Sigma Announces 18-35 f/1.8 DC HSM Art for APS-C

[Wildfire]

Right, the FF lens "gathers" twice as much light, and then spreads the "gathered" light over a much larger sensor. The crop lens "gathers" less light, but then focuses the light onto a small sensor, creating a brighter exposure than the full frame lens despite having "gathered" less actual light.

 

[TrumpetPower!]

Right, the FF lens "gathers" twice as much light, and then spreads the "gathered" light over a much larger sensor. The crop lens "gathers" less light, but then focuses the light onto a small sensor, creating a brighter exposure than the full frame lens despite having "gathered" less actual light.

...but only when shot at an f/number sufficiently smaller such that the actual physical aperture is the same.

In other words, the APS-C camera shot with a 50mm lens at f/1.8 gathers the same amount of light as a 135 camera with an 80mm lens at f/2.8, and framing and depth of field and the rest are also comparable.

50 / 1.8 = 27.8
80 / 2.8 = 28.6

Cheers,

b&

 

[Kit.]

I've decided to offer some help to Sigma with an FAQ set for potentially confused customers.

Nope, comparing shutter speeds of 70/2.8 to shutter speeds of 35/1.8 is not helping at all.

You seem to think that all that matters is the amount of light falling on a square micrometer of the sensor.
While it doesn't. We don't shoot sensors.
What really matters is the amount of light that is gathered from a square millimeter of the subject.

 

[AdamJ]

Nope, comparing shutter speeds of 70/2.8 to shutter speeds of 35/1.8 is not helping at all.

I didn't compare it specifically with 70mm, I compared it with a 24-70mm, because that was the comparison others made that started the debate.

You seem to think that all that matters is the amount of light falling on a square micrometer of the sensor.

If you read the thread, you'll understand that my objective was to dispel the notion that f/1.8 on APS-C is in effect the same as f/2.8 on full frame. That notion is actually an unhelpful and circuitous way of saying that APS-C sensors are, because of their typically smaller pixels, usually noisier than full-frame sensors - simple as that.

 

[Mantanuska]

[quote author=AdamJ]

If you read the thread, you'll understand that my objective was to dispel the notion that f/1.8 on APS-C is in effect the same as f/2.8 on full frame. That notion is actually an unhelpful and circuitous way of saying that APS-C sensors are, because of their typically smaller pixels, usually noisier than full-frame sensors - simple as that.
[/quote]


^^ THIS

And that FAQ table. All that needs to be said.

 

[Kit.]

Nope, comparing shutter speeds of 70/2.8 to shutter speeds of 35/1.8 is not helping at all.

I didn't compare it specifically with 70mm, I compared it with a 24-70mm, because that was the comparison others made that started the debate.

While you should be comparing pictures/scenes shot with both and then presented on comparable mediums (screen or paper).

Otherwise the comparison is meaningless.

You seem to think that all that matters is the amount of light falling on a square micrometer of the sensor.

If you read the thread, you'll understand that my objective was to dispel the notion that f/1.8 on APS-C is in effect the same as f/2.8 on full frame. That notion is actually an unhelpful

Just the opposite. The end result is about the same, because it is mostly limited by purely optical considerations. No matter what technical solutions are employed between the subject and the final image on paper or on monitor.

and circuitous way of saying that APS-C sensors are, because of their typically smaller pixels, usually noisier than full-frame sensors - simple as that.

Nope. They aren't noisier "because of their typically smaller pixels". Given the same size of the lens' entrance pupil (i.e. the same amount of light captured by the lens from an area on the subject), the same angle of view and enough of pixel well depth, they all will provide about the same amount of noise on the final medium.

They are noisier because the entrance pupils of their smaller lenses are generally smaller than on the lenses that provide the same angle of view for larger sensor formats. There are physical limitations in making large entrance pupils for smaller focal lengths, as pupils bigger than f/0.5 are impossible even theoretically.

