How to compare camera film ISO number to quantum efficiency?

[Khashishi]

The efficiency of a scientific camera like http://www.princetoninstruments.com/products/imcam/proem/ is specified by the quantum efficiency vs wavelength. The meaning is simple enough. Assuming a fill factor of 1, it's the fraction of incident photons that generate a charge.

But photographer cameras are specified by some digital-equivalent ISO film speed. For example, this camerahttp://www.usa.canon.com/cusa/about_canon/newsroom?pageKeyCode=pressreldetail&docId=0901e02480fb7db6 boasts an equivalent 4000000 ISO. How can I compare the sensitivity of this camera to the above? Is this number useful for anything? Can one simply increase the ISO by increasing the ADC gain (but not actually improving signal to noise)? I have a nagging suspicion that this is just a marketing number which is far removed from actual physical meaning.

Of course, the scientific camera doesn't capture color information, which probably can't be done at high efficiencies.

 

[Drakkith]

As I understand it, the ISO rating for a digital camera says little to nothing about its quantum efficiency. You can actively change the ISO of many digital cameras by adjusting the ADC gain, which just means that the signal generated at readout is amplified more or less than normal. You might be able to search for the specifications of the actual sensor used in the camera, which should then give you a quantum efficiency curve.

How can I compare the sensitivity of this camera to the above?

Hmm... not sure. Looking around, it appears to be fairly complicated. Perhaps someone else knows more.

Is this number useful for anything? Can one simply increase the ISO by increasing the ADC gain (but not actually improving signal to noise)? I have a nagging suspicion that this is just a marketing number which is far removed from actual physical meaning.

I don't have a lot of experience with handheld digital cameras, as I mostly do astrophotography and use specialized cameras, but my understanding is that while the ISO setting has no effect on the SN ratio, it will still directly impact the image. In general, a higher ISO will require less exposure time, but will have more noise than a lower ISO (because the amplification process amplifies both noise and signal, and a high ISO has more amplification). A lower ISO will have less amplification, but the image will be dimmer compared to a higher ISO for any given amount of illumination.

The key to digital ISO, as far as I know, is that it is an easy to use 'processing technique'. In astrophotography with my specialized cameras, I have to manually process the pictures using image processing software in order for them to look good. This is, obviously, undesirable for the overwhelming majority of camera users who want a simple 'point and shoot' camera. The ISO of the camera is a simple setting that you can adjust to automatically process images a certain way to give you decent quality with little to no software adjustment. See this link: https://en.wikipedia.org/wiki/Film_speed#The_ISO_12232:2006_standard

Hopefully I haven't gotten something blatantly wrong here.

 

[Drakkith]

Of course, the scientific camera doesn't capture color information, which probably can't be done at high efficiencies.

With a single-sensor camera, no. You either have to use a bayer-array to get color from a single exposure, or take multiple exposures through different color filters, both of which will greatly reduce the overall efficiency since they are blocking large portions of the light from reaching the sensor.

 

[Andy Resnick]

<snip> Of course, the scientific camera doesn't capture color information, which probably can't be done at high efficiencies.

Echoing Drakkith, there's no clear way to equate ISO and QE. QE generally refers to the physical conversion of photons into electrons, while ISO (and gain in general) refers to the electrical signal after conversion. It's worth pointing out that there can be several amplification stages, and each are associated with slightly different noise characteristics. CMOS chips can have on-chip amplification, for example.

 

[Khashishi]

With film, it makes sense to talk about a sensitivity in terms of how much the negative darkens in response to photon flux. But for the digital realm, it seems like ISO is talking about how much voltage rises in response to photon flux. I'm not sure since the definition isn't broadly given. But this number seems to be ultimately meaningless since you can amplify the voltage as much as you want. Ultimately, what matters is how many photons get counted on the ADC, and how much noise there is. So digital ISO just a marketing term?

 

[scientific601]

Found this. It may help.
http://www.strollswithmydog.com/effective-quantum-efficiency-of-sensor/

 

[Andy Resnick]

<snip>So digital ISO just a marketing term?

I wouldn't go that far. Recall, ISO makes more sense when comparing one film to another: ISO 400 is 2x as sensitive as ISO 200, as opposed to "ISO 400 film will become 50% opaque when the absorbed energy reaches (just to make up a number) 0.001 J."

