8.0mp vs. new 10.1mp

celter wrote in post #2347676
They did a test that was shown on TV the other day. It was a picture printed to about 6x8. They had one version with 10 mpix,the other was 1mpix. The audience could not see the difference.

While I tend to agree that in the vast majority of instances there's only very minor difference between the appearance of an 8mp image, and a 10mp image, I wouldn't place too much emphasis on what the audience in the aforementioned test perceived.

Just as casual music listeners taking part in tests of audio quality often perceive no real difference between a $39 FM radio and a $3900 audio salon set-up; and casual tv viewers perceive no real difference between "regular" tv playing a movie on a VCR, and the finest HDTV playing the movie from a DVD; so do most audiences notice little real difference between 10mp and 1mp. Or (a decade or so back) between an 8x10" print from an ISO 100 negative, and an 8x10" print of the same scene from an ISO 800 negative.

None of this means that there are no differences. Just that casual inspection - the kind done by most people, most of the time - doesn't immediately pick up on it.

Again, though - the difference between 8mp and 10mp is such that even careful inspection by a serious viewer may not reliably pick up on it. In some situations, yeah. But in most situations, quite possibly not. I mean, it's the difference between around 3400x2300 pixels, vs. 3800x2600 pixels, more or less. (I'm doing the numbers from memory here, so I may not be exact to the very pixel.) Call it around a 12% difference in linear resolution. By most standards, that's pretty minor. Oh, I'll take it if I can get it, sure. But I won't lose any sleep over it, if I can't get it. Most of the time it's probably less significant than the difference between shooting with good glass vs. very good glass. Or handheld (at relatively fast shutter speed) vs. tripod.

12345Michael54321 wrote in post #2349108
... Most of the time it's probably less significant than the difference between shooting with good glass vs. very good glass. Or handheld (at relatively fast shutter speed) vs. tripod.

Take 24-70mm f/2.8 as an example. If a GOOD lens is defined as Sigma 24-70mm f/2.8, and if a VERY GOOD lens is defined as Canon 24-70mm f/2.8. Then, the difference between a good lens and a very good lens is smaller than the difference between 8MP and 10MP. (See photozone.de for lens resolutions).

So if one is seeking highest resolution, the combination of a good lens + 10MP will produce better results than the combination of a very good lens and 8MP.

As many pointed out, the difference between 8MP and 10MP is small, I like to share my experience of how small that is.

I just printed out two 8"x12" photos. One had 6.3MP, the other 2MP due to heavy cropping. To my surprise, I could just barely tell the difference in quality, and others around me could not tell any difference.

pieq314 wrote in post #2350054
Take 24-70mm f/2.8 as an example. If a GOOD lens is defined as Sigma 24-70mm f/2.8, and if a VERY GOOD lens is defined as Canon 24-70mm f/2.8. Then, the difference between a good lens and a very good lens is smaller than the difference between 8MP and 10MP. (See photozone.de for lens resolutions).

So if one is seeking highest resolution, the combination of a good lens + 10MP will produce better results than the combination of a very good lens and 8MP.

The differences in measured lens resolution are completely subordinate to enlargement factor -- and this is true even of the enlargement factor difference between APS-C and full frame. In fact this is true of megapixel resolution as well: it is far less important than enlargement factor. This is because the smaller the sensor (physically), the more desperately dependent the camera becomes on top glass quality (and this is largely independent of pixel count).

The reason for this is as follows (and if you'd like a reference to this stuff, PM me, I have an article written by a physicist on the subject):

Also keep in mind that the example I raise below is assuming an image with extremely fine detail.

