I've been researching DSLRs for a while now and I see a lot of criticism of the megapixel trend in DSLR manufacturers where more megapixels is regarded as not necessarily a good thing.
I'm interested in a thread on this. My experience with astro CCD cameras would indicate more megapixels is usually a good thing (within reason but I see the KAI1600 chip is not that popular and that is essentially a 16mp STL11 sized chip). I also don't see a lot of people racing to get a new 29mp KAI29050 camera (29mp STL11 sized sensor).
Certainly matching pixel size to seeing and optics is a good thing.
I don't know that matching pixel size to terrestial type photography is as important. There is no seeing to get in the road and normally scenes are fairly well lit.
It seems it is usually an argument of low noise, high ISO performance verus resolution (where small pixel size gives "better resolution").
Some posters act as if you are an idiot if you want large numbers of megapixels.
Normal CCD theory has you wanting signal to rise above the noise floor. DSLR engineers seem to be more advanced than Astro CCD makers in reducing noise with firmware, filters and upgrading their chips with their built in circuitry.
That is another difference between astro and terrestial. Terrestial uses CMOS technology usually and astro uses CCD usually. Each has its advantages. CMOS has the advantage for DSLRs because the amplifiers and circuitry is on the sensor itself which takes up space leaving less room for the pixel to collect light but makes it all very compact and cheaper.
Given this extra degree of freedom of on-sensor circuitry, filters, firmware smoothing it seems DSLR engineers routinely improve noise performance of small pixels to the point where it seems you can have the best of both worlds within a certain range of performance ie; large number of pixels and low noise performance. There must be an optimum number of pixels for full frame sensor with ideal low noise and high ISO performance. What do you think that may be? It hard to say for sure as engineers keep improving the existing but they must hit a wall at some point.
How long before Nikon/Canon put out a medium format DSLR? It must be getting close to that being the next step up for them. The Nikon D800 is already being seen as an alternative now to medium format digital cameras.
The two things for me in regard to mp in a DSLR are how big do you want to print , I never print smaller than A3 . Having the extra pixels also means when chasing wildlife, distant subjects or macro I can crop 1/2 the image blow it up and still print big . The more pixels the easier it is.
I suppose you want as many pixels are you can before noise starts to reduce image quality in shadows and dim areas.
Although I must admit I took some 40D shots with my TEC110 (wildlife) and also a 5D Mk ii. The 5D images were not as good as I expected them to be. The 40D images were not far behind at all. Hardly conclusive as it was only several images though.
Sometimes a crop sensor is better in that it gives that extra zoom factor
but if you can use a full frame and it will shoot also in smaller format then you have the best of both.
Have you experienced much an improvement in images from jumping up a lot in mp?
I have gone from 6mp - 12mp APS and that was a big imprevement in spacial resolution . Now have 12mp full frame and while I can shoot much
easier in low light with much better image quality I have noticed a drop
in image resolution not to mention lens reach . My 400mm has gone from an effective 600mm back to 400mm again without the resolution to blow it back up. This is why I have been waiting for something like the D800 so I can get a bit more out of my long lens and macro. I still use my 12mp APS for macro as it still gives better fine detail . My next printer will be an A2 and even now I can do panorama where the more resolution the better .
How long is a piece of string?
Computer technology is increasing exponentially. When will it stop?
I think as far as CCD's go we are in the early stages of development. Look how far they have come in the last ten years or so. What will they be like in the next ten years?
You only need to compare a CCD with the human eye. The resolution and the exposure latitude of the eye is amazing. I am not sure of the resolution of the eye or what its theoretical mega pixel range would be but I sure it is very high. Granted you can not do a long exposure with our eyes but imagine if you could. Maybe this is were the technology is going something that will replicate the eye.
A system where you could control each individual pixel rather than one blanket exposure.
I think there will be a shift in the way CCD's are created more towards the biological. Same with computers, instead of a binary code which is on or off, why not go to an analog system with various levels of on and off?
