photogerald
Senior Member
Do you have the lp/mm figure for the DA70?
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
25 MP APS-C coming?
- Thread starter wildkat2
- Start date
Anders W
Forum Pro
It seems that some contributors to this thread and its sibling
http://forums.dpreview.com/forums/readflat.asp?forum=1036&thread=37437421
are still under the impression that improving the resolution of the sensor has no benefits (or might even be detrimental to image sharpness) if the sensor already "outresolves" (i.e., has higher resolution than) the lens. This is simply wrong.
According to Norman Koren ( http://www.normankoren.com/Tutorials/MTF.html ), the resolution of an image system consisting of a lens and a sensor can be approximated by the formula
1/l + 1/s = 1/i
where l is the resolution of the lens, s that of the sensor and i that of the image. It follows from this forumula that an improvement in the resolution of one component will always improve the resolution of the image no matter how good or bad the other component is (except in the practically irrelevant case where its performance is zero).
Let us assume, for example, that the resolution of the lens is 1000 line pairs per sensor height whereas that of the sensor is 2000. We then have
1/1000 + 1/2000 = (2 + 1) / 2000 = 1/667
i.e., the image resolution is approximately 667. If we now increase sensor resolution to 3000, we have
1/1000 + 1/3000 = (3 + 1) / 3000 = 1/750
so that image resolution has now increased to 750 although the sensor "outresolved" the lens from the outset.
What is true, however, is that improving the weaker link in the chain (in this case the lens) makes more sense than improving the stronger one, if we have to choose between them, and if the cost per unit of improvement is the same. Thus, if in the above scenario, we would have improved the resolution of the lens rather than that of the sensor by 1000 units, we would have
1/2000 + 1/2000 = (1 + 1) / 2000 = 1/1000
so that the image resolution now increases by a greater margin.
Given the above considerations, I have a hard time seeing any down sides to further improvements in sensor resolutions (other than the marginal drawback of bigger files) as long as such improvements are not the only objective of sensor development. And judging from the evidence so far, it is not. The Pentax K-5, for example, has not only higher sensor resolution than any previous Pentax DSLR but also lower noise, even on a per-pixel basis, and greater dynamic range.
http://forums.dpreview.com/forums/readflat.asp?forum=1036&thread=37437421
are still under the impression that improving the resolution of the sensor has no benefits (or might even be detrimental to image sharpness) if the sensor already "outresolves" (i.e., has higher resolution than) the lens. This is simply wrong.
According to Norman Koren ( http://www.normankoren.com/Tutorials/MTF.html ), the resolution of an image system consisting of a lens and a sensor can be approximated by the formula
1/l + 1/s = 1/i
where l is the resolution of the lens, s that of the sensor and i that of the image. It follows from this forumula that an improvement in the resolution of one component will always improve the resolution of the image no matter how good or bad the other component is (except in the practically irrelevant case where its performance is zero).
Let us assume, for example, that the resolution of the lens is 1000 line pairs per sensor height whereas that of the sensor is 2000. We then have
1/1000 + 1/2000 = (2 + 1) / 2000 = 1/667
i.e., the image resolution is approximately 667. If we now increase sensor resolution to 3000, we have
1/1000 + 1/3000 = (3 + 1) / 3000 = 1/750
so that image resolution has now increased to 750 although the sensor "outresolved" the lens from the outset.
What is true, however, is that improving the weaker link in the chain (in this case the lens) makes more sense than improving the stronger one, if we have to choose between them, and if the cost per unit of improvement is the same. Thus, if in the above scenario, we would have improved the resolution of the lens rather than that of the sensor by 1000 units, we would have
1/2000 + 1/2000 = (1 + 1) / 2000 = 1/1000
so that the image resolution now increases by a greater margin.
Given the above considerations, I have a hard time seeing any down sides to further improvements in sensor resolutions (other than the marginal drawback of bigger files) as long as such improvements are not the only objective of sensor development. And judging from the evidence so far, it is not. The Pentax K-5, for example, has not only higher sensor resolution than any previous Pentax DSLR but also lower noise, even on a per-pixel basis, and greater dynamic range.
GordonBGood
Veteran Member
But I don't see that increasing the APS-C sensor photosite density to 25 Mega Pixels (MP's) or higher is some "imaginary marketing pitch".
No, I don't believe that, nor is it necessary to go to higher APS-C MP's than we have now by a large factor or say 10 or more.
