Is Full Frame sharper than APS-C? My answer here

[TrumpetPower!]

We could not tell any difference, apart from the DoF.

I'm actually not at all surprised. You even could have equalized the DoF by shooting the 5DII at f/11 and ISO 200.

In good light with good glass at small apertures, you'd have to be printing large and looking very critically to tell the difference.

But in dim light, or if you need to stretch the dynamic range by pulling up the shadows...that's when the 5DII is going to start to pull away. Or if you're making prints bigger than 24" x 36".

And still, poor technique can easily obliterate the differences.

b&

 

[tortilla]

Once again, all else is clearly not always equal. If all you've got is a 300mm lens and you want to shoot the full disc of the moon, yes, of course, you're going to get better results from that lens with a 7D than a 5DIII. But the guy next to you with a 5DIII and a 1200mm f/5.6 is going to get an image of the moon that puts yours with your 7D to shame.

300 mm vs. 1200 mm... so now we're talking about lenses, not sensors anymore?

But I agree with you, when comparing FF vs crop it's not only about resolution. It's also about dynamic range, noise, cost, weight etc.

 

[Don Haines]

Once again, all else is clearly not always equal. If all you've got is a 300mm lens and you want to shoot the full disc of the moon, yes, of course, you're going to get better results from that lens with a 7D than a 5DIII. But the guy next to you with a 5DIII and a 1200mm f/5.6 is going to get an image of the moon that puts yours with your 7D to shame. And the guy the next hill over, the one with a 32" Dobsonian? Well....

yeah..... but is the guy on the last hill hand-holding the 32" Dobsinian? And if he is, who's going to argue with him?

 

[TrumpetPower!]

Once again, all else is clearly not always equal. If all you've got is a 300mm lens and you want to shoot the full disc of the moon, yes, of course, you're going to get better results from that lens with a 7D than a 5DIII. But the guy next to you with a 5DIII and a 1200mm f/5.6 is going to get an image of the moon that puts yours with your 7D to shame. And the guy the next hill over, the one with a 32" Dobsonian? Well....

yeah..... but is the guy on the last hill hand-holding the 32" Dobsinian? And if he is, who's going to argue with him?

The <a href="http://hubblesite.org/">gang down south</a> might be able to give him a good run for the money....

b&

 

[RLPhoto]

The bigger the format, the less critical of the lens that's infront of it. That's why my MF film shot with a beat up scratched 85mm 2.8 is sharper than 35mm cameras of the day.

The same holds true with current digital MF and 35mm systems.

 

[TrumpetPower!]

Let us say that you have a APS sensor 24Mp with the same S/N as a 24Mp 24x36 sensor then it is an optical question, it is hard to make a APS lens 1,5 1,6 times better which is require compared to 24x36mm lens.

This is incorrect.

Because there will be less absolute magnification for a same-sized print with the larger format, even if the sensors have the same pixel dimensions, the larger format will be sharper and have less noise. Again, always assuming all else is comparable, including a longer focal length lens for the larger format.

Cheers,

b&

 

[helpful]

Let us say that you have a APS sensor 24Mp with the same S/N as a 24Mp 24x36 sensor then it is an optical question, it is hard to make a APS lens 1,5 1,6 times better which is require compared to 24x36mm lens.

This is incorrect.

Because there will be less absolute magnification for a same-sized print with the larger format, even if the sensors have the same pixel dimensions, the larger format will be sharper and have less noise. Again, always assuming all else is comparable, including a longer focal length lens for the larger format.

Cheers,

b&

Assuming M.R. means signal to noise ratio by S/N, then he is correct. Actually, the linear resolving power of the lens needs to be increased by the crop factor, and the absolute resolving power of the lens needs to be increased by the square of the crop factor. But this is all based on an imaginary hypothesis, an APS sensor with the exact same S/N ratio as full-frame. Anything is true if the hypothesis itself is false.

 

[TrumpetPower!]

Let us say that you have a APS sensor 24Mp with the same S/N as a 24Mp 24x36 sensor then it is an optical question, it is hard to make a APS lens 1,5 1,6 times better which is require compared to 24x36mm lens.