 

[Mantanuska]

They are noisier because the entrance pupils of their smaller lenses are generally smaller than on the lenses that provide the same angle of view for larger sensor formats. There are physical limitations in making large entrance pupils for smaller focal lengths, as pupils bigger than f/0.5 are impossible even theoretically.

You still get more light from a smaller focal length (at the same pupil size) . Thats why we describe the amount of light coming through the lens as a function of the focal length and pupil size, or in other words we use F-stop notation.

a 10mm aperture on an 18mm lens that gives a certain fov on APS-C lets in more light than a 10mm aperture on a 28mm lens that would produce the same fov on FF.

Thats why zoom lenses that aren't constant aperture get darker when you zoom in. The pupil isnt getting smaller, you just have less light entering the lens because you are constricting the FOV.

An F-stop is and F-stop regardless of format or focal length

 

[wickidwombat]

personally i see the f1.8 light gathering ability and effective wider depth of field as a benefit not a negative
the only issue is having a good high iso APS-C sensor to pair it with currently there isnt one in canons line

 

[9VIII]

Nope, comparing shutter speeds of 70/2.8 to shutter speeds of 35/1.8 is not helping at all.

I didn't compare it specifically with 70mm, I compared it with a 24-70mm, because that was the comparison others made that started the debate.

You seem to think that all that matters is the amount of light falling on a square micrometer of the sensor.

If you read the thread, you'll understand that my objective was to dispel the notion that f/1.8 on APS-C is in effect the same as f/2.8 on full frame. That notion is actually an unhelpful and circuitous way of saying that APS-C sensors are, because of their typically smaller pixels, usually noisier than full-frame sensors - simple as that.

Unfortunately I think that may actually still hold true.

http://www.geofflawrence.com/inverse_square_law.html

To get equivalence you need to account for the smaller surface area and the extra distance from the subject. Using a shorter focal length to make up for the tighter FOV effectively puts you farther away from the subject.


If someone would actually make a lens that gives the same FOV at the same focal length instead of throwing all the extra light away in FOV reduction, then we'd have an amazing short focal length lens. Right now wide angle on crop cameras just seems like a waste.

I do agree that everything would be better off with smaller pixels though, and that people should stop equating MP count with noise.

 

[Kit.]

They are noisier because the entrance pupils of their smaller lenses are generally smaller than on the lenses that provide the same angle of view for larger sensor formats. There are physical limitations in making large entrance pupils for smaller focal lengths, as pupils bigger than f/0.5 are impossible even theoretically.

You still get more light from a smaller focal length (at the same pupil size) . That's why we describe the amount of light coming through the lens as a function of the focal length and pupil size, or in other words we use F-stop notation.

a 10mm aperture on an 18mm lens that gives a certain fov on APS-C lets in more light than a 10mm aperture on a 28mm lens that would produce the same fov on FF.

You don't get get more light from the subject at the same entrance pupil size. What you get is higher illuminated sensor, but you get it only because the same amount of light is concentrated on smaller sensor area. However, the shot noise on the image of an area of the subject is exactly the same, because the amount of photons originated from this area and reaching the sensor is exactly the same.

The amount of light the lens gathers through the same entrance pupil from the same solid angle of the scene is exactly the same no matter if the lens is going to focus this light on a 5 sq. micron pixel or on a 50 sq. micron pixel. As long as it's the same pixel on the output device, it will carry the same amount of shot noise.

Think about it.

Thats why zoom lenses that aren't constant aperture get darker when you zoom in. The pupil isnt getting smaller, you just have less light entering the lens because you are constricting the FOV.

But they are shooting different subjects!

There is even a more "ideal" object for comparison than zoom lenses. It's teleconverters. They increase the focal length keeping the entrance pupil constant. If you shot the same sporting event (i.e. the same shutter speed) with the wide open lens with and without a teleconverter ("push" the darker lens shots with higher sensor ISO) you will find the noise less apparent on a full frame of the shot without the teleconverter. But that's because you are shooting different subjects. Crop your non-TC shot to cover the same subject as your TC shot, convert both images to the same pixel count and see if the difference in the noise still exists.