So using ISO in digital cameras occurred because rather than tell photographers "the electronic gain can vary between 0 and 100", in order to 'speak their language' there are ISO-equivalent gain settings: normalized to ISO 100, it makes more intuitive sense what ISO 200, ISO 320, etc. mean for different lighting conditions. Also recall that film response and f-stops are logarithmic scales- the basic unit is 'to double' or 'to half'.

That said, when a camera manufacturer claims ISO ratings of 10 kajillion are possible, that's marketing.

 

[sophiecentaur]

Isn't the important thing Signal to Noise Ratio? You can whack the gain up as high as you like but the pictures just get grainier and grainier, The ISO number is not enough to tell you how good or bad the pictures will look.

 

[davenn]

Isn't the important thing Signal to Noise Ratio? You can whack the gain up as high as you like but the pictures just get grainier and grainier,

yes that's true, But then you can take into account how different sensors and assoc. electronics handle higher ISO's ( read higher ADC gains)

for example, my Canon 5D Mk3 full frame sensor, blows my Pentax K5 crop sensor out of the water on ISO settings above 1000.
I do a lot of low light event photography in large auditoriums running anywhere from 2400 - 5000 ISO and still get good performance. The K5, a good quality DSLR, just cannot handle those shooting conditions.

CMOS chips can have on-chip amplification, for example.

This is a contributing factor to better signal to noise ratio ... The high end Sony sensors do this as used in The Sony A7, 7r, 7s
and the Nikon D810 (a)
Compared to my Canon 5D3 where the amplification and ADC is off-chip. This gives the Sony sensors an edge over the Canon ones
in the low noise performance gameDave

 

[Drakkith]

Isn't the important thing Signal to Noise Ratio? You can whack the gain up as high as you like but the pictures just get grainier and grainier, The ISO number is not enough to tell you how good or bad the pictures will look.

As best as I can tell, you can assign a 'base' ISO for any given sensor that is determined by the quantum efficiency of the sensor, the size of the photosites, and the noise generated by reading out and amplifying the signal from the sensor. So a higher base ISO would mean that the sensor gives a higher quality image compared to a lower ISO sensor for exposures at the same time, at the same luminance, with the same optical system.

However, this base ISO is not what is generally advertised. Part of it is because different processing and compression methods add different amounts of noise to the images. Plus the same sensor can be put in different cameras with different optical properties. In addition, there are multiple possible 'speeds' to which the ISO rating can refer to, which are: saturation based speed, noise based speed, and standard-output-sensitivity (which is still a type of 'speed' in this context). There's also the fact that you can arbitrarily increase the ISO on the camera.

I also found this from the wiki article on ISO: Despite these detailed standard definitions, cameras typically do not clearly indicate whether the user "ISO" setting refers to the noise-based speed, saturation-based speed, or the specified output sensitivity, or even some made-up number for marketing purposes.

So unless the ISO number between two different cameras is wildly different, I'd take it with a grain of salt. And probably some lime.

 

[sophiecentaur]

As best as I can tell, you can assign a 'base' ISO for any given sensor that is determined by the quantum efficiency of the sensor, the size of the photosites, and the noise generated by reading out and amplifying the signal from the sensor. So a higher base ISO would mean that the sensor gives a higher quality image compared to a lower ISO sensor for exposures at the same time, at the same luminance, with the same optical system.

However, this base ISO is not what is generally advertised. Part of it is because different processing and compression methods add different amounts of noise to the images. Plus the same sensor can be put in different cameras with different optical properties. In addition, there are multiple possible 'speeds' to which the ISO rating can refer to, which are: saturation based speed, noise based speed, and standard-output-sensitivity (which is still a type of 'speed' in this context). There's also the fact that you can arbitrarily increase the ISO on the camera.

I also found this from the wiki article on ISO: Despite these detailed standard definitions, cameras typically do not clearly indicate whether the user "ISO" setting refers to the noise-based speed, saturation-based speed, or the specified output sensitivity, or even some made-up number for marketing purposes.

So unless the ISO number between two different cameras is wildly different, I'd take it with a grain of salt. And probably some lime.