This is incorrect.

Because there will be less absolute magnification for a same-sized print with the larger format, even if the sensors have the same pixel dimensions, the larger format will be sharper and have less noise. Again, always assuming all else is comparable, including a longer focal length lens for the larger format.

Cheers,

b&

is it, if the S/N and Mp is the same from the two sensor areas, then it must be an optical question , and there the APS lenses must be 1,5 1.6 better than the 24x36mm lens.

As I already explained, it is a question of how much enlargement is necessary for a print.

If you're making a 24" x 36" print, the APS-C image will be enlarged 41x from the sensor's original size of 14.8mm x 22.2 mm, but the full-frame will only be enlarged 25x from the sensor's original size of 24mm x 36mm.

Whether you measure S/N in noise per pixel or noise per square mm of sensor, because you've got two and a half times as many square mm of sensor per pixel with the larger format, you're getting that much better of an overall signal to noise ratio.

Imagine you were back in the days of film. You have a 35mm camera loaded with Velveeta on one tripod and an 8x10 view camera loaded with the exact same film on a second tripod. The 35mm camera has a 50mm lens and the view camera has a 400mm lens, both of which give the same normal field of view on the respective cameras.

You'd agree that, since it's the exact same film, the exact same chemistry, that the S/N ratio is exactly the same, right?

Now, let's say we're making an 8" x 10" print from the exposure. The 35mm negative needs to get enlarged 7.5x to make the print, but, for the view camera, it's a contact print.

If you don't agree that the contact print from the view camera will be dramatically sharper and have far less grain than the enlargement from the 35mm camera, then you truly are hopeless.

Now, you might still argue that the pixel dimensions are relevant, but I'll show they're not.

Let's not use traditional printing methods, but rather scan the film. But we want to wind up with the same megapickle files for both, to simulate this contrived example. We're going to scan the 8x10 negate at a lowly 300 ppi, and we're going to scan the 35mm negative at a whopping 2250 ppi. We're still going to make an 8x10 print, and we're going to do it at 300 ppi. And, because of the resolutions I picked, it "just happens" that no interpolation of either file is necessary; both will still print at 300 ppi at 8" x 10".

Once again, if you still don't think that the print from the 300 ppi scan of the view camera's negative will blow away the 2250 ppi scan from the 35mm camera, you're hopeless.

Cheers,

b&

 

[East Wind Photography]

This assumes that you also always print the full frame. For some cropping is always required. If you crop the same image to the same composition your results may vary. We don't always have the luxury of filling the frame especially when using a prime lens in less than adequate quarters.

Let us say that you have a APS sensor 24Mp with the same S/N as a 24Mp 24x36 sensor then it is an optical question, it is hard to make a APS lens 1,5 1,6 times better which is require compared to 24x36mm lens.

This is incorrect.

Because there will be less absolute magnification for a same-sized print with the larger format, even if the sensors have the same pixel dimensions, the larger format will be sharper and have less noise. Again, always assuming all else is comparable, including a longer focal length lens for the larger format.

Cheers,

b&

is it, if the S/N and Mp is the same from the two sensor areas, then it must be an optical question , and there the APS lenses must be 1,5 1.6 better than the 24x36mm lens.

As I already explained, it is a question of how much enlargement is necessary for a print.

If you're making a 24" x 36" print, the APS-C image will be enlarged 41x from the sensor's original size of 14.8mm x 22.2 mm, but the full-frame will only be enlarged 25x from the sensor's original size of 24mm x 36mm.

Whether you measure S/N in noise per pixel or noise per square mm of sensor, because you've got two and a half times as many square mm of sensor per pixel with the larger format, you're getting that much better of an overall signal to noise ratio.

Imagine you were back in the days of film. You have a 35mm camera loaded with Velveeta on one tripod and an 8x10 view camera loaded with the exact same film on a second tripod. The 35mm camera has a 50mm lens and the view camera has a 400mm lens, both of which give the same normal field of view on the respective cameras.

You'd agree that, since it's the exact same film, the exact same chemistry, that the S/N ratio is exactly the same, right?

