Is 8MP the sky?

[jlobb]

I know that people are always talking about how digital will get more

and more hi resolution until you can go to best buy and buy a 50

gigapixel camera for 35 cents.....but lets be reafreakinlistic.

 

The first 8MP consumer cameras came out over a year and a half ago.

Now what do we have? Camera companies arn't going higher, infact most

of the new cameras that are comming out for consumers are in the 7.1

megapixel range it seems.

 

Did people finally wise up and realize that 8mega pixel cameras are

just plain pointless?

[sl attanapola]

It is due to the fact that the 8MP sensors had problems with noise.

 

When noise issues have been solved then higher megapixel cameras will eventually come out.

 

At the moment the new generations of digital compact sensors are tiny.My Optio 450 has a 1/1.8" sensor which is 4MP.The new 4MP sensors are 1/2.5" and much noiser.

 

I predict that 1/2.5" sensors with 6MP,7MP or 8MP will soon comeout.This will enable production costs to drop too.

 

Also better in camera processing and manufactering techniques will lead to reduction in the amount of noise seen in the final picture.

 

Look at the Fuji F10.It supposedly has the same sensor as the Fuji F810 or a slightly modified version with processing technology from the S3pro dSLR.Noise levels are remarkably lower.

[aaron_johnston]

I don't know that there is any clear manufacturer incentive for consumer models to go higher than 7.0/8.0 megapixels right now. Given that there are already 16+ megapixel pro cameras, it certainly is technologically possible. Whether they can get this resolution into a tiny image sensor is the question.

 

For the average consumer, I think size and design are bigger overall selling points than megapixels. Considering that many of these people still tend to see digital cameras more as gadgets, I don't see manufacturers compromising the sleek and cool element just to accommodate larger image sensors and megapixel counts.

 

Also, with such a wide variety of product out there, manufacturers also need to maintain a good spread to maintain visible differences between consumer, prosumer and professional cameras.

[gsbhasin]

Not ncessarily. People used to say the same when cameras leaped from 4MP to 5MP for the 1/1.8 chip and everyone cringed when they jumped to 6MP or 7MP.

<p>There is no empirical formula to yield the max MP and the noise in dB (SNR) at a given ISO for a chip of certain size with specified pixel size.

<p>I am amazed to see too, glowing reviews of Sony P200(7MP) and Fuji F10.

<p>Regards,<br>Gurpreet

[joeb]

I think economics will limit higher pixel counts. The prices of DSLR's are dropping and the market for a $600+ digicam is getting smaller. High end digicams are being released in the $5-600 range. When the DSLR prices were in the $1500 range, a $1000 highend digicam made sense. As sensor size increases camera cost and quality of images increase. The smaller sensors use cheaper lenses too. The 2/3" sensors are much better than the smaller sensors. The 2.0 to 1.5 senors of the current DSLRs are better still. As the DSLR prices drop and customers are educated on the benifits of the larger sensors the market for highend digicams shrinks. I would not be suprised to see the next generation of highend digicams with 2.0 to 1.6 sensors.

 

I don't think you will see a new generation of the highend digicam's for a year or more. (Highend = 2/3" sensor or better with 8 MP or better sensor)

[johnb]

Jonathan, <br>

Well this is how I look at it:<br>

<br>

Before my 6mg digital I shot film only and<br>

my biggest print was 8"x10"<br>

<br>

Now with my 6mg digital my average print is 13"x19"<br>

and they are better or equal to my best of 8"x10"<br>

even when looked at close in a sunny room.<br>

<br>

So 6mg is enough for me! Now lets get more stops of light<br>

recorded and I will be much happier. (ex. 11 instead of 5)

[jon_noble]

I think they realised that their sensors did not have good signal/noise ratios mainly because of small each pixel's dimensions were getting. Thats whyu i didnt buy a p'n's with a high MP count and a dSLR with less MP.

 

A way around this is to make each photovoltaic site slightly larger instead of improving the circuitry which lead to smaller pixel counts. As soon as they get a way around this signal/noise ratio problem i wouldnt be surprised that MP would go up again. Or they could make full frame sensors instead (wahay!)

