The *real* reason dSLRs are better in low-light

I didn't mean to cause such a fire storm.

Low-light performance is related primarily to light-collection area (entrance pupil diameter).

You could then, quite reasonably, ask why dSLR lenses have larger entrance pupil diamters than lenses for smaller sensors. The answer would be because it's actually cheaper to produce a longer lens (needed for a larger sensor) with a given entrance pupil diameter than it is to produce a shorter lens with the same entrance pupil diameter and optical quality.

You could then ask why that is. The answer is because of the limitations of the materials. The materials used in lenses have limited refractive index and non-zero dispersion (that causes CA). If those limitations didn't exist, we could produce arbitrarily fast lenses economically (assuming the material wasn't outrageously expensive).

As a result, larger sensors that require longer lenses for the same FOV can have lenses with larger entrance pupils for an economical price than can their smaller-sensor cousins.

Nevertheless, it comes down to the fact that the larger sensor cameras do indeed have larger light-collection areas in their lenses and therefore have better low-light performance simply because they have more light to work with.

--
Lee Jay
(see profile for equipment)

Doug, thanks for clarify this matter. I was confused among these debates. Now it's much clear that Lee Jay is right.

Everybody is discussing with their own point of view - over the same thing.

Just my 0.2 cent.

--
Thadoe Hein
http://www.pbase.com/thadoehein
(see profile for equipment lists - current & sold)

ljfinger said:

Think of it like this. How are telescopes described? By their
physical aperture. The 10 meter Keck observatory (both of them)
has a 10 meter aperture (primary mirror) not a 10 meter sensor (!).
They don't even talk about the sensor size because it is basically
irrelevant. The light collection area is all that matters (we're
not talking about resoltion here for which baseline matters and not
area, we're just talking about imaging a faint object).

Click to expand...

With this analogy, it finally dawned on me where you are coming from. If you are saying the physical aperture size of a telescope is like the physical aperture size (entrance pupil, if you will) of a camera lens and the bigger the better(physically) always, you are wrong.

The telescope is trying to collect as much light as possible to increase the brightness and resolving power always as the signal is always weak. (Same is true for a parabolic microwave dish-the bigger is always better). This is not true for the aperture size of a camera lens. You adjust the aperture size up AND down with the scene brightness to MATCH the sensitivity of the sensor to keep the histogram out of the underexposed and the blown out regions. You set the sensor sentitivity by setting the ISO which determines the noise for a given illumination. You don't always try to collect as much light as possible as you do with a telescope.

Nuff said.

--mamallama

the illumination of both lenses is THE SAME over a specified area. The ONLY reason the lens is bigger is to illuminate a larger area. you could put that lens on a compact and use a tiny sensor and it would make NO Difference at all to the exposure (although severly affect the FOV).

So it has to do with the medium's size and it's ability to detect light. when dealing with 1.6 ad FF the diffence due to size is minimul. when dealing with P&S's the difference is large.

If your theory is true the why does your 5D and my 1D (4mp) not function the same in low light. They both use the same lens

obviously I am not saying that a larger aperture makes no difference - it does, bot only in relation to the sensor it covers.

FF f1.8 is the same illumination accross a specified area as compacts are at f1.8

I agree with your statement. I already provided several times the explanation (as others did) as you can see there:
http://forums.dpreview.com/forums/read.asp?forum=1000&message=15037948

However, people seem to love the 'larger sensor is better' just as they enjoyed 'more MP is better' years ago: it's simple and they won't accept easily a change to something they understood and agreed on.

Good luck,

Olivier

If you were to cram 40 megapixels into a 1.6 crop SLR sensor, would you still have a low noise picture?

The answer is of course, a resounding NO!

The reason is, of course as everyone here has been saying, that the real attribute that contributes to a low noise image here is the size of the potosite, given the same ISO and f-stop. any other way of looking at is nonse since we are all looking for a relative noise free image on a our digicams, not our 10 metre telescopes for heavens sake.