Now, let's say we're making an 8" x 10" print from the exposure. The 35mm negative needs to get enlarged 7.5x to make the print, but, for the view camera, it's a contact print.

If you don't agree that the contact print from the view camera will be dramatically sharper and have far less grain than the enlargement from the 35mm camera, then you truly are hopeless.

Now, you might still argue that the pixel dimensions are relevant, but I'll show they're not.

Let's not use traditional printing methods, but rather scan the film. But we want to wind up with the same megapickle files for both, to simulate this contrived example. We're going to scan the 8x10 negate at a lowly 300 ppi, and we're going to scan the 35mm negative at a whopping 2250 ppi. We're still going to make an 8x10 print, and we're going to do it at 300 ppi. And, because of the resolutions I picked, it "just happens" that no interpolation of either file is necessary; both will still print at 300 ppi at 8" x 10".

Once again, if you still don't think that the print from the 300 ppi scan of the view camera's negative will blow away the 2250 ppi scan from the 35mm camera, you're hopeless.

Cheers,

b&

 

[TrumpetPower!]

This assumes that you also always print the full frame. For some cropping is always required. If you crop the same image to the same composition your results may vary. We don't always have the luxury of filling the frame especially when using a prime lens in less than adequate quarters.

If you're shooting side-by-side with a 7D and a 20mm prime, a 5DIII and a 35mm prime, and an 8x10 view camera and a 250mm prime, you're going to be cropping away the same proportional amount to get the same composition from each, rendering the cropping point moot.

If you're distance-limited, sure, it can make a difference...but generally not at much as most people tend to think. If that little birdie only fills an eighth of your frame with the 800mm f/5.6 on the 5DIII, you'll get better results putting the lens on the 7D, but not hugely better. The real answer is to improve your tracking skills so you can get closer and fill the frame, with whatever camera / lens combination.

Cheers,

b&

 

[AlanF]

Let us say that you have a APS sensor 24Mp with the same S/N as a 24Mp 24x36 sensor then it is an optical question, it is hard to make a APS lens 1,5 1,6 times better which is require compared to 24x36mm lens.

This is incorrect.

Because there will be less absolute magnification for a same-sized print with the larger format, even if the sensors have the same pixel dimensions, the larger format will be sharper and have less noise. Again, always assuming all else is comparable, including a longer focal length lens for the larger format.

Cheers,

b&

is it, if the S/N and Mp is the same from the two sensor areas, then it must be an optical question , and there the APS lenses must be 1,5 1.6 better than the 24x36mm lens.

As I already explained, it is a question of how much enlargement is necessary for a print.

If you're making a 24" x 36" print, the APS-C image will be enlarged 41x from the sensor's original size of 14.8mm x 22.2 mm, but the full-frame will only be enlarged 25x from the sensor's original size of 24mm x 36mm.

Whether you measure S/N in noise per pixel or noise per square mm of sensor, because you've got two and a half times as many square mm of sensor per pixel with the larger format, you're getting that much better of an overall signal to noise ratio.

Imagine you were back in the days of film. You have a 35mm camera loaded with Velveeta on one tripod and an 8x10 view camera loaded with the exact same film on a second tripod. The 35mm camera has a 50mm lens and the view camera has a 400mm lens, both of which give the same normal field of view on the respective cameras.

You'd agree that, since it's the exact same film, the exact same chemistry, that the S/N ratio is exactly the same, right?

Now, let's say we're making an 8" x 10" print from the exposure. The 35mm negative needs to get enlarged 7.5x to make the print, but, for the view camera, it's a contact print.

If you don't agree that the contact print from the view camera will be dramatically sharper and have far less grain than the enlargement from the 35mm camera, then you truly are hopeless.

Now, you might still argue that the pixel dimensions are relevant, but I'll show they're not.

Let's not use traditional printing methods, but rather scan the film. But we want to wind up with the same megapickle files for both, to simulate this contrived example. We're going to scan the 8x10 negate at a lowly 300 ppi, and we're going to scan the 35mm negative at a whopping 2250 ppi. We're still going to make an 8x10 print, and we're going to do it at 300 ppi. And, because of the resolutions I picked, it "just happens" that no interpolation of either file is necessary; both will still print at 300 ppi at 8" x 10".