[greg s]

"Is 8MP the sky?"

 

No. There is still quite a ways to go in regards to improvemnts for a number of dgital camera metrics, and a willing market for those improvements. The pace of technological advancment has been rapid and it ain't over. You will need a good lens to take advantage of them.

[scott_eaton]

It's a dumb question.

 

Image quality with a digital camera is a correlation between sensor size and total sensor count. I don't care how many pixels you cram into that little sensor - the image quality will still suck and most of us haven't seen much difference in the print quality.

[jlobb]

"Image quality with a digital camera is a correlation between sensor size and total sensor count. I don't care how many pixels you cram into that little sensor - the image quality will still suck and most of us haven't seen much difference in the print quality."

 

exactly what I was thinking. I'll put an 8x10 from an image I take with my D1 up against an image taken with a Sony F828(i think that's the number) anyday.

[gsbhasin]

To Scott, everything is dumb because he holds pre-conceived notions. Give me the equation that demonstrates that you can pack ONLY so many pixels in a given size.

<p>Note I am not saying that you can pack an infinite number of pixels in a sensor of a given size. I am saying it is too early to say whether we have reached the upper limit.

<p>Digital cameras are still following <b>their own version</b> of Moore's law. Have we reached the limit? I dont know. Scott maybe does have an equation that gives SNR for a given wafer with specified photosites.

<p>I am not disputing that such an equation exists. I am in the dark about it. Scott can of course say 'yea that's a dumb statement'.

<p>Scott, they might have made you a hero, but why you are riding a high horse like Don Quixote, all the time? C'mon man get off it, give the horse a break sometime, he must be tired now...

<p>Peace<br>g

[jlobb]

"To Scott, everything is dumb because he holds pre-conceived notions. Give me the equation that demonstrates that you can pack ONLY so many pixels in a given size.

 

Note I am not saying that you can pack an infinite number of pixels in a sensor of a given size. I am saying it is too early to say whether we have reached the upper limit. "

 

Actually, you can only fit them on until your photosensors reach the size of the wavelenghth of blue light. I don't know what that is, but i'm sure if you get on google you could find it.

[michael_spencer3]

Jonathan Lobb wrote: "... Actually, you can only fit them on until your photosensors reach the size of the wavelenghth of blue light".

 

Hmmm. I though the photosensors reacted to the presence of photons. Do photon diameters vary with the wave length? Very confusing.

[jlobb]

You can't have a photosensor smaller than the largest wavelength of light you're trying to catch, and your "photon" is a packet of energy vibrating within a set wavelength. Different wavelengths of light create different colors, that's how your eye knows what color you're looking at, same goes for digital cameras. If the photosensor has a smaller diameter than the wavelength of violet (not ultra violet) light then the light won't hit the sensor all the time, it'll bleed into other sensors and get distorted. You're colors will be all kinds of shitty wrong.

[mark u]

Small sensors have several inherent limitations that are defined by the physics of light. That means that technology won't overcome those limitations, any more than it can create a perpetual motion machine.

 

The first limitation is simply that a sensor that has say 1/16th the area of a larger one can ony capture 1/16th of the number of photons for a given exposure (aperture and shutter speed). For concreteness, compare a roughly APS size sensor of 24x16mm (384 sq mm) with a 1/2.5" digicam sensor (about 24.7 sq mm, almost exactly 1/16th the area). If it has the same number of pixels as the larger sensor, then each pixel also can capture at best 1/16th as many photons, so colour and luminosity are less well defined, even before we take account of the wave/particle features of the quantum physics of light. Remember that a pixel is trying to capture at least 256 distinguishable shades of grey, rather than just representing binary on/off values.

 

The second limitation is due to the inescapable consequences of diffraction. A simple equation that measures this is known as Raleigh's diffraction limit, which states that to resolve adjacent points of light, they must be separated by 1.22 times the light wavelength multiplied by the f-stop. Visible light ranges from 0.4-0.7 microns roughly (blue->red). Some sensors are already using pixels that are less than 2.5 microns across, which implies that beyond f/3 they can add nothing to resolution of red, and beyond f/5 not even to blue, even if there is no other degradation from the lens or sensor optics (IR/UV and anti alias filters and microlens layer) - patently a false assumption. Thus although it is possible to make smaller pixels optical limitations make this a pointless exercise. It also explains why small sensor digicams do not offer apertures smaller than f/8 - images would actually become noticeably less sharp all over.

