sharper images with larger sensor?

[OutsideTheMatrix]

yes i learned now, that MTF50 lp/mm is a better criteria than DXO sharpness rating

lp/mm is not a good criteria when comparing against different sized sensors. Remember, that is line pairs per mm at the sensor. What you care about is line pairs in the final displayed image. To get to a given display size you will need to magnify the resolution at the m43 sensor by a factor of two compared to the FF sensor.

Many sites use "line pairs per picture height" or some other similar metric so that you can compare across different sensor sizes better without having to scale the numbers yourself. Lenstip is one of the few that reports in lp/mm instead.

If a test site is giving you lp/mm and you want to compare across two different sensor sizes then you can divide lp/mm by the crop factor to give you a sensible comparison. That will give you "line pairs per picture diagonal" actually which is of course still a bit different than "line pairs per picture height" when systems have different aspect ratios.

If you want to compare between different test sites to see how comparable their results are you can convert lp/mm to lp/ph by looking up the height of the given sensor (13mm for m43 and 24mm for FF) and multiplying lp/mm by that dimension.

, and the Olympus seems better here.

Indeed, at the sensor the Olympus lens is better in lp/mm. Of course a cell phone camera lens would be even better still, but of course would give substantially worse sharpness in the final image than the Olympus because again the cell phone sensor is so much smaller.

If we instead account for the crop factors the Tamron/Sony absolutely trounces the Olympus. Peak resolution of the Tamron/Sony being about twice that of the Olympus.

But as already pointed this has more to do with the fact that the m43 camrea used in the test is a 12MP sensor while the FF camera used in the test is a 42MP sensor.

But it seems very complicated and i'm not sure yet about what it means if i take a picture with System A vs B.

https://www.lenstip.com/564.4-Lens_review-Tamron_28-75_mm_f_2.8_Di_III_RXD_Image_resolution.html

https://www.lenstip.com/497.4-Lens_...ED_12-100_mm_f_4_IS_PRO_Image_resolution.html

In this particular case the A7Rii is just going to completely walk all over the E-PL1 almost no matter what lens you put on it. 12MP just can't get anywhere close to 42MP unless you put an incredibly weak lens on the 42MP camera.

It is quite complicated and confusing to be sure.

Just remember that all of these "lens test" sites are actually "lens/camera system test" sites. They aren't actually testing the lens by itself.

I'm confused, why compare sensor sizes? I dont look at an image to compare sensor sizes, I compare resolution and field of view. What puts the most pixels on a given target and how good do those pixels look?

I have seen detail from a Nikon P900 and P1000 camera on the planets that you could never get from a FF or M43 camera even with a huge telephoto lens attached.

 

[Stizzu68]

To answer your original question, no a FF sensor does not necessarily give sharper images. A higher resolution (i.e. more MP) sensor typically will give sharper images unless a lens is really poor and you can purchase FF cameras with significantly more resolution than m43 cameras.

So if you shoot a 45MP FF system against a 20MP m43 system indeed with any decent lens the 45MP FF system is going to give you sharper images. Whether you print or display them large enough to notice that difference is another story.

If on the other hand you shoot a 24MP FF system against a 20MP m43 system there will be very little difference in sharpness again assuming you haven't done something pathological like selected an excellent lens for one and a poor lens for the other.

A reasonable example of this is the Z6/24-200 vs say E-M1II/12-100. These two combinations are about the same size and weight with the nearly the same sensor resolutions and both sport superzoom lenses regarded to be some of the best every made. It is very difficult to say one is "sharper" than the other if you compare images.

Now put the 24-200 on a Z7 and it is immediately clear the that you get a sharper result just because the Z7 has far more sensor resolution.

Now turn on pixel shift "hi-res" mode on the Olympus and almost all of the difference with the Z7 result is made up and once again the results are about equally "sharp". (There remains a small advantage to the Z7 sensor over the pixel-shifted Olympus in general as far as detail resolving goes).

you are perfectly right.

I would leave the high-resolution mode out of the choice as it is something that you cannot always use. If having high-resolution files is important I would go for a camera with a native high-resolution sensor.

Stefano

 

[ahaslett]

To answer your original question, no a FF sensor does not necessarily give sharper images. A higher resolution (i.e. more MP) sensor typically will give sharper images unless a lens is really poor and you can purchase FF cameras with significantly more resolution than m43 cameras.

So if you shoot a 45MP FF system against a 20MP m43 system indeed with any decent lens the 45MP FF system is going to give you sharper images. Whether you print or display them large enough to notice that difference is another story.

If on the other hand you shoot a 24MP FF system against a 20MP m43 system there will be very little difference in sharpness again assuming you haven't done something pathological like selected an excellent lens for one and a poor lens for the other.

A reasonable example of this is the Z6/24-200 vs say E-M1II/12-100. These two combinations are about the same size and weight with the nearly the same sensor resolutions and both sport superzoom lenses regarded to be some of the best every made. It is very difficult to say one is "sharper" than the other if you compare images.