Interesting info. Thanks for that view of the world. I imagine is was the same - but much worse - when film was the thing but measuring things like SNR was much harder so it's very likely that one company's 400ASA was in no way the same as another's. Just near enough for people not to bother. Then there was the idea of processing for higher or lower film speed . . . . . .Still, all the 'best' photographers used to have a favourite film and could tell if they were using a different one.
The way that the camera processor applies 'noise reduction' must be very relevant too.

 

[William White]

Interesting info. Thanks for that view of the world. I imagine is was the same - but much worse - when film was the thing but measuring things like SNR was much harder so it's very likely that one company's 400ASA was in no way the same as another's. Just near enough for people not to bother. Then there was the idea of processing for higher or lower film speed . . . . . .Still, all the 'best' photographers used to have a favourite film and could tell if they were using a different one.
The way that the camera processor applies 'noise reduction' must be very relevant too.

there wasn't much difference between different films rated at the same ASA in regard to the "correct" exposure of an 18% grey card.

Velvia, for example, was an exception: rated at ASA50 but really was ASA40 or even lower. Not a great deal of difference, but enough to ruin a shot on a high contrast day. Bracketing really was essential then.

the choice of colour film was really down to the strength of the dyes. Velvia was particularly saturated, which some liked for high-colour images, but was absolutley useless for realism, particularly portraiture (and particularly black faces, which ended up looking decidely magenta). Provia was better for "realistic" colour.With digital cameras, there is no real equivalence with ISO of film (you can't always rely on the sunny 16 rule!)
http://www.dxomark.com/About/In-depth-measurements/Measurements/ISO-sensitivityNoise reduction is important too. Very few cameras produce raw-RAWS; that is there is always something going on before the file is written. Some cameras do produce raw-RAWs, but you need to hack the camera and use fiddly software to open the file.

Interesting for camera nerds, not so much for people wanting to process a decent image, where a RAW with "zero" noise reduction is just fine. But even then, an image opened in, say ACR with zero noise reduction, will look different to the same file in, say CaptureOne, with zero noise reduction. There are things happening all along the chain.

 

[sophiecentaur]

RIGHT. So, yet again, we find that a single number is not much good for characterising anything complicated.

A very non-linear response and, when you got down to it, buckets of dye and a splash of chemicals. . . . . But some people still love film. Also there are vinyl loves.

 

[William White]

RIGHT. So, yet again, we find that a single number is not much good for characterising anything complicated.

A very non-linear response and, when you got down to it, buckets of dye and a splash of chemicals. . . . . But some people still love film. Also there are vinyl loves.

in film, the ISO let you know that shooting a grey card to correct exposure would require half as much exposure time if you doubled the ISO, or let you use the sunny16 rule correctly etc...

Negative films, especially in the 200-400ASA range had so much latitude you didn't even need a light meter. With a bit of practice you could estimate the exposure, and even if you were up to 2 stops out, you got a nice clean image. Photojournalists called it F8 and be thereBut, a single number is all that is needed for the photographer. 'Correct' exposure is not complicated.

 

[sophiecentaur]

But, a single number is all that is needed for the photographer. 'Correct' exposure is not complicated.

A single number is all that's effectively available to a photographer, aamof. With film, 'correct' really means 'least worst', when you consider how poor the colourimetry of colour film is. A grey scale is the least of your worries when you want good fidelity. Because the non linearity of the colour signal channels in film can't be corrected for, film always had / has a characteristic 'filmy' look. Nice if it's what you happen to want but a nightmare if you want naturalistic pictures. For the last 40+ years, colour TV imaging has improved and improved, and one big reason is that the three channels are separate and individually controllable. It's a massive, fundamental advantage with electronic signals over photo-chemical signals, I guess.

 

[William White]

film always had / has a characteristic 'filmy' look.

Are you talking about when used for motion pictures? Or photographic prints?Prints (on paper etc) use dyes/toners, I'm not sure its always possible to tell the difference between a print from a "good" negative (especially large format) and a "good" digital camera.

High-resolution films like velvia have an effective resolution of about 160 lines/mm. I've scanned lots of velvia and that seems about right. Obviously, this is only attained with the best optics. Portrait objectives are often low contrast (to reduce perceived sharpness); macro objectives high contrast.

