[Pentaxor]

wow, that's a very refreshing idea.

I hope Pentax/Hoya is listening and start making big glasses for the APS-C sensor dslrs. now I'm more than content and believe that an FOV, DOF and HIGH ISO noise difference between APS-C and FF cameras can be resolved by having an equivalent lens for each system. the power lies within the glass.

frame sensor size + corresponding lens focal length/size and speed equivalent = same output for both different camera system.

if when Pentax Hoya releases these large glasses in the near future (hopefully soon), consider it the beginning of the end for FF cameras.


anyway, to support or to test this hypothesis or theory, I would start shooting any available large fast glasses that I could find and compare Noise results at different ISO levels.

is anyone here in particular did a test already?

---

Last edited by Pentaxor; 09-03-2009 at 11:24 AM.

[WMBP]

Kit lenses illustrate this tradeoff too. A basic 28-80/3.5-5.6 lens for FF is indeed the "larger" version of the 18-55/3.5-5.6 needed to get the noise advantage, but again, zoom lens design being what it is, there isn't necessarily much of a size difference in practice. On the other hand, an f/2.4-4 standard zoom for APS-C would match the FF kit. Suggesting that Sigma is really onto something with their 17-70 - that's the equivalent of a 28-105/4-6.7 for FF. Again, though, zoom design being what it is, the 17-70 for APS-C is actually a *larger* lens than the typical 28-105.

Marc,

I'm not sure what you're saying about the Sigma 17-70. Could you elaborate? Is it that the Sigma 17-70 goes to f/2.8 at the wide end? Wouldn't the Pentax DA* 16-50 f/2.8 be even better?

I also don't quite understand what you mean when you say that the 17-70 is larger than the typical 28-105. I have some big Sigma lenses (for example, the 28 f/1.8) but the 17-70 doesn't seem all that large to me. What am I missing?

Will

[arpaagent]

Hey all, just wanted to chime in and say I appreciate Haakan for putting this thread out there and the rest for contributing their time and thoughts to helping everyone else understand it. After reading through all the posts and reading some of that "Equivalence" paper, it makes pretty good sense from a theoretical standpoint. As others have noted, there are certain realistic limits that we need to keep in mind, but this has been a very interesting topic to read about over the past few days

Thanks! I'm even more content with my APS-C now , particularly for macro + wildlife (where more DoF and reach are benificial).

[Haakan]

Sorry for being out, but sitting in a different time zone makes it sometimes difficult to keep up.

I appreciate the nice feedback, and I also appreciate the support both falconeye and especially Marc has done in bringing much more clarity to this topic than I was capable to. And Will (WMBP) for progressing the discussion an make sure no stone was left untouched.

I also think Marc's summary did a very good job in consolidating the discussions, so I will not try to do that myself (I guarantee, it will only start confusing the issue now when we finally have started to get a grip of it)

So I will only add an illustration to what Marc said about that it is really the physical size of the lens that give the low light performance (noting that this is in practice true only for lenses longer than "normal"). (I have not stuck to just one brand, it was really decided by which images looked good on BH Photo)

I have kept APS-C on left size, On the right size I have put the lens that gives about equal FOV. Where it say a +, that is where the FF have a benefit in low light performance, and I have put a = on the side where they are similar in low light performance. I think it gives a quite good illustration about that equal lens diameter gives similar performance when looking at the longer lenses.

It is also obvious that when getting to wider lenses, the "theory" breaks down in practice. E.g. there is no 33mm f/1.0 lens to give equal performance on APS-C as the 50mm f/1.4 on FF.

But for me, that is OK. I tend to have the most problems with low light when I use longer lenses (harder to keep steady while hand holding, shooting sports or nature shots, or as Marc, shooting indoor towards a scene performance). And in these siutations the theory seems to hold also quite well in practise.

Now I only have to convince my wife that I can buy the SMC Pentax-A* 135mm F1.8

Again thank you all for contributing.

Best regards,
Haakan

Attached Images

 

---

Last edited by Haakan; 09-03-2009 at 10:50 PM.

[edl]

Haakan,

Excellent thread, I've enjoyed reading through it.

Might I suggest adding 85/1.4 on FF, and 58/1.0 (not available) on APS-C to your chart? The 85/1.4 never leaves my D700 for low-light street shooting, and it's a perfect focal length for head and shoulders shots.