Lens resolution is measured in lpm (line pairs per mm). 100 lpm means a lens can resolve 100 line pairs per mm (as recorded on the gold standard of sensor / film resolution, which is ISO 25 technical film). A lens with 50 lpm resolves exactly that, 50 line pairs per mm. So if you take a 100 lpm lens and show it a 100 lpm target, it will resolve 100 lpm. If you take a 50 lpm lens and show it a 100 lpm target, it will only record 50 lpm. In other words, an exceptional 100 lpm lens will project twice the fine detail of an above average 50 lpm lens. (Keep in mind, parenthetically, that not a single one of the Canon L-lenses makes it to 100 lpm -- we're talking Mamiya, Zeiss, Leica, Schneider, Rodenstock territory -- but the best L-lenses make it to 80 lpm).

Now look at the other end of the spectrum here, which is output. If you have an APS-C sized sensor, which is 15x24mm (0.6 x 0.9 inches), and you want to make an 16x20 inch print, you need to enlarge the captured image 27-fold. But you're enlarging the baseline image information 27 times.

So say you have an outstanding lens at 75 lpm. This means that in every mm of your sensor you can resolve up to 75 line pairs. When enlarged to 16x20 inches, these 75 line pairs are no longer 75 per mm, but rather 75 per 27mm (because you have a 13-fold linear enlargement, but you're not capturing any new detail in the enlargement). So your final print, even if you have a billion megapixels, will be 75/27 = 2.8 lpm.

When viewed from 10 inches a way, the human eye (with normal vision under normal lighting) will regard as sharp any image that is 7-12 lpm. So this enlargement from an APS-C image will not look sharp to the normal human eye until you are minimally 25 inches away. This is physiologic, and it has to do with the physical spacing between cells in our fovea (the part of the retina with highest visual acuity).

If this same image were taken on a 5D or 35mm instead of APS-C, you would decrease the enlargement factor by 1.6 times, so the print resolution would be 4.5 instead of 2.8, and the minimum viewing distance would be 15.6 inches away. So just by using APS-C instead of full frame, you will have to back nearly twice as far away for the image to look sharp (again, assuming an image that had at least 75 lpm of real world detail when it hit the lens).

How much better would a lens have to be to match this difference? Simple: 75 lpm x 1.6 = 120 lpm.

120 lpm

Zeiss claims a lens with this resolution (140 in fact), but this claim is highly controversial. But in the real world, there is no commercial photographic lens with resolution like that -- maybe the Hubble telescope has lenses like that. Your best lenses hit 80 lpm in the center at f/8 if you're lucky.

So the point is not to plug the 5D (which I don't own) over APS-C (which I do own). The point is that many of these lens resolution and sensor resolution discussions we have are really absurd given that it would take impossible improvements in lens quality to make up for the inherent size-imposed limitations of the sensor.

Again, PM me if you want references for this.

sandpiper wrote in post #2347899
Yes, if you think that those added features are important to you.

There isn't going to be any significant difference in the image quality, so it's the features you are paying for.

If you want to improve the image quality, you might do better with a used XT and put the saving towards better glass. Top quality lenses will make far more difference to IQ than the extra pixels.

Only you know how much difference the new features will make to you.

Yeah I've read a few reviews where they said the lens that comes with it isn't very good, so putting the difference in cost onto a better lens is a good suggestion.

Here's a list I got of the new stuff, would appreciate comments on their worth:

1) New 10.1 effective Megapixel CMOS sensor (versus 8.0 on the Rebel XT)
>> so we've already determined it's not a big difference

2) New dust reduction system
>> sounds like a good thing to have, does dust affect the life of the camera or just the image quality? (ie. spots on the picture) And how would you clean the dust without this feature?

3) Large 2.5" LCD display (versus 1.8" on the XT)
>> this seems like a nice feature, being able to see things a little more clearly

No more LCD info display (the main LCD is used instead)
>> well this creates room for a bigger LCD, anyone who's experienced both types of info displays, does this make it easier to use?

4) 9-point autofocus system (versus 7-point)
>> this one I don't know anything about can someone explain what exactly this means, and if 2 more points makes much of a difference?