Once again, if you still don't think that the print from the 300 ppi scan of the view camera's negative will blow away the 2250 ppi scan from the 35mm camera, you're hopeless.

Cheers,

b&

I think S/N based on area (where number of photons captured per unit time is proportional to area) varies fundamentally as the square root of ratio of areas. So increasing area by 2.5x improves S/N only by 1.6x. (I do know other factors do come in, but the basic statistics of S/N is a sqrt relationship).

 

[klickflip]

Thanks an really interesting test. As you've managed to show that there really isn't much difference comparing 2000px wide Jpgs that show some compression.
If thats all your images are ever going to be finalised as and viewed on screen at 50% then the 650 will be just great.

However... The view selected isn't the best for a proper test I think. Try subjects with smooth tones and darker gradients plus interesting detail and then you'll see the noise & sharpness when especially viewing or working on them in RAW or at least tiffs in photoshop viewed 200%.
Try some nicely lit portraits or some well composed architecture and there will be a big difference.

But its horses for courses really, if you're just going be shooting what we might call messy scenes like this and kids running around and holiday pics then it might not well be distinguishable to FF instantly.

More than anything it will them more come down to your technique to get the best out of FF, framing, catching beautiful light and the moment together whilst selecting the 'optimum' shutter / aperture, iso and lens plus raw/post processing which can make a huge difference to noise & sharpness plus the grade to push a feel which can be more important than anything.

A good photographer can make APS-C look like FF or MFDB and a bad photographer can make FF look like P&S.

And the previous post that mentioned what about cropping.. well if you are constantly cropping your FF that much then
you shouldn't have FF or a DSLR in my eyes anyway

 

[TrumpetPower!]

I think S/N based on area (where number of photons captured per unit time is proportional to area) varies fundamentally as the square root of ratio of areas. So increasing area by 2.5x improves S/N only by 1.6x. (I do know other factors do come in, but the basic statistics of S/N is a sqrt relationship).

You're correct about the actual numbers, of course. I didn't mean to imply a specific numerical relationship, which is why I tried to phrase it as loosely as I did.

With each standard increase in format, you gain an effective (roughly) one stop of ISO performance. If you're happy with ISO 400 on 4/3, you'll be happy with ISO 800 on APS-C, and with ISO 1600 on full frame, and ISO 3200 on medium format -- with a big, honkin' caveat that different manufacturers are using different technologies with their sensors in each segment. In the real world, it's more like ISO 100 on APS-C has half the noise as ISO 100 on 4/3, and so on...your MF back might not actually do all that well at ISO 3200, but its ISO 200 will be at least as clean as ISO 100 on full frame.

And, of course, we're also talking about actual prints, not pixel peeping or other meaningless measures.

Cheers,

b&

 

[TrumpetPower!]

"As I already explained, it is a question of how much enlargement is necessary for a print."

this is about pixels, no film, as long you have 1,6 times better lenses and use f8 and the signal/noise is equal
from the APS 24 mp and the 24 mp 24x36 there is no enlargements advantages with 24x36
It is a pure optical question= we need 1,6 better lenses on the APS

Mikael, if what you wrote were true, then a 4000 ppi, 22 megapickle scan of a 35mm negative would produce just as good an 8" x 10" print as a 500 ppi, 20 megapickle scan of an 8" x 10" negative from a view camera. Or, that a 5DIII would make as good a print as a 20 megapickle 8x10 view camera back.

It would also mean that this 18 megapickle toy:

http://www.amazon.com/Cybersnap-1018-Micro-Digital-Camera/dp/B007PVMKPU/ref=sr_1_35?s=photo&ie=UTF8&qid=1358787121&sr=1-35

would beat the pants off of the original 1Ds with its wimpy 11 megapickles.

And that assertion is so laughably idiotic that there's no point in further discussion with you.

Seriously, dude. Get a clue. You're in more need of one than any I've seen here in a looooooooong time.

