Do not miss: Understanding ISO and Your Camera's Sensor

[Detail Man]

As most know, I hold a contrary view about pixel size vs SNR.

... to make sure I have not missed anything, I did a Google on "noise camera pixel size" and went to the sites that had some level of authority:

Stanford paper - larger pixels = better SNR

http://white.stanford.edu/~brian/papers/pdc/pixelSize_SPIE00.pdf

From the above-referenced paper (in Section 2.1, PDF Page 3, journal Page 453) - where the authors are clearly speaking very specifically about "pixel-level" temporal measuremenrs):

DR increases roughly as the square root of pixel size, since both C and reset noise (kTC) increase approximately linearly with pixel size.

True only to the extent that the spectrum of all internal readout/ADC noises are random in nature.

SNR also increases roughly as the square root of pixel size since the RMS shot noise increases as the square root of the signal. These curves demonstrate the advantages of choosing a large pixel.

True only to the extent that the Photon Shot Noise existing within the light itself dominates noise.

.

I attempt (for the Nth time, it seems) to inspire you to squarely face the following "equivalence". In the quoted text from the Stanford paper above, I substitute every occurance of the term "pixel" with the phrase "image-sensor active-area" ([bolded] below to show the changes made):

DR increases roughly as the square root of [image-sensor active-area] size, since both C and reset noise (kTC) increase approximately linearly with [image-sensor active-area] size.

SNR also increases roughly as the square root of [image-sensor active-area] size since the RMS shot noise increases as the square root of the signal. These curves demonstrate the advantages of choosing a large [image-sensor active-area].

Remaining (for the moment) in the world of temporal noise measurements that such single-photosite analysis necessitates, explain to us why the above [substitutions] would in any way be different when applied to image-sensor active-area sizes of arrays of multiple photosites.

.

Enter the spatial (inter-photosite measurement) domain:

With microlens array assemblies, 100% optical fill-factor is not an unreasonable assumption.

In the spatial domain (as opposed to the temporal domain - the only domain possible with single photosite analysis), explain to us why the above [substitutions] would in any way be different when applied to image-sensor active-area sizes of arrays of multiple photosites.

You have never responded to that query (previously specifically made to you several times) ...

If you cannot identify what differences demonstrably exist between "individual photosite size" and "image-sensor active-area sizes of arrays of multiple photosites" (as those phrases are used in the original, and in the by me modified, quotes appearing above), then there exists no meaningful case that can be made whatsoever for any sort of unique "primacy" of single-photosite analysis.

What matters in analyzing image-sensor performance is spatial (inter-photosite measurements performed over some given measurement time) - not temporal single photosite measurements.

RAW photosite data relates to image-data formed from an image-sensor active-area consisting of an array of multiple photosites. It is not (in applications of interest) about individual photosite output - that is, unless we are discussing a single-photosite ("cyclops") imaging device using one photodetector.

Dear Edward,

Well, it looks like you will never answer the above query. I do understand why. There is no answer that differentiates all that "monster pixel worship" from "monster sensor worship" (and the dreaded "total light"). Since (for 100% fill-factor, anyway, down to around 2 Micron pixel size, or so) that fact is indeed true, "pixel size" pretty much relates to spatial frequency resolution only.

Note that each and every image-sensor design is a different creation. Thus, Read Noise and Full Well Capacity (the determinants of "engineering Dynamic Range") depend more on the particular design and process variables than on individual photosite-aperture size itself.

Google is more than just "your friend". It also has the ability to humble one considerably, and reveal just how little one (self included) may know about subjects until they think long and hard about what is really going on. Yours truly has been thus humbled many, many times.

.

Demonstrable realities:

There are, in fact, a couple of tangible (as in "real") areas where "big pixel enthusiasts" can actually "hang their hat" on solid ground, and real, actual concerns (as opposed to vaporous floobydust).