5) Can now take more shots in a row in burst mode (27 vs 14 JPEG, 10 vs 6 RAW)
>> always been interested in having this feature on a camera, it nearly has double the amount of pictures in a row

6) New Picture Styles feature, first seen on the EOS-5D and 30D
>> I guess this makes it a little more user friendly for someone who doesn't know the best manual settings for landscape, portrait etc

7) Battery life is 10% lower than on the XT
>> I think I read this is due to the larger LCD, 10% doesn't sound like a lot but when actually using it do you notice it more? Also read the battery grip adds more battery life.

Any other comments welcomed

oh and after I've researched the two a little more I will check them out in the store to get a better feel for them too.

I'd love to see lots of scientific tests on this issue as well as a few others. Real tests! Same with Raw vs jpeg shooting, real scientific tests where folks were tested if they could actually see any differences. Kind of like a blind test, coke versus pepsi where so many folks swear they can tell a huge difference and then fail to be able to tell in a blind taste test. This question, as like many others is tainted by personal investment in the answer instead of neutral observations.

vicereine wrote in post #2350589
7) Battery life is 10% lower than on the XT
>> I think I read this is due to the larger LCD, 10% doesn't sound like a lot but when actually using it do you notice it more? Also read the battery grip adds more battery life.

Any other comments welcomed

oh and after I've researched the two a little more I will check them out in the store to get a better feel for them too.

While I cannot comment on a comparative, as I own the XTi, but have never owned the XT, battery life on the XTi is quite good. I was at the zoo today at 0c temps, and shot ~200 pictures all with IS on, and the battery level never moved the entire time.

To me, the AF was reason enough to go for the XTi over the XT. Take this with a grain of salt however, as I picked up the XTi for only $20 more than the XT ($849 Cdn with kit lens at a retail store)

vicereine wrote in post #2350589
Yeah I've read a few reviews where they said the lens that comes with it isn't very good, so putting the difference in cost onto a better lens is a good suggestion.

Here's a list I got of the new stuff, would appreciate comments on their worth:

1) New 10.1 effective Megapixel CMOS sensor (versus 8.0 on the Rebel XT)
>> so we've already determined it's not a big difference

2) New dust reduction system
>> sounds like a good thing to have, does dust affect the life of the camera or just the image quality? (ie. spots on the picture) And how would you clean the dust without this feature?

3) Large 2.5" LCD display (versus 1.8" on the XT)
>> this seems like a nice feature, being able to see things a little more clearly

No more LCD info display (the main LCD is used instead)
>> well this creates room for a bigger LCD, anyone who's experienced both types of info displays, does this make it easier to use?

4) 9-point autofocus system (versus 7-point)
>> this one I don't know anything about can someone explain what exactly this means, and if 2 more points makes much of a difference?

5) Can now take more shots in a row in burst mode (27 vs 14 JPEG, 10 vs 6 RAW)
>> always been interested in having this feature on a camera, it nearly has double the amount of pictures in a row

6) New Picture Styles feature, first seen on the EOS-5D and 30D
>> I guess this makes it a little more user friendly for someone who doesn't know the best manual settings for landscape, portrait etc

7) Battery life is 10% lower than on the XT
>> I think I read this is due to the larger LCD, 10% doesn't sound like a lot but when actually using it do you notice it more? Also read the battery grip adds more battery life.

Any other comments welcomed

oh and after I've researched the two a little more I will check them out in the store to get a better feel for them too.

1)I agree.Only in extreme conditions(LARGE prints) and heavy cropping it might be helpfull but not a big deal.

2)I see some difference between my 10D and my new 400D(many lens changes) with more than 1200 pics and practicly NO dust even at f32.

3)It is good to be bigger and brighter and higher resolution.I think that the info at the back is good(big,clear to see letters).It does not bother me.

4)The MOST significant upgrade in my opinion.The difference between 7 and 9 AF-points is that the center point is more accurate especialy with fast glass 2.8 and faster.Better focus period.