Look at it this way, if you increase the size of the light capturing part of the lens at the front, you must be changing the f-stop as well, otherwise you will not be metering the same as everyone else for a given ISO, shutter speed and f-stop, which is planily nonsense.

I can see what you are sayign but it is obvious to me that the real technology, and therefore the changes to the level of noise we endure are with the design of the sensor, and the relative size of the photosite.

great argument though :O)

mamallama said:

The telescope is trying to collect as much light as possible to
increase the brightness and resolving power always as the signal is
always weak. (Same is true for a parabolic microwave dish-the
bigger is always better). This is not true for the aperture size of
a camera lens.

Click to expand...

I don't see why not. After all, stopping down is always an option Chosing f1 aperture on an f2 lens on the other hand can turn out to be a little difficult...

mamallama said:

You adjust the aperture size up AND down with the
scene brightness to MATCH the sensitivity of the sensor to keep the
histogram out of the underexposed and the blown out regions.

Click to expand...

Not really. IMO aperture should primarily be used to dial in preferred DOF. After all, in most cases theres 2 more settings to fix yer histgram thingy; exposure time and ISO. Current shutterspeeds go up to 1/8000 or more, so in a lot of situations theres plenty of room for the imaginary

Of course in camera/lens design too there's an everlasting quest for those ever so precious light photons.

Dear L.J.

You description is correct. However, you know very well that all these parameters (FOV, pupil diameter, size of photo-sites, etc.) are closely related to each other. What is confusing about your statement is that you have taken the --pupil diameter-- as the starting point to calculate everthing else. In my opinion, to design a lense we normally should start differently. Take this example:

For a full-frame sensor, calculate the focal lenght of a lens system that would give us a field of view of 46.8° (same as the eye).

Solution: for a full-frame sensor, we have

d = sqrt(24*24 + 36*36)

we know that for a FOV of alpha, the focal lenght can be calculated by this formula

f = d / (2 * tan (alpha 2))

as such, we'll have

f = 49.99 or the same good old 50mm lens.

So in conclusion, the starting piont was really the size of the given sensor followed by the desired FOV. Obviously, the pupil diameter is never mentioned or needed in the above calculation. IF you really need to know what is the size of the pupil, you should know what F value you're talking about. For a F = 1/4 the pupil will be 50 / 4 = 12.5mm.

So my personal conclusion is that primary reason for better low-light performance of DSLRs is the size of the sensor. A bigger sensor needs more light (for the same FOV) which according to the above formula, obviously, needs a larger lens.

Regards,
Arash

ljfinger said:

I didn't mean to cause such a fire storm.

Low-light performance is related primarily to light-collection area
(entrance pupil diameter).

You could then, quite reasonably, ask why dSLR lenses have larger
entrance pupil diamters than lenses for smaller sensors. The
answer would be because it's actually cheaper to produce a longer
lens (needed for a larger sensor) with a given entrance pupil
diameter than it is to produce a shorter lens with the same
entrance pupil diameter and optical quality.

You could then ask why that is. The answer is because of the
limitations of the materials. The materials used in lenses have
limited refractive index and non-zero dispersion (that causes CA).
If those limitations didn't exist, we could produce arbitrarily
fast lenses economically (assuming the material wasn't outrageously
expensive).

As a result, larger sensors that require longer lenses for the same
FOV can have lenses with larger entrance pupils for an economical
price than can their smaller-sensor cousins.

Nevertheless, it comes down to the fact that the larger sensor
cameras do indeed have larger light-collection areas in their
lenses and therefore have better low-light performance simply
because they have more light to work with.

--
Lee Jay
(see profile for equipment)

Click to expand...

I was thinking some more about this whole issue. It seems to be really a chicken and egg situation. Essentially, phisics tells us that pupil diameter and the sensor size have a "linear" relation ship. i.e.

p = d * C

where p is the pupil and d is the sensor diameter and C is a constant which is dependant on the F number and FOV ( C = 1 / (2 * F * tan(FOV/2)).

