[johnmflores]

The biggest advantage that APS-C has over full frame is that it shares a mount with full frame cameras.

Actually, the biggest advantage that APS-C has over full frame is price.

[Nicolas06]

Actually, the biggest advantage that APS-C has over full frame is price.

+1. The biggest advantage is that this is the de-facto standard for consumer DSLR with very competitive prices. You can get camera and kit for 500$ and get good gear that last years and give you lot of nice pictures.

The gear isn't the limiting factor.

[Rondec]

Actually, the biggest advantage that APS-C has over full frame is price.

I guess I should have been clear that I meant the biggest advantage APS-C has over micro four thirds or one inch sensors is that there is an upgrade path to a larger sensor if one decides he/she needs or just wants to move up. If you feel like you have maxed out your Olympus or Panasonic equipment or, your V1, your only solution is to buy all new gear with a different mount. If you have a K30 and a DA 40 and DA 70 and a DA 50 f1.8, you could easily move to a K-1 and continue using the same lenses. The same is true for Canon, Nikon and Sony.

APS-C may eventually become the "gateway drug" to full frame.

[biz-engineer]

I think 10 years from now, aps-c will only be available in mirrorless and hi-end compact camera's. Dslr will be full frame only and mainly for pro shooters, so in lower volume then mirrorless is.

Well, I think this will happen before the 10 years mark.

[dtmateojr]

I guess I should have been clear that I meant the biggest advantage APS-C has over micro four thirds or one inch sensors is that there is an upgrade path to a larger sensor if one decides he/she needs or just wants to move up. If you feel like you have maxed out your Olympus or Panasonic equipment or, your V1, your only solution is to buy all new gear with a different mount. If you have a K30 and a DA 40 and DA 70 and a DA 50 f1.8, you could easily move to a K-1 and continue using the same lenses. The same is true for Canon, Nikon and Sony.



APS-C may eventually become the "gateway drug" to full frame.

That's one way of looking at it. The apparent doubling of focal lengths going from m43 to ff is too much. You lose a lot on the wide end. m43 is good for those instances when you need something small to bring wherever you go without sacrificing image quality. It's the mobile camera for photographers. It's not meant to be a replacement for FF but if you are not invested in camera gear yet then m43 is a very good option and you won't really gain much by going bigger.

Then again, who doesn't want to go bigger? Liking something is reason enough to buy it. If I didn't have too many cameras I would want to try the K1 too, not because it will make better photos but simply because it's a really good piece of hardware. That thing is more usable than the Nikon and Canon counterparts. CaNikon users don't know what they are missing.

[wildman]

A slight digression...

...if the market for dedicated cameras becomes smaller manufactures might become much more specialized, high end and expensive.
Thus, for instance, Canon might be the place to go if you are a nature or sports photographer and the street shooter looks to Fuji for small fast responsive unobtrusive scene grabbers - sort of a digital M3 for instance. But no single manufacturer will try to cover the whole range from PS to FF and all the possible iterations thereof any longer. The logical outcome of a shrinking market?

Just a thought.

[dtmateojr]

You explain us you couldn't see the pixel on a 4K on a 85" display as you explain. But 4K is 8MP. And it was enough for 85". That pretty huge. How big really do you plan to print? 200"? And how far away would you be from the picture?

If you have more pixels but your eyes can't see the difference, this just mean they will average the noise. You eyes are the limitation that made DxO choose this resolution.

The thing is if you go for that 85" and look near to it, the D810 will have better details thanks to added resolution at low isos... And that both will have far less than 8MP worth of detail at high iso like 6400 anyway...

I think you missed the whole post or maybe I'm not understanding you.

What I was saying is that printing a FF 36Mp at 8x10 is not a measure of its real performance. A 16Mp APS-C at 8x10 is not a measure of real performance. It just so happened that 8x10 is a standard that has been chosen by dxomark and it is a fact that this arbitrary standard favours higher megapixels. 8x10, mind you, is also used for certain calculations other than SNR. Circle of confusion, for example, which dictates depth of field is also based on a 8x10 print viewed from a distance of 10".

If for example they normalised at something larger than 8x10 (8Mp), say, at 16Mp, then the PRINT snr graphs of the D4 vs D800 will change dramatically because they would no longer be equal. The D4 will come out on top. So in this case alone we could already see the unintended bias resulting from that arbitrary choice of standard.

And then there's the human factor. As per my own experience, a 85" 4K tv (8Mp) looks very good at a distance of 1.5m. Which means that anything larger than 8Mp printed at 85" will probably not make any difference. You might notice something different if they were put side by side but on their own then would pass image quality requirements. So using PRINT snr as basis for sensor performance is not a good indicator, especially not with a tiny 8x10.

[Nicolas06]

I think you missed the whole post or maybe I'm not understanding you.