No one has talked about the increase in ISO sensitivity capability, or at least nothing that is germane to the discussion as per the OP.
We are a long way from that. Using current technology, each photosite has a full well electron capacity of about 40,000 electrons at ISO 80. If each sensel were to only capture one electron on average, we could look at having photosite densities of 40,000 times as much. Not that I am a proponent of such techniques, ust pointing out that electron capacities aren't the limit to photosite density for APS-C sensors.
Again, an APS-c sensor that can out resolve the most sharp lenses at their sharpest and at their optimum aperture would eliminate the need for Anti Aliasing (AA) filters. For APS-C sensor sizes, this probably happens at about 50 MP or maybe a little higher at 100 or 120 MP for Bayer Pattern sensors using a Colour Filter Array (CFA) . Thus, 25 MP for an APS-C sensor is no problem at all, and likely will produce a resolution benefit for good lenses used where they are sharp.
Regards, GordonBGood
Robert Hoy
Senior Member
That should be changed to "wold enter previous medium format territory". Medium format territory now is 40 MP and higher. There might be a couple of MF backs still being made "new" at 22 MP, but those are essentially low end/entry level models. The guts of MF now is 50 MP and higher.
Would you call Sony's new 16 MP P&S cameras "entering digital SLR territory"? Not a chance.
--
http://roberthoy.zenfolio.com/
http://www.photographybyhoy.com
Robert Hoy
Senior Member
They resolve pixels, sure but how many pixels of real resolution do they record? In the 7 to 8 MP would be about right. They also have a small glass lens designed just for that image sensor and focal range.
Yes Canon is slightly smaller than Sony aps-c. Canon is 1.62x and Sony is 1.55x now and both brands have gotten slightly smaller over the past several years.
The thing I've seen and read about Canon's 18MP aps-c camera is with such small pixels, faster shutter speeds are needed to avoid blurred pixels from camera shake. This eats away at gains in high ISO by needing 1/2 to 2.5 times focal length for sharp photos. Going to 25 MP aps-c would almost require 1/3 x focal length for hand held shooting which isn't much of a problem in good light but lower light will suffer from using high ISO sooner.
--
http://roberthoy.zenfolio.com/
http://www.photographybyhoy.com
MightyMike
Forum Pro
Lets look at this for a moment, 8mp resolved on a 1/2.33 sensor would be like 200mp on a FF sensor or 89mp on a 1.5x APS-C... lets not get into the 450mp potential of current MF sensors...
I think we've got a long way to go before we hit the theoretical limit
--
Mike from Canada
'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote
'
http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=
1/L + 1/S = 1/I
if you have a lens that can resolve 1000 and a sensor that can resolve 3000, 1/1000 + 1/3000 = 4/3000 = 1/750.
Let's say you increase the resolution of the sensor to infinity. The overall resolution will become that of the lens.
So it still holds true that the overall resolution is limited by its weakest component. In theory, the "resolution" always increase, but if a lens is blurry, increasing the sensor resolution will only give you a higher resolution of the blurry image.
The question is, is it of any practical value? My opinion is no, after all you can resample and then smooth out the jaggies of a low res blurry image to get a similarly useless high resolution image that look like that from a better sensor. I can do that in the infinitesimally improbable case where I need a high resolution blurry image. Almost 100% of the time, these detail-less images are just taking space.
if you have a lens that can resolve 1000 and a sensor that can resolve 3000, 1/1000 + 1/3000 = 4/3000 = 1/750.
Let's say you increase the resolution of the sensor to infinity. The overall resolution will become that of the lens.
So it still holds true that the overall resolution is limited by its weakest component. In theory, the "resolution" always increase, but if a lens is blurry, increasing the sensor resolution will only give you a higher resolution of the blurry image.
The question is, is it of any practical value? My opinion is no, after all you can resample and then smooth out the jaggies of a low res blurry image to get a similarly useless high resolution image that look like that from a better sensor. I can do that in the infinitesimally improbable case where I need a high resolution blurry image. Almost 100% of the time, these detail-less images are just taking space.
Cideway
Veteran Member
Actually just looking at DXo results for the lens, it appears that there may be some sample variation going on as the characteristics of the lens have completely changed for one of the cameras, on the other two there is a strong peak of centre sharpness and then it drops off severely to the corners, while the other show no real peak and less fall off to the corners.
