Pixel race will never stop

No, it will not, sure about that.
Amount of pixels on a sensor will grow every year.
But, how much pixels do we really need ?
What if sensor gets interchangable. to choose your own set-up ?

Comments are welcome, thanks in advance.

Have a nice day.

Best regards, Herman

Yes, it will !

Look at the GF1...

Canon ? Look at the G11...

A lot less pixels then G10 but a better sensor.

Yip, we are all looking for a superior 10MP sensor.
But I believe that pixel race has started and it will never stop.

Have a nice day.

Best regards, Herman

I don't believe so, at least not for the average camera user. It is not quite the same situation as with computers (Moore's law). If the number of pixels continues to rise across the board it means bigger hard drives, faster, more efficient operating systems and photo applications, ever faster processors, interfaces and front side buses in computers, faster servers, faster networks, and all the rest. It all depends what you need to get the job done to the level required. There are many other areas manufacturers can explore to improve quality, particularly in smaller cameras, since it is clear that with the small sensors being used cramming on ever more pixels is not the answer, even if has been seen, until now, as a good sales pitch for the average, relatively unknowledgeable, buyer. The Canon G11 example (having fewer pixels than the model it replaces) is indeed one which shows that a measure of sanity is returning to the situation.

Herman Dijkhuis said:

No, it will not, sure about that.
Amount of pixels on a sensor will grow every year.
But, how much pixels do we really need ?
What if sensor gets interchangable. to choose your own set-up ?

Comments are welcome, thanks in advance.

Have a nice day.

Best regards, Herman

Click to expand...

Charlie Le Denn said:

No, it will not, sure about that.
Amount of pixels on a sensor will grow every year.

Click to expand...

For which the just-announced batch of main-stream consumer p&s cams by some of the major companies is proof.

But also on the DSLR side, megapixels seem to be increased with each generation. Just have a look at the Canon 7D, which has more megapixels than the 50D or 500D (which already have ENOUGH mp), but look at the noise ...

I believe that with current technology (standard Bayer-pattern CCD and CMOS sensors), there is a limit as to what can be done. If you ask me, the future is a 6 or 10 mp Foveon sensor, or the technique that Hasselblad uses for their latest medium format cameras, sensor rotating, to give an effective resolution that equals the megapixel count. Imagine 10 mp with a REAL 10 mp resolution ...

Simply adding more pixels, especially on small sensor cameras, apparently does not add quality to the image, but on the other hand, on both the FF and DX end, the cameras with more pixels tend to resolve finer detail than their predecessors. More pixels, with technology that also reduces noise and helps to capture more light, seems to be part of the "better sensor" equation. It's just that all 14 (or 12 or 24) MP sensors and their accompanying hardware/software are not created equal. The limit, to me, is where the sensors are out-resolving even the finest lenses that can be made. At that point, more pixels does not add more quality, no matter how well engineered. Until someone comes up with lens designs that resolve even more detail, of course!

I think its a matter of simply advance. Back in film days, the grain was increasingly improved. Infact they were so improved that by the mid 80's even consumer neg film of nominal ISO 100 and 200 are well up to the typical 8X10 enlargement, and even ISO 400 would do. But still film did get improved all the time and still do today.

Similar logic is about the sensor MP, but lets not forget as the film goes, its not only the grain structure got improved, but the chemistry and all the things associated. As long as there's way to made it perform better and ( remain commercially viable and attracting ), the improvement will continue ...

So if say 10 years down the road, there is a sensor in 4/3 that will deliver 8KX6K resolution, and also imaging , signaling, and processing firmware / software had made it so the image quality can be maintained right up to the high ISO range .. then why not. No we won't need all that, but just as many things , having the capability is something. Needing them is something else ...

--

• Franka -

Does the lens resolving power can catch up with the megapixel? I read somewhere that Olympus lens max resolving power around 18megapixel

Herman Dijkhuis said:

No, it will not, sure about that.
Amount of pixels on a sensor will grow every year.
But, how much pixels do we really need ?
What if sensor gets interchangable. to choose your own set-up ?

Comments are welcome, thanks in advance.

Have a nice day.

Best regards, Herman

Click to expand...

--
malim
http://www.flickr.com/photos/akupunyahal

It's not like more MP doesn't have it's downsides: larger file size, poor performance at higher ISO's, brings out flaws in the lenses...

Seems to be stabilizing in the 12-14mp range. Beyond that, it doesn't benefit most people, as most people typically don't print over 8x10.

Also, we're seeing a shift in display methods as photos are increasingly published for web displays, and LCD photo displays drop in price. These devices are emissive while photo prints are reflective, in that one method emits light while the other reflects it. Emissive devices are far more precise than reflective, so their resolution requirements are much lower, less than half that of ink/dye/toner on paper.

