APS-C IMAGE QUALITY VS fULL fRAME AT TESTING

Can someone explain why is it that a lens that is just so so at the center and borders on a APS-C sensor , the same lens on a Full Frame is sharp . By what I have understand the center that is soft with a APS-C sensor , that same lens on a Full frame , the center should also be soft but I have seen lens when tested for sharpness having a sharp center with FF but soft with a APS-C sensor . Whats happening here ? I'm I missing something ?

I'd say it's mostly to do with the fact that many APS-C cameras are higher resolutions (smaller, more densely packed pixels) than their full-frame counterparts. It's all to do with pixel size vs sensor resolution.

I'm guessing you are being tricked a bit by assessing "sharpness" at 100% pixel view too.

You'll probably find that, at image level and after suitable sharpening, the APS-C images will be very close to the full frame images in terms of sharpness. I do.

Very simple. The image from the physical crop sensor is enlarged 1.6 times more.

If the pixel counts (not densities) are similar, the main factor is the crop factor. The resulting advantage is not 60% more though since the lens resolution (per sensor height) is only one of the factors.

You're not enlarging the FF image as much, so it isn't as "hard" on the lens in question. This allows an FF sensor to provide more resolution than APS-C even if it has less MP (as long as you frame the same/are not FL limited). For example, I've noticed that the 13MP 5D will give slightly better resolution than the 18MP sensor in my 7D if framed the same.

That's why I find it funny that people often say FF requires better lenses: the opposite is true both in a resolution and DOF standpoint. Even the much maligned 28-135IS on a 6D or 5D3 will give better results than a 17-55 on my 7D.

--
-Scott

Sort of - though I fear you may be making the same error that a few others do here.

The 12.8MP 5D definitely does not resolve more than the 18MP 7D. The whole idea is ridiculous as the sensor has about 5 million pixels less resolution.

But sharpness (which is NOT resolution) is a factor of more than resolution. The other big contributor is acutance. And it is this parameter that the 5D is ahead in, because the big pixels, which resolve less, provide better acutance. So overall, the sharpness may appear better even though the sensor is resolving less detail. This difference can largely be eliminated with use of proper sharpening though, leaving more detail resolved and very similar sharpness in the 7D image.

Yes - but then both the 6D and the 5D3 have higher resolution sensors than the 7D. Just as the 7D has a higher resolution sensor than the 5D. 

not according to dxo and gianp2k!

did we ever hear from Andy W.?

Though I think you could be correct, we need a "proper test"

To take a 50 and physically move closer to get the same framing - duh - moving closer my eyes see more resolution, more fine print, when I move closer.

The farther way to go to magnify from a small sensor vs the lesser way to go to magnify from a big sensor also has some impact.

the lens behaves differently on different cameras

the aa is different

lots of differences

I wouldn't use "definitely"

We need to see a proper test!

sharpening ain't a good thing - > more noise

better acutance is noticeable. others discount my claim of weak aa. probably weak aa AND those nice Big Pixels

...but how big does one need to go with this stuff? I make 20x30's and below. my 5dc, 60d, T4i are fine. One would be hard pressed to tell the difference iso100-400 between my 60d and your 5d3 at these sizes unless we are talking Dramatic Shallow DoF of your siggy F1.4 @ F1.4 - which I'll have soon for my 5dc - and then it will be difficult to tell the difference iso 3200 and below - except for the faster capture rate advantage of 5d3 of course

Just curious about ...

1. What was wrong with the test I posted earlier in that other thread? Do you consider it to be not "a proper test"?

2. If you think something was wrong, how would you design a test for comparing the resolving abilities of two different sensors?

Yeah - I didn't mean to be dismissive. The AA filter would have some effect (if, indeed, it does have a weak AA filter).

I agree completely. But stating that all these cameras will deliver great quality images seems to be quite controversial around here! Hehe!

The only problem is that it does, most of the time (DXO). It is not that ridiculous at all. We are talking about systems, not sensors. And BTW, "some posters" said that with the same pixel count, not densities, the dominating factor is the crop factor. With different counts, you have two factors fighting against each other - the enlargement factor and the pixel count. With soft lenses like the 50mm ones wide open, the 5D will resolve much more. With super sharp ones, well stopped, the 7D may have an a advantage.

