Comparing lenses for full frame vs. APS-C sensors

[xpatUSA]

Originally Posted by xpatUSA Is it possible to clarify the above statement? I ask because I have never had a problem showing "detail" on my humble 24-inch 1920x1200px sRGB monitor. Every app that I have can zoom in as much I want. One image pixel can be for example 40 pixels on my screen.

I think the statement makes sense if you refer to the image as a whole, not a portion blown up.

David's statement did not include a "fit to screen" criterion but - yes - if an image is down-sampled to fit a screen or a print for that matter, "detail" will be lost.

For example, the 27" NEC monitor I use for photo editing has roughly 3.7 MP.

I can counter your monitor example with my own - my most-used camera is 2268x1512px and my monitor is 1920x1200px. Not a huge difference in dimensional terms and I normally view at 100% zoom which, as you know, makes a one-to-one relationship between sensor pixels and screen pixels.

On my Canon printer, a 13 x 19 print is 22.2 MP, and I can get that with no upsampling. A 17 x 22 print is 33.7 MP, although with my camera, that requires a little upsampling even without cropping--that is, some of the detail is created by software.

David's statement was about "display devices" not just The Print, and he made no mention of re-sampling.

By the way, are you really suggesting that up-sampling creates detail? That, for example, an ant previously invisible in the original will somehow appear?

In any case, on my screen, when I zoom in my viewer is set to Nearest Neighbor which does not do anything to original pixels other than make them bigger because I only zoom at integer ratios.

Since the sub-discussion is about "detail" which is undefined so far perhaps we should address that first. However that is undesirable because it would probably open another can of worms.

So my point remains that my "display device" can show sensor pixels in a one-to-one relationship with screen pixels and therefore "detail" is shown as-is with no loss of detail. To avoid further controversy, that is my only point and it makes David's statement questionable.

[xpatUSA]

... 24mm is not quite wide enough for some landscapes on an APS-C camera and secondly the lens is optimised across the full "image circle" of a full frame camera. You could get a cheaper EF-S lens that's been optimised for APS-C with a wide end of 16mm for landscapes ...

Hello Dave,

There are some who say that when a full-frame lens is used on an APS-C camera, the image is made from the "sweet spot" of that lens and thereby is of higher "quality".

If "quality" means MTF, one could for example expect sharper corners given similar-looking lens charts between a full-frame and an APS-C lens, I reckon.

Not talking about relative weight or ergonomic factors ...

[dje]

Hello Dave,

There are some who say that when a full-frame lens is used on an APS-C camera, the image is made from the "sweet spot" of that lens and thereby is of higher "quality".

If "quality" means MTF, one could for example expect sharper corners given similar-looking lens charts between a full-frame and an APS-C lens, I reckon.

Not talking about relative weight or ergonomic factors ...

Yes Ted I guess that’s possible, you’d have to look at specific models. It’s also possible that the APS-C lens could have comparable or even better mtf figures in the centre for a lower cost. Comparisons would need to be made on Lppmm units rather than Lpph units which have the sensor ht factored in.

Dave

[xpatUSA]

Originally Posted by xpatUSA Hello Dave,

There are some who say that when a full-frame lens is used on an APS-C camera, the image is made from the "sweet spot" of that lens and thereby is of higher "quality".

If "quality" means MTF, one could for example expect sharper corners given similar-looking lens charts between a full-frame and an APS-C lens, I reckon.

Yes Ted I guess that’s possible, you’d have to look at specific models. It’s also possible that the APS-C lens could have comparable or even better mtf figures in the centre for a lower cost. Comparisons would need to be made on Lppmm units rather than Lpph units which have the sensor ht factored in.

All good points, Dave. If I recall correctly, I've read that smaller diameter lens are easier to make well, compared with large. Or was it the other way round? The lens cognoscenti here would know more than I about that.

Just found this for what it's worth:

https://www.lensrentals.com/blog/cat...charts/page/1/

Everybody likes Roger! I just looked at their Sigma lens charts and they're all 'DG' (full frame) ho hum so off to look at others ...

[DanK]

I agree with Dave E.: It doesn't make sense to plan what focal lengths you will buy until you know what sensor size you will buy. The 24-105 f/4 II is my most used lens on my FF, but I agree with Dave that it doesn't go wide enough to be a good walk-around lens on an APS-C camera.

I also can't say strongly enough that I agree with Bill that all of this depends on what you are going to shoot. I'll give an example. Given what I shoot, f/2.8 lenses are largely a waste for me. They are heavier and more expensive (bigger glass). I almost never need the narrower depth of field, and with modern sensors, I would rather raise ISO by a stop when I occasionally need the extra speed then lug around a lot of expensive glass that I rarely need. And given what I shoot, the depth of field at apertures bigger than f/4 is usually too shallow for me. Also, again given what I shoot, using one of the several optically superior 24-70 f/2.8 lenses would leave me changing lenses very frequently; I go back and forth across that 70mm line all the time. So for me, the 24-105, which is a very good lens but not as good as several of the 24-70 lenses, is the best choice. None of this applies to many other people who shoot differently.

