[Haakan]

Sorry for the perhaps controverisial heading, but I thought that it might catch the attention, as well as it does convey something that, at least to me, came somewhat as a surprise.

When considering FF vs APS-C I always took as absolute fact that if all things equal, the FF will have about one stop better low light performance. This based on that with double the area, it will have one stop better ISO performance.

But then I started to think what parameters I play with when constructing an image. The parameters I control are FOV, DOF (aperture) and shutter time. The ISO setting comes as a consequence of these parameters (and sometimes as a limitation if noise becomes too high).

So then I thought that in order to have more of an "apple to apple" comparison, I should compare the noise performance under the same image parameters, i.e. under the same FOV, DOF and shutter time.

For a given FOV and DOF, the required f-number is proportional to the format size. Thus with FOV, DOF and same shutter time, the FF will require a one stop higher ISO compared to the APS-C.

In order to see how a SNR curve between APS-C and FF would look like if comparing images taken under the same image conditions (same FOV, same DOF and same shutter speed), I took the DxOMark numbers for Nikon D90 and D700 (I used these cameras since they have the same pixel count and was released roughly at the same time, so they should be similar in technology evolution status, but I do not want to trigger a brand vs brand discussion).

I have attached the curve I ended up with. Of course, the comparison holds only within the limits that the aperture of the lens can be changed.

A very practical situation would be e.g. in sports or in nature shooting using long focal length lenses. Using a lens of 400mm f/2.8 for APS-C and 600mm f/4.0 would actually result in images having the same noise performance. But the 600mm lens would set one back an additional $2000 dollars (using Nikon prices)

I am sorry if I have stated something that is old news to you all, but too me the result came somewhat as a surprise and was not what I had intuitively thought before. If I have made any logical sommersaults in my thinking, I will be happy to be corrected. As for now, it made me realize that for the way I shoot images, a FF would actually not give me any benefit in terms of low noise performance.

Best regards,
Haakan

Attached Images

 

[mithrandir]

Thank you for thinking outside of the box. Apples to Apples is the way to go. I am not an optical or electrical engineer so I can't comment on the technical aspects, but I think you are on to something. Those wiser than I will be able to confirm it for us.

[StigVidar]

But then, Pentax should start making a 16-50/2 zoom instead of f/2.8. Olympus understand this fact, but not Pentax. And we are also missing a lot of f/1 prime lenses.

[Haakan]

But then, Pentax should start making a 16-50/2 zoom instead of f/2.8. Olympus understand this fact, but not Pentax. And we are also missing a lot of f/1 prime lenses.

I fully agree. The other end of the FF equation is on the short focal length, and there, if you want to have a small DOF, the FF have its benefits using current set up of lenses. E.g. a 35 mm f/1.0 or 1.2 would be a perfect APS-C lens, matching the 50mm lenses for FF. But for me, the low light performance is more important for the longer focal lengths (my wide lenses I can easily handhold, and I shoot mostly landscape with them and there I want somewhat larger DOF)

Best regards,
Haakan

[StigVidar]

I fully agree. The other end of the FF equation is on the short focal length, and there, if you want to have a small DOF, the FF have its benefits using current set up of lenses. E.g. a 35 mm f/1.0 or 1.2 would be a perfect APS-C lens, matching the 50mm lenses for FF. But for me, the low light performance is more important for the longer focal lengths (my wide lenses I can easily handhold, and I shoot mostly landscape with them and there I want somewhat larger DOF)

Best regards,
Haakan

Yes, theory is one thing and practical use is another. I am perfectly happy with aps-c and my current lens line up. I don't think I would switch to 24x36 even if I had the option.

[snogglethorpe]

It's an interesting way to look at it, but whether it's relevant seems to strongly depend on situation -- in many cases the subject is sufficiently distant that even wide open, the DOF is "enough", and the reduced DOF from a FOV-equivalent FF camera simply wouldn't matter (or would be desirable!).

[Gooshin]

noise is not the issue

BANDING is

[falconeye]

This is just another side of the same medal we have been talking about in the other thread.

Let me jump straight to conclusion now:

Two cameras taking images with
- same distance to subject
- same field of view
- equal diameter lens (same lens aperture in mm, possibly different f-stops)
- same shutter speed
- same quantum efficiency (same CMOS process)

will produce identical (i.e., indistinguishable!) results when considering the following aspects:
- image composition
- noise
- dynamic range
- depth of field
- resolution limited by diffraction

To be explicit: the sensor size doesn't come into play here!

