What a larger sensor is NOT about...

[Roland Karlsson]

Do they? I've seen prints from an 828, I've never seen anything
from an R1.

Silly. Substitute R1 with any APS sized camera - i.e. any DSLR.

Roland

 

[lingsb]

No-one. The point of using 'effective ISO' is for comparing
different camera systems, with different sensor sizes--not for
calculating exposure. People compare focal lengths at 35mm
equivalents. Why not ISO ratings and f-numbers?

Because those terms are already in use for something totally
different. Why confuse people?

Focal length is a physical quality of a lens system, yet it can be
multiplied by a crop factor to compare the field of view with
different sensor sizes. What is being suggested that the same is
done for f-number and ISO. (To compare equivalent light-gathering,
DoF, etc)

Focal length is focal length - often meassured in mm. It is
sometimes used as a comparision to 35 mm FF. This is not all that
useful IMHO. The 4:3 ratio (often used for digital cameras) is not
comparable to the 2:3 ratio used for 35 mm FF. Depending on how you
crop you get different results. I surely hope that this practice
will just die out.

How would you suggest comparing cameras with different sensor sizes? Bare in mind that the quantities of interest include field of view, depth of field and sensitivity.

 

[Joe0Bloggs]

For comparison between different camera formats, I use the term 'ISO multiplier'--this can be derived from theoretical calculations i.e. ISO ratio is equal to area ratio of the two sensors, or from observation, by looking at different ISO images from two cameras and finding the point at which the noise is comparable.

Then there is the f-number multiplier, which is the multiplier you can use on the small sensor camera's lens to obtain the f-number required to obtain an exposure with the same shutter speed and using the ISO multiplier. E.g. if the ISO multiplier is 4, the f-number multiplier is 2, because multiplying ISO by 4 means increasing sensitivity by two stops, which means using a 2 stops slower lens, which means multiplying f-number by (sqrt(2))^2 = 2.

 

[Joe0Bloggs]

I don't know how they defined 'effective ISO' (the post was too long and I already understood the concept involved) but unless you use a standardized size to define the usual ISO numbers things will get very confusing very quickly. I think my system, which uses ratios exclusively for comparing two cameras, is easier to understand

http://forums.dpreview.com/forums/read.asp?forum=1000&message=15001315

 

[Roland Karlsson]

How would you suggest comparing cameras with different sensor
sizes? Bare in mind that the quantities of interest include field
of view, depth of field and sensitivity.

This problem is not new.
There are several formats for film based cameras also.
Here are some useful meassures.

Sensor size: 10x15 mm
Focal length: 11 - 33 mm
Horizontal FOV: x - y degrees
ISO: 100 - 800
Aperture: 2.8 (4.5 at max focal length) - 16
Exposure time 1/8000 - 24 sec
Max dynamic range: 12 stops
RAW bits per pixel: 16
Sensor type: Bayer CFA

The only one missing is FOV. Add that - and you are done.

Roland

 

[lingsb]

How would you suggest comparing cameras with different sensor
sizes? Bare in mind that the quantities of interest include field
of view, depth of field and sensitivity.

This problem is not new.
There are several formats for film based cameras also.
Here are some useful meassures.

Sensor size: 10x15 mm
Focal length: 11 - 33 mm
Horizontal FOV: x - y degrees
ISO: 100 - 800
Aperture: 2.8 (4.5 at max focal length) - 16
Exposure time 1/8000 - 24 sec
Max dynamic range: 12 stops
RAW bits per pixel: 16
Sensor type: Bayer CFA

The only one missing is FOV. Add that - and you are done.

So, how do I know what the likely noise is going to be at a given ISO (assuming similar sensor technology)? What about depth of field?

 

[Michael_KD]

• A larger sensor is NOT about less noise in your picture

To be correct the noise depends on pixel size and not of sensor size.
http://www.dpreview.com/learn/?/Glossary/Digital_Imaging/Noise_01.htm

Regards

 

[Roland Karlsson]

So, how do I know what the likely noise is going to be at a given
ISO (assuming similar sensor technology)? What about depth of
field?