(1) For sub-Micron photosites (unless specific processes are employed in fabrication, which do not appear to have been implemented, it seems), there is the matter of Random Telegraph Noise in the MOSFET source-follower amps which actually *does* present the most significant technological barriers for the "honey, I shrunk the pixels again" bit. Read Eric Fossum's posts as well as Wang's paper (quoted by me) in the thread where this post exists:

http://www.dpreview.com/forums/post/52427150

(2) Lens-system Diffraction effects do indeed "exact a price" where it comes to spatial frequency resolution (the magnitude of the MTF response) for smaller photosite-apertures:

http://www.dpreview.com/forums/thread/3475094

However, note that "big pixels" themselves exact a very similar spatial frequency resolution "cost". Therefore, pixels can just as well be "too big" as they can be "too small". Think about it sometime.

.

That's all, folks,

DM ...

 

[Tony Beach]

Thanks for the links--interesting, informative reading, most of which I hadn't seen before.

Google is your friend

You just have to learn how to use it. I developed the following doing research for work and school:

• Try several combinations of key words about your question - in this case "noise camera pixel size".
• Filter the responses based on the source and your goal. In this case I ignored blogs, personal web sites, advertising, etc. I concentrated on camera manufactures (e.g. Nikon)

Funny, because when I checked out the comments at the Nikon site the one knowledgeable photographer who responded took exception to it while three people who were not knowledgeable (one even asked if it was possible to make ISO 1600 shots into ISO 400 shots) found the article useful. Another article had to do with specific sensors for scientific applications, and I didn't bother with any of your other random hits.

--
http://imageevent.com/tonybeach/twelveimages

 

[Detail Man]

- but in order for users to effectively maximize Exposure, they require the benefits of reliable on-camera RAW channel level indicators - preferably in "preview" (in order to avoid having to "chimp" by first shooting, and then reviewing).

There is another way.

First, taking the lead from Adams, "calibrate" your "system". I do it by taking a series of images of a gray card and adjusting the exposure from above saturation to below the 0 dB noise point. I plot the resulting ADU vs exposure and measure the number of stops from suggested exposure from the internal meter. I do this for the various qualities of light (sunny, shade, tungsten) and for each lens I have. Turns out I need only nine numbers - 3 each for each quality of light and each class of lens (modern Pentax, legacy Pentax, and other manufacturers). To simplify maters I have wrote software to un-compress the raw files, measure the average ADU values and the SNR, then plot the values. Software also measures computes number of stops from saturation to meter reading. Saturation is where the standard deviation goes down. I identify the suggested exposure by taking two images at the suggested exposure. The software locates the images with the duplicate exposures and uses that as the meter suggested exposure. The major limitation is the need to insure that the light stays relatively constant during the test exposures. I measure and report the variation in LV reported by the camera so I can spot a bad run. Here is an example of the output:



Second, I shoot raw.

Third, for those scenes with troublesome subject brightness range, I use the internal meter in spot mode and determine the recommended exposure for the brightest area where I need to avoid saturation. I adjust that exposure with the calibration factor for the lens and light quality. For my Pentax K-3 and most lenses it is 2 2/3 to 3 2/3 stops depending on quality of light.

It sounds like you have managed to employ some rigors to help you keep RAW-levels linear. Cool.

This is also called "placing" in the zone system.

What meaning does "zone system" have for a solid-state image-sensor that is essentially linear ? The "zones" exist in the non-linear mind's eyes of the beholder (at the "presentation/perceptual end").

Finally, in post processing I adjust black level, brightness and contrast to suit the image.

Ah yes, "black level" - the savior of many a noisy image in processing. Trims that print/display "DR" right down. From there, one's eyes may imagine some vast and deep "DR". Very sophisticated stuff.

In other words I "expose for the highlights and develop for the shadows".

Keep that RAW image-data linear "on top", indeed - especially helpful if one resigns themselves to "throwing away the bottom end" as a "brute-force binary" method of "noise reduction". Such an approach seems to be a justifiable "absolute last ditch effort to save a ship swamped in image-noise".

DM

 

[JimKasson]

Dear Edward,

Well, it looks like you will never answer the above query. I do understand why. There is no answer that differentiates all that "monster pixel worship" from "monster sensor worship" (and the dreaded "total light"). Since (for 100% fill-factor, anyway, down to around 2 Micron pixel size, or so) that fact is indeed true, "pixel size" pretty much relates to spatial frequency resolution only.