I think more pixels are better, the sharper, the more information that is captured the more pleasing it is to the eye. I can imagine that one day you will be able to look at a computer screen it will have a photo of a landscape and you will not be able to tell the difference between the real thing. It would appear as if you were looking through a window.
As long as the consumer demands higher res it will be produced.
The main drawback with going more and smaller pixels (apart from noise)
is that it starts showing up flaws in your lens . If you aren't using the top
pro lens then its a bit of a waste but even now , getting over 30+ mp you will start to see CA that you didnt think existed before. Lets hope lens tech keeps pace with the sensors.
The main drawback with going more and smaller pixels (apart from noise)
is that it starts showing up flaws in your lens . If you aren't using the top
pro lens then its a bit of a waste but even now , getting over 30+ mp you will start to see CA that you didnt think existed before. Lets hope lens tech keeps pace with the sensors.
That is a good point Mark
So for the improvement of ccd's to continue optics will need to evolve just as fast, or what would be the point?
Are there any new and exciting developments in lens technology?
Phil apart from nano coatings and better designs , not a lot of change .
We may see more exotic and expensive glass like fluorite used I guess.
I do wonder how long the 35mm standard will be used untill larger formats
start appearing . They have to keep coming up with selling points some how or other.
As I understand it, diffraction limiting is the one factor that will ultimately limit sensor size increases. For the EOS 5DII, the aperture above which softening occurs due to diffraction limiting is F11. There's a trade off between pixel size and the loss of resolution at small apertures, the laws of physics govern that, ultimately the only way around is bigger sensors, i.e. medium format.
Last edited by acropolite; 11-02-2012 at 07:33 PM.
Nice discussion. I rarely use many megapixels. The 10MP of the 40d is enough for me. Mostly I shoot 2.5MP (I never primt them) and the full 10MP (raw) for astrophotos.
But another limitation is far more evident: dynamic range. Cameras and displays do usually not go beyond 8 bits while our eyes can distingcy 12-14 bits. Simple example: looking into a bright red light looks RED but taking a photo of it (even in the best possible way in RAW) the center is white with a red halo around it. Reason: on a sensor all three RGB pixels are saturated while in our eyes only the red pixels are fully saturated.
This is something I've done a lot of reading about and testing of various cameras. As I understand it, DSLRs a few years ago (say ~ 3-5 years) got to the point where they are operating near theoretical limits. there is very little noise originating in-camera that can be reduced much further (apart from smoothing algorithms which must sacrifice detail).
thermal noise is clearly an issue in longer exposures. but for my primary focus of night landscapes with say 30 second exposures, shot noise totally dominates. read noise, thermal noise and other in-camera sources are negligible.
i tested the original 5D (13MP) verse my 5DmkII (21MP). the original 5D had a few more hot pixels which are easily removed, but otherwise it's performance in low light with short exposures was better than the mkII. many people will find that counter-intuitive but it is purely a statistical function of using bigger pixels to gather the light.
for my type of night photography, the resolution gain is of no benefit.. noise swamps the detail and the lens is working wide open or close to it so it can't deliver the detail anyway.
for long exposure imaging through scopes etc there are a whole different set of criteria and you match the sensor to the optics based on what you want to achieve. but for me.. pixel size rules.
Interesting replies. By the way Phil what do you think of the time lapse feature of the HD video of the new Nikon D800? Its in-camera time lapse but its HD quality (you may be shooting at beyond HD quality).
For low light scenarios for sure I can see the value of larger pixels. It would depend on what your main use of the camera will be. Perhaps its like telescopes - no one camera is best for all uses. The DSLR models do seem to split not only in price but for features best suited for certain applications.
Small pixels usually means smaller wells, more likely to be noisy (although my FLI ML8300 has 5.4 micron pixels - smaller than most DSLRs and it is not noisy at all). I am not sure what effect smaller wells has on daytime photography.
Also is 14 bit or 16 bit camera processing more important to the resulting image more than some of these factors?