This shift in display technology from reflective to emissive has other ramifications: the difference between the large sensors and the smaller sensors is becoming increasingly irrelevant.

Herman Dijkhuis said:

No, it will not, sure about that.
Amount of pixels on a sensor will grow every year.
But, how much pixels do we really need ?
What if sensor gets interchangable. to choose your own set-up ?

Comments are welcome, thanks in advance.

Have a nice day.

Best regards, Herman

Click to expand...

How many pixels do we need? I'm sure it depends on the circumstances of usage. For myself, 12 MP are probably enough most of the time. I discovered recently that I can make quite adequate 30"x60" panoramas from single shots with that as long as I don't have to crop too much (if you get right up on the image, 6-8" away, and look carefully at the areas of strong contrast, you can see the individual pixels, but overall it is impressive to me and the others who have seen it.)

Choose my own? hmm...

I'd love to have Foveon pixel by pixel sharpness/contrast transitions. If we could get a 6-8 MP Foveon in a modern (responsive) body at u4/3 size, it'd be a great choice for any well lit application.

I would really like to see an electronic shuttered EXR style sensor (16-20 MP? - which for most purposes means 8-10 MP, with the super CCD benefits in resolution and artifact challenges) in someting in this size class, too. The research I did when the Fuji u4/3 rumors started running around got me excited about the potential for this. Having simultaneous still and video shooting, being able to set the DR of the camera to whatever I want it to be to capture the scene. A complete dataset for focus and exposure during continuous shooting.

But as much as I'd like both of the above, I really am pretty happy with the GH1 sensor - the multi-aspect ratio nature of it is nice, and it delivers excellent images. I should go take more pictures...
Walter

Personally, I would be willing to sacrifice high ISO for better detail, so I'd be happy with a 24 or 40 MP sensor limited to ISO 400 -- maybe 200. Of course, that's not for everyone, but seems like there is enough market for the camera makers to offer some choices.

Just how much we can really use is anybody's guess, but some testing with my Epson printer indicates it can use at least 360 ppi resolution, which works out to a little over 10 MP for an 8x10 -- after crop -- or 40 MP for a 16x20. There's a very good argument that the sensor needs to outresolve the lens, though no one really knows by how much. I've seen numbers anywhere from 2.5x to 8x, but we really won't know until such sensors become available.

The pixel race may slow or plateau, but it's not over.

Gato

--
Street Fashion and Alternative Portraits:
http://www.silvermirage.com

Yea, your 150% right... the trouble is what Joe Average consumer wants, more pixels is easy... big numbers am alwayz betta! Thats why I think Canon digging their heels in with the G11 was SO great and such a breath of fresh air, it shook a huge section of the prosumer market by the scruff of their necks and forced them to rethink a stupid race that crammed as many pixels as possible into as small a sensor as possible while burning exposure quality and flexability to do it.

Gato Amarillo said:

There's a very good argument that the sensor needs to outresolve the lens, though no one really knows by how much. I've seen numbers anywhere from 2.5x to 8x, but we really won't know until such sensors become available.

Click to expand...

Outresolve the lens? Interesting notion.... I guess one or the other will bottleneck if unmatched but the overall effects of the bottleneck would look way different visually... hmmm... I am sure they would have bumped up against this to HUGE depth and complexity in astronomy... hmmm... yea... professional optical astronomers would know the real deal on this even if it’s a fringe topic to photographers... they would probably have entire volumes of lore on it.

If you wanted to simply look at the image and did not study it scientifically with a microscope there would have to be a threshold where differences in image/sensor quality would be undetectable to the eye because it had reached its resolution limit.... like doing a 1:1 scale image reproduction of a sensor or piece of film... you would simply not know what ISO the film was or what resolution the sensor was even if the values varied by huge amounts because the level of detail exceeded the ability of your eye to resolve it.

Glass will be the limit now ... 4/3 glass will top out about 18 or so Mpix and full frame maybe 60Mpix - at the moment the sensor sweet spot is about 12Mpix for 4/3 and will corrospondingly be about 48Mpix for FF....

beyond this glass will have to get much more pricy - and its not cheep now!

The main improvements will be ISO - imagine geting clean pure 25,000 ISO.... on 4/3... this would also alow very low light video still keeping shutter as 60fps...

Read speed - 120fps on 12Mpix image... recorded as a 4k video stream....

Forveon (or whatever that fuji thing is called) sensor quality - @ 15Mpix (4/3)....

Once sensors can be made clean with good ISO on say a 1/3" chip you will see more acurate lenses made but smaller - ok DOF will suffer, but you could make somthing to rival a 5Dmk2 with a tiny F0.95 penny size lens with a range of 18mm - 200mm.... still would be expencive cus that galss would have to be somthing - but far more sensable than the huge chunk of glass for a FF camera!