5)Good upgrade

6)Well for me not a big deal but why not

7)Same processor, same battery but more futures it is normal to have less battery power.I have the grip and 3 batteries.No power problem.

cloose wrote in post #2350745
While I cannot comment on a comparative, as I own the XTi, but have never owned the XT, battery life on the XTi is quite good. I was at the zoo today at 0c temps, and shot ~200 pictures all with IS on, and the battery level never moved the entire time.

To me, the AF was reason enough to go for the XTi over the XT. Take this with a grain of salt however, as I picked up the XTi for only $20 more than the XT ($849 Cdn with kit lens at a retail store)

Hey that's good news about the battery life, one thing I hate is when my current digi runs out. Well sounds like you got a good deal lol..I can get the XT for $675CAD or XTi for $850CAD, with the kit lens..would definately be an easier decision if it were only 20 bucks more

DrPablo wrote in post #2350315
Again, PM me if you want references for this.

MP sent.

I am not sure if you agree with me in the following or not (I got the impression that you will not agree).

The final resolution in the photo is a convolution of the lens and sensor. If you have a 80 lines/mm lens. And now you have two sensors: one at 80 lines/mm and the other at 100 lines/mm. The 100 lines/mm sensor will still record better resolution than the 80 lines/mm sensor. See dpreview.com XTi test.

Canon XTi/400D has something like 85 lines/mm on sensor. So combined with a good or very good lens, it will record better resolution than the 8MP cameras.

DrPablo wrote in post #2350315
How much better would a lens have to be to match this difference? Simple: 75 lpm x 1.6 = 120 lpm.

120 lpm

Zeiss claims a lens with this resolution (140 in fact), but this claim is highly controversial. But in the real world, there is no commercial photographic lens with resolution like that -- maybe the Hubble telescope has lenses like that. Your best lenses hit 80 lpm in the center at f/8 if you're lucky.

Minor details:

Lenses on some point and shoot cameras have this kind of resolution (surprise!). But their sensor size is so small, the total # of lines they can resolve is smaller than DSLRs.

I have some flawless 10x8s from my 3mp compact.

The only difference between my 8mp dSLR and my 3mp compact is that the colours are better on my dSLR.

rhys wrote in post #2351495
I have some flawless 10x8s from my 3mp compact.

The only difference between my 8mp dSLR and my 3mp compact is that the colours are better on my dSLR.

You are correct if uncropped 8x10s are all you want to print. You could probably have saved yourself a lot of money an bought a used D30 instead of a newer 8 MP camera.

However, some of us want to make prints much larger the 8x10 and/or crop dramtically. For example I have made 24 x36 prints and portrait oriented 13x19 crops from a photo that was taken in landcape mode. At those enlargement factors you can see the limitations of an 8 mp camera and 3 mp would have noticeably less resolution. Unfortunately 10 or 12 mp does not give that much more resolution. Going from 3 to 8 you more than doubles the number of pixels. To get the same improvement from 8 mp you would have to go to 21 mp. Canon has not released a camera with that many mp, yet.

How many people are going to make prints big enough or crop so much that they need a camera with that many pixels? There are some and they may find themselves looking at medium format rather than 35 mm based camera systems.

pieq314 wrote in post #2351428
The final resolution in the photo is a convolution of the lens and sensor. If you have a 80 lines/mm lens. And now you have two sensors: one at 80 lines/mm and the other at 100 lines/mm. The 100 lines/mm sensor will still record better resolution than the 80 lines/mm sensor. See dpreview.com XTi test.

I've never seen formal reviews of sensors that describe them in lpm, but that makes sense. The gold standard for testing lens resolution (which is done by the lens makers themselves) is to use the lenses with ISO 25 technical film, which has a resolution that exceeds 140 lpm. Fine grained slide films like Velvia also have resolutions greater than 120 lpm.

If you have a sensor that can record 100 lpm and a different one that records 80 lpm, the 100 lpm will only record better if your lens transmits greater than 80 lpm of detail. In other words, if the better sensor is lens-limited and the inferior sensor is sensor-limited, then there will be more detail recorded by the better sensor (assuming an lpm target or a finely detailed subject).