Now put the 24-200 on a Z7 and it is immediately clear the that you get a sharper result just because the Z7 has far more sensor resolution.

Now turn on pixel shift "hi-res" mode on the Olympus and almost all of the difference with the Z7 result is made up and once again the results are about equally "sharp". (There remains a small advantage to the Z7 sensor over the pixel-shifted Olympus in general as far as detail resolving goes).

you are perfectly right.

I would leave the high-resolution mode out of the choice as it is something that you cannot always use. If having high-resolution files is important I would go for a camera with a native high-resolution sensor.

Stefano

That is a personal choice, which I have made. However, HiRes can be useful in some situations, for example indoor still life and similar unmoving subjects.

Andrew

 

[Interceptor121]

yes i learned now, that MTF50 lp/mm is a better criteria than DXO sharpness rating

lp/mm is not a good criteria when comparing against different sized sensors. Remember, that is line pairs per mm at the sensor. What you care about is line pairs in the final displayed image. To get to a given display size you will need to magnify the resolution at the m43 sensor by a factor of two compared to the FF sensor.

Many sites use "line pairs per picture height" or some other similar metric so that you can compare across different sensor sizes better without having to scale the numbers yourself. Lenstip is one of the few that reports in lp/mm instead.

If a test site is giving you lp/mm and you want to compare across two different sensor sizes then you can divide lp/mm by the crop factor to give you a sensible comparison. That will give you "line pairs per picture diagonal" actually which is of course still a bit different than "line pairs per picture height" when systems have different aspect ratios.

If you want to compare between different test sites to see how comparable their results are you can convert lp/mm to lp/ph by looking up the height of the given sensor (13mm for m43 and 24mm for FF) and multiplying lp/mm by that dimension.

, and the Olympus seems better here.

Indeed, at the sensor the Olympus lens is better in lp/mm. Of course a cell phone camera lens would be even better still, but of course would give substantially worse sharpness in the final image than the Olympus because again the cell phone sensor is so much smaller.

If we instead account for the crop factors the Tamron/Sony absolutely trounces the Olympus. Peak resolution of the Tamron/Sony being about twice that of the Olympus.

But as already pointed this has more to do with the fact that the m43 camrea used in the test is a 12MP sensor while the FF camera used in the test is a 42MP sensor.

But it seems very complicated and i'm not sure yet about what it means if i take a picture with System A vs B.

https://www.lenstip.com/564.4-Lens_review-Tamron_28-75_mm_f_2.8_Di_III_RXD_Image_resolution.html

https://www.lenstip.com/497.4-Lens_...ED_12-100_mm_f_4_IS_PRO_Image_resolution.html

In this particular case the A7Rii is just going to completely walk all over the E-PL1 almost no matter what lens you put on it. 12MP just can't get anywhere close to 42MP unless you put an incredibly weak lens on the 42MP camera.

It is quite complicated and confusing to be sure.

Just remember that all of these "lens test" sites are actually "lens/camera system test" sites. They aren't actually testing the lens by itself.

I think we are drifting here lpmm and lpph are linked by a very simple linear relationship and are not a way to compare across formats in fact they are not even a way to compare within the same format if the cameras used for testing are different. Even at same pixel count two cameras can produce different resolution on the same lens.

The DxOMark perceptual megapixels was an attempt to create a layman unit of measure for sharpness but ultimately it has the same issue of any other method

In that respect DxOMark metric is confusing as it does not relate to any other standard while lpmm lpph are within ISO standards and can be converted into another taking well into account that you need to keep using the same camera to test a lens relative to another.

This is the reason why I referenced lpmm which can be converted to any other ISO standard there is no measure better than another and nobody is shooting line pairs in any case so lines per image or mm means nothing in practical terms to anyone

What we are interested is relative not absolute performance of two lenses on the same format. If you use the same ancient camera everything is fine as long as the lens does not resolve the full resolution of the sensor

So the idea of looking at a resolution map to choose a camera is fundamentally flawed

--

instagram http://instagram.com/interceptor121
My flickr sets http://www.flickr.com/photos/interceptor121/
Youtube channel http://www.youtube.com/interceptor121
Underwater Photo and Video Blog http://interceptor121.com

 

[kenw]

A genuine question. Tripod based pixel shift produces a different effect on colour sampling than demosaicing a Bayer CFA file with higher nominal pixel resolution.

That's a great point, and I almost mentioned it but decided the post was already getting a bit long!

Pixel shift does a great job at avoiding color moire of course which depending on the subject can make it "better" than a higher resolution native Bayer image that might produce objectionable color moire.

A native higher resolution sensor on the other hand has a smaller pixel size (again relative to picture height if we've got different sized sensors) which means the sensor MTF is actually still higher in a single shot than compared to the shifted lower resolution sensor. This difference can be mostly made up for by some additional sharpening of the pixel shifted image (and it appears that Olympus actually does this when they make their "pseudo-RAW" files for pixel shift likely so that the image is still well suited to the typical defaults of RAW converters).