Thanks,
Edmund

[newarts]

.....
The parameters I control are FOV, DOF (aperture) and shutter time.
.....

A lot of good discussion has been devoted to this subject, but an important physical limit has been glossed over. It is the limit imposed by the number of photons per unit area it is possible to collect in a given time.

When doing scaling mathematics one must be very careful of divide operations (to avoid the equivalent of "divide by zero" type errors.)

The problem here is that the stated logic leads to a conclusion that it is possible to have acceptable noise quality for a vanishingly small sensor. This conclusion is based on an equivalent "divide by" error described below.

The noise/signal limits identified in the ISO definition for photographic speed are 1:40 (excellent) and 1:10 (acceptable). This means on average 100 photons per display pixel must be collected (1:10)^2 for "acceptable" and 1600 (1:40)^2 photons collected for "excellent" results. Let's take 1000 photons/display pixel to be "good enough" noise behavior for 18% reflectivity.

It turn out that a silicon based pixel saturates at around 1000 photons absorbed per square micron. Of course this varies with technological detail, but not by much.

These two facts lead to the conclusion that it is barely possible to take a photo with 1000 photons collected for mid-range illumination per display pixel using silicon collection pixels as small as 2 square micrometers.

The thought-experiment scaling undertaken in this thread involving DOF, FOV, Sensor Size, time, etc was correct so far as it went but glossed over the collection limit.

I'll come back with numbers.

Dave in Iowa

[nostatic]

assume a spherical cow...

[Marc Sabatella]

I'm not sure what you're saying about the Sigma 17-70. Could you elaborate? Is it that the Sigma 17-70 goes to f/2.8 at the wide end? Wouldn't the Pentax DA* 16-50 f/2.8 be even better?

Sure. But you'd also expect a constant f/2.8 zoom to be even larger and more expensive than a basic kit lens "should" be. Actually, the perfect FF-matching kit lens for APS-C would be an 18-55/2.8-4. Sigma kind of overshot the mark by making a 4X instead of 3X zoom (which would have been smaller and cheaper).

I also don't quite understand what you mean when you say that the 17-70 is larger than the typical 28-105.

Exactly that. The 17-70 might not be big compared to the 28/1.8, but it is larger than most 28-105's.

[falconeye]

glossed over the collection limit.

Dave, I considered your point in my post #93, answer (3) appending to my post #33. I wrote:

FF at ISO100 performs like APS-C at ISO50 (with a lens big enough) and ISO50 may not be available for APS-C. But D700 starts at ISO200, so equal again.

The collection limit is there but can be more easily rephrased in saying that the APS-C system must offer a lower low ISO limit (by a factor of 2) than the FF system in order not to break the equivalence.

You are right, in the extreme limit, in may turn out that say ISO 1 cannot be built anymore. But in practice, the smaller sensors have a lower low ISO limit too. Moreover, it is a concern for bright light performance, not low light performance.

So, for the sake of this thread, your consideration is of no concern.

[RBellavance]

Sigma kind of overshot the mark by making a 4X instead of 3X zoom (which would have been smaller and cheaper).

Ah, but the added range is what made me choose the Sigma over the DA18-55 and DA16-45 when I bought my K10D (almost three years ago already... time flies!). And the Sigma is not much larger than the DA16-45 anyway (although it is 100g heavier).

[GLXLR]

Sorry for being out, but sitting in a different time zone makes it sometimes difficult to keep up.

I appreciate the nice feedback, and I also appreciate the support both falconeye and especially Marc has done in bringing much more clarity to this topic than I was capable to.

I also think Marc's summary did a very good job in consolidating the discussions, so I will not try to do that myself (I guarantee, it will only start confusing the issue now when we finally have started to get a grip of it)

So I will only add an illustration to what Marc said about that it is really the physical size of the lens that give the low light performance (noting that this is in practice true only for lenses longer than "normal"). (I have not stuck to just one brand, it was really decided by which images looked good on BH Photo)

I have kept APS-C on left size, On the right size I have put the lens that gives about equal FOV. Where it say a +, that is where the FF have a benefit in low light performance, and I have put a = on the side where they are similar in low light performance. I think it gives a quite good illustration about that equal lens diameter gives similar performance when looking at the longer lenses.