--
Chris.
A weather sealed ultra wide, is that too much to ask?
http://www.pentaxphotogallery.com/chriside
GMT +9.5
Pentax SLR talk FAQ
http://forums.dpreview.com/forums/read.asp?forum=1036&message=23161072
--
Chris.
A weather sealed ultra wide, is that too much to ask?
http://www.pentaxphotogallery.com/chriside
GMT +9.5
Pentax SLR talk FAQ
http://forums.dpreview.com/forums/read.asp?forum=1036&message=23161072
Anders W
Forum Pro
Precisely!
Certainly. But my main point is that it is not limited by the weaker component alone (unless the stronger is infinitely strong, which is a practically irrelevant case). Furthermore, we are currently far from a scenario where sensor resolution outstrips lens resolution by orders of magnitude. On the contrary, there are clear reasons to think that improvements in both components will make a practical difference.
Nope. Increasing the sensor resolutions will make the image less blurred than it initially was because the sensor's contribution to the total amount of blur will decline.
Yes, it is of practical value. Smoothening out the jaggies will of course not increase the resolution of the image. Improving the resolution of the sensor, by contrast, will reduce the total amount of blur.
That said, and as I took care to point out in the post to which you respond, it makes more sense to improve the weaker component than the stronger, if (and that's an important if) you have to choose between the two and the per-unit cost of the improvement is the same.
ogl
Senior Member
- Messages
- 1,858
- Solutions
- 1
- Reaction score
- 1,032
approx. f/5.6 for the best lenses. the picture with cheap zoom lenses will be like P&S camera's results.
TheLastMan
Forum Enthusiast
Just out of interest, would there be any difference between the depth of field in a "30mm" equivalent crop from the centre of a 25mp image taken with a 15mm lens, and an image taken on a 6.25mp camera with a 30mm lens?
(assuming the same aperture and shutter speed of course).
Intuitively I would assume that DOF in the crop would be greater - but I have no scientific justification for that assumption.
Cideway
Veteran Member
Well considering that the SDXC standard offers a faster maximum speed then Compact Flash... The fact is what devices actually operate at the speeds that these cards a capable of writing at?
--
Chris.
A weather sealed ultra wide, is that too much to ask?
http://www.pentaxphotogallery.com/chriside
GMT +9.5
Pentax SLR talk FAQ
http://forums.dpreview.com/forums/read.asp?forum=1036&message=23161072
--
Chris.
A weather sealed ultra wide, is that too much to ask?
http://www.pentaxphotogallery.com/chriside
GMT +9.5
Pentax SLR talk FAQ
http://forums.dpreview.com/forums/read.asp?forum=1036&message=23161072
flektogon
Veteran Member
This formula is valid only for analog systems, like film cameras, not for digital systems.
flektogon
Veteran Member
For the digital systems (like digital cameras) you need to use 2-times higher sampling frequency (= 2-times more pixels/mm) than the resolution of the lens in line pairs / mm. So, if the lens resolution is 100 lp/mm, the sensor should have 200 pixels/mm density.
HonourableTyr
Senior Member
Really that depends on the CoC size being used and how large you print. The larger you print the less resolution you have.
1MP would give you a very poor image. I never see much point in digital zoom other than for small PP crops/straightening.
--
Regards,
Tyr
My Photos:
http://www.pentaxphotogallery.com/danielfranco
MightyMike
Forum Pro
the 30mm crop of a shot that was taken at 15mm F4 would have the same DOF as a shot at 15mm F4, but it also depends of print size and CoC
--
Mike from Canada
'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote
' http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=
Thank u so much for the explanation! Its really useful. Now the dxomark and photozone data make sense.
Lets say dxo has sample variation in this particular case. But dxo shows improvement of resolution by putting it on a higher MP sensor in every case . This is true for any camera and lens combo. Photozone also shows similar result. See Anders W's explaination how it happens.
Robert Hoy
Senior Member
You are assuming that an entire 35mm lens glass can be the same quality has the small P&S lens glass is and that is probably asking a lot.If you crank up the pixels to 89 MP on aps-c, the result will be P&S image quality which means ISO 200 or higher will be near useless!
Theoretical maybe, but that won't yield very good image quality.
--
http://roberthoy.zenfolio.com/
http://www.photographybyhoy.com
Similar threads
- Locked
- Locked
Keyboard shortcuts
- f
- Forum