Additionally, cost is mostly determined by weight of glass and would be almost independent of sensor size too (same size of lenses, same weight, same cost).

Where sensor size comes into play is here:

- possible number of pixels
- possible resolution as limited by classical optics for a given cost
(a 50-135/2.8 APS-C 100 lp/mm zoom may be harder to build than a 75-200/4 FF 67 lp/mm zoom)

So, when comparing FF with APS-C, available lenses (and their diameters in mm) and their cost must be compared. Not the sensor sizes.

It is the availability of bigger glass which gives FF a head (like 75-200/2.8 for FF but no 50-135/2.0 for APS-C). And the inavailability is not marketing (only). It is partly due to technical obstacles which would make such offers expensive. Like FourThirds already is more expensive than APS-C now, when including the cost of glass.

So, what's this third graph supposed to be (white, faintly visible)? Did you manually move one graph from one place to another and that's the "remainder"?

[Haakan]

This is just another side of the same medal we have been talking about in the other thread.

Let me jump straight to conclusion now:

Two cameras taking images with
- same distance to subject
- same field of view
- equal diameter lens (same lens aperture in mm, possibly different f-stops)
- same shutter speed
- same quantum efficiency (same CMOS process)

will produce identical (i.e., indistinguishable!) results when considering the following aspects:
- image composition
- noise
- dynamic range
- depth of field
- resolution limited by diffraction

To be explicit: the sensor size doesn't come into play here!

I am happy that you have the same conclusion, that makes me feel that I did not do too many logical faults.

So, what's this third graph supposed to be (white, faintly visible)? Did you manually move one graph from one place to another and that's the "remainder"?

Yes, the white is where the original D700 graph was when having the x axis displaying ISO.

Best regards,
Haakan

[alohadave]

If you are interested in detailed analysis read this:

Noise, Dynamic Range and Bit Depth in Digital SLRs

This chapter details noise in different sized sensors and how ETTR really works:

Noise, Dynamic Range and Bit Depth in Digital SLRs -- page 3

[Torphoto]

Soo what your saying is the pics from my 5d are not better than those from an aps-c cam? that I wasted my cash for it? that there really is not a 1-2 stop advantage in noise?

hummm Try and explain that to me again.

[GLXLR]

I guess it makes sense for Super long lenses, but what about normal lenses?

If you put that 400mm f/2.8 on a FF body, how would it be possible to get the same bokeh as it on APS-C with the same FL? A 280mm or 300mm f/2 would be the only option and as of now there are no production 300mm f/2s.

Or a normal prime. To get the bokeh of a 50mm f/1.4 on FF, you would need a 35mm f/1.0 on digital sensors. Of course, 50mm f/1.4 bokeh on FF is (IMO) TOO creamy and f/2 would be enough for most people, but that would STILL require a 35mm f/1.4 (canon sells one for $1300)

[pingflood]

Soo what your saying is the pics from my 5d are not better than those from an aps-c cam? that I wasted my cash for it? that there really is not a 1-2 stop advantage in noise?

hummm Try and explain that to me again.

God damn it, you need to stop taking pictures and sit around and theorize instead!

[Haakan]

Soo what your saying is the pics from my 5d are not better than those from an aps-c cam? that I wasted my cash for it? that there really is not a 1-2 stop advantage in noise?

hummm Try and explain that to me again.

I also was somewhat perplexed from the result of the excercise. But when I thought of it it did make some sense. I was too used of looking at low noise performance by comparing noise at different ISO levels. But when you think of it, ISO is a not a parameter you set when you construct on image. You want to control the FOV, the DOF and Shutter Speed to achieve your wanted "look". Myself I mostly use aperture priority to get my wanted DOF, then I set the longest time I can use depending on if the object moves, or if I use handheld. The ISO I have to use is then a consequence of those parameters. And it so happens that the relation between FOV, DOF and sensor size is such that the gain I have in better performance at a given ISO for the FF (roughly one stop) is exactly opposite to the loss I have due to having to go to a smaller aperture to achieve the same DOF.

Of course the relation assumes that you are within the limits set by the lenses, but this is likley to be the case when looking at longer focal lenses where max aperture is typically getting smaller as focal length increases. For wide lenses the FF will give you the possibility to go to a smaller DOF than what is typically possible with current set of lenses. But for me the problem with having to go to high ISO is most often occuring on the slightly longer focal lengths (sports, nature/animals etc.) and then the benefit with FF in terms of noise performance seems more or less gone.

Best regards,
Haakan