And how about the weather in Ulan Bator?

Seriously -

You could divide the maximum ISO with the square of the largest aperture, i.e. 3200 / (2.8*2.8) = 408. You could call that sensitivity index. This is a questionable number, but it is the best you can get.

Ragarding DOF. Most people don't really care to get a number. Why? Because a number is totally unintiutive. You learn your camera and what DOF it gives - thats all. Moreover - different cameras with different sized sensors are so totally different regarding DOF - so it is just not possible to do any comparison - withou having at least a set of numbers. One example - small sensor cameras can't get a shallow DOF. So - how shall you compare to another camera that both can get shallow and large DOF?

Roland

 

[lingsb]

So, how do I know what the likely noise is going to be at a given
ISO (assuming similar sensor technology)? What about depth of
field?

And how about the weather in Ulan Bator?

Seriously -

You could divide the maximum ISO with the square of the largest
aperture, i.e. 3200 / (2.8*2.8) = 408. You could call that
sensitivity index. This is a questionable number, but it is the
best you can get.

Ragarding DOF. Most people don't really care to get a number. Why?
Because a number is totally unintiutive. You learn your camera and
what DOF it gives - thats all. Moreover - different cameras with
different sized sensors are so totally different regarding DOF - so
it is just not possible to do any comparison - withou having at
least a set of numbers. One example - small sensor cameras can't
get a shallow DOF. So - how shall you compare to another camera
that both can get shallow and large DOF?

My point is that by converting ISO and f-number to '35mm equivalent' quantities, you can get a good idea of both of these. A lot of people are very familiar with the characteristics of 35mm system, including what sort of DoF they would expect with a certain focal length at a certain f-stop. Say you have an 85mm f/1.8 lens on a DX camera. Are you more likely to have a feel of the DoF and FoV if I said it would be like a 135mm lens at f/2.8 on 35mm film?

 

[Ronald S]

ISO does NOT exist (I have not read the whole thread or the threads referred at, so sorry if I repeat some things).

CCD buckets can be filled with electrons which stem from converting photons. How well this conversion is done (i.e. the sensitivity / quantum-efficiency) cannot be controlled externaly. It is an intrinsic part of the the CCD design/fabrication itself. You can therefore produce different CCDs with different sensitivities but once your chip exist the number of electrons you create will always be the same (within statistical limits of course) for the same amount of light.

The only thing you can do is count the electrons as precise as possible, i.e. 10-12-14 bit A/D AND give the A/D convertor ample time to measure as correctly as possible. So from a RAW file point of view, ISO does not exist, and is totaly sensless.

I would much more prefer to have figures like

• quantum efficiency,
• full well capacity,
• dark and other noise,
• A/D conversion and accuracy,
• etc ... so I will be able to compare two sensors.

I would also like more exact figures about the geometry of the pixels (corrected for the microlensses used) so we can calculate the

• FOV (in rad.) between two pixels and
• FOV (in rad.) for the whole sensor, giving us a better way to compare systems as a whole (lens+sensor), also the
• equivalent lens diameter is of importance together with the
• real focal length (no 35mm equivalent stuff etc...).

As the front lens diameter does not tell the whole story (for short focal lenghts, internal masking), this is why the equivalent diameter is usefull.

Somewhere photograpy has stepped in and tried to make things easier to understand (marketing), but at the expense of creating uncomparable things. Which is maybe the intention ;-)

Ronald

 

[Roland Karlsson]

ISO does NOT exist (I have not read the whole thread or the threads
referred at, so sorry if I repeat some things).

Oh yes - ISO exists Its a standardisation body.

Seriously - you are able to choose ISO settings on your camera. This setting is used (at least if you use an external light meeter) just as it would be used for film based cameras. It is what you got when you shall determine the exposure.

You are perfectly right that this is a great simplification. You are also right that there are much better meassures if you want an exact characterisation of the photon collecting system. But the simplification is useful and not something evil marketing people have invented just to annoy and you It is actuall quite old - long before modern marketing.