 

You might think that you can gain by making faster lenses (with a low f-number), but the economics make this a much more expensive route for increasing system sensitivity than increasing pixel and sensor dimensions long before you reach the theoretical maximum achievable widest f-stop of f/0.5. For tiny digicam sensors, yields are already extremely high, and sensor cost in wholesale quantities is no more than a few dollars. Indeed, yields on APS and larger sensors have continued to increase, but still have some way to go to catch up with yields on small sensors. This is evidenced by the trends in prices for cameras ranging from pocketable digicams up to digital medium format backs. Technology will continue to increase yields and cut costs for larger sensors, but the further cost reductions that can be expected for small sensors are now an insignificant proportion of the price of a camera.

 

Another major limit on pixel size that derives from the quantum physics of light is due to photon shot noise. Basically, photons are not transmitted in an even flow, but instead arrive at random intervals that are described by the Poisson statistical distribution. A property of this distribution is that it has a standard deviation that is equal to the square root of its mean (average). If you consider our comparison between sensors, then the "large" APS sensor might capture say 48,000 photons to represent a fully exposed pixel, while the digicam sensor would capture 1/16th or just 3,000 in the pixel from the same exposure. A 5 stop darker area of the image would capture on average 1/32nd of these amounts. However, the standard error due to photon shot noise will be 4 times higher for the small sensor relative to the average number of photons collected. In this example, photon shot noise will be on average over 10% of the signal in the dark area for the digicam sensor. Incidentally, allowing for other sources of noise in the sensor/camera, you probably need at least 3,000 photons captured at full exposure to make full use of a standard 8 bit jpeg file.

 

Things get even worse if you have to use a higher ISO (e.g. trying to keep a high enough shutter speed to freeze action or avoid camera shake in low light) - each extra stop (doubling of the ISO) implies halving the number of photons captured. This basically explains why digicams with tiny pixels produce noisy images in low light conditions. There are several other sources of noise in the sensor and camera electronics that aggravate matters, especially with darker areas of an image in low light. Some further reduction of these is likely, but it gets harder and more costly to achieve. Even using longer exposure times to try to capture more photons does not help greatly, since one source of noise (known as dark current) increases in direct proportion to exposure time, offsetting and eventually overwhelming the benefit of the extra photons (where the noise is only reduced by a square root factor). You (or the camera) can of course apply a noise reduction algorithm to your image, but because noise reduction is a sophisticated form of averaging the effect is to reduce the resolution that the smaller pixels were supposed to produce, again making the smaller pixels pointless.

 

To summarise: for small sensors, pixel sizes have already run into limitations from the physics of light. Although reducing pixel sizes still further is certainly technically feasible even with current technology, there is no point. Improving the quality of images from small cameras can be achieved most cheaply by increasing sensor and pixel sizes, although there is another optical limitation that prevents e.g. an APS size sensor being fitted to a camera as pocketable as an APS film camera such as a Canon Ixus/Elph: such a camera requires a much more bulky lens, since digital cameras need light projected close to perpendicular to the sensor (a limitation that does not apply to film, but also affects wide angle capabilities for digital). Technology may solve this limitation through improved sensor optics and pixel design that will avoid the need for the bulky lens by allowing the sensor to capture more angular light.

 

Meantime, DSLRs are going to get more affordable - $500 new is not far off now and that won't be the lower limit. There are already several (from Canon, Pentax and Olympus) which are similar in size to "prosumer digicams". For those who want to print really big and need more than 8MP, this will come from larger sensor cameras where pixel counts can be increased albeit at the expense of low light capability and dynamic range and probably speed of operation due to the extra data. For those that value the creative control and low light capabilities that come with even larger sensors the cost of entry will fall more rapidly.