As such, I think the discussing if the chicken (sensor size) came first or the egg (pupil diameter) is not very productive.

Regards,
Arash

Dazaau said:

the illumination of both lenses is THE SAME over a specified area.
The ONLY reason the lens is bigger is to illuminate a larger
area. you could put that lens on a compact and use a tiny sensor
and it would make NO Difference at all to the exposure (although
severly affect the FOV).

Click to expand...

Of course it WOULD effect exposure: the largest part of that big sensor would be totally unexposed. That is the very basis of this long discussion.

Dazaau said:

So it has to do with the medium's size and it's ability to detect
light.

Click to expand...

No it has to do with the medium's size and its ability to COLLECT light. A sensor is a collector, not a detector.

Dazaau said:

when dealing with 1.6 ad FF the diffence due to size is
minimul. when dealing with P&S's the difference is large.

Click to expand...

Minimal? apsc sensors are less than half the size of FF sensors. Further more, a lens' F-value being the same on a 1.6 and and FF is only because on the 1.6 over 50% of the incoming light photons are wasted, illuminating nothing but the sides of the mirror box. But the area the lens illiminates is identical, over half is just not used. However, if it would be possible to redirect all incoming light to the 1.6x crop sensor, that lens (not the sensor!) would in fact be more sensitive on a 1.6 crop camera than on a FF.

Its really simple, a big projection area needs more light than a small one. Ever seen the light source used in movie theater projectors?

cantspellfortoffee said:

The reason is, of course as everyone here has been saying, that the
real attribute that contributes to a low noise image here is the
size of the potosite, given the same ISO and f-stop. any other way
of looking at is nonse since we are all looking for a relative
noise free image on a our digicams, not our 10 metre telescopes for
heavens sake.

Click to expand...

I do not agree with your "same ISO and f-stop". From a photographic point of view, what matters is what you get in the image (Field Of View, Depth Of Field, noise, exposure duration...), and not the ISO value set in your camera.

Check also this:
http://forums.dpreview.com/forums/read.asp?forum=1000&message=17724607
(and previous discussion)

Olivier

Arash Salarian said:

As such, I think the discussing if the chicken (sensor size) came
first or the egg (pupil diameter) is not very productive.

Click to expand...

The thing is that the pupil diameter - for a given Field of View - will determine the Signal/Noise ratio whatever the sensor size... (whereas the sensor size alone will provide NO indication).

Olivier

Eh, I forgot to add that (1) the only reason exposure is the same is becasue using a large slens on a little sensor so much light is wasted

And (2) in order for your theory of 'expure being the same' to be true, reversing the situation (eg put a small lens before a large sensor) should also be true, which it isnt

If that lens was unaltered it would effect exposure as the largest part of that big sensor would be totally unexposed. So total area exposure would be dramatically less

If that little lens WAS altered (to provide a larger circle image), it would effect exposure too, because that larger image circle would be much dimmer

Olivier GALLEN said:

Arash Salarian said:

As such, I think the discussing if the chicken (sensor size) came
first or the egg (pupil diameter) is not very productive.

Click to expand...

The thing is that the pupil diameter - for a given Field of View -
will determine the Signal/Noise ratio whatever the sensor size...
(whereas the sensor size alone will provide NO indication).

Olivier

Click to expand...

As I said before, pupil diameter and sensor size are linearly related. As such, the following two statements are equaly valid:

1- The thing is that the pupil diameter - for a given Field of View -
will determine the Signal/Noise ratio whatever the sensor size...

2 - The thing is that the sensor size - for a given Field of View - will determine the Signal/Noise ratio whatever the pupil diameter...

Regards,
Arash

Arash Salarian said:

For a full-frame sensor, calculate the focal lenght of a lens
system that would give us a field of view of 46.8° (same as the
eye).