Note that each and every image-sensor design is a different creation. Thus, Read Noise and Full Well Capacity (the determinants of "engineering Dynamic Range") depend more on the particular design and process variables than on individual photosite-aperture size itself.

Google is more than just "your friend". It also has the ability to humble one considerably, and reveal just how little one (self included) may know about subjects until they think long and hard about what is really going on. Yours truly has been thus humbled many, many times.

.

Demonstrable realities:

There are, in fact, a couple of tangible (as in "real") areas where "big pixel enthusiasts" can actually "hang their hat" on solid ground, and real, actual concerns (as opposed to vaporous floobydust).

(1) For sub-Micron photosites (unless specific processes are employed in fabrication, which do not appear to have been implemented, it seems), there is the matter of Random Telegraph Noise in the MOSFET source-follower amps which actually *does* present the most significant technological barriers for the "honey, I shrunk the pixels again" bit. Read Eric Fossum's posts as well as Wang's paper (quoted by me) in the thread where this post exists:

http://www.dpreview.com/forums/post/52427150

(2) Lens-system Diffraction effects do indeed "exact a price" where it comes to spatial frequency resolution (the magnitude of the MTF response) for smaller photosite-apertures:

http://www.dpreview.com/forums/thread/3475094

However, note that "big pixels" themselves exact a very similar spatial frequency resolution "cost". Therefore, pixels can just as well be "too big" as they can be "too small". Think about it sometime.

.

DM, how about a camera with a really big pixel?

Sloof Lirpa announces high dynamic range camera - the last word

As is their wont this time of year, the Sloof Lirpa team have produced another blockbuster: this year it's a camera that provides far greater Engineering Dynamic Range (EDR) than anything on the market today. DxO is currently reporting the best EDRs of a bit over 14 stops. The Sloof Lirpa...

blog.kasson.com

Jim

 

[Detail Man]

DM, how about a camera with a really big pixel?

http://blog.kasson.com/?p=9329

Very clever, indeed ! Such gives new meaning to the amorphous phrase, "pixel-level sharpness".

.

Here is a recent street shot taken with my own Cyclops 2000 that exemplifies the modern condition:



Best viewed at 100% in a fairly dark viewing environment, wearing all black clothing, with a goatee.

 

[JimKasson]

DM, how about a camera with a really big pixel?

http://blog.kasson.com/?p=9329

Very clever, indeed ! Such gives new meaning to the amorphous phrase, "pixel-level sharpness".

.

Here is a recent street shot taken with my own Cyclops 2000 that exemplifies the modern condition:



Best viewed at 100% in a fairly dark viewing environment, wearing all black clothing, with a goatee.

Touche!

Jim

--

the last word the last word - Photography meets digital computer technology. Photography wins -- most of the time.

Photography meets digital computer technology. Photography wins -- most of the time.

blog.kasson.com

 

[roadtrpr]

http://www.adorama.com/alc/0015174/article/AdoramaU-Understanding-ISO-and-Your-Cameras-Sensor

Looks like what is needed is some good people to get off the island and venture out into the real world and spread the word about photography according to science.

"Exposuristas" and their ilk have not only been banished here to "the island" - they appear to have infiltrated the "inner sanctum" of DPReview. Indeed, they seem to be lurking "behind every bush".

I thought they migrated here.

There are eight million stories in the Naked City.

Now if you can just persuade Gollywop to make that into a viral, scratch that, a popular you tube video.

"The Resolution Will Not Be Televised" ...

You expect a resolution?

(The market-dominant) JPG-shooters are often unable to reliably determine sensor-level Exposure (by knowing that the status of the RAW channel ADC outputs is a high, yet linear, non-saturated one), even if they would like to. True also for RAW shooters (without their determining reliable "calibration" methods for a specific camera under specific conditions).

It is common for JPG-shooters to have an encouraged faith that their camera metering systems will act effectively to prevent such conditions, while still maximizing sensor-level Exposure. Not so ...

Given the effects of the non-linear gamma-correction (as well as compression in the non-linear RGB transfer-function "shoulders"), I think that it (may) be that some JPG-shooters find themselves preferring to err on the side of lower (sensor-level) Exposure levels - reducing the amount of loss of highlight-details within the in-camera encoded JPGs, while at the same time having the adverse effect of reducing (sensor-level) Signal/Noise Ratios.