It is also obvious that when getting to wider lenses, the "theory" breaks down in practice. E.g. there is no 33mm f/1.0 lens to give equal performance on APS-C as the 50mm f/1.4 on FF.

But for me, that is OK. I tend to have the most problems with low light when I use longer lenses (harder to keep steady while hand holding, shooting sports or nature shots, or as Marc, shooting indoor towards a scene performance). And in these siutations the theory seems to hold also quite well in practise.

Now I only have to convince my wife that I can buy the SMC Pentax-A* 135mm F1.8

Again thank you all for contributing.

Best regards,
Haakan

Yup. APS-C and FF basically = excluding limits of lens. Only problem is will Pentax listen? Will they make a 50-135mm f/2.0 (no need for it though) as a true 70-200mm f/2.8, a 200mm f/2 for a true 300mm f/2.8, or a TS lens? Until they do, I doubt people will take Pentax seriously.

Ohh... I really want a 30mm f/1.4 or f/1.0 (I doubt the f/1.0 )

[Haakan]

Haakan,

Excellent thread, I've enjoyed reading through it.

Might I suggest adding 85/1.4 on FF, and 58/1.0 (not available) on APS-C to your chart? The 85/1.4 never leaves my D700 for low-light street shooting, and it's a perfect focal length for head and shoulders shots.

Thanks,
Edmund

Hi Edmund,
thanks for the feedback.
You can get a 55mm f/1.2 though which is quite close. I thought of having also that combination in the chart since this seems to be kind of the transition zone in focal length for this discussion (approaching f/1.0).

One important part to remember though is that these lenses (85/1.4 and 55mm/1.2) have a very thin DOF fully open, actually to the extent that in many cases I might step it down somewhat to a smaller aperture to gain DOF. And as soon as I have stepped it down, even just from 1.4 to 1.7, I have lost the benefit with FF again (in terms of noise).

Best regards,
Hakan

[pingflood]

A lot of good discussion has been devoted to this subject, but an important physical limit has been glossed over. It is the limit imposed by the number of photons per unit area it is possible to collect in a given time.

When doing scaling mathematics one must be very careful of divide operations (to avoid the equivalent of "divide by zero" type errors.)

I did sort of stumble into that realm in post #34:
https://www.pentaxforums.com/forums/726563-post34.html

[ricardobeat]

Yup. APS-C and FF basically = excluding limits of lens. Only problem is will Pentax listen? Will they make a 50-135mm f/2.0 (no need for it though) as a true 70-200mm f/2.8, a 200mm f/2 for a true 300mm f/2.8, or a TS lens? Until they do, I doubt people will take Pentax seriously.

I'm rooting for the TS lens

[newarts]

So Falk, basically you are saying that the people clamoring for lower resolution "lower noise" sensors are chasing a mirage? There've been some very heated debates on FM over this, with folks firmly entrenched in either camp.

One thing though... if you have a noise floor in the circuitry, once you get to a certain pixel pitch won't simply the low # of photons hitting each photosite cause the signal to entirely disappear in the noise? If you disregard the photon being a singular entity and simply look at it as a distribution of light then yeah, I am with you the entire way; however since we are ultimately down to individual photons then there has to be a breaking point somewhere. Right?

E.g. if the photosites receive say 200 photons each (no clue what the real numbers are) and the circuitry noise is equivalent to 20 photons, things are fine, but increase the pixel density significantly and each photosite might get 20 or 30 photons, and at that point, how do you sort it out from the noise? Or is the noise "orderly" enough that you can effectively get past it?

Pingflood, you are right about the effect of circuitry noise, but there is a more fundamental limit having to do with light itself (hence cannot be escaped by technological improvements.) The arrival of light is always a probabilistic process; conceptually, the sensor captures a stream of discrete photons.

For an image from an 18% reflective surface to be "excellent" from a noise standpoint, each display pixel must correspond to the collection of at least 1600 photons (40^2). (See the ISO specification on film speed: Film speed - Wikipedia, the free encyclopedia.) It is true even in the absence of circuitry noise.

Simply put, this puts a lower limit on sensor size for an "excellent" image of a particular scene for display at a particular size with constant f-number and exposure time (even in the absence of "equivalence" requirements.)

Dave in Iowa