 

[AdamJ]

Nope, comparing shutter speeds of 70/2.8 to shutter speeds of 35/1.8 is not helping at all.

I didn't compare it specifically with 70mm, I compared it with a 24-70mm, because that was the comparison others made that started the debate.

While you should be comparing pictures/scenes shot with both and then presented on comparable mediums (screen or paper).

Otherwise the comparison is meaningless.

Which of those FAQs are you challenging?

You seem to think that all that matters is the amount of light falling on a square micrometer of the sensor.

If you read the thread, you'll understand that my objective was to dispel the notion that f/1.8 on APS-C is in effect the same as f/2.8 on full frame. That notion is actually an unhelpful

Just the opposite. The end result is about the same, because it is mostly limited by purely optical considerations. No matter what technical solutions are employed between the subject and the final image on paper or on monitor.

This doesn't advance the debate at hand.

and circuitous way of saying that APS-C sensors are, because of their typically smaller pixels, usually noisier than full-frame sensors - simple as that.

Nope. They aren't noisier "because of their typically smaller pixels". Given the same size of the lens' entrance pupil (i.e. the same amount of light captured by the lens from an area on the subject), the same angle of view and enough of pixel well depth, they all will provide about the same amount of noise on the final medium.

They are noisier because the entrance pupils of their smaller lenses are generally smaller than on the lenses that provide the same angle of view for larger sensor formats. There are physical limitations in making large entrance pupils for smaller focal lengths, as pupils bigger than f/0.5 are impossible even theoretically.

From my personal experience of 5D II vs 7D, that simply isn't the case.

 

[Kit.]

Nope, comparing shutter speeds of 70/2.8 to shutter speeds of 35/1.8 is not helping at all.

I didn't compare it specifically with 70mm, I compared it with a 24-70mm, because that was the comparison others made that started the debate.

While you should be comparing pictures/scenes shot with both and then presented on comparable mediums (screen or paper).

Otherwise the comparison is meaningless.

Which of those FAQs are you challenging?

Everything that relates to shutter speed.

Because it's not like people do (or at least should) pick their shutter speeds based on some fancy digits printed on their lenses. People should pick their shutter speeds based on the subject movements they need to freeze. If you cannot afford some part of the subject to move more than 5mm (in the subject plane) during the time of exposure, you will need to estimate the speed of this part (in mm/s) and divide it by 5 - then you will get (in 1/s) the lowest shutter speed you could afford to use in your shot, and then you will need to live with whatever amount of shot noise you are getting from the object get at this speed.

Now let's look how much noise we are actually getting. On our final image (printed or displayed on the monitor) we choose the smallest scale we need to resolve. It's historically - from the film era - called "circle of confusion" when reduced to length units in sensor plane; by at the moment we don't need it in length units, we need the solid angle that comprises this circle, the solid angle of the feature we need to resolve, the solid angle we need to get our photons from. And then the number of photons caught by this circle is the number of photons coming to our lens from that solid angle during the time of exposure - and then passing through the entrance pupil of our lens.

Then the absolute shot noise will be the square root of that number.
And the relative shot noise will be (absolute shot noise)-1.

See? No fancy digits printed on the lens. No sensor size. Even no sensor's pixel size. Only the subject (under the lighting), the desired end result and the absolute effective lens opening (in square millimeters or similar area units). Once you get this fixed, you know what noise you will get for your shutter speed, and you know what shutter speed corresponds to your noise.

Of course, you may want to boost your sensor's ISO to reduce the in-camera noise (and the right setting of ISO usually makes in-camera noise irrelevant when you need higher shutter speeds). But that's what you can afford to control. What you cannot afford to control is the amount of photons coming from your subject.