But - you are right that it in some cases is slightly confusing. One of the things that makes it confusing is matrix meetering. Matrix meetering knows more about your camera than it reveals. Therefore - if you use matrix meetering - you might be surprised when you look at the histogram. Why is the histogram not centered? Why is it situated in the lower parts? And why is it in higher parts in other situations?

The answer is that the camera knows about the dynamic range and tries to optimize the picture. If it is darker - it puts the picture lower to avoid large apertures and long exposure times. And if it is brighter it puts the picture higher to optimise noise. So - even if you choose ISO 200 - the camera will do its own creative interpretation of ISO 200.

So - you are right. But - most don't care

Roland

 

[Michael_J]

Most likely the 1DMKII n - because I already have the 1D and a bunch of Canon lens.

Lots of good info on this thread - most of it over my head and useless to me, but very informative.

 

[Shaun Williamson]

Well thats an interesting observation and something I have wondered
about.
The only way to improve small photodiode performance is to deliver more
light to the photodiode during exposure. However if they can't register
extra light beyond a certain point then that may not be possible.
But at present most small digicams have lenses that start at F2.8. If it is
possible to make high quality lenses that start at F2.0 then even this would
offer a worthwhile increase in noise performance especially when combined
with a sensor like the one in the Fuji F10.
By an increase in lens performance I mean acceptable sharpness and less
purple fringing at wide apertures and wide FOVs.

 

[Shaun Williamson]

I agree that at present we cannot construct an F0.5 lens but at least let
us get a clear view of what the problem is. At some time in the future we
may discover optical materials that allow us to get closer to this ideal.
Let me give you a practical example of what it may be possible to achieve.
The first digital camera I bought had a Canon designed zoom with an
aperture of F2.0 at the wide (35mm) end to F3.0 at the tele (105mm)
end. I present I use a Fuji F10 which has a good low noise 6mp sensor.
Its lens starts at F2.8. This makes it difficult to achieve a small DOF and
limits its low light performance.
A Fuji F10 sensor combined with the Canon lens would allow a whole F stop
improvement in the noise performance and this would be a big
improvement. Opening the aperture up by 1 stop delivers twice as much
light to the photodiodes on the sensor and this is a worhtwhile
improvement in performance.
Lets keep in mind that what every photographer really want is a small
pocket sized high performance camera that they can carry everywhere.
The DSLR is still too large and too heavy to meet that need.

 

[Shaun Williamson]

Well now you can look at and print Phil's samples on this site.

 

[Shaun Williamson]

I agree although I would say photosite size.

 

[Alan Martin256506]

One of the problems with this discussion is that all sorts of
confusions creep in. So we have people wanting to redefine
aperture and ISO etc. In terms of their redefinition they
then want to claim that the Sony 828 is a superior camera to
the R1. This is just silly since everyone knows that the R1
can take better pictures with lower noise than the 828.

I made no such claim: I said that the R1 has the faster lens (in the sense that it gathers image-forming light faster, and has shallower DOF), as is apparent from these cameras' 35mm equivalent maximum apertures as I defined them. I also said that larger sensors, such as the one in the R1, have more dynamic range (at http://forums.dpreview.com/forums/read.asp?forum=1000&message=14968333 .

It is the naive use of the standard specifications that would lead to the false conclusion that the F828 lens is superior (f/2.0-2.8 vs. f/2.8-4.8).

Defining alternate versions of aperture and sensitivity isn't such a big problem: we already use 35mm equivalent focal length as well as actual lens focal length, without too much confusion about which is which. Digital photographers are already familiar with what "35mm equivalent" means, and where it is not appropriate (e.g., discussions of lens design), it is easy enough to obtain the actual lens specifications. My proposal is merely to take this approach one step farther.

But why should we do such a thing? I gave what I consider to be good reasons in my post at http://forums.dpreview.com/forums/read.asp?forum=1000&message=14976936 ;. One of the key advantages is that the depth-of-field properties of a lens, in terms of the final printed result, are contained in a single number - the 35mm equivalent aperture.