[DanK]

By the way, are you really suggesting that up-sampling creates detail? That, for example, an ant previously invisible in the original will somehow appear?

Indeed, that's exactly what happens, although one has to be careful about what one means by detail. Upsampling obviously doesn't somehow find detail that isn't there, but it does try to estimate what detail might have been there, and it imputes those data. How well it does this depends on the algorithm, how much imputation is needed, and characateristics of the image. These tools are getting better; for some images, the machine-learning based enhancement in Adobe software works better than the mathematical algorithms the software offers for upsampling. I've had images that aren't helped at all by Adobe's new enhancement tools, but I've had others where it helped quite a bit. Recently, someone sent me a photo of someone in a race that was about 150K. It had sentimental value, and he wanted 8 x 10 prints. An 8 x 10 print with my printer is 7.2 MP. Using Adobe's Super Resolution, I was able to print adequate prints--clearly not what I would have liked, but good enough that the recipients were delighted. You can do the arithmetic to estimate how many imputed data points that required.

[xpatUSA]

Just found this for what it's worth:

https://www.lensrentals.com/blog/cat...charts/page/1/

Everybody likes Roger! I just looked at their Sigma lens charts and they're all 'DG' (full frame) ho hum so off to look at others ...

Here , Roger says:

" ... Also, when you do use a full-frame lens on a crop sensor, the camera is shooting through the “sweet spot” of the lens: the weakest part of most lenses are the edges and corners. The crop sensor image circle doesn’t use them."

Doesn't really resolve the issue though because of performance variance between lenses of any size or model. So if I were to post an example where the center area of an FF lens is "better" than the whole area of some APS-C lens, that would not be proof of the general statement.

So, in the end, the luck of the draw rules ...

[DanK]

And as a practical matter, using a FF lens on a crop-sensor camera in order to use only the sweet spot generally means spending more money and lugging around a larger and heavier lens.

I did use one FF lens when I shot primarily with an APS-C camera: the 70-200 f/4 IS L (first generation). It was superior to the alternatives at the time, and being an f/4 lens, it wasn't all that big or heavy.

[ajax]

That may leave you with quite a small choice of cameras, David, IMO. I know of no digital camera that simply converts the sensor per-pixel output voltage via the ADC to a per-pixel 16-bit number in the raw data segment of a "raw" file. Anybody?



Again, I claim that what you call "the real raw data" only exists briefly in camera's buffer memory and that what ends up on the memory card has had some pre-processing applied. That would be in spite of what dumbed-down articles say.



Now that I agree with.

Just to clarify in case it matters. A program that I use a lot for post processing is called Rawtherapee. As open source software it does NOT favor any particular camera maker. I believe it supports just about, if NOT, all camera makes that produce so-called raw files (whatever those are). This is true even though these files are created in a proprietary format particular to the make of camera. I believe Adobe, which as far as I know does NOT make cameras, has defined a format of their own which might then be considered a standard. Last I checked they provide software that converts the proprietary files, for lots of camera makers, to use their (?standard) format should that be desired as might be expected for users of Photoshop using cameras from different makers. My thinking would be that this imposes some limits on the extent to which camera makers can very and still be supported by this software.

As best I can tell “demosaic” is not a word in the English language. Likewise, there is no mention of it in the documentation provided with software supplied by Canon for developing their raw files. Apparently, it doesn’t matter to Canon. My use of the term is entirely based on the documentation that Rawtherapee provides at this link.

My opinion is that what I’ve said herein when referring to demosaicing and raw files is quite consistent with Rawtherapee’s usage of the terminology. My apologies if that differs from how others might use the same terms. It might be worth pointing out that Rawtherapee is open source software designed & developed by a very large community of people many of whom do not speak English as their native language. I think it is fair to say they agree with this definition. However, I’d have to admit that such a term may have NO official meaning and could mean something else to others.

Of course it would be fair to say that pre-processing (whatever that might mean) of some kind by the camera is involved in producing so-called raw files.

[xpatUSA]

Just to clarify in case it matters. A program that I use a lot for post processing is called Rawtherapee. As open source software it does NOT favor any particular camera maker. I believe it supports just about, if NOT, all camera makes that produce so-called raw files (whatever those are). This is true even though these files are created in a proprietary format particular to the make of camera. I believe Adobe, which as far as I know does NOT make cameras, has defined a format of their own which might then be considered a standard. Last I checked they provide software that converts the proprietary files, for lots of camera makers, to use their (?standard) format should that be desired as might be expected for users of Photoshop using cameras from different makers. My thinking would be that this imposes some limits on the extent to which camera makers can very and still be supported by this software.