But in practical terms it's not so simple. For instance, people who shoot medium format and especially large format are always lens limited. Most LF lenses are outstanding, but they are still the limiting factor in the amount of recorded detail. And yet people who shoot ultralarge formats (like these 11x14, 7x17, 16x20, and 20x24 inch cameras) are happy to shoot with banged up antique lenses that are 75 years old.

Why is this? It's not as simple as 'the film is huge, it just records so much detail'. The answer is in the circle of confusion, which is the size of a 'detail' or 'point' projected onto the sensor. Because all lenses have at least some aberrations, even a laser pointed at the lens will still record as a small disc on the sensor -- there's a lower limit to the size of this recorded detail. And this is simply the reciprocal of lpm -- a lens that projects 100 lpm will create circles of confusion that are 1/100 = 0.01mm. A lens that projects 20 lpm will create CoC that are 1/20 = 0.05 mm. Just like lpm, there is a physiologically-derived CoC that will appear sharp or in focus at a given viewing distance.

Now the following example, again, is not a LF plug. It just makes it easier to illustrate the difference between lens and sensor resolution in a mathematically more tangible way.

So say you spend $1500 on a nice used Deardorff 8x10 camera, and feel like getting a crappy old 25 lpm lens on the cheap. Most people with 8x10 cameras just contact print, they don't enlarge. So an 8x10 image will still have 25 lpm of detail or CoC of 0.04mm.

Now you take a Nikon D2x with a heroic 120 lpm lens. So on the APS-C sensor you project CoC of 1/120 = 0.008 -- better, right? I mean the CoC is 1/3 the size of that on the LF image. The thing is, no one looks at APS-C, full frame, or even medium format images as a contact print, i.e. at the actual size of the sensor -- they are always enlarged. You don't have many 15x24mm pictures on your wall.

So you enlarge the APS-C image to 8x10, which is a 13.5x linear enlargement (15mm = 0.59 inches). So the CoC of 0.008 in the originally captured image are also enlarged, of course, which produces CoC of 0.108 when enlarged to 8x10.

So at this enlargement, the smallest recordable details from the D2X will be 0.108mm, ar 2.5 times the CoC from the 8x10 camera that remains 0.04mm. This means that the minimum viewing distance for the D2X image is 2.5 times farther from the print than the 8x10 image.

This is an extreme example, in which the LF camera has a crap lens and the D2X has a great lens. But realistically most people in both SF and LF worlds who use good lenses are hanging out in the 60-80 lpm world. Obviously if both cameras use equivalent lpm lenses, then at any given output size the 8x10 would have 13.5x finer detail.

You could also think about CoC as a function of total film size. And given the same CoC, a full frame 24x36 frame will have a 1.6-fold smaller CoC per unit film area than a 15x24 APS-C camera. The role of megapixels in this is trivial, because all increasing megapixels will do is increase the precision with which you record this detail, but fundamentally, by doubling the megapixels, you're still just doubling the pixels used to record an 0.05 or whatever detail -- but that doesn't mean that the pixels actually see any more detail than that.

The whole point of these anecdotes is to show how it's not as simple as being lens limited or sensor limited, and in a sense all cameras are lens limited. You have to make many enlargements to get small format images to an acceptable viewing size. And the resolution of the sensor or film doesn't change the size of the projected CoC.

Regarding the lens resolution of P+S cameras, I've never seen tests of them, but if true it's probably because there are so few lens elements and air/glass interfaces.

I think unless you are pixel peeping at poster sized print outs, then you are unlikely to notice differences between 8 and 10MP in the real world

But most folk dont pixel peep posters . . . most folk will stand 3 or 4 ft away and admire the image as a whole.

In such cases, your better going for the best lens possible so that you eliminate ( as best you can ) issues such as colour fringing and bad bokeh which influence much more than individual pixels.