If my exposure meter says that this scene requires F2.8 at 1/60th
at ISO 100 then no matter which camera I am using I can set the
correct exposure. That is the real use of these ratios and that
would be destroyed by redefining them.

Not so. The factors applied to the F-number and the ISO cancel out in the exposure formula, so your exposure meter will work just as well with 35mm equivalents as with the actual F-number and ISO sensitivity. If this were not the case, I would not have recommended such a system, nor called it "35mm equivalent".

It has also been claimed that the larger absolute diameter of the
lens opening at an aperture of F2.8 on a large camera allows it to
direct more light onto the sensor. This is true of course but that
light has to be directed over a larger sensor area than it does in
a compact digicam. So just saying that the larger lens collects
more light does not in itself explain anything about the better
noise performance of an APS v a compact sensor.

Perhaps... but then, neither does the difference in sensor size in itself explain the difference in noise. To obtain better noise performance at the same light level and shutter speed requires the combination of a larger sensor and a larger lens. (I would say a faster lens, in terms of 35mm equivalent maximum aperture.) The sensor only needs to be larger to keep the required lens F-number from becoming infeasibly (or even impossibly) low.

This kind of noise performance - essentially, low-light performance - should be kept distinct from noise performance on images properly exposed at minimum sensitivity, which is really equivalent to dynamic range (if contrast is kept the same). It is well-known that large-sensor cameras have an advantage in dynamic range, regardless of the speed of their lenses. I guess I should have been clearer about what I meant by noise performance.

Construct a small sensor with large photosites and you will get the
same noise (but not much resolution) on your compact digicam. Put
small photosites on a large sensor and you will get the same noise
as you get in a compact digicam.

This is true enough, but it is not noise itself that I claim to be relatively invariant at a given 35mm equivalent sensitivity, but the tradeoff between noise and resolution in megapixels. There is a certain amount of light available, and it can be spent either on resolution or on reducing noise. Pushing this tradeoff to extremes can significantly increase or decrease noise, but that's not very relevant to actual photography.

The more important question is how can we get a small sensor to
perform as well as a large sensor. One answer to this would seem to be
to make lenses with a larger maximum aperture. In theory a compact
small sensor with a F1 lense would give lower noise than a F2.8 lens
on an APS camera. The F1 lens would also give the same DOF on our
compact as a F4 gives on an APS camera (approximately). However
constructing such wide aperture sharp lenses would not be easy and it
is not clear to me that the small photodiodes on a small sensor could
cope with such higher levels of light.

You're absolutely right about this, and the difficulty of making low-F-number lenses (which actually becomes an impossibility below f/0.7, if transmission losses are factored in) is one of the main reasons why large-sensor cameras have superior optical performance.

In 35mm equivalent terms, the result is that larger sensors can have faster lenses. This might sound strange when some lenses are interchangeable between cameras with different sensor sizes. For instance, the Canon 50mm f/1.4 on the 1.6x crop 350D (or 20D) has 35mm equivalent specifications of 80mm f/2.2. However, when you consider that the light that falls outside the sensor area doesn't count toward 35mm equivalent lens speed, this makes sense.

35mm equivalent specifications deliberately abstract away from what happens between the capture of light by the lens and the production of the final image. I think that is a good thing. It is far easier to deal with different sensor sizes if their effect is to change the range of settings available for sensitivity and aperture, rather than to change the meaning of each setting.

--
Alan Martin

 

[Olivier GALLEN]

Small photodiodes on a small sensor can not cope with such higher
levels of light due to small full-well capacity.
Thus, bright lens can not improve SNR at base ISO because of small
full-well capacity.

I am fairly optimistic about the Dynamic Range issue, and hope that unlimited DR will be available in the mid/long term.

As an example, here is an article based on variable integration time on a per pixel basis: http://www.us.design-reuse.com/articles/article7411.html

Olivier

 

[Olivier GALLEN]

As many people already explained, you can compute more than just the "focal length equivalent in 35mm" (ie: get the same Field of View).