What I was saying is that printing a FF 36Mp at 8x10 is not a measure of its real performance. A 16Mp APS-C at 8x10 is not a measure of real performance. It just so happened that 8x10 is a standard that has been chosen by dxomark and it is a fact that this arbitrary standard favours higher megapixels. 8x10, mind you, is also used for certain calculations other than SNR. Circle of confusion, for example, which dictates depth of field is also based on a 8x10 print viewed from a distance of 10".

If for example they normalised at something larger than 8x10 (8Mp), say, at 16Mp, then the PRINT snr graphs of the D4 vs D800 will change dramatically because they would no longer be equal. The D4 will come out on top. So in this case alone we could already see the unintended bias resulting from that arbitrary choice of standard.

And then there's the human factor. As per my own experience, a 85" 4K tv (8Mp) looks very good at a distance of 1.5m. Which means that anything larger than 8Mp printed at 85" will probably not make any difference. You might notice something different if they were put side by side but on their own then would pass image quality requirements. So using PRINT snr as basis for sensor performance is not a good indicator, especially not with a tiny 8x10.

A D4 has the same issue as D810... If you actually look at measurement the screen measurement (so 16MP) give that for SNR18% at 32db is a bit better than iso 800, around iso1000.

If we use the normalized "print" setting the SNR18% is a bit better than 1600 iso, arround 2000 iso actually. Reducing the resolution by a factor of 2 allowed to double the iso for the same SNR18% at 32db that physically make lot of sense.

Now if we look at D810, the print result is around 1800-1900 isos while the screen SNR at "screen" resolut of 36MP is at iso 400.

It doesn't take a math genius to understand that if you were reducing to 16MP instead of 8MP the result would be between 400 and 1800-1900. Likely arround iso900 so rougly the D4 performance for screen, full 16MP pixel resolution.

The idea that the results would be very different look to be quite overrated. The hypothesis was that only the D810 would improve by reducing the resolution but that the D4 was so badly designed that it performance stayed constant and didn't improve on reduced resolution. This of course make no sense.

What is true is if I check at very high isos, like iso 25600, the D4 is actually better than the D810. The larger pixels do help, but only in extreme cases. This may be useful for sports but with current technology if you get 1MP worth of real detail at 25600 you'd be lucky anyway. You'd want to print no more like 4x6" max and you'll see the print is far from perfect still.

There no problem with DxO if you take the time to understand what you read. And if you don't, really if you just take the single score as a rough indication as they explain ignoting variations lowers than 5 point you get quite good approximation of the sensor performance anyway. This is not like there were very bad cameras out there.

[BrianR]

If for example they normalised at something larger than 8x10 (8Mp), say, at 16Mp, then the PRINT snr graphs of the D4 vs D800 will change dramatically because they would no longer be equal. The D4 will come out on top. So in this case alone we could already see the unintended bias resulting from that arbitrary choice of standard.

Choosing a different reference resolution would just shift their print graphs up or down by exactly the same constant amount, the D4 and D800 would still lie pretty much on top of one another and probably most people wouldn't even notice.

Their print normalization calculations are here, you can check for yourself how they do it: Detailed computation of DxOMark Sensor normalization - DxOMark

[dtmateojr]

Choosing a different reference resolution would just shift their print graphs up or down by exactly the same constant amount, the D4 and D800 would still lie pretty much on top of one another and probably most people wouldn't even notice.



Their print normalization calculations are here, you can check for yourself how they do it: Detailed computation of DxOMark Sensor normalization - DxOMark

Nope. You can do the math yourself:

nSNR = SNR + 20 x log10(sqrt(N1/N2))

N1 is the sensor image size and N2 is the chosen print size. You can get N1 from the SCREEN snr.

At N2 = 16Mp or higher, the D4 beats the D800.

And you can read about that here

https://dtmateojr.wordpress.com/2014/05/19/megapixel-hallucinations/

Cue in the admins that would flag me again for posting a link to my blog. They would rather that I post a very long reply here than point a link to a more detailed explanation. So click on that link before a trigger happy admin deletes it.

[BrianR]

Nope. You can do the math yourself:

nSNR = SNR + 20 x log10(sqrt(N1/N2))

If you change N2 from 8 to 16, both print graphs get shifted down by 10*log(16/8)=3dB.

What you normalize to has zero impact on which comes out on top. The shapes of the print graphs and relative positions to one another is not in anyway changed by different choices of N2.

[Nicolas06]

Nope. You can do the math yourself:

nSNR = SNR + 20 x log10(sqrt(N1/N2))

N1 is the sensor image size and N2 is the chosen print size. You can get N1 from the SCREEN snr.

At N2 = 16Mp or higher, the D4 beats the D800.

And you can read about that here

https://dtmateojr.wordpress.com/2014/05/19/megapixel-hallucinations/

Cue in the admins that would flag me again for posting a link to my blog. They would rather that I post a very long reply here than point a link to a more detailed explanation. So click on that link before a trigger happy admin deletes it.

The SNR of D4 at real iso 100 is 43db at 8MP for print size. You formula say at 16MP it would be 40 and that is what on the graph for screen size...