As best I can tell “demosaic” is not a word in the English language. Likewise, there is no mention of it in the documentation provided with software supplied by Canon for developing their raw files. Apparently, it doesn’t matter to Canon. My use of the term is entirely based on the documentation that Rawtherapee provides at this link.

My opinion is that what I’ve said herein when referring to demosaicing and raw files is quite consistent with Rawtherapee’s usage of the terminology. My apologies if that differs from how others might use the same terms. It might be worth pointing out that Rawtherapee is open source software designed & developed by a very large community of people many of whom do not speak English as their native language. I think it is fair to say they agree with this definition. However, I’d have to admit that such a term may have NO official meaning and could mean something else to others.

Of course it would be fair to say that pre-processing (whatever that might mean) of some kind by the camera is involved in producing so-called raw files.

I too use RawTherapee, David. It is the best global (whole image) editor on the planet, IMO. As you know, it even offers several types of raw conversion.

As to that word, I greatly prefer "demosaicing" to "demosaicking" which looks quite ugly to my eye ..

[xpatUSA]

And as a practical matter, using a FF lens on a crop-sensor camera in order to use only the sweet spot generally means spending more money and lugging around a larger and heavier lens.

I did use one FF lens when I shot primarily with an APS-C camera: the 70-200 f/4 IS L (first generation). It was superior to the alternatives at the time, and being an f/4 lens, it wasn't all that big or heavy.

For some us, you get no choice if you want a proper 1:1 macro lens. In other words, the lens manufacturer (Sigma) only makes full-frame true macro lenses for my DSLR.

[ajax]

The mention of lens correction has caused me to take a look at my Rawtherapee Version 58 to see what it supports. Interestingly, there are lots of Canon lenses that can be selected but none are RF lenses. Any ideas about why? If I bought one I would want Rawtherapee to support it.

[xpatUSA]

The mention of lens correction has caused me to take a look at my Rawtherapee Version 58 to see what it supports. Interestingly, there are lots of Canon lenses that can be selected but none are RF lenses. Any ideas about why?

If you're asking me, I'm not a Canonista so I can't say, sorry.

If I bought one I would want [RawTherapee] to support it.

I understand.

[DanK]

The mention of lens correction has caused me to take a look at my Rawtherapee Version 58 to see what it supports. Interestingly, there are lots of Canon lenses that can be selected but none are RF lenses. Any ideas about why? If I bought one I would want Rawtherapee to support it.

I'm guessing it's because RF lenses are new. RawTherapee is open source, and it only is updated when someone gets around to it. The most recent version was released almost a year ago, according to the web page. That's a drawback of open source software. Adobe added support for either two or four more RF lenses, depending on how you count, in the last month alone. That's part of what you pay for: they update camera and lens profiles all the time, as new gear hits the market.

[Abitconfused]

I have decided to sell almost all of my DSLR gear for mirrorless. I have seen the resale price of DSLR cameras & lenses beginning to fall significantly on the used market. I recently bought a Z 24-70mm f/2.8 S that I fit to my (APS-C) Z 50. Here, the lens holds the camera...well almost. If I need an enlargement, I use Topaz Gigapixel (reasonably priced). The equivalent FOV is, of course, 36-105mm which works well for me as a street photographer (The Street Photographer, Take to the streets with your camera) by Ed Ruth, a Kindle Book).

[xpatUSA]

I have decided to sell almost all of my DSLR gear for mirrorless. I have seen the resale price of DSLR cameras & lenses beginning to fall significantly on the used market. I recently bought a Z 24-70mm f/2.8 S that I fit to my (APS-C) Z 50.

Congratulations, Ed, a big step for anyone. Multiple cameras is OK but, at my age, every time I pick one up I have to remember all it's workings - quite difficult at times.

Recently, I sold all my Sigma digital cameras except the original DSLR (SD9) which I will keep forever ...

[Abitconfused]

Ain't that the truth! The link between user and machine is acute in digital photography. Reestablishing that interface with new gear can be a challenge. I have recently discovered that the extra length of the 24-70mm f/2.8 lens makes it a good candidate for use by simply placing it on a counter, perhaps with a small block of wood under the front of the lens, and taking a photo in reduced light...no tripod needed. So now I carry around two small pieces of wood just for that purpose.

[pnodrog]

To summarise.

Regardless of equipment nobody is going to take a photo worth enlarging to any size while online discussing it....

[Abitconfused]

Truly...I have become a bit desperate lately as I have covered about everything in the vicinity. We go to the beach next month so there will be some target rich areas to photograph within.

[lunaticitizen]

The mention of lens correction has caused me to take a look at my Rawtherapee Version 58 to see what it supports. Interestingly, there are lots of Canon lenses that can be selected but none are RF lenses. Any ideas about why? If I bought one I would want Rawtherapee to support it.

I remember RawTherapee uses an external lens correction profile called Lensfun. A quick look at their database indicates there are only 3 RF lenses supported currently.

You also have to update this Lensfun database on RawTherapee manually.