So it is sort of half a dozen on one hand and six on the other in the sense of resolving things. With "pixel shift" you can about double the megapixels compared to a native sensor. It isn't totally apples to apples but it is a reasonable comparison.

Have you direct experience of comparing images where tonal gradation matters as between a pixel shifted MFT image and a high resolution ordinary FF image?

Yes, I did some comparisons between 16MP high-res on a E-M5II and a native 45MP Z7. The "hi-res" shots are very comparable in tonal gradation to the Z7 but this has little to do with the pixel shifting and more to do with the fact that the 8 shots of the hi-res combined have effectively lowered the base ISO noise of the E-M5II to match the base ISO noise of the Z7.

The biggest difference in "tonal gradation", at least as far as what I consider "tonal gradation", is actually just down to exposure more than anything. Being able to distinguish subtle gradations in tone (or color for that matter) is actually more of a SNR question than anything else. Collect more photons you get better SNR in the midtones and you can see more subtle gradations.

This is where much of the "FF advantage" for landscape shooting comes from - an "equivalent" lower base ISO on the FF sensor. If we don't care at all about shutter speed (which we usually don't in landscape) so that we can shoot both systems at the same ISO for a given depth of field then a FF sensor at say ISO 100 will collect four times as many photons across the image as a m43 sensor also at ISO 100. That difference is actually quite obvious in smooth gradations like in the sky.

Here for example is a crop of a scene with some sky shot with m43 pixel shift, m43 single shot and the Z7 (as usual click to see the orignal at higher resolution):



The clouds moved between the exposures but you can clearly see that just because of the increased midtone noise of the single exposure m43 shot that you just can't see subtle tonal variations that you can see in the HR and Z7 shots. Same thing if you look at the pavement in the parking lot.

So certainly if I zoom in to 100 or 200% then I can start to pick out differences in how demosaicing and moire behave between a native sensor with an AA filter, a native sensor without an AA filter or a pixel shifted image. But at least for how I think of "tonal gradations" none of that really matters. What really matters, even in down sampled images, is just what was the amount of noise in the midtones and on modern sensors that is entirely down to just how many photons you "counted" in the entire image. Pixel shift combines 8 images and so that gets better. But you could also just take 8 regular unshifted images and and average them in post processing and get the same improvement in "tonal gradation".

Apologies if my definition of "tonal gradation" is different than yours and perhaps I answered the wrong question!

--
Ken W
See profile for equipment list

 

[FingerPainter]

You can safely ignore the perceptual megapixes of DxOMark

Check for lpmm on lenstip to have an idea. Usually resolution improves with smaller pixels so lpmm for MFT are around double of full frame

Why would you check a lp/mm rating? It may give you a bit of an idea of how sharp your lens is, but that doesn't tell you how sharp you pictures will be. As such it is useful for measurebating but not for practical photo production.

Much more useful is LP/PH, this tells you how much detail you can see in an image of any particular size.

Fortunately it is easy to convert from lp/mm to LP/PH. Just multiply the lp/mm figure by the sensor height in mm - 13mm for MFT and 24mm for FF.

A MFT lens/sensor combination that measures 72 lp/mm will produce an image that is only as sharp as an FF lens/sensor combination that measure 39 lp/mm.

 

[ahaslett]

You answered the right question. I have an A7R2 and an EM1.2 and shoot landscape. Your answer makes perfect sense. If only Olympus had done base ISO 25, switched capacitance at ISO 200, 28Mpix sensor and RAW exposure aids....

Andrew

 

[Interceptor121]

You can safely ignore the perceptual megapixes of DxOMark

Check for lpmm on lenstip to have an idea. Usually resolution improves with smaller pixels so lpmm for MFT are around double of full frame

Why would you check a lp/mm rating? It may give you a bit of an idea of how sharp your lens is, but that doesn't tell you how sharp you pictures will be. As such it is useful for measurebating but not for practical photo production.

Much more useful is LP/PH, this tells you how much detail you can see in an image of any particular size.

Fortunately it is easy to convert from lp/mm to LP/PH. Just multiply the lp/mm figure by the sensor height in mm - 13mm for MFT and 24mm for FF.

A MFT lens/sensor combination that measures 72 lp/mm will produce an image that is only as sharp as an FF lens/sensor combination that measure 39 lp/mm.

lpmm or lpph or other measures are convertible as per ISO standard none is better than another and none is meaningful as nobody can visualise what it means to shoot 2000 lines or 3000 lines anyway better than 100 lines per mm

Your reasoning is also flawed as you cannot generalise conclusions on tests across formats as those are carried out with different sensors within the same format with different sensor resolution field of view also comes into the mix

There is no way to simplify something that is complex you need to understand it which is what I was trying to do explaining the logic of dxomark

There seem to be some people that just take a number from a review and try to extrapolate or generalise other numbers but it is not that easy. You need to understand how it works.

 

[FingerPainter]

I was just checking out the DXO Lens database and all the M43 lenses have much lower sharpness ratings than even the most inexpensive FF lenses.

