In theory pixel count can go as high as technology and cost allows us, the lenses won't become the limiting factor. In practice, if resolution of a sensor goes up one would like to have lenses with higher resolution too. Good L glass from now is good enough to give us satisfying results even at a futuristic 100Mp. When we are there, probably new optical techniques and new sensor technologies will make the kind of extrapolations we are making from the moment of now anachronisms that will make us smile then
The lens and the sensor each limit the total imaging system resolution. Even though current DSLR sensors are "more limiting" in terms of resolution than lenses, we can clearly see the differences in lens performance today.
The 1Ds is already brutal in revealing poor lens performance - this will get worse as the MP count increases. I don't know about Peter's 100MP claim - at some point we'll reach the diffraction limit for resolution in the 35mm
(or APS) sensor size. My guess is somewhere around 40 MP for a 35mm sensor.
Peter de Weerdt wrote:
In theory pixel count can go as high as technology and cost allows us, the lenses won't become the limiting factor. In practice, if resolution of a sensor goes up one would like to have lenses with higher resolution too. Good L glass from now is good enough to give us satisfying results even at a futuristic 100Mp.
Peter
100Megapixels--I guess I better get a bigger hard drive.
I think as the years go on we'll see fewer and fewer restrictions on size.
Imagine when a terrabyte is used to describe a single image...600mb isn't so much really...it would actually be a remarkable image...
Personally I'm looking forward to an infinite hard drive...hopefully one day the limitations of how much "space" we have will be gone...
With all respect I disagree. My line of thinking is more like Norman Koren's http://www.normankoren.com/Tutorials/MTF7.html rather than Michael Reichmann's opinion. A basic misundertanding on sharpness, very often made is that either the lens or the sensor (or film if you like) limits the other. But the the truth is that lens and medium together create the image and and resolving power is the average of both. In a formula I have given here more often it reads like:
1/ resolving power of the lens + 1 / resolving power the medium = 1 / total resolving power.
That means future sensors with higher resolutions will add to the total resolving power no matter how good or bad the lens. I am always amazed how easily people, even knowledgeable and highly respected photographers, forget this while everyone is able to see it themselves in a low resolution newspaper photo: it is easy to recognize the high quality optics shot from the bad lens of a simple boxcamera for instance, even though the resolution of the newspaper is below the resolution of the boxcamera lens. The above equation explains simply why, but the "one component limits the other" thinkers forget it all the time.
Well Peter, I think Shannon's Sampling Theorem will prove you wrong. According to your formula, if you have a 100 lp/mm lens and 100 lp/mm sampling device (200 samples/mm), the system resolution is 50 lp/mm.
But Shannon's Sampling Theorem states that if you encode a signal at more than twice its maximum frequency (with infinite precision and accuracy), then you will be able to decode the exact original signal. A lot of things we use every day, such a landline phones and the GSM phone you carry everywhere, rely on this.
I think I'll place my bet with the latter -- a 100 lp/mm lens in front of a 200 samples/mm capture device will produce a 100 lp/mm encoding. Every time.
Coming at the from another perspective; Are the lenses we use today going to be the considered the 'Golden Age' in the years to come? I can forsee, especially with products like DO Lab's Optic Pro, that lens quality might just take a back seat to software correction of lens faults.
If you take computer hardware as a model, DO's software could eventually be integrated into the camera's processor. Match the correction algorithms to specific optics, and do the correction and enhancement in camera.
How does this tie in with the original question? With software, the sensor and lens will make a processing triad. The lens won't outresolve the sensor or vice versa. Taken to extrremes; Why build fantastic lenses if the camera can correct the faults?
To paraphrase Sony/Philips: Perfect lenses forever?
With all respect I disagree. My line of thinking is more like Norman Koren's http://www.normankoren.com/Tutorials/MTF7.html rather than Michael Reichmann's opinion. A basic misundertanding on sharpness, very often made is that either the lens or the sensor (or film if you like) limits the other. But the the truth is that lens and medium together create the image and and resolving power is the average of both. In a formula I have given here more often it reads like:
1/ resolving power of the lens + 1 / resolving power the medium = 1 / total resolving power.
That means future sensors with higher resolutions will add to the total resolving power no matter how good or bad the lens. I am always amazed how easily people, even knowledgeable and highly respected photographers, forget this while everyone is able to see it themselves in a low resolution newspaper photo: it is easy to recognize the high quality optics shot from the bad lens of a simple boxcamera for instance, even though the resolution of the newspaper is below the resolution of the boxcamera lens. The above equation explains simply why, but the "one component limits the other" thinkers forget it all the time.
I think Reichmann might be talking about lens performance wide open or in the corners, where the sensor is currently better than the lenses. A good lens stopped down is certainly better than the sensor at this time.
warp_foo wrote:
If you take computer hardware as a model, DO's software could eventually be integrated into the camera's processor. Match the correction algorithms to specific optics, and do the correction and enhancement in camera.
Mike
You should read Mike Johnston's collumn for this week. No ammount a software processing will allow a imaging system to deliver more detail than can make it though the lens in the first place.
bouch wrote:
You should read Mike Johnston's collumn for this week. No ammount a software processing will allow a imaging system to deliver more detail than can make it though the lens in the first place.
All lens designs are tradeoffs. Tradeoffs between things like resolution, sharpness, contrast, all kinds of aberrations, maximum aperture (speed), and above all cost. The most obvious, sharpness, is a non-issue with digital since there are so many easy ways to enhance local contrast (sharpen). Some aberrations, like pincushion/barrel and chromatic aberrations, are fairly easily corrected. Just like an image is sharpened in processing, it could be corrected for, say, barrel distortion. But this obviously requires the lens to communicate some sort of parameters or characteristics to use in this process. A lens that doesn't need to be highly rectilinear and exhibits chromatic aberrations less than 2 pixels wide could have other benefits -- like improved flatness of field or resolution. Contrast also perhaps isn't as critical, as long as it's "good enough". The weaknesses are then reduced to a point where the lens provides a lot more apparent optical quality for the money. Obviously, this doesn't help where money is no object, or for scientific or instrument-grade applications, but in the digital photo realm it's a great idea. Digital photography, like in audio, isn't about raising the high-end bar, but about dramatically improving quality at the more casual end.
Jan - I was only talking about resolution, not any other aspect of lens performance. I think you would probably agree that you can't improve resolution through software. I agree with you that you can use software to correct some other types of lens defects. I don't really have too much trouble correcting CA & vignetting in PS now. The problem with a lens feeding this kind of aberation info to the camera is that the degree of the apparent aberation is subject & exposure dependent.
bouch wrote:
Jan - I was only talking about resolution, not any other aspect of lens performance. I think you would probably agree that you can't improve resolution through software. I agree with you that you can use software to correct some other types of lens defects.
I see your point and agree of course. I was just trying to make the argument that if a lens doesn't need to be so well corrected for aberrations, then designer can focus on resolution -- and so we'll get better technical quality for the same money.
what is not to like about this camera! soon I will put my review in the mix , but in short it is maybe the finest thing I have ever held in my hands , intuitive, smart, transparent, finely engineered, and the images are sublime , so sharp, rich and flexable in post if needed but really I am doing less in post than I ever thought possible
thanks for asking
J
Edited by jmcfadden on Mar 17, 2004 at 03:06 PM GMT