 

[sophiecentaur]

Are you talking about when used for motion pictures? Or photographic prints?

In either case, there is very little control of 'levels and curves' with film and printing with correcting filters gives only a limited amount of control / correction. Some people like old technology and I have no problem with that. New technology can deliver some very dodgy results - JPegs can be terrible for some subjects and some automatic 'correction can do just the wrong thing with some pictures. I think the notion that there's more information of a good 35mm negative than on a digital picture, has been put to bed by now.
Of course, Printing is a problem for both technologies and there are some truly poor digital printer systems around.
But the thread has been hijacked by our conversation and I think we should let it go. Suffice to say, afaiaa, there is very little commercial imaging done with film, these days, and that really says it all.

 

[Andy Resnick]

I should take back a little of what I said earlier- the ISO scale is a standard scale, and so it must be based on some physical parameters. Otherwise, there's no way to guarantee that films from different manufacturers will behave identically and similarly, to ensure that different digital sensors will produce consistent output. And light meters- they all need to produce outputs that are consistent with film properties and digital sensor properties.

Here are the film standards:
http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=34533
http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=3580
http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=11948

And the digital sensor standard:
http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=37777

 

[William White]

Andy, further up I posted the link ot DXOMark that measures sensors and lenses.


"As tests show, the ISO settings reported by camera manufacturers can differ significantly from measured ISO in RAW. This difference stems from design choices, in particular the choice to keep some “headroom” to avoid saturation in the higher exposures to make it possible to recover from blown highlights."My digital camera needs to be set to about ISO 160 to return an exposure that would be equivalent of "true" ISO 100.

Add to that, its base ISO is 200, which means that when set to ISO 100 the camera is over-exposing by a stop and then pulling back the exposure in post processing (so its not a raw-RAW).

The upside is less noise. The downside is a (potential) loss of dynamic range.

Most photographers now "expose to the right" anyway increasing exposure until the highlights are almost blown and then under-exposing in post - processing. (when with positive film one would often under-expose by 1/3 stop to saturate the colours)

 

[Andy Resnick]

Andy, further up I posted the link ot DXOMark that measures sensors and lenses.

"As tests show, the ISO settings reported by camera manufacturers can differ significantly from measured ISO in RAW. This difference stems from design choices, in particular the choice to keep some “headroom” to avoid saturation in the higher exposures to make it possible to recover from blown highlights."

My digital camera needs to be set to about ISO 160 to return an exposure that would be equivalent of "true" ISO 100.

Add to that, its base ISO is 200, which means that when set to ISO 100 the camera is over-exposing by a stop and then pulling back the exposure in post processing (so its not a raw-RAW).

The upside is less noise. The downside is a (potential) loss of dynamic range.

Most photographers now "expose to the right" anyway increasing exposure until the highlights are almost blown and then under-exposing in post - processing. (when with positive film one would often under-expose by 1/3 stop to saturate the colours)

Right- and that's why a 18% grey card can be so important (side note- an edge-on view of a stack of paper is close to an 18% gray card). For myself, I typically have to 'fine tune' the auto exposure setting on my camera (by as much as 1 1/3 stop) in a lens-dependent fashion to get a correct exposure. I figured that's an error in the auto expose algorithm.

But this is drifting off-topic: the OP asks about relating ISO/ASA/DIN to QE, and I think we all agree that there is no clear relationship. Parameterizing electronic gain in terms of ISO/etc. is useful, but not quantitative.

Personally, I would be interested to know more about using film ISO as a radiometer, since film is often used to gauge exposure to ionizing radiation. I don't feel like buying those ISO specs...

ps- have you ever played around with film like SO-192 or PFG-01?

 

[William White]

when I still used film, the most exotic films I used were Kodak colour InfraRed which was quite fun, but difficult to expose correctly. You had to compose the picture, focus, then screw in the IR filter (which was almost totally opaque to visible light - the viewfinder was black); then re-focus for infra red (or stop right down); then guestimate the amount of infra red light in the scene. The film was used by forestry commissions and botanists to take aerial photos of forests - healthy trees showed up bright red, less healthy tress pinker and more and more white.
Your point about different lenses giving different exposure for the same f-stop is about transmission. Its why in filming, lenses have t stops rather than f stops.