I see 5D mark ii and D800 and perhaps others are 14 bit with D800 also processing at 16 bit (not sure what that means if the A/D converter is running at 14 bit but processing at 16bit - not quite the same as A/D running at 16 bit is it?).
But DSLR engineers seem to find little things to enhance the performance of the sensor like better transmission of the Bayer matrix filters, better microlenses, better image processors for less noise.
I did not know of the F ratio limitation for resolution for pixels. Interesting. How is that worked out?
Also what about these Foveon sensors. They don't have Bayer matrixes and no filters at all and all pixels are in the one spot not 2 x 2 grid as in Bayer. Perhaps they have a bright future?
As long as Foveon is in Sigma's hands, it will go nowhere (as far as 35mm goes). You might want to see what Hasselblad (medium format) have done with their 200 megapixel system.
While boasting impressive technology, the cameras are next to useless for astrophotography as they're not designed for high ISO usage. ISO-1600 (if it is even possible) on the particular brand of medium format system you buy, would be so noisy. As a result, a lot of them are crippled at ISO-400, sometimes at ISO-800.
Bert is correct, re: CoC. The CoC value changes depending on the output/size of the print you wish to enjoy. thereby changing the f-/ratio you /should/ theoretically shoot at.
Personally, I use f/16 for maximum depth of field in my landscape work. It's tack sharp, and, I print 30x20s, and the rare 36x24.
I'm not entirely sure how far I can push my medium format system, though. Needs more investigation.
Apologies for the FCW link. It was intended in humour.
Diffraction
Diffraction limits the ultimate detail an imaging system can deliver. This is a fundamental limit due to the wave nature of light.
The trend in digital cameras is to increase the pixel count, and the easiest way to do this is to decrease pixel size. However, practical limits have already been reached! The diffraction spot size in the focal plane is a function of f/ratio and is independent of focal length. The diffraction spot diameter is:
Diffraction spot diameter = 2 * 1.22 w * f / D = 2.44 * w * f_ratio,
where w = wavelength, f = focal length, D = aperture diameter, and f_ratio is the f/ratio of the optical system. The diffraction spot size is given in the table below for green light of 0.53 microns (5300 angstroms) in wavelength:
F/ratioDiffraction
Spot Size
(microns)2 2.6 2.8 3.6 4 5.2 5.6 7.2 8 10.3 11 14.2 16 20.7 19 24.6 22 28.5 32 41.4 45 58.2 64 82.8 A simple way to remember the diffraction size in the above table is to multiply the f/ratio by 1.3 to get the spot diameter in microns.
So using that, this new Nikon D800 with 36.3mp and just slightly larger than 35mm sized sensor would need to be shot at F9 or less and for a telescope probably not more than F8-F11.
So if these cameras get any smaller pixels then we will be limited to imaging lower F ratios below F8 and they would be only useful on fast telescopes or lenses F stops that may not suit the image you want to take.
So Nikon taking their new camera to 36.3mp is probably pretty close to the ceiling where you start to lose flexibility in imaging and diffraction softening of images starts to become an issue. 45mp is probably a workable maximum number of pixels in a 35mm sized sensor.
So that only leaves making the sensors larger if the megapixel race is to continue.
It also puts another perspective on why these F1.2 prime lenses area valuable thing.
Thanks Bert and H. H- I misunderstood your link, I took it as a Nikon bash. I'm neither for or against Nikon or Canon or anyone else just merely interested in the cameras that will get me the best results.
I also see Pentax has a medium format digital camera - the 645 that is bringing the price lower (still very high at around $11,000).
Yes Sigma cameras don't get a good review - its like great sensor, camera acceptable type rating. But the sensor design seems to have the advantage of our LRGB type imaging sharpness. But high ISO noise performance seems backward. One advantage though of the Sigma is you can remove the UV/IR filter easily and replace it easily which could make it very very good for astrophotography - poorish high ISO performance notwithstanding.