Herman Dijkhuis said:

No, it will not, sure about that.
Amount of pixels on a sensor will grow every year.
But, how much pixels do we really need ?
What if sensor gets interchangable. to choose your own set-up ?

Click to expand...

The problem is the more pixels you have per square mm of sensor, the more diffraction becomes a problem. I would imagine that just about all of the new p&s cameras that were announced at CES are diffraction limited at all focal lengths. This means while you may have 14MP of pixels, you may only have 8MP or so of useful information. Now, when you reduce to web size viewing or print at 4x6 or even 8x10, you average several single pixels to get one output pixel. So it looks ok when you put it to normal use, but if you need to do extreme crops or print large, it won't hold up all that well.

For instance, I plugged in the numbers for a 1/2" sensor, which is slightly larger than the 1/2.33 sensors in the recent crop of cameras. According to the calculator, it says diffraction will start having an effect at any lens that is f/2.5 or slower. Now, in practice, I think you can probably get away with a stop or so before it becomes too apperent, so lets say the diffraction starts becoming apperent at f/4.

Picking an arbitrary Olympus p&s camera that was recently announced, the Olympus muju 9010 has a f/3.2-5.9 lens. So ok, if you are shooting at the camera's wide angle position (28mm equivalent field of view to a film camera), it will give you a good picture, but if you zoom to 280mm equivalent field of view, the image probably isn't as sharp as it could be and needs more sharpening to be acceptable.

Here is the calculator I used, which better explains diffraction, etc.:
http://www.cambridgeincolour.com/tutorials/digital-camera-sensor-size.htm

Here is the dpreview information on sensor sizes:
http://www.dpreview.com/learn/?/Glossary/Camera_System/sensor_sizes_01.htm

So why do manufacturers keep bumping up the number, when most of them probably know it will not give better pictures? Because it is only 3 numbers most sales people hawking cameras quote (what is the price, what is the zoom range, and what is the megapixels), and what consumers have to compare against. Remember, camera makers do not sell directly to us consumers, they sell to the corporate buyers of Walmart, Newegg, Best Buy, etc. and those buyers buy on what they think their sales staff can sell.

So if your sensor has higher resolution than your optics can resolve, you can get rid of the anti-aliasing filter, making the camera LESS expensive to produce. I would suspect that is some of the drive behind cranking up the MP.

I agree though, that beyond a certain point, the extra MP do nothing for anyone but the marketers. It will be interesting to see how the back-lit sensors shake up marketing, if at all.

--
My pictures...
http://picasaweb.google.com/wymanfamily3
http://www.markwyman.com/photos/default.asp

Agree for the most part, but physics still applies. The always lusted after shallow depth of field is nearly impossible on small P&S sensors and still a little hard on M43rds.

Heat and thus noise will be a bigger problem on a small 12 megapixel P&S sensor than a 12 megapixel FF sensor because of simple physical density of the imaging elements.

Herman Dijkhuis said:

No, it will not, sure about that.
Amount of pixels on a sensor will grow every year.
But, how much pixels do we really need ?
What if sensor gets interchangable. to choose your own set-up ?

Comments are welcome, thanks in advance.

Have a nice day.

Best regards, Herman

Click to expand...

I would prefer better iso than more pixels. As for an example, I prefer the D700 over the A850.

Michael Meissner said:

The problem is the more pixels you have per square mm of sensor, the more diffraction becomes a problem. I would imagine that just about all of the new p&s cameras that were announced at CES are diffraction limited at all focal lengths.

Click to expand...

There is still various aliasing with a 12Mp 1/2.3" like this one:

http://www.imaging-resource.com/PRODS/SD940IS/FULLRES/SD940IShRES.HTM ; [12 Mp file!]

which suggests we'll still see an improvement from going higher.

Michael Meissner said:

This means while you may have 14MP of pixels, you may only have 8MP or so of useful information.

Click to expand...

14 Mp Bayer of course means 7Mp green and 3.5Mp each of blue and red. Seen that way, it's not that much.

Michael Meissner said:

Here is the calculator I used, which better explains diffraction, etc.:
http://www.cambridgeincolour.com/tutorials/digital-camera-sensor-size.htm

Click to expand...

It's a bit too simplified. The criterion is based on a certain contrast, but contrast can be enhanced by PP. More on that here:
http://forums.dpreview.com/forums/read.asp?forum=1018&message=30447131

I'm sure 4/3" would still see a benefit from many more than 12 Mp, maybe even 100, though it obviously depends on the lenses.

It would be nice to not see file size grow with Mp count, so hopefully the manufacturers will get better at offering reduced bit depth, visually lossless compression using level thinning, or even lossy compression and binning (the result of which would need new demosaic algorithms).

The nice thing about high pixel counts is that the pixel contrast gets lower so the demosaic algorithms don't have to try so hard to extract every ounce of detail and mazing and other artefacts will be gone.