All shooters (JPG/RAW) can simply adjust F-Number (for minumum desired Depth of Field) and Shutter Speed (given camera/subject stabilities) prior to adjusting the camera's "ISO" setting value (to the lowest feasible ISO setting in ISO-invariant cameras) - but in order for users to effectively maximize Exposure, they require the benefits of reliable on-camera RAW channel level indicators - preferably in "preview" (in order to avoid having to "chimp" by first shooting, and then reviewing).

DM

I enjoy your writing style, but I suggest that if you join The missionaries that go out to spread the good word in the real word you leave the direct communication with the photo taking masses to others.

 

[Detail Man]

Am unable to find any (out of a of a bevy of) "instructional videos" on YouTube that do *not* equate "ISO setting" to "sensor sensitivity" as well as to "exposure". "When in Rome ...", I suppose ?

ISO confused now, everything somehow seems crystal clear. ISO causes "noise" (mostly random). This phenomena seems to truly be a gender-neutral one within human thought and discourse.

The utmost abstractions are the true weapons with which to control our thought of concrete fact.
- Alfred North Whitehead

DM

 

[Detail Man]

http://www.adorama.com/alc/0015174/article/AdoramaU-Understanding-ISO-and-Your-Cameras-Sensor

Looks like what is needed is some good people to get off the island and venture out into the real world and spread the word about photography according to science.

"Exposuristas" and their ilk have not only been banished here to "the island" - they appear to have infiltrated the "inner sanctum" of DPReview. Indeed, they seem to be lurking "behind every bush".

I thought they migrated here.

There are eight million stories in the Naked City.

Now if you can just persuade Gollywop to make that into a viral, scratch that, a popular you tube video.

"The Resolution Will Not Be Televised" ...

You expect a resolution?

I assumed that ... *you* ... were the one who here dreams of some sort of imagined "resolution" ?

(The market-dominant) JPG-shooters are often unable to reliably determine sensor-level Exposure (by knowing that the status of the RAW channel ADC outputs is a high, yet linear, non-saturated one), even if they would like to. True also for RAW shooters (without their determining reliable "calibration" methods for a specific camera under specific conditions).

It is common for JPG-shooters to have an encouraged faith that their camera metering systems will act effectively to prevent such conditions, while still maximizing sensor-level Exposure. Not so ...

Given the effects of the non-linear gamma-correction (as well as compression in the non-linear RGB transfer-function "shoulders"), I think that it (may) be that some JPG-shooters find themselves preferring to err on the side of lower (sensor-level) Exposure levels - reducing the amount of loss of highlight-details within the in-camera encoded JPGs, while at the same time having the adverse effect of reducing (sensor-level) Signal/Noise Ratios.

All shooters (JPG/RAW) can simply adjust F-Number (for minumum desired Depth of Field) and Shutter Speed (given camera/subject stabilities) prior to adjusting the camera's "ISO" setting value (to the lowest feasible ISO setting in ISO-invariant cameras) - but in order for users to effectively maximize Exposure, they require the benefits of reliable on-camera RAW channel level indicators - preferably in "preview" (in order to avoid having to "chimp" by first shooting, and then reviewing).

I enjoy your writing style, but I suggest that if you join The missionaries that go out to spread the good word in the real word you leave the direct communication with the photo taking masses to others.

I shall gracefully "bow out" now, so that those more eloquent than I can now take the podium:

DM ...

 

[Great Bustard]

DM, how about a camera with a really big pixel?

http://blog.kasson.com/?p=9329

Very clever, indeed ! Such gives new meaning to the amorphous phrase, "pixel-level sharpness".

.

Here is a recent street shot taken with my own Cyclops 2000 that exemplifies the modern condition:



Best viewed at 100% in a fairly dark viewing environment, wearing all black clothing, with a goatee.

I'm disappointed to find that you're spying on me. ;-)

 

[roadtrpr]

http://www.adorama.com/alc/0015174/article/AdoramaU-Understanding-ISO-and-Your-Cameras-Sensor

Looks like what is needed is some good people to get off the island and venture out into the real world and spread the word about photography according to science.