When I was learning photography, I used a grey card, but you soon get to know what things pass for 18% grey : bright blue sky; shaded grass; paving slabs; rocks. The matrix metering of new digital cameras is absolutely fantastic though. Its quite difficult to get it wrong.

 

[Andy Resnick]

<snip>
Your point about different lenses giving different exposure for the same f-stop is about transmission. Its why in filming, lenses have t stops rather than f stops.
<snip>

Interesting... I never thought of that. I figured it was field-of-view related, like the bright sky tended to get included in the metering if I used my wide angle...

 

[William White]

More complex objectives have many more lens elements, with each lens element reducing transmission.

The old Zuiko lenses, for example, used to be prefixed E. Zuiko, F.Zuiko, G Zuiko etc. The prefix was the number of elements, E meant 5, F meant 6 and so on to give the photographer an idea of transmission.

This prefixing never carried on to the automatic-everything digital era (sadly, neither did useful depth scales which are hardly ever seen on still-photography objectives anymore)

Another factor is whether each element is coated, mult-coated, the type of coating, and of course the type and quality of glass. Glass ages and transmission suffers, especially with lenses that contain thorium oxide and/or rare Earth's which causes the lens to discolour (I've read that 'bleaching' the lens with UV for several weeks can restore brightness, depending on the rare earth).

In practice, this makes the square root of no difference in still photography; but it IS a problem in filmography where constant 'brightness' for a given exposure is demanded. Hence the very expensive and exotic t-stop objectives.For a given exposure, what is in your field of view is irrelevent.
If you say, had 100ASA 1/100th sec F4 then changed your lens to a wider angle to include a bright sky; your exposure *should* not change if the transmission of the lenses are the same - the dimmer part of the image (say the land) would have exactly the same exposure as before..

A problem would arise if you had your camera in shutter priority or aperture priority or automatic, and you were not aware that the fstop of shutter speed had changed.

 

[sophiecentaur]

Digital has made such a difference to the business of exposure. I always look at the histogram of a shot and that tells me how the initial exposure sits with respect to the highlights and darks, better than the ever so smart exposure control that the camera does. A second or third exposure costs me nothing unless the rabbit has disappeared down its hole, of course. But the time to assess the light conditions is way before you want to take the important pictures.

in filming, lenses have t stops rather than f stops

OMG, life's too short. That's another reason for going digital.

 

[William White]

OMG, life's too short. That's another reason for going digital.

t stops are used in digital cinematography

http://www.usa.canon.com/cusa/professional/products/lenses/cinema_lenses/cinema_prime_lenses

If you had two (or more) cameras shooting a scene, you need to ensure that the exposure from each camera is the same, otherwise it would look terrible as the editor flipped between shots! You cannot use f-stops in when filming.
(histograms are taken from a lo-res preview the camera generates; and they are worse bet than trusting the camera's light meter. The histograms are not an accurate description of the exposure of the raw file. It's a safer bet to use the light meter, spot meter and then expose to the right, (or use the fancy matrix metering and adjust the exposure depending if the scene is abnormally bright (+2 stops) or dark (-1.5 stops).

 

[sophiecentaur]

t stops are used in digital cinematography

http://www.usa.canon.com/cusa/professional/products/lenses/cinema_lenses/cinema_prime_lenses

If you had two (or more) cameras shooting a scene, you need to ensure that the exposure from each camera is the same, otherwise it would look terrible as the editor flipped between shots! You cannot use f-stops in when filming.
(histograms are taken from a lo-res preview the camera generates; and they are worse bet than trusting the camera's light meter. The histograms are not an accurate description of the exposure of the raw file. It's a safer bet to use the light meter, spot meter and then expose to the right, (or use the fancy matrix metering and adjust the exposure depending if the scene is abnormally bright (+2 stops) or dark (-1.5 stops).

Every day I learn something interesting on PF. So that histogram display is barely worth having? I know that histograms can vary quite a lot from shot to shot, with only a small change of zoom or pan direction so that would make sense. I will try doing without it and asses my results.
Regarding t stops, it was common practice to line up all the (studio) TV cameras against a grey scale and adjust their channel gains to eliminate the variation between cameras as much as possible. That was my experience with analogue TV cameras. I would have thought that the equivalent could easily be done with digital; something that film directors could never get away with.