DXO doesn't publish lens sharpness ratings. They publish sharpness ratings for combinations of lenses and sensors. Their default list gives data for the highest pixel-resolution sensor on which the lens has been tested, Since some FF sensors have much higher pixel counts than the highest pixel-count MFT sensors, you will inevitably get much higher sharpness ratings for the FF combinations. It might be more instructive to look at sharpness ratings for combinationa of FF lenses and FF sensors that have about 20MP.

However, when a FF lens has the same optically measured centre sharpness as a MFT lens, the sharpness rating of the FF combination will be much higher, because of the larger sensor size. E.g., if the MFT combo produces 50 line pairs per millimetre over a 1mm high patch, this means it is producing 650 line pairs over the picture height. If the FF lens and sensor also produce 50lp/mm over a 1mm patch, it means the combo is producing 1200 line pairs over the picture height.

So an 8x10 displayed image produced from the FF combo will look much sharper than an 8x10 displayed image produced from the MFT combo.

Fortunately for MFT users, MFT lens/sensor combos often produce significantly higher lp/mm measures than the corresponding FF lens/sensor combos when sensor pixel count is similar. However, in many cases the MFT advantage in lp/mm is not quite the 24:13 advantage it would need to match FF image sharpness.

Can anyone explain this to me?

Im still looking for a new camera system so does this mean when i take pictures with

Olympus OM-D E-M1 Mark II + Olympus M.Zuiko 12-100mm F4 IS Pro

vs

Sony A7 iii + Tamron 28-200mm f/2.8-5.6 Di III RXD

the images on the FF are much sharper (cant find the lenses on DXO)?

 

[Interceptor121]

You answered the right question. I have an A7R2 and an EM1.2 and shoot landscape. Your answer makes perfect sense. If only Olympus had done base ISO 25, switched capacitance at ISO 200, 28Mpix sensor and RAW exposure aids....

Andrew

A smaller sensor will always be incapable of taking as much light as a larger one. This (together with a 12 bits ADC) is the reason why base ISO on MFT is 200 and not 100 or 50.

At the very extremes of the range (lots of light and very little light) sensor size does matter. Equivalence only exists at constrained depth of field when that no longer holds larger format wins

So if you are lurking at the extremes of the spectrum you should get a larger format camera. I shoot underwater where we shoot small apertures and there is limited benefit to larger sensor and a lot more costs I am happy where I am and I know the limits of the set up as I have shot full frame and apsc before.

If I was purely a land shooter I would not shoot an MFT format but all considered my image give a run for their money to other formats in local competitions and this is because sensor performance is just one factor in photography

No matter the progress sensor area counts even if smaller sensors are more efficient intrinsically than larger one and read faster which is a benefit for example in video.

 

[FingerPainter]

You can safely ignore the perceptual megapixes of DxOMark

Check for lpmm on lenstip to have an idea. Usually resolution improves with smaller pixels so lpmm for MFT are around double of full frame

Why would you check a lp/mm rating? It may give you a bit of an idea of how sharp your lens is, but that doesn't tell you how sharp you pictures will be. As such it is useful for measurebating but not for practical photo production.

Much more useful is LP/PH, this tells you how much detail you can see in an image of any particular size.

Fortunately it is easy to convert from lp/mm to LP/PH. Just multiply the lp/mm figure by the sensor height in mm - 13mm for MFT and 24mm for FF.

A MFT lens/sensor combination that measures 72 lp/mm will produce an image that is only as sharp as an FF lens/sensor combination that measure 39 lp/mm.

lpmm or lpph or other measures are convertible

Then why present the data using units that disguise which combination gives sharper images?

People will naturally think a higher number is better than a lower number. Since photographers mostly care more about how their images look than how their lenses perform on some arbitrarily chosen metric, it is better to use the units that more easily let them determine relative performance with respect to what they care about: image sharpness.

 

[kenw]

I'm confused, why compare sensor sizes? I dont look at an image to compare sensor sizes, I compare resolution and field of view. What puts the most pixels on a given target and how good do those pixels look?

Exactly. Sensor size in of itself tells you very little.

 

[Eric Nepean]

You have asked a complex question. As others have said, forget about DXO.

The magnification from sensor to to final size is different for FF and M43, so lpm, or any measure per mm isn't a doo comparison.

You want to use per Image height. Knowing that the the height of the M43 sensor is 13.3mm and and for the FF sensor is 24mm, you can compare Lenstip measurements with other Lenstip measurements.

Measurements of sharpness and resolution correlate very poorly between different organizations, so comparing for example between Optical Limits and Lenstip is a minefield.

Right now I don't know of any online service that has evaluated both Olympus 12-100mm F4 and Tamron 28-200mm f/2.8-5 for Sony

I happen to own both Olympus OM-D E-M1 Mark II + Olympus M.Zuiko 12-100mm F4 IS Pro and Sony A7 iii + Tamron 28-200mm f/2.8-5.6 Di III RXD. and I will try to post a comparison.

