor, a limit of physics. After that, they'll have to figure out how to shoot a bunch of quick exposures and average out the noise, without too much subject blur, or some other trick... physics gets the final say.
Sony's a big proponent of backside illumination photo sensors. That's probably why Apple's interested -- the 1/4" CMOS sensor is just too frickin' tiny to be useful at 8Mpixel without employing nearly all of the magic tricks these days. And back illumination is one of the best ones.
They used Micron sensors for the earlier phones... don't know offhand who made the iPhone 4 sensor.
They're going to 8Mpixel because every other high-end smartphone has had an 8Mpixel sensor since last summer or so. That doesn't make it a good idea.
Take the current iPhone... 5Mpixel 1/4" sensor, f3 fixed aperture lens. If you do the math, you find the diffraction limit on that sensor is just above f3. The resolution limit is the point at which the circle of confusion of the lens matches the size of the Airy disc (the disc image formed due to diffraction). Basically, once you're there, the image gets fuzzier as you stop the lens down, and adding more megapixels just lowers the f-number (eg, same sized Airy disc, smaller sensor site). So unless Apple spends a little more, risking esthetics and thickness, the move to 8Mpixel may do nothing, because the lens they're using now will, if anything, look a bit fuzzier on an 8Mpixel sensor. And they're sticking to 1/4", of course, to keep things otherwise the same.
Not that Apple's above selling you "marketing pixels" anyway. And of course, being a single sensor, the image is already softened due to Bayer interpolation. So I'm sure they're not expecting anyone to notice, and expecting all Apple Fanbois to praise this new sensor endlessly. And they're probably honing their image enhancement algorithms (already well at work on camera phones, "Flip" cameras, P&S, and any other imager used by consumers as a replacement for real camera or camcorder), so everyone will agree this is a better camera.
I switched from Sony to Panasonic camcorders a year and a half back... but primarily because Panasonic simply made a better product. And year, part of that reason is that Sony, more than anyone, was pushing some of their corporate vision on the features of different models, rather than just doing what they could do with the technology. That ultimately kills market share... it has with Sony on many fronts, and it will with Apple too.
And those who leave probably don't come back, even when if they fix those problems.
Still like my PS3, and I still use Sony Media Software. You can't actually blame every division of a company for the evils of some. But I certainly do look elsewhere before even considering a Sony product. And I will never own another Apple product (I did, once, it was a gift). It's hard to say for certain if, across the board, Apple or Sony has been more evil. I was more directly affected by Sony... they felt it necessary to protect the copyrights of my own live MiniDisc recordings by not allowing direct digital transfers... even after they built a device that allowed direct digital transfers TO MiniDisc. But over all, Apple has been far more successful, lately, at locking people in, at controlling their use of their own hardware, pushing completely proprietary formats as "standard" (like the still-mysteriously-growing use of Apple's totally proprietary ProRes video CODEC in professional video -- you can't even create a ProRes file on a non-MacOS PC), putting DRM on nearly everything, etc. That doesn't mean Sony isn't as evil in their collective heart, but if they are, Apple's still much better at it. Particularly in areas that actually matter.
Yup. And Apple is a product integrator, not a component manufacturer. They don't make a single thing on their own. But they sell so many, this isn't a problem. And in fact, much of their strategy is to maximize volumes by selling only a very limited number of models of a product, and then reusing these parts as much as possible. Thus the single version of the iPhone, iPod Touch, and iPad each year, using many of the same parts in each.
And in fact, Apple's increasingly buying parts from direct competitors. Samsung made mode of the expensive ICs in the iPhone 4 and iPad 1, not sure if they've found a different SOC foundry for the iPad 2's A5 chip. Of course, Samsung is the second largest IC company in the world, and a primary supplies of memories of all kinds... like Sony with imaging products, they're hard to avoid. Only a few companies make LCDs... Samsung (again), Sharp, Toshiba, and some relatively unknown companies in China make up the bulk of them. Samsung also makes nearly all of the AM-OLED displays used in volume production
And you can't get away from any one company so easily, anyway. Sony has invested heavily in joint projects with Samsung, Toshiba, and Sharp in displays. Apple's done similar deals with LG, Sharp, and Toshiba. Panasonic outright bought Sanyo in 2009, primarly to make themselves the king of all rechargeable batteries, at least outside of China. These things happen.
Sony had been supplying components and subsystems to Apple for decades. Back when I worked at Commodore, I took various Macs and Mac IIs apart, and they usually had large modules, even power supplies, from Sony. They used Sony picture tubes in their monitors in the old days (as did anyone selling a "Trinitron" tube.