 

[William White]

Every day I learn something interesting on PF. So that histogram display is barely worth having?.

you could try adjusting the camera's JPEG setting (say from "mute" to "natural" to "saturated" or whatever it is called on your camera); adjust the gamma or the colour space, and see what that does to the histograms - and the light coming into your camera has nothing to do with what you have done.

 

[sophiecentaur]

you could try adjusting the camera's JPEG setting (say from "mute" to "natural" to "saturated" or whatever it is called on your camera); adjust the gamma or the colour space, and see what that does to the histograms - and the light coming into your camera has nothing to do with what you have done.

I use RAW mostly and let Aperture do the Jpegging on the final result but I guess the Image Engine on the camera must be on a reduced set of image data. Jpeg certainly takes a lot of liberties with images - on the grounds that you won't spot what it's done when it's chucked away Megs and Megs of data. The RAW data is not affected by the 'colour balance' setting but the image on the screen is - another reason to support what you were saying.

 

[Andy Resnick]

<snip>
For a given exposure, what is in your field of view is irrelevent.
If you say, had 100ASA 1/100th sec F4 then changed your lens to a wider angle to include a bright sky; your exposure *should* not change if the transmission of the lenses are the same - the dimmer part of the image (say the land) would have exactly the same exposure as before..

A problem would arise if you had your camera in shutter priority or aperture priority or automatic, and you were not aware that the fstop of shutter speed had changed.

What i meant is that since the camera's light meter covers a subregion of the frame, using a wide angle lens (as best I can tell, the transmission is about 90%), the subregion can contain both bright sky and shadowed ground; while using my normal lens (transmission is 95%, not much different), the subregion is either sky or ground.

 

[William White]

what you are saying is that the auto-exposure is an "average" of the light across the frame and if the average light changes, auto-exposure changes

which is obvious, is it not?

so the angle of the lens is not the factor, the average brightness of the scene is.

What you are better doing is taking a spot-exposure from a point in the frame that is "18% grey" and either locking the exposure or using manual exposure.

However, unless you are very lucky and your two lenses have the same transmission (they wont) the exposure will be different - up to a 1/2 stop.This is the point of t-stop lenses in filming.

With two lenses of the same transmission, say a t-2 t24mm wide angle and t-2 100mm telephoto, with the camera exposure fixed; then the exposure will be the same. You can have multiple cameras, with multiple focal length objectives, all set the same exposure settings, and the final exposure will be, for all intents and purposes, the sameTrying to do this with f-stop lenses is a nightmare (almost impossible) because the transmission can be very different (its easier to talk in stops rather than %). DxOmark measure the transmission of lenses, which are all online and its the norm for there to be up to 1/3 stop difference between even the most expensive lenses of the same given f-stop. Cheaper lenses can be even worse.

 

[parinaz]

how Iso connect to potential barier??

 

[sophiecentaur]

Interesting... I never thought of that. I figured it was field-of-view related, like the bright sky tended to get included in the metering if I used my wide angle...

Most cameras provide Spot Metering so that you can avoid that problem but it is hard to use it any better than just using the auto setting and compensating later. It's something else to learn about and can be avoided if you are prepared to use bracketing shots. Also, if you shoot with Raw, you have a fair amount more tolerance than when using JPEG files. (But I now see I have made a similar point before on this rather elderly thread)

 

[davenn]

how ISO connect to potential barrier??

that was a bit difficult to understand what you are referring to ?? I will assume you are referring to the individual pixels on the sensor

The ISO setting doesn't affect that. All the ISO setting is, is an amplifier that takes the signal from the sensor and increases/decreases the signal level by changing the gain setting electronically.
Increasing the ISO = increasing the gain ... it amplifies signal + noise. This is why, particularly on lower end cameras, greatly increasing the ISO is pointless as the noise level becomes unacceptable even at moderate ISO settings 800 - 1600. High end DSLR cameras, like
the Canon 5D Mk3, 6D, Nikon D810 and the like have much higher quality sensors and amplifier electronics to reduce the noise level so that the ISO (gain) can be increased higher before noise become so much of a problem.Dave