Before anything, cameras need to be optimized for shooting raw. Not only they are not, most of the application/user level programmers in the industry have no clue what are the challenges.

Same is true about shooting accessories. No colormeter will tell you what filter you need to use to balance raw. No exposure meter will allow you to enter the dynamic range in a direct and easy way or, G-в forbid, to enter the table of DR depending on ISO setting. No studio lights or flashes are optimized for the spectral sensitivity of a typical sensor, and no filters are offered.

 

[Iliah Borg]

Soon they will start preaching one can change the sensitivity of a mike using volume control. Not to mention sensitivity vs. responsivity.

 

[roadtrpr]

http://www.adorama.com/alc/0015174/article/AdoramaU-Understanding-ISO-and-Your-Cameras-Sensor

Looks like what is needed is some good people to get off the island and venture out into the real world and spread the word about photography according to science.

"Exposuristas" and their ilk have not only been banished here to "the island" - they appear to have infiltrated the "inner sanctum" of DPReview. Indeed, they seem to be lurking "behind every bush".

I thought they migrated here.

There are eight million stories in the Naked City.

Now if you can just persuade Gollywop to make that into a viral, scratch that, a popular you tube video.

"The Resolution Will Not Be Televised" ...

You expect a resolution?

I assumed that ... *you* ... were the one who here dreams of some sort of imagined "resolution" ?

(The market-dominant) JPG-shooters are often unable to reliably determine sensor-level Exposure (by knowing that the status of the RAW channel ADC outputs is a high, yet linear, non-saturated one), even if they would like to. True also for RAW shooters (without their determining reliable "calibration" methods for a specific camera under specific conditions).

It is common for JPG-shooters to have an encouraged faith that their camera metering systems will act effectively to prevent such conditions, while still maximizing sensor-level Exposure. Not so ...

Given the effects of the non-linear gamma-correction (as well as compression in the non-linear RGB transfer-function "shoulders"), I think that it (may) be that some JPG-shooters find themselves preferring to err on the side of lower (sensor-level) Exposure levels - reducing the amount of loss of highlight-details within the in-camera encoded JPGs, while at the same time having the adverse effect of reducing (sensor-level) Signal/Noise Ratios.

All shooters (JPG/RAW) can simply adjust F-Number (for minumum desired Depth of Field) and Shutter Speed (given camera/subject stabilities) prior to adjusting the camera's "ISO" setting value (to the lowest feasible ISO setting in ISO-invariant cameras) - but in order for users to effectively maximize Exposure, they require the benefits of reliable on-camera RAW channel level indicators - preferably in "preview" (in order to avoid having to "chimp" by first shooting, and then reviewing).

I enjoy your writing style, but I suggest that if you join The missionaries that go out to spread the good word in the real word you leave the direct communication with the photo taking masses to others.

I shall gracefully "bow out" now, so that those more eloquent than I can now take the podium:

DM ...

Sorry, I forget the smiley face.

 

[Iliah Borg]

"ISO isn’t even an acronym, but a shortname for “International Organization for Standardization..."

Well, we know it is not - it stands for Greek iso- ≈ equal (ἴσος)

If a group of missionaries do ever get off the island they might find it effect to write one or more Wikipedia articles. A title that could serve as bait would be "Exposure Triangle". It might take a team to fend off the revisions from the exposure triangle worshippers.

Couple of months ago I wrote privately to one of the chief triangulistas, suggesting to change the course, especially in regards to ISO settings explanation. He told me that

- it is too complicated for the average user (he actually went much stronger than that)

- it is of no practical use

- it means all the voluminous work of evangelists needs to be re-written, and their names may be smeared because of 10+ years of misstating the matter.

 

[Detail Man]

DM, how about a camera with a really big pixel?

http://blog.kasson.com/?p=9329

Very clever, indeed ! Such gives new meaning to the amorphous phrase, "pixel-level sharpness".

.

Here is a recent street shot taken with my own Cyclops 2000 that exemplifies the modern condition:



Best viewed at 100% in a fairly dark viewing environment, wearing all black clothing, with a goatee.