Cheers,

b&

 

[AlanF]

I have been using a 7D for 3 years for bird photography and earlier this month I bought a 5D III, with which I am absolutely delighted. But, the old 7D can hold its own at the limits of resolution. This is not scientific as only two photos are compared under slightly different conditions, but they do illustrate the advantages of pixel density at 100% crops for fine detail. Both were in raw, 100% crops and with the 300mm f/2.8 II + 2xTC at f/8. The 7D was at iso 320 and /1250, the 5D III at 800 and 1/500 (so same amount of light but the 7D was given an iso advantage). The 7D shot is less pixillated (1332x1218 vs 809x726) and there is slihgtly better resolution of the feathers on the back. Last night, I took a snow scene lit by lamplight at 1/40 sec and iso 12,800 with the 5D III - I couldn't have done this on on the 7D. It's horses for courses.

 

[TrumpetPower!]

The 7D shot is less pixillated (1332x1218 vs 809x726) and there is slihgtly better resolution of the feathers on the back.

The 5DIII shot is also a bit underexposed, which isn't doing it any good. Indeed, the loss of IQ because of that underexposure is roughly comparable to the resolution gain of the 7D.

The two are close enough that I wouldn't make the decision of which camera to pick up based on raw resolution. Instead, I'd go with the 5DIII for its superior autofocus performance -- and, presumably, for the not-yet-real 7DII for its autofocus equal to the 5DIII but higher resolution and faster framerate. But I still might pick a 1DX instead of either, despite its lower pixel density, for its superior metering ability plus its even marginally better autofocus performance plus its absolutely insane framerate.

Cheers,

b&

 

[NormanBates]

Indeed, film is different, it's as if you were comparing a 18 mpix APS-C sensor with a 50 mpix fullframe sensor.
My test is of that which is relevant to me: close to 20 mpix on both

Also, my test is of that which is relevant to me, which means it's basically useles for birding fans

 

[RLPhoto]

Indeed, film is different, it's as if you were comparing a 18 mpix APS-C sensor with a 50 mpix fullframe sensor.
My test is of that which is relevant to me: close to 20 mpix on both

Also, my test is of that which is relevant to me, which means it's basically useles for birding fans

.

Just saw your tests, Nice job. FF takes a wet dump on APS-C for sharpness, and MF would do the same for FF 35mm.

 

[AlanF]

The 7D shot is less pixillated (1332x1218 vs 809x726) and there is slihgtly better resolution of the feathers on the back.

The 5DIII shot is also a bit underexposed, which isn't doing it any good. Indeed, the loss of IQ because of that underexposure is roughly comparable to the resolution gain of the 7D.

The two are close enough that I wouldn't make the decision of which camera to pick up based on raw resolution. Instead, I'd go with the 5DIII for its superior autofocus performance -- and, presumably, for the not-yet-real 7DII for its autofocus equal to the 5DIII but higher resolution and faster framerate. But I still might pick a 1DX instead of either, despite its lower pixel density, for its superior metering ability plus its even marginally better autofocus performance plus its absolutely insane framerate.

Cheers,

b&

I had made a decision to buy the 5D III, it is a phenomenal piece of kit and, most importantly with incredible autofocus consistency. The 7D is 3 years older technology. I showed those shots to show that even with outdated technology, pixel density can be important - I am not arguing to buy a 7D. We'll see what a more modern APS-C can do if ever the 7D II arrives.

 

[Sporgon]

"As I already explained, it is a question of how much enlargement is necessary for a print."

this is about pixels, no film there you are enlarging grain etc , as long you have 1,6 times better lenses and use f8 and the signal/noise is equal from the APS 24 mp and the 24 mp 24x36 there is no enlargements advantages with 24x36
It is a pure optical question= we need 1,6 better lenses on the APS

Whoa !! Come on guys, Mikael is right here. I'm amazed how many people still confuse format with digital image size. It must be all us old gits who come from film.

To use 18 MP as example, 18MP of digital info is 18MP - it doesn't matter what format it's in ! A 5D mk1 file has to be "enlarged" more than that from a 7D. Talking of enlarging an APS file "40" times is wrong - who ever said that does not understand what is happening to enable them to view this digital data as a picture. Enlarge an APS Film negative 40 times - yes.

I'm surprised Neuro hasn't jumped in and put that one straight !