Now let's apply the same formula to D810. it get 44db at 8MP for print size. From this point the formula say it would be 41 for 16MP. The D810 is a bit better than D4 at iso real iso 100 at 8MP... It stay a bit better than D4 at real iso 100 at 16MP.

You could do the other way, let's start from the 36MP score and apply your formula... At 36MP the SNR is 37db at real iso 100 for D810... The formula give us 40.5 db. A tiny bit better than D4 still.

The D810 at low iso manage to keep the same SNR as D4 at same resolution. But this was to be expected. If you want to keep same SNR at higher resolution, either you sensor technology got better or you increase the sensor size. This doesn't even need a test.

So the question basically is the 39.7db of D810 SNR ar 36MP iso 47 good enough to be used while the D4 manage 42.3 db at iso 75, 16MP?

Well we can do it differently. Do you think the image the D4 take at real iso 100-120 terrible or would you rate them really good in term of noise level? Because this is the iso setting you have to put the D4 to get the same SNR as D810...

If you tend to trash all the shoots that could not do it at iso 75 because iso 120 too bad, yeah you can't really benefit of 36MP of D810... if iso 125 on D4 is acceptable quality then with D810 you can choose. Downscale to 16MP for same high iso performance or keep 36MP for more details.

Honestly man all of this is very borring. Both camera are great and if you have looked at a 36MP picture you seen it can be sharp and high quality. Thanks.

Besidre the SQRT might confuse you to thing you get only 1.4 for a 2X factor but you can of course remove the SQRT from a log:

nSNR = SNR + 20 x log10(sqrt(N1/N2)) is equivalent to
nSNR = SNR + 10 x log10(N1/N2)
and of course if you simply decide to shift the actual N2 size from 1 factor, say 2, then that just substracting 3db.

Want 16MP results snr for all camera? Remove 3db to all results. Done.

[dtmateojr]

If you change N2 from 8 to 16, both print graphs get shifted down by 10*log(16/8)=3dB.

What you normalize to has zero impact on which comes out on top. The shapes of the print graphs and relative positions to one another is not in anyway changed by different choices of N2.

Did you calculate it? Because the second term drops to zero for N1=N2. Read my blog if you are not scared of the truth.

---------- Post added 03-20-16 at 09:02 ----------

The SNR of D4 at real iso 100 is 43db at 8MP for print size. You formula say at 16MP it would be 40 and that is what on the graph for screen size...

Now let's apply the same formula to D810. it get 44db at 8MP for print size. From this point the formula say it would be 41 for 16MP. The D810 is a bit better than D4 at iso real iso 100 at 8MP... It stay a bit better than D4 at real iso 100 at 16MP.

You could do the other way, let's start from the 36MP score and apply your formula... At 36MP the SNR is 37db at real iso 100 for D810... The formula give us 40.5 db. A tiny bit better than D4 still.

The D810 at low iso manage to keep the same SNR as D4 at same resolution. But this was to be expected. If you want to keep same SNR at higher resolution, either you sensor technology got better or you increase the sensor size. This doesn't even need a test.

So the question basically is the 39.7db of D810 SNR ar 36MP iso 47 good enough to be used while the D4 manage 42.3 db at iso 75, 16MP?

Well we can do it differently. Do you think the image the D4 take at real iso 100-120 terrible or would you rate them really good in term of noise level? Because this is the iso setting you have to put the D4 to get the same SNR as D810...

If you tend to trash all the shoots that could not do it at iso 75 because iso 120 too bad, yeah you can't really benefit of 36MP of D810... if iso 125 on D4 is acceptable quality then with D810 you can choose. Downscale to 16MP for same high iso performance or keep 36MP for more details.

Honestly man all of this is very borring. Both camera are great and if you have looked at a 36MP picture you seen it can be sharp and high quality. Thanks.

Wrong! SNR is SCREEN snr NOT print snr. Read my blog.

[Nicolas06]

Did you calculate it? Because the second term drops to zero for N1=N2. Read my blog if you are not scared of the truth.

---------- Post added 03-20-16 at 09:02 ----------




Wrong! N1 is SCREEN snr NOT print snr. Read my blog.

You assume I have to always use 36MP for my D810 and 16MP for the D4. But that's plain wrong. I can at any time decide to resize my D810 to 16MP if I see fit and remove all advantages the D4 had in you dreams.

Now if I choose a low iso setting, I can have more pixels, more resolutions and print larger and be happer to get more details with my manifying glass when staring at the print on the D810.

And I loose 3db. Yeah but that still good enough. Exactly like D4 loose 3db at 16MP vs 8MP. How edifying is it !

Now if the picture are good enough, that completely irrelevant.

[BrianR]

Did you calculate it?

Yes - it's basic logarithm stuff.

Because the second term drops to zero for N1=N2.

Of course, but that is in no way contradicting what I just said.

Read my blog if you are not scared of the truth.

If you rewrite it in a non-inflammatory tone, I'd be happy to.