This is not such a shock... the two companies have been very similar in many ways, again for decades. Their idea of product design, shooting for the high end based on marketing and reputation as much as actual features, good and interesting mechanical, design, etc. And yes, their ideas of product lock-in and the ability of the consumer to control only so much of the hardware, these ideas were invented by Sony and perfected by Apple. Apple has been doing it better, lately... Sony's actually conceeded on the device lock-in (Sony cameras now offer standard memory card slots along with the proprietary slot) and DRM lock-in (Sony still does DRM, but it's cross platform... Apple's still locking you into Apple-only products).
If you have seen any real different between Apple and Sony on their policies, you've been looking only at Sony Music, and ignoring Apple entirely. They're cut from the same cloth. And it's not an accident; Sony, and in particular the Sony of the 70s and 80s was one of Steve Jobs' primary models for business, to this day.
I just replaced an 8Mpixel camera ith an 18Mpixel camera, and the image quality is dramatically better. That's because these are DSLRs with large (APS-C) sensors. On a much smaller sensor, you're already diffraction limited... more pixels will do nothing to deliver a sharper image. Before you factor in the blurring due to Bayer interpolation, a 1/4" sensor is already diffraction limited at f2.8 and up... at 2Mpixels. Increasingly, sensor resolution boosts are just "marketing bits".
Sure, you would. But when was the last time you recall Apple making even one concession to function over form? They'll put the 8Mpixel sensor in because that's the largest any reputable company makes in 1/4", same size as the iPhone 4 camera.
The camera in a smart phone is a useful tool for scanning, augmented reality, videophone, and "better than nothing" snapshots. Apple needs 8Mpixel because other top tier smartphones have had them since last summer, and more than ever, Apple actuqlly has to compete with features,not just hype. But if you want a camera, buy a camera.
We have your Geese. If we don't get some trees soon, it's dinner time.
The problem with this arrangement is that, now with the Nokia deal, they're in a special position. Thus, every other hardware vendor is second tier in the Windows 7 Phone world. Sure, Google picks favorites, but only for a single new software release (Xoom with Android 3.0, Nexus One for 2.1, Droid for 2.0, etc).
This is similar to the "strategy" Microsoft and Toshiba used to rule the market with HD-DVD. Oh, wait, Toshiba's special position there kept all other hardware people away. Maybe MS was slight smarter, waiting until they had a few other companies signed up for W7P before the Nokia deal was announced. But I really question whether those guys will stick around in the long term. Even if W7P amounts to little more than a software port and name change for the Android phone the companies were going to build anyway.
We barely knew ye!
Learning the language is not the same as learning to program. But take my case... I taught myself BASIC when I was 12, then FORTRAN and Assembler. In college, I learned Pascal and LISP. I didn't get into C until I was 20, working the summer at Bell Labs. That took a day to learn. My C code was and still is more formal than most of what you found in the K&R days, thanks to the Pascal that came before it. And I'm sure I've improved over the years -- but that had nothing to do with learning the language -- your ability to program is independent of the language.
So back to this kid... if he's got some supercharged brain for mathematics, and already basically had a calculus-sized hole in this mathematical understanding, maybe he took this in very fast. Calculus itself is conceptually very easy -- the hard part is the algebra, and understanding how to apply it.
I taught myself C, C++, Python, and Java each in a day... not the same week. LISP and assembler took longer, but they were only my 5th and 3rd languages, respectively. And I was 14 for assembler (Z-80), 18 for LISP, 19 for C.... I only taught myself BASIC when I was 12. That took about a month, but I didn't have an actual manual on the BASIC I was using, and no real book on BASIC either, only old copies of "Kilobaud" microcomputing. And I was learning how to program at the same time.. but that was more of an on-going process. And my IQ tested at 161, peak.. and that was under the older systems that allowed scores of 200 or more.
Of course, once you know how to program in several languages, you're just learning the language when you do this. It's very different to claim you leaned to program complex systems in a week. And while I certainly leaned Java in a day... and was actually producing code for a project before the end of that day, I don't claim to have learned, much less mastered, all of the standard library functions. Much less LISP... I'm not sure anyone (ok, maybe Guy Steele Jr.) knew most of MacLisp or InterLisp back in the day, much less Common Lisp. I'm not sure anyone tried -- every LISP programmer carried around books of functions. When you needed something, you looked it up, outside of the general purpose stuff.
There's a big difference between learning the language and learning everything that goes into that ecosystem as a standard resource.
As for leaning calculus, I learned some over the course of ten months in High School... and more in four more advanced college courses. And a good six or more in Mathematics and Electrical Engineering (I double majored at CMU) that applied those methods. So when did I actually "learn calculus"? I do find it hard to image he took in five or six college level courses in a week.
The Microsoft thing is BS.. not the idea in general.