 

[kenw]

yes i learned now, that MTF50 lp/mm is a better criteria than DXO sharpness rating

lp/mm is not a good criteria when comparing against different sized sensors. Remember, that is line pairs per mm at the sensor. What you care about is line pairs in the final displayed image. To get to a given display size you will need to magnify the resolution at the m43 sensor by a factor of two compared to the FF sensor.

Many sites use "line pairs per picture height" or some other similar metric so that you can compare across different sensor sizes better without having to scale the numbers yourself. Lenstip is one of the few that reports in lp/mm instead.

If a test site is giving you lp/mm and you want to compare across two different sensor sizes then you can divide lp/mm by the crop factor to give you a sensible comparison. That will give you "line pairs per picture diagonal" actually which is of course still a bit different than "line pairs per picture height" when systems have different aspect ratios.

If you want to compare between different test sites to see how comparable their results are you can convert lp/mm to lp/ph by looking up the height of the given sensor (13mm for m43 and 24mm for FF) and multiplying lp/mm by that dimension.

, and the Olympus seems better here.

Indeed, at the sensor the Olympus lens is better in lp/mm. Of course a cell phone camera lens would be even better still, but of course would give substantially worse sharpness in the final image than the Olympus because again the cell phone sensor is so much smaller.

If we instead account for the crop factors the Tamron/Sony absolutely trounces the Olympus. Peak resolution of the Tamron/Sony being about twice that of the Olympus.

But as already pointed this has more to do with the fact that the m43 camrea used in the test is a 12MP sensor while the FF camera used in the test is a 42MP sensor.

But it seems very complicated and i'm not sure yet about what it means if i take a picture with System A vs B.

https://www.lenstip.com/564.4-Lens_review-Tamron_28-75_mm_f_2.8_Di_III_RXD_Image_resolution.html

https://www.lenstip.com/497.4-Lens_...ED_12-100_mm_f_4_IS_PRO_Image_resolution.html

In this particular case the A7Rii is just going to completely walk all over the E-PL1 almost no matter what lens you put on it. 12MP just can't get anywhere close to 42MP unless you put an incredibly weak lens on the 42MP camera.

It is quite complicated and confusing to be sure.

Just remember that all of these "lens test" sites are actually "lens/camera system test" sites. They aren't actually testing the lens by itself.

I think we are drifting here lpmm and lpph are linked by a very simple linear relationship and are not a way to compare across formats in fact they are not even a way to compare within the same format if the cameras used for testing are different. Even at same pixel count two cameras can produce different resolution on the same lens.

Agree. That was my point that the 12MP vs 42MP was dominating the apparent differences here. I mentioned lp/mm and lp/ph conversion in case you wanted to compare results between two sites that had measured the same lens on the same camera (or at least same sensor resolution). Any "cross-system" comparison is going to be fraught.

The DxOMark perceptual megapixels was an attempt to create a layman unit of measure for sharpness but ultimately it has the same issue of any other method

Exactly. Not to mention DxOMark frequently botches it badly along the way.

In that respect DxOMark metric is confusing as it does not relate to any other standard while lpmm lpph are within ISO standards and can be converted into another taking well into account that you need to keep using the same camera to test a lens relative to another.

DxOMark frequently likes to apply some sort of black box to standard measurements and give us some single number to rule them all - which often in the process ends up stripped of any real value.

This is the reason why I referenced lpmm which can be converted to any other ISO standard there is no measure better than another and nobody is shooting line pairs in any case so lines per image or mm means nothing in practical terms to anyone

Yes, but again this does put a burden on the user and the OP already was confused looking at lp/mm between a FF measurement and a m43 measurement. That's why I responded to their post to point out lp/mm on m43 compared to lp/mm FF tells you nothing in itself by the raw number lp/mm. The OP cares about which one is going to look the sharpest in their displayed image after all.

What we are interested is relative not absolute performance of two lenses on the same format.

That's not what the OP is interested in though or the question they are asking.

What LensTip and others can show relatively accurately though is what you just stated - the relative performance of two lenses on the same sensor. Beyond that things are fraught - as I think the OP is discovering from the length of the thread

If you use the same ancient camera everything is fine as long as the lens does not resolve the full resolution of the sensor

But of course pretty much every lens does so the old cameras mask the differences between lenses and then the test site feels they need to change the camera they are using to test and then... well... yeah.

So the idea of looking at a resolution map to choose a camera is fundamentally flawed

I mean it can tell you *something* if you are very careful about considering what it is actually telling you. But *mostly* it is just telling you what the resolution of the sensor is...