Actually, the same considerations also hold true for background blur, DoF and - in some sense - bokeh. All what you would ideally see on the final picture was already determined when the light coming from the subject crossed the (virtual) entrance pupil (the rendering of diaphragm visible through the front lens), then anything lying after the entrance pupil only added its distortions to it.

When you want to apply the same to bokeh, you will need to take into consideration the non-uniformity (and dependence on the distance) of the entrance pupil's transparency for out-of-focus objects. It's uniform for lenses with well-corrected spherical aberrations, darker in center (and lighter at the blades) for nearer objects in under-corrected lenses and for farther objects in over-corrected lenses, and lighter in center (and darker at the blades) otherwise.

For example, an unobscured bright dot at infinity will be rendered on the final picture exactly like the entrance pupil moved (without change of its physical size) into the subject's plane of focus in the direction of that dot. It will be flatly lighted for lenses with well-corrected spherical aberrations, nicely round for lenses with under-corrected spherical aberrations and doughnut-shaped for lenses with over-corrected spherical aberrations. Other than that, nothing should happen behind the entrance pupil in a good lens (of course, there could be ghosts and stuff if the lens is not so good). So, if your shoot with EF 50/1.4 wide open, a bright dot on your (far enough) background will be rendered as a flat octagon of ~36 mm (in subject's plane of focus) diameter, no matter what subject you shoot, no matter what your magnification is (well, if it's true macro, there could be exceptions), no matter what sensor size your camera has.

You seem to think that all that matters is the amount of light falling on a square micrometer of the sensor.

If you read the thread, you'll understand that my objective was to dispel the notion that f/1.8 on APS-C is in effect the same as f/2.8 on full frame. That notion is actually an unhelpful

Just the opposite. The end result is about the same, because it is mostly limited by purely optical considerations. No matter what technical solutions are employed between the subject and the final image on paper or on monitor.

This doesn't advance the debate at hand.

Well, I could not just go and say that most guides that teach photography by numbers do it in the wrong way (not that their numbers are wrong, but they just focus your attention on secondary details, letting you fail to see what really happens with the picture). I needed some pretext for that.

Those guides pay too much attention to what happens in the camera, although it's what happens outside the camera is what forms your images. The camera just adds its own technical limitations. These technical limitations were big enough during the "35mm film rolls" era, big enough to limit your thinking about your subject and about your end result. Those limitations weren't as big for sheet film, those limitations aren't as big for current large photo sensors, you can focus now on what happens outside the camera, paying only secondary attention on what happens inside.

They aren't noisier "because of their typically smaller pixels". Given the same size of the lens' entrance pupil (i.e. the same amount of light captured by the lens from an area on the subject), the same angle of view and enough of pixel well depth, they all will provide about the same amount of noise on the final medium.

They are noisier because the entrance pupils of their smaller lenses are generally smaller than on the lenses that provide the same angle of view for larger sensor formats. There are physical limitations in making large entrance pupils for smaller focal lengths, as pupils bigger than f/0.5 are impossible even theoretically.

From my personal experience of 5D II vs 7D, that simply isn't the case.

Maybe you did it wrong then.

 

[AdamJ]

For willful obfuscation, you take the biscuit (and the competition has been intense). ;D

 

[indigo9]

For willful obfuscation, you take the biscuit (and the competition has been intense). ;D

Whoah this is still going? A helpful reminder about what rs actually said initially:

..In other words ISO 10,000 on crop is pretty much equal to ISO 25,600 on FF in terms of noise. So feel free to shoot smaller apertures on FF and use higher ISO's to get the same light gathering and noise.
...
I see this Sigma lens as being a crop alternative to the 24-70 II on FF in just the same way as the 17-55 IS is a crop alternative to the 24-105L on F

Fixed the FAQ for you:

The fact is, f/1.8 is f/1.8 whether mounted on FF or crop.