I'm disappointed to find that you're spying on me. ;-)

It was decided by "scrambled-eggs" that such covert activities were a matter of "national security". It seems that you have been into something of a "coma" recently - perhaps due to a space/time "field curvature", where gravity-waves are pulling you into dark-matter event-horizons ? Astigmata ! ...

 

[Detail Man]

... especially in regards to ISO settings explanation. He told me that:

- it means all the voluminous work of evangelists needs to be re-written, and their names may be smeared because of 10+ years of misstating the matter.

So far, only a few of the "illuminati" have prostrated themselves before the "cathedral of science":

 

[Iliah Borg]

... especially in regards to ISO settings explanation. He told me that:

- it means all the voluminous work of evangelists needs to be re-written, and their names may be smeared because of 10+ years of misstating the matter.

So far, only a few of the "illuminati" have prostrated themselves before the "cathedral of science":

We often encounter a different phenomena, too. Personal feelings (reed: hatred) force people to support the point of view they know to be wrong. They are doing that in an attempt to hurt the feelings. As if anybody cares, really.

BTW, it is interesting to watch how many people spread that ISO is derived from iso- myth.

--
http://www.libraw.org/

 

[roadtrpr]

"ISO isn’t even an acronym, but a shortname for “International Organization for Standardization..."

Well, we know it is not - it stands for Greek iso- ≈ equal (ἴσος)

Not only that but if it is were to be what they say then it is an acronym. How else ISO be construed to be "a shortname" for International Organization for Standardazation"?

If a group of missionaries do ever get off the island they might find it effect to write one or more Wikipedia articles. A title that could serve as bait would be "Exposure Triangle". It might take a team to fend off the revisions from the exposure triangle worshippers.

Couple of months ago I wrote privately to one of the chief triangulistas, suggesting to change the course, especially in regards to ISO settings explanation. He told me that

- it is too complicated for the average user (he actually went much stronger than that)

- it is of no practical use

- it means all the voluminous work of evangelists needs to be re-written, and their names may be smeared because of 10+ years of misstating the matter.

--
http://www.libraw.org/

Missionary work is not easy and you can't expect to convert all the innocents. Maybe in addition to Wikipedia articles a popular blog could be created that would administer curative medicines surreptitiously.

I will quit now before my coefficient of distraction gets out of hand.

 

[Iliah Borg]

Maybe in addition to Wikipedia articles a popular blog could be created that would administer curative medicines serriptiously.

I know a lot of folks submitting the corrections to Wiki only to see their edits to be reverted in a matter of minutes.

A collective blog may be a better option, with editors elected among those who are qualified to speak on the matters. Promoting such a blog may be a difficult task however.

--
http://www.libraw.org/

 

[Detail Man]

Maybe in addition to Wikipedia articles a popular blog could be created that would administer curative medicines serriptiously.

I know a lot of folks submitting the corrections to Wiki only to see their edits to be reverted in a matter of minutes.

I, for one (as described in a recent post on this forum) once "tilted at" such "Wiki windmills" ...

A collective blog may be a better option, with editors elected among those who are qualified to speak on the matters. Promoting such a blog may be a difficult task however.

A browse through some of the "livelier" comments submitted regarding this DPReview article:

http://www.dpreview.com/articles/2666934640/what-is-equivalence-and-why-should-i-care

... gives a taste of the joys in store for (any kind of even mildly complicated) "truth tellers". The perceived value of making efforts to know/understand how stuff (actually) works seems virtually nil.

"Lord Bustard" reminds me of a battleground medic with no anesthetics, running out of bandages.

"Part and parcel" of the evolution of a "consumer state" - where money substitutes for knowledge.

.

Invention is the mother of necessity.

A point of view can be a dangerous luxury when substituted for insight and understanding.

Many a good argument is ruined by some fool who knows what he is talking about.

We become what we behold. We shape our tools and then our tools shape us.

Environments are not just containers, but are processes that change the content totally.

- Marshall McLuhan

.

We are so constituted that we believe the most incredible things; and, once they are engraved upon the memory, woe to him who would endeavor to erase them.
- Goethe

.

The fact that an opinion has been widely held is no evidence whatsoever that it is not utterly absurd.
- Bertrand Russell