 

[Interceptor121]

You cannot disguise anything because only a fool would compare resolution across formats

There are two different things

1. Sensor resolution

2. Lens resolution

As 2 needs 1 to be measured it follows that you can't compare 2 lenses using different sensors resolutions. If you do your comparison is flawed

Example for you:

Canon 50mm F1.8 STM has a maximum resolution of 46 lpmm or 2208 lpph because it is measured on the Canon 6D

Panasonic 25mm F1.4 MFT has a maximum resolution of 75 lpmm or 1950 lpph because is measured on the Olympus E-PL1

The canon 6D has a pixel size of 6.54 microns and therefore a sensor resolution of 76.45 lpmm or 3669 lpph

The Olympus E-PL1 has 116 lppm or 3037 lpph

Which lens is better? The Panasonic because it resolves more of the available pixels (64% vs 53%)

Which of the two cameras has more resolution? The Canon 6D irrespective of any lens test

Current MFT sensor have a nyquist limit at 151 lpmm or 3939 (which is the reason why the sensors resolution are what they are)

You don't need to do any lens test to determine which sensor has more theoretical resolution this is given by the megapixel count.

So the Sony A7RII of the op at 45 megapixels will produce sharper image than a camera at 20 megapixels and less sharp than a camera at 50 megapixels if they use the same lens

This is so obvious one does not need to go and read lens reviews to find out.

If you keep the megapixel count constant within formats larger sensor produce higher quality image in virtue of larger area and therefore light gathering. Also this is very easy to understand.

What I was trying to explain to the op is that there is no need to go and read DxOMark lens test to find out something that is already known. In some cases a good sensor with a poor lens will produce a lower quality of a smaller format with a great lens but those are exception to the rule. As most of the photos are consumed at relatively low resolution sensor size and megapixel count become not so interesting once you pass a manageable overall resolution.

--
instagram http://instagram.com/interceptor121
My flickr sets http://www.flickr.com/photos/interceptor121/
Youtube channel http://www.youtube.com/interceptor121
Underwater Photo and Video Blog http://interceptor121.com

 

[knickerhawk]

You cannot disguise anything because only a fool would compare resolution across formats

There are two different things

1. Sensor resolution

2. Lens resolution

As 2 needs 1 to be measured it follows that you can't compare 2 lenses using different sensors resolutions. If you do your comparison is flawed

Example for you:

Canon 50mm F1.8 STM has a maximum resolution of 46 lpmm or 2208 lpph because it is measured on the Canon 6D

Panasonic 25mm F1.4 MFT has a maximum resolution of 75 lpmm or 1950 lpph because is measured on the Olympus E-PL1

The canon 6D has a pixel size of 6.54 microns and therefore a sensor resolution of 76.45 lpmm or 3669 lpph

The Olympus E-PL1 has 116 lppm or 3037 lpph

Which lens is better? The Panasonic because it resolves more of the available pixels (64% vs 53%)

Which of the two cameras has more resolution? The Canon 6D irrespective of any lens test

Current MFT sensor have a nyquist limit at 151 lpmm or 3939 (which is the reason why the sensors resolution are what they are)

You don't need to do any lens test to determine which sensor has more theoretical resolution this is given by the megapixel count.

So the Sony A7RII of the op at 45 megapixels will produce sharper image than a camera at 20 megapixels and less sharp than a camera at 50 megapixels if they use the same lens

This is so obvious one does not need to go and read lens reviews to find out.

If you keep the megapixel count constant within formats larger sensor produce higher quality image in virtue of larger area and therefore light gathering. Also this is very easy to understand.

While I'm generally sympathetic to the point you're trying to make (to the OP), the above paragraph needs to be qualified. Within the equivalence envelope, the larger format does not have a "light gathering" advantage. To get to that advantage, you must go outside of the equivalence envelope, which normally implies shallower depth of field which, in turn, affects the image "sharpness" being discussed here (we don't usually shoot flat resolution charts in the real world). This is yet another wrinkle to cross-format comparisons that makes singular numeric ratings so problematic.

What I was trying to explain to the op is that there is no need to go and read DxOMark lens test to find out something that is already known. In some cases a good sensor with a poor lens will produce a lower quality of a smaller format with a great lens but those are exception to the rule. As most of the photos are consumed at relatively low resolution sensor size and megapixel count become not so interesting once you pass a manageable overall resolution.

 

[LightCameraAction]

To answer your original question, no a FF sensor does not necessarily give sharper images. A higher resolution (i.e. more MP) sensor typically will give sharper images unless a lens is really poor and you can purchase FF cameras with significantly more resolution than m43 cameras.

So if you shoot a 45MP FF system against a 20MP m43 system indeed with any decent lens the 45MP FF system is going to give you sharper images. Whether you print or display them large enough to notice that difference is another story.

If on the other hand you shoot a 24MP FF system against a 20MP m43 system there will be very little difference in sharpness again assuming you haven't done something pathological like selected an excellent lens for one and a poor lens for the other.

A reasonable example of this is the Z6/24-200 vs say E-M1II/12-100. These two combinations are about the same size and weight with the nearly the same sensor resolutions and both sport superzoom lenses regarded to be some of the best every made. It is very difficult to say one is "sharper" than the other if you compare images.

Now put the 24-200 on a Z7 and it is immediately clear the that you get a sharper result just because the Z7 has far more sensor resolution.