I think everyone agrees that you would use the same ISO on both cameras.

I think everyone agrees that if you were physically cutting up film then crop f/1.8 would be the same as full frame f/1.8.


In the real world though, digital crop sensors are not simply full frame sensors with the edges chopped off. Manufacturers keep the same photo resolution -- ie shrink the photosites -- at the expense of high ISO noise. For an identical production process and same resolution ISO 3200 FF is not the same as ISO 3200 crop: for cameras made within a few years of each other, crop will always be noisier.


So in low light situations, how would you get the same quality images on crop as full frame? First you'd lower the ISO. But then it would be underexposed -- so you'd either lower shutter speed (something that every camera can do and has nothing to do with the lens) or ... open the aperture, just like rs said at the very start.

f/1.8 is f/1.8 only if you think that image quality doesn't matter.





I approve of Radiating's summary:

So while for a given area of a light sensetive material, ISO provides a set level of amplification, cameras do not have a consistent area that they absorb incoming light on

...

Simply put APS-C cameras have more dense sensors with more pixels, even with identical resolution. Because ISO is dependent on a given volume of light passing through a given area, if we increase the number of pixels in that area each pixel will have a stronger amplification level.

Let's say Canon releases the new 70D with a next-generation crop sensor... Now let's say this new sensor has noise performance as good as the 5D Mark II. (A stretch, but I think it's possible)

A crop camera with better ISO performance than the same manufacturer's previous full frame? Show where this has ever happened in history.




It seems that AdamJ and Wildfire are semantically correct, but are deliberately ignoring what Kit, rs and I are all trying to say: you can't look at a lens in isolation.

Parallel:
50mm lens X has a resolution of 500 lp/mm, but the only camera you can put it on is a 640x480 VGA webcam.
50mm lens Y has a resolution of 50 lp/mm but will mount on a 5d3.

Which is a better lens?

Sure, you're semantically and theoretically correct to say the f/1.8 = f/1.8 but if you care about actual image quality then you can't ignore the sensor.

 

[thomeos]

Hey folks, first post and hopefully something usefull..

FYI:

A british shop just listed the lens for pre-order with a pricetag of GBP 999,- for both Nikon and Canon mount.

Don't let the URL fool you:
http://www.ukdigital.co.uk/sigma-30mm-f1-4-ex-dc-hsm-lens-canon-2966.html
http://www.ukdigital.co.uk/sigma-30mm-f1-4-ex-dc-hsm-lens-canon-2967.html

It seems my anticipated MSPR of 999 [regional currency] seems to become reality!

Cheers
Thomas

 

[Wildfire]

Let's say Canon releases the new 70D with a next-generation crop sensor... Now let's say this new sensor has noise performance as good as the 5D Mark II. (A stretch, but I think it's possible)

A crop camera with better ISO performance than the same manufacturer's previous full frame? Show where this has ever happened in history.

You're right, it's never happened in history. But maybe it will soon. (Yes, that's a big maybe!).

I'd guess that the noise performance of the new Rebel SL1 can't be too much worse than the original 5D's. I think the 7D Mark II has a good chance of having excellent noise performance. It will never beat the latest full frame bodies, but maybe it can give the previous generation of FF cameras a run for their money.

And what about using Sigma's f/1.8 zoom on a full frame camera? Sure, it's not designed for FF, but from the Korean site with the samples it appears to mount and work properly on a 5D Mark II (though with heavy vignetting at all FLs below 28mm). If you're willing to lose resolution by cropping out the vignetting, then you actually can achieve FF-level noise performance with this lens, and still shoot at an f/1.8 exposure. (AND get the full f/1.8 depth of field as well!)

I'm not trying to say that crop will be better than FF -- it never will be. I just want to clarify for the people who may not know or understand the facts: at the same ISO/ shutter speed, an f/1.8 crop exposure is always brighter than an f/2.8 FF exposure.