Click to expand...

Where did you get that?

http://forums.dpreview.com/forums/read.asp?forum=1018&message=17075832

--
Lee Jay
(see profile for equipment)

cantspellfortoffee said:

If you were to cram 40 megapixels into a 1.6 crop SLR sensor, would
you still have a low noise picture?

Click to expand...

Yes you would. It's called "reduced enlargement" (a lousy term - too much like "jumbo shrimp" or "military intelligence").

cantspellfortoffee said:

The answer is of course, a resounding NO!

Click to expand...

Not so. Tell me, why is an ISO 100 piece of 4x5 film much lower in "noise" (grain) than an ISO 100 piece of 35mm film?

cantspellfortoffee said:

Look at it this way, if you increase the size of the light
capturing part of the lens at the front, you must be changing the
f-stop as well, otherwise you will not be metering the same as
everyone else for a given ISO, shutter speed and f-stop, which is
planily nonsense.

Click to expand...

Not so. I can have a 6mm f2.7 lens with an entrance pupil of 2.22mm for the S3 and a 38mm f2.7 lens with an entrance pupil of 14.1mm for the 5D. Larger pupil diameter, same f-stop.

cantspellfortoffee said:

I can see what you are sayign but it is obvious to me that the real
technology, and therefore the changes to the level of noise we
endure are with the design of the sensor, and the relative size of
the photosite.

great argument though :O)

Click to expand...

Light collection area is what matters. What you do with that light, be it spread it thinly across a large sensor or concentrate it intensely on a small one, matters little (not at all except for 2nd-order effects).

--
Lee Jay
(see profile for equipment)

Olivier GALLEN said:

I agree with your statement. I already provided several times the
explanation (as others did) as you can see there:
http://forums.dpreview.com/forums/read.asp?forum=1000&message=15037948
However, people seem to love the 'larger sensor is better' just as
they enjoyed 'more MP is better' years ago: it's simple and they
won't accept easily a change to something they understood and
agreed on.

Click to expand...

I just didn't expect that such an obvious argument would meet with any resistance at all.

--
Lee Jay
(see profile for equipment)

mamallama said:

With this analogy, it finally dawned on me where you are coming
from. If you are saying the physical aperture size of a telescope
is like the physical aperture size (entrance pupil, if you will) of
a camera lens and the bigger the better(physically) always, you are
wrong.

Click to expand...

I'm right under the conditions I stated ("low light" as stated in the subject of the thread).

mamallama said:

The telescope is trying to collect as much light as possible to
increase the brightness and resolving power always as the signal is
always weak. (Same is true for a parabolic microwave dish-the
bigger is always better). This is not true for the aperture size of
a camera lens. You adjust the aperture size up AND down with the
scene brightness to MATCH the sensitivity of the sensor to keep the
histogram out of the underexposed and the blown out regions. You
set the sensor sentitivity by setting the ISO which determines the
noise for a given illumination. You don't always try to collect as
much light as possible as you do with a telescope.

Click to expand...

Of course. But if low-light performance is the metric, as I stated in the subject of the thread, then more light is always better since it wouldn't be low-light if you had to stop down to prevent blow-outs at your lowest ISO.

--
Lee Jay
(see profile for equipment)

LJ, please, as you very well know the discussion was not about the true FOV of human eye. The discussion was about how pupil size and sensor size were related to each other for given FOV. Now, if you don't like my example, take any other example you wish.

And yes, I agree, probably we can't give a set, standard value for the FOV of the eye.

Regards,
Arash

ljfinger said:

Arash Salarian said:

For a full-frame sensor, calculate the focal lenght of a lens
system that would give us a field of view of 46.8° (same as the
eye).

Click to expand...

Where did you get that?

http://forums.dpreview.com/forums/read.asp?forum=1018&message=17075832

--
Lee Jay
(see profile for equipment)

Click to expand...