Now turn on pixel shift "hi-res" mode on the Olympus and almost all of the difference with the Z7 result is made up and once again the results are about equally "sharp". (There remains a small advantage to the Z7 sensor over the pixel-shifted Olympus in general as far as detail resolving goes).

you are perfectly right.

I would leave the high-resolution mode out of the choice as it is something that you cannot always use. If having high-resolution files is important I would go for a camera with a native high-resolution sensor.

Stefano

I think it depends on what question on is trying to answer. If one is curious about the potential of micro 4/3 lenses (e.g., if a 40 megapixel sensor is used in the future), test with high resolution modes may be useful.

If one is asking what you can get today, I agree with your assessment.

 

[poppyjk]

Here is an informative video of cinematographers discussing the advantages of high resolution in their craft. They cover the range of 2k, 4k, 8k, and 11k and what the combination of the eye and brain does in interpreting images beyond apparent "sharpness".

The video: https://www.dpreview.com/forums/post/64133810

They demonstrate with examples the advantages of 8K large format to their work and describe why they are anxious for future technology to bring them closer to getting Hasselblad 100mp images at 30 fps.

Starting at 25:30 in the video is a 25 minute detailed discussion of the physiology of human vision. I hope you find it as fascinating and informative as I did.

At about 35mp resolution (8k in video) and above with high quality pixels (i.e. color accuracy, tonality, dynamic range, absence of digital noise), the smoother tonal and luminance transitions start to look realistic to the brain. This gives the brain enough cues to signal depth and dimensionality. This phenomenon is separate from and in addition to the classic lens 'depth of field'.

Here are a couple of screenshots from the video.

Their paradigm. The elements are interrelated, not separate. They are keys to the image acquisition and manipulation. The points they make show that this is independent of how the image is viewed whether a 4k big screen, 1080 monitor, or smartphone.



View: original size

Perspective and magnification differences vary by resolution.



View: original size

Resolution affects depth perception, color, luminosity, tonality, and transitions dramatically. It is about more than just perceived sharpness. On the left is a 35mp(8k) image. On the right is a 100mp(11k) image. (The 8k and 11k "equivalents" are from the presenters.)



View: original size

Again 35mp left; 100mp right. The cinematographers hope that their technology continues to move toward 100mp(11k in video) because that is what science demonstrates that the eye/brain can see and interpret (this claim is by the presenters). Their opinion: who wouldn't want to see movies that matched the eye's capability and provide an almost completely immersive visual experience.

As still photographers we can enjoy the benefits of high resolution one frame at a time.



View: original size

--
"Well done is better than well said" - Benjamin Franklin
You have my express consent to edit any of my images that I post on DPR.

 

[knickerhawk]

A genuine question. Tripod based pixel shift produces a different effect on colour sampling than demosaicing a Bayer CFA file with higher nominal pixel resolution.

That's a great point, and I almost mentioned it but decided the post was already getting a bit long!

Pixel shift does a great job at avoiding color moire of course which depending on the subject can make it "better" than a higher resolution native Bayer image that might produce objectionable color moire.

Indeed...and it's not just reduced color moire. Aliasing in general is much better (as your comparison below very nicely shows with all of the diagonally presented windows and other edges).

A native higher resolution sensor on the other hand has a smaller pixel size (again relative to picture height if we've got different sized sensors) which means the sensor MTF is actually still higher in a single shot than compared to the shifted lower resolution sensor. This difference can be mostly made up for by some additional sharpening of the pixel shifted image (and it appears that Olympus actually does this when they make their "pseudo-RAW" files for pixel shift likely so that the image is still well suited to the typical defaults of RAW converters).

An interesting comment about Oly applying some (pre)sharpening to its pseudo raw files. I've seen what I believe is good evidence of Panasonic doing that to its pseudo raws, but not much evidence of Oly doing it. If nothing else, the Oly HiRes files require more input sharpening than the Panny ones. What's your reason for saying that Oly is also doing the (pre)sharpening?

So it is sort of half a dozen on one hand and six on the other in the sense of resolving things. With "pixel shift" you can about double the megapixels compared to a native sensor. It isn't totally apples to apples but it is a reasonable comparison.

Have you direct experience of comparing images where tonal gradation matters as between a pixel shifted MFT image and a high resolution ordinary FF image?

Yes, I did some comparisons between 16MP high-res on a E-M5II and a native 45MP Z7. The "hi-res" shots are very comparable in tonal gradation to the Z7 but this has little to do with the pixel shifting and more to do with the fact that the 8 shots of the hi-res combined have effectively lowered the base ISO noise of the E-M5II to match the base ISO noise of the Z7.

The biggest difference in "tonal gradation", at least as far as what I consider "tonal gradation", is actually just down to exposure more than anything. Being able to distinguish subtle gradations in tone (or color for that matter) is actually more of a SNR question than anything else. Collect more photons you get better SNR in the midtones and you can see more subtle gradations.

So true.