The whole FCC idea of "Whitespace" is that we have a huge chunk of the best overall spectrum put aside for OTA television. But in most areas, most of that spectrum isn't used.. even given the losses due to original cellular (channels up to 83) and the more recent 700MHz auction for 4G (channels in the 60's on UHF).
So the idea of whitespace radio is simple: treat it as ISM radio (like 900MHz and 2.4GHz in the USA) once you acertain that the channel (in 6MHz chunks, just like TV, in the USA) is not used.
The problem is, just using sensing, you can't know if the channel you pick is clear. Your receiver can go into spectrum analyzer mode and not see a thing, but it's still very possible your transmitter is going to interfere with the guy down the street. who for whatever reason (rooftop antenna with 40dB LNA) can actually get that OTA channel.
Thus, the current plan for whitespace radiio... radios need to be location aware, and only use channels legal for that specific location. This is trivial to do, and it pretty much just works. Nothing MS is doing here improves this, far as I can tell. You can't be correct about the usability of a channel from a single monitoring point, whether you spend $100 or $100,000 on that spectrum analyzer. And so, given the need for one node in the network to have a separate internet connection, nothing MS does online is an improvement over the basic idea -- we absolutely know where the licensed radio is, because it's LICENSED! That license is for a certain areas, and no army of MS spectrum analyzers can be certain that your neighbor can't receive that channel, within the licensed area. Beyond that area, it just doesn't matter -- you get to use that channel anyway.
Years back anyway, CMU taught one "Introduction to Programming" course for Freshmen, anyone in Engineering or Science took this. The gateway to computer science was a largely theoretical course, with very little actual programming (after I took this, 15-211 back in 1980, they split it into two courses a year or two later). The goal was to teach the foundations of computer science, not lock you into any given methodology.
I think the key is to teach different methodologies, and not to rely too heavily on any specific language.
When I was at CMU, they didn't really get into parallel programming. Later, working at Commodore on hardware and learning AmigaOS programming, I caught on that "thread" is just another regular programming construct, just like using "class", "function", "structure", or "method".
And yet, the fact that mainstream software is still not up to that level is a shame. In programming Windows some years back, I found that the Windows API was actually working hard to serialize thing (via the single application message queue) that I had imagined were happening in parallel, as they would have been on any reasonable OS. Linux doesn't get a pass either ... it's a royal pain in the butt to do asynchronous I/O in Linux, something that, as an Amiga programmer, I believe fundamental enough it should be a literal no-brainer.
The real critical thing here... what gets taught, or not taught, at the college level today ultimately sets the scale of things like APIs we're all going to be working with in the future. As much as you'd kind of hope that actually innovations in computer science make it into new OSs, it's ultimately going to be those two 2nd year programmers writing that one module you're interacting with. If they don't understand the need for multithreading, much less the fact it should be ubiquitous, they will screw things up in ways that will at least give you a headache (as a user or programmer), at worse have negative impact on the quality and evolution of software.
Time have changed.
When I was at CMU, Math/CS majors (they didn't have specialized CS/CE degrees at the time... I did "computer engineering" by double-majoring in EE and Math/CS) took "Introduction to Pascal" as a Freshman course (then, as now, you declared your major when you applied). At least the course I took was extremely well done, It taught everything from a modular programming perspective -- at least in as much as that's possible, given late 70s/early 80s Pascal running on large computers (DEC-20s). Having known only BASIC, FORTRAN, and Assembly as an in-coming Freshman, this was a very good course.
The second level course, 15-211, was hardcore computer science. It started with FSMs (Finite State Machines), went though lambda calculus, all sorts of things that might be possible in programming languages... and all book and paper work, until near the end. The point was not to teach any programming language, but to give you the fundamentals for any language that might follow.
And since those days, I've done at least one major project in over 40 different languages. The idea that OOPL was required is a bit odd... beyond the 15-211 course (which had a 50% attrition rate in 1980, and got splt into to course covering the same material a year or two later... I got an "A", and yeah, it was hard enough to make me brag about that "A", even now), there weren't many specific requirements. You have to have certain number of courses at certain levels... I took CAD and Comparative Languages (Pascal, APL, LISP, and SNOBOL4 were the main languages... we spent a day on FORTAN and a day in InterCal), I could have taken Computer Engineering (the large project course... but we did that in CAD).
Languages themselves are often fads. There's no reason a university CS course should get too language specific, but you do need to program in order to learn how to program, and to really get the higher level ideas of CS. The best thing all that extra money I spent at CMU (and paid back over the next ten years) bought me was improved tools for taking on new things. So I picked up Java in a day when I needed it, same with Python and Ruby... RF Hardware Engineering took me a bit longer. But you really don't know where you're going to end up after school... 5, 10, 25 years later.