If you consider a system wich has an Equivalent Factor "EF" = 2 (aka: cropping factor): when you use a 50mm at f/2.0 and 100 ISO on this system, you will get the same Field of View, Depth of Field and Signal (as in "Signal to Noise Ratio" which leads to noise in your pictures) as with a 100mm at f/4.0 and 400 ISO on a 35mm system (aka: 24x36).

This is based on optical geometry (-> FOV), DOF formulae (EF^2 due to real focal to keep same FOV, divided by EF due to enlargement to get the same size output on which 'acceptable' blur is measured) and Quantity of Light received (x EF^2 for f-number, and / EF^2 for sensor area) - while understanding the various types of noise (Read Noise, Photon Noise, Dark Noise...).

So I have to multiply focal length by EF, f-number by EF and divide ISO by EF^2 to get the Equivalent in the 35mm system (and know how to get the same image). This is quite a convenient way to compare various systems and take into account the sensor size and the lens f-numbers...
So, here it is: an excel sheet to compare various systems with this Equivalence!
http://olivier.gallen.mageos.com/Temp/Sensors-Equivalence.xls

Summary:

'SENSORS' page:

• Size of various formats (active area when available, source quoted)
• Factor (Equivalent Factor) based on diagonal ratio to 35mm
• Influence of the smaller format stated in f-stop reduction
• What ISO I should select to get the same Signal as with 3200 in 35mm

(blue is changeable)

'SYSTEMS' page:

• Groups of Systems with selection of lenses
• On the left: real range of values (REAL focal lengths, f-numbers)
• Middle: Format name and Equivalence Factor used for calculations
• Right: 35mm equivalents focal lengths and f-numbers used to make comparisons

(you can change systems and values: you should change description on the left, change real values in blue and check Format and Factor. You can also copy/paste lines to add new elements)

'CHARTS' page:

• Charts showing several categories (DSLR Pro, Prosumer, Amateur...)
• Horizontal: equivalent focal length (FOV) in a log scale (10 to 1000mm)
• Vertical: equivalent f-number (DOF, Signal) in a log scale
• For DSLR, lines are colored by Brand (Canon=Red, Nikon=Yellow, Olympus=Blue)
• Systems with stabilization are in doted lines

=> The LOWER the lines, the better for DOF control and Signal (low light ability)

NOTES:

• if you see any error (sensor size, calculations, logics, translation...), please send me an email with description, correction and source (if possible)
• if many people want to enhance it (formats, systems, lenses, charts), I will try to help
• if you have any suggestion to improve this tool: please tell (I am not that happy with the charts...)
• For the experts: OK it is not 100% true nor accurate (I know that blur will be different depending on low-pass filter, sharpening, etc... and affect DOF; that Signal will depend on microlenses, Fill Factor, Q and Noise is even a trickier issue... especially in the anti-noise department; that focal/f-number are never accurate; that different ratio(3/2; 4/3; 16/9) are not properly handled; etc...) BUT my point is to start by changing from the usual "1=2" we see everywhere and get closer to "1=1"... even if there is a 10% approximation here Therefore, be kind to me
• good luck and have fun

Olivier

 

[Shaun Williamson]

I agree with everything you say. It is easier to make a camera with a large
sensor. However all photographers including professionals would like a smaller
camera. The quest for the smaller lighter camera led to the invention of the
35mm film camera which initially could only produce noisy grainy images.
My Fuji F10 easily out performs the original Leica cameras and the films that
were available at that time.
All that I am saying is that lenses with apertures in the range of F1.0 to
F8.0 would enable small sensor cameras to perform as well as APS cameras
since they would get that extra light to the sensor in low light situations.
There are many small digicams that perform as well as DSLRs but only at
ISO 100. The challenge for camera makers is to increase this to ISO 400 and
beyond that.
If you carry your DSLR with you everywhere, everday then perhaps you
won't be interested in this.
On the other hand if we can make smaller cameras with larger sensors then
that might be another solution.