This is where much of the "FF advantage" for landscape shooting comes from - an "equivalent" lower base ISO on the FF sensor. If we don't care at all about shutter speed (which we usually don't in landscape) so that we can shoot both systems at the same ISO for a given depth of field then a FF sensor at say ISO 100 will collect four times as many photons across the image as a m43 sensor also at ISO 100. That difference is actually quite obvious in smooth gradations like in the sky.

Here for example is a crop of a scene with some sky shot with m43 pixel shift, m43 single shot and the Z7 (as usual click to see the orignal at higher resolution):



The clouds moved between the exposures but you can clearly see that just because of the increased midtone noise of the single exposure m43 shot that you just can't see subtle tonal variations that you can see in the HR and Z7 shots. Same thing if you look at the pavement in the parking lot.

This is a nicely done comparison. Thanks for sharing. For anyone reviewing these three crops, check out the sky tones as suggested but also be sure to check out the various manmade straight edges (e.g., along the windows, corrugated siding, building panels, wires). This is where the aliasing lives in the single frame shots. It often gives rise to the false detail that we accept as plausible or real detail. However, it's also this "detail" that quickly becomes problematic when applying processing adjustments.

So certainly if I zoom in to 100 or 200% then I can start to pick out differences in how demosaicing and moire behave between a native sensor with an AA filter, a native sensor without an AA filter or a pixel shifted image. But at least for how I think of "tonal gradations" none of that really matters.

Agreed, but where it DOES matter is at the adjustment stage. More often than not I'd rather start with a (soft) HiRes or AA filtered raw than the "comparable" AA-less high megapixel raw. It depends significantly on the subject, but I find it easier to add plausible digital detail in the form of sharpening or application of contrast strategies than to eliminate the harsh baked-in digital "detail" of aliasing/moire.

What really matters, even in down sampled images, is just what was the amount of noise in the midtones and on modern sensors that is entirely down to just how many photons you "counted" in the entire image. Pixel shift combines 8 images and so that gets better. But you could also just take 8 regular unshifted images and and average them in post processing and get the same improvement in "tonal gradation".

Apologies if my definition of "tonal gradation" is different than yours and perhaps I answered the wrong question!

 

[kenw]

You answered the right question. I have an A7R2 and an EM1.2 and shoot landscape. Your answer makes perfect sense. If only Olympus had done base ISO 25, switched capacitance at ISO 200, 28Mpix sensor and RAW exposure aids....

If you can wade through the technical minutiae there is a good thread over in PS&T on the current limits of full well capacity:

Limitations to FWC for a given pixel size?

Hi, I sort of assumed that there is a limitation of Full Well Capacity for a given pixel size. I have seen some descriptions that there is a capacitor connected to the photodiode to extend FWC. Is the size of that capacitor limited by design parameters or is it a design compromise to balance...

www.dpreview.com

Long story short, whatever decisions one might make in using dual conversion gain (DCG) to optimize low and high ISO DR you still hit a present technology limit on the maximum number of photons you can count per unit area (i.e. base ISO or FWC/area). This is down to the design of the pinned photo diode (PPD) that is at the heart of every image sensor and DCG doesn't do anything to improve this. What DCG lets you do is get the best of both worlds on a single sensor (i.e. use as much of the FWC of the PPD as possible at base ISO but also get the best read noise at high ISO) whereas before DCG you had to make a choice as to whether you optimized for one case or another regardless of the FWC of the PPD.

Right now this FWC limit is roughly around 3000e-/um^2. You can nudge it a bit one way or another depending on particulars of the design but you can't get arbitrarily low. There are limits to the diffusion profile of the PPD and the voltages you can run the chip at. You can also choose just how much of FWC you use as it becomes less linear. But there isn't much wiggle room here. Seems like around half a stop, maybe a full stop at best.

The E-M1II already has a FWC of 3012e-/um^2 from DxO's data (GH5 about the same). The Nikon Z7 is at 3104e-/um^2 (the other high resolution FF sensors about the same). Of course the Z7 claims a base ISO of 64 while the E-M1II says its base ISO is 200 but fundamentally that's down to choices other than what the actual FWC of the sensor is. Despite their different stated ISOs they actually have essentially the same FWC and it is about at the limit of what current sensor architectures can do.

So basically at the moment around 3000e-/um^2 is the rough limit for FWC regardless of sensor size. So we couldn't improve the FWC of m43 much beyond what the E-M1II and GH5 already achieve in a single exposure.

Now proper RAW exposure aids and modes - that would be a win for any camera! Why are forced to kludge this basic thing???!?

Also Olympus has gotten quite close to implementing what we really could use - a "synthetic" low ISO where we do in camera averaging to effectively increase the FWC. The LiveND feature is essentially like this, but it is only in a few cameras and biased towards only being used for very long exposures. The reduction in noise is almost an unintended side effect! Of course the issue is that for high shutter speeds this would be averaging non-contiguous exposures and so there are some limitations to how one would use it. Still for a landscape photographer would be very useful. And as sensor read rates improved and stacked architectures allow for more on chip processing this could become a very functional feature on almost any camera - dial the ISO arbitrarily low via repeated exposures read fast and accumulated on the sensor.