Those are just different words for different MTF values. DXOmark used to publish the whole (measured) MTF chart, and you could see that the 5D was better (most of the time) in any part of it - acutance or whatever. The way they present their data now makes it impossible to extract the actual measurements, unfortunately.

It is largely irrelevant. The 5d2/3/6D have about 5-8% higher linear resolution but the measurements show MUCH higher resolution, same lens, same f-stop. The pixel count is a factor but just one of them. The enlargement factor is much stronger.

needs same distance, same fov. So can't use same prime lens

70-200 II, same fov, resolution chart looking for line pairs

then, after seeing which is better, describe what it means in how large one would need to go to see the difference in a gallery when viewed from 5 feet:-D

it does:-D

yeah, I hear ya!

No. It's impossible for a sensor with less resolution to resolve more using the same lens.

The only way that can happen is if one presents more detail to the sensor with less pixels to begin with - such as is often the case with lens testing (such as DxO).

The mistake that is made is that people use lens testing data to draw conclusions about sensor resolving power. This approach, one you'll see here a lot by those who haven't though about it or don't understand, is flawed when one attempts to compare the same lens on different sized sensors.

Once again - see above.

Yes - see above.

In these lens tests, a significantly higher amount of optical detail is projected onto the sensor plane when testing the FF cameras. The figures show this - what they don't show is the comparative ability of the different sized sensors with respect to resolving detail.

Do you have a 5D series/6D and an 18MP crop? If so, you can do the test yourself. Set up a tripod about ten feet from a highly detailed subject and then use your best lens to take a shot with both. I guarantee you that the crop will out-resolve the full frame when both sensors are presented with the same image detail.

Big problem here is that you are presenting different amounts of detail to the sensor plane. Doing so pretty much kills the ability to draw any useful conclusions. (I'm sure you know this MAC).

The only way the determine which sensor is capable of resolving more detail is to present exactly the same image detail to both and then compare them at 200% or 300%. Any other approach is flawed when trying to determine resolving power.

The reason you don't see test sites do this is simple - it's obvious that the higher resolution sensor will resolve more. No need to test this - but it's a pity they don't because we have armchair experts running off to DxO et al using lens test data to falsely determine sensor resolving abilities.

Obviously the AA filter will play some part. And, though I haven't bothered, it'd be easy to find out how much by doing some poking around on the net to find some decent tests between the D800 and D800e.

Of course, weak or non-existent AA filters introduce their own issues, so it's not all champagne and caviar.

You have a funny smarta$$ way (no offence) to twist this topic. Yes, if I am under a restraining order not to choose the proper lenses for the format, I would discover that the crop camera with the 10-22 at 10mm resolves more of the dollar bill in front of me that an FF body with the same lens at 10mm does (yes, you can mount it on FF). Nice way to confuse innocent people coming here for advice.

But that was not the question the OP asked. Read the OP again. It was answered a long time ago. You would expect that the thread would die but ... this is dpreview.

How come nobody ever presented enough detail to the 18mp Canon sensor so that it could fall just 8% short of the 21mp 5D2 one? Let me guess - wide spread FF conspiracy? Every possible measurement shows considerably more than 8% advantage.

Examples (all in the center)

here (about 25% advantage)

here (about 40%)

here (about 35%)

It must be different processing, right? Wait... what about

this (about 18%)

It's difficult to see how calling someone that would not cause even a glimmer of offence.

Besides, breaking the forum rules so obviously is not a wise thing to do, an you've put me in a difficult position. So, you are warned about attempting to defeat the profanity filters. Don't do it again.

Well, I rather see it the other way around.

When talking about comparative resolving power of different sensors, it's confusing to present the conclusion framed around data derived from an unequal comparison.

It seems to me it'd be a lot clearer if people understood this and didn't incorrectly claim that lower resolution sensors actually resolve more!

Correct. And my response with respect to this resolution discussion was not to the OP, but to poster who claimed that the 5D outresolves an 18MP crop. It doesn't - not unless you present it with more detail to resolve to begin with (which, admittedly, the poster did point out).