If crop performance can approach FF performance at only a fraction of the cost, then that's good news for photographers everywhere. That's why this new lens is a big deal and that's why I want to get one. (I don't even own any crop cameras right now!)

 

[Pi]

I think everyone agrees that you would use the same ISO on both cameras.

I do not.

f/1.8 is f/1.8 only if you think that image quality doesn't matter.

What is "f" in either case?

 

[indigo9]

Let's say Canon releases the new 70D with a next-generation crop sensor... Now let's say this new sensor has noise performance as good as the 5D Mark II. (A stretch, but I think it's possible)

A crop camera with better ISO performance than the same manufacturer's previous full frame? Show where this has ever happened in history.

You're right, it's never happened in history. But maybe it will soon. (Yes, that's a big maybe!).

I'd guess that the noise performance of the new Rebel SL1 can't be too much worse than the original 5D's.

Based on the fact that the 5D was slightly ahead of the 7D (which if I'm not mistaken is the same sensor as SL1 but different firmware and/or QC?) I would think you're right -- dxo says the 5D is slightly ahead but I've tried shooting the same shot with both in low light and it's pretty hard to pick which is better (once you can actually get the 5D to focus!).

I'm not trying to say that crop will be better than FF -- it never will be. I just want to clarify for the people who may not know or understand the facts: at the same ISO/ shutter speed, an f/1.8 crop exposure is always brighter than an f/2.8 FF exposure.

In fairness, I've assumed that if you were wanting wide apertures it was usually for low light shooting -- if you're shooting low ISO and just want short depth of field then it really doesn't matter (both approach the human SNR perception limit)

I think everyone agrees that you would use the same ISO on both cameras.

I do not.

I'm quite happy to be shown wrong, but was extrapolating from what would happen with physical film, where physically cropping film does nothing to ISO (in which case ISO would indeed be the same on both full frame and crop).

Unless I've misread https://www.google.com.au/search?q=ISSCC_ISO_BAER.pdf then for digital cameras ISO speed is still based on sensor response to a given intensity of light, but simply uses a different measurement protocol.

 

[Pi]

I'm quite happy to be shown wrong, but was extrapolating from what would happen with physical film, where physically cropping film does nothing to ISO (in which case ISO would indeed be the same on both full frame and crop).

Cropping decreases the total light and increases the noise (because you enlarge).

The equivalent settings on FF requires the ISO to be multiplied by 2.56 (with the same QE anyway).

 

[indigo9]

I'm quite happy to be shown wrong, but was extrapolating from what would happen with physical film, where physically cropping film does nothing to ISO (in which case ISO would indeed be the same on both full frame and crop).

Cropping decreases the total light and increases the noise (because you enlarge).

The equivalent settings on FF requires the ISO to be multiplied by 2.56 (with the same QE anyway).

I think we're talking about something different.


What I was talking about is a scenario i just tested:

Full frame camera (5d)
- looking at a uniform white wall (possibly a cloudless sky is easier to make uniform?)
- 50mm EF lens
- f/4.0
- 1/30sec
- ISO for correct exposure is 1600

Same scene on an APSC camera (60d)
- looking at a uniform white wall
- 50mm EF lens [yes, it's cropped and not the same image -- that's why I specified a uniform wall]
- f/4.0
- 1/30sec
- ISO for correct exposure is still 1600. Histograms are virtually identical centered slightly above 50%. ISO 4000 is very much overexposed.

In this situation, a crop camera is the same as physically cropping film and the ISO is unchanged.

The fact that the total light reaching the crop sensor is less doesn't change the camera's ISO setting, the fact that less total light reaches the crop sensor changes the amplifiers to meet the ISO sensitivity definition.

Edit: changed the scenario settings to be real settings rather than hypothetical.

More amplification in the crop means more noise; when you say "equivalent settings" I assume you actually mean "equivalent settings to reach the same SNR"?