But, your point is taken. And it was not my intention to confuse people, but rather to get them thinking more about how resolution, acutance, subject distance and the like effects the final outcome. And perhaps point out to them the folly in claiming that higher resolution sensors out-resolve lower resolution ones - because they don't. It's an oxymoron. The way one uses them may lead to more final image detail or perhaps sharper images (may), but that is not because the sensor is resolving more detail - only that it's been presented with more detail to resolve to begin with.

Those are acutance maps - they don't measure resolution!

Which brings us nicely back to the OP - as it's this that he is seeing.

There still sadly remains quite a bit of confusion out there regarding the difference between sharpness and resolution.

My intention was to attempt to clear the confusion a bit. But posting acutance maps as evidence  of sensor resolving power is not helping.

You are too sensitive. This is a commonly accepted "urban" term in the US with no good substitute. I believe it would not be even censored on daytime TV.

But we were not talking about resolving power of sensors. You were the one who turned it into such a discussion. Whether the 5D resolves or does not more than the 7D cannot be taken out of the context in which this question appeared in the first place. And that was - as a part of a system.

They do. The measure the whole MTF range. They call the MTF-5 figure resolution but they say that this is hard to measure and too sensitive to noise. They used to present the MTF-20 figures as a measure for resolution but now they decided that they need to scramble the data. PZ measures and reports the MTF-50 data on processed file. DPR, which uses the DXO data, labels the graph "Sharpness". All those fancy words that you (and now, DXO) use are just different MTF values.

But that is not the main point. You are in denial about the enlargment factor and its role in the measurements and in what we see in reality. You never agreed that this was a factor, with the same or a similar lens anyway.

Yeah - I agree. Same over here TBH.

But, if the profanity filters don't agree, then neither do the site administrators. And, in that case, it's breaking the rules regardless.

Fair enough - I did so because the term "resolution" was being used in a confusing manner. I personally think "sharpness" would be a better term for these situations - as it makes no particular claim regarding a single image parameter, but is rather more encompassing and befitting of image quality.

Sharpness (which is what the OP was enquiring about) is a factor of many things. Resolution is only one of them.

I know you already know this DD, but for the sake of clarity for any following, final image sharpness is affected by ...

• The optical qualities of the lens.
• The distance to the subject.

Both these affect the amount of detail available for the sensor to resolve to begin with. They are driven by the actual lens and sometimes by the size of the sensor (but not always).

• Sensor resolution.
• Pixel size.
• AA filter.

These govern how small the detail can be and still be recorded by the sensor. The higher the resolution, the higher the resolving power and the smaller the size of detail can be. They also affect the acutance, or edge contrast, between pixels with respect to the detail recorded. This is the trickier one to understand, but Wikipedia has a short, concise explanation worth looking at if someone is interested.

Then, there are things such as subject and camera movement, the aperture setting, diffraction effects and some others (such as scene lighting) that all contribute too.

And that's before you start manually altering sharpness (by changing the acutance) in post!

As you may be able to tell from this, resolution is only one factor that contributes to image sharpness. And it's entirely possible (in fact it happens all the time) that you can have a camera that resolves more but does not deliver sharper final images in many situations. It's worth noting though, that in some situations, it almost always will - think macro photography and reach limited situations as examples of where higher resolution will be a big bonus, only eventually negated by noise when light levels drop (which is not a hard limit, but rather a product of the current technology used).

Sorry if I confused things - it was not my intention. As I said, I was attempting the opposite. Hopefully the little bit above makes it a bit clearer.

No, I'm not at all.

The "enlargement factor" has little to do with the abilities of the sensor to resolve detail. That, in a nutshell, is my point!

It only affects the final image detail in so much as it will affect the subject distance. And, by doing so, it allows one to present more detail to be resolved!

And, when one sensor has more detail to resolve to begin with, it's then not possible to make comparisons about which sensor is capable of resolving more detail.

If your point is that the full frame resolves more because it allows you to present more detail to it to record, then I can only say that, in this case, you aren't really comparing the resolving capabilities of the sensor at all, just the difference in the way sensor sizes utilise a given focal length.

You are the only one comparing the resolving capabilities of different sized sensors, nobody else asked for that. It is a way to avoid comparing systems. As a whole, without going into specifics, FF systems resolve more. Would you deny that? Need more resolution - go FF. Need even more - get MF or LF. More - just go there and see it with your own eyes.