Nerd Wars...Part 1

chuxter · May 12, 2010

There is another thread in this Forum. It's both terrible and instructive. This thread was started by 2am_strobist who asked 2 simple questions:

" I noticed that when i over or under expose by + or -1EV there is noticeable noise in the image. What causes that? Can it be avoided? "

His questions were answered rather quickly (like the very first reply from Will Rushmore):

" If you have Auto ISO enabled, the camera might switch to a higher ISO setting, and you will get more noise. "

2am_strobist immediately acknowledged that this was the problem:

" You know, I think thats it. I stopped using Auto ISO a while ago and enabled it recently. I forgot that it was on. Man, its always the little things! "

That should have been the end... but NO !!!

2am's thread served as a platform for a scientist-in-training, who insisted that Will's suggestion was not supported by good scientific evidence. thomasnb verbosely proclaimed that:

" Increasing the ISO doesn't increase noise, it reduces it. "

[Out of compassion, I stripped most of his verbosity.]

At this juncture, we were only 6-posts deep. If we had stopped at 5, much of the angst and gnashing would have been avoided. But his above provocative statement did as expected...generated a lot of counter arguments.

Thomas disagreed with everyone. While perhaps not loosing the debate, he was at least out numbered! His "mentor", Emil Martin popped in occasionally to offer support. [Actually, he joined the party because I used one of Emil's published statements to refute Thomas]

In his first post, Emil said:

" If however, exposure is fixed and ISO increases, N/S goes down:"

And offered this "proof":

Which was credited it to:

http://www.guillermoluijk.com/article/iso/index.htm

Well, that set of images is confusing, plus Guillermo's website is in Italian. It's confusing because it appears to show that Emil and Thomas were correct! In spite of the Google translator I did finally understand what Guillermo did and with some help from Emil, I even understood what he meant when Emil wrote, " ...exposure is fixed..." . Following my epiphany, I offered an explanation...

He set up seven objects and took an exposure reading at 1600 ISO, adjusting the camera until it gave a "proper" exposure. He didn't tell us exactly what the histogram looked like...assume he was good enough to generally expose to the right? He then took 5 pictures, varying the sensitivity setting from 100 ISO to 1600 ISO. He did not show us the RAW files, so we have to presume that the 100 ISO picture was massively underexposed. He then used CDRAW to adjust the exposure, so that all 5 pix would look OK. He then made small crops of 3 of the 7 objects that were, according to him in "lighter areas" and assembled them into the matrix (above).

I then suggested that if a similar "experiment" was conducted where the illumination was varied as the sensitivity/gain/ISO changed (instead of PPing them to correct the underexposures), the result would be different. I hereby submit MY "experiment", in Part 2...

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham

chuxter · May 12, 2010

I set up my Nikon D300 on a tripod and provided light from the right side with an SB-900 bounced off a white umbrella, located slightly above the scene. I used a 60mm f/2.8 macro lens. I set the aperture at f/4.5 and the exposure time at 0.0005 seconds in M-mode. WB was set on Auto. I focused on the cap embroidery in LV mode and then locked it. I took two series of pix...one with a fixed illumination (1/32 power) and one with a variable illumination. [For those that don't know, Nikon has a CLS feature that "closes-the-loop" on illumination...I used this feature to have the camera "properly" expose each shot as the sensitivity/gain/ISO was varied...I did not apply any exposure compensation in the camera] I used 5 different sensitivity/gain/ISO settings: 200, 400, 800, 1600, & 3200. [The D300 doesn't have a 100 ISO setting]

The following compilation shows the results. On the right are the 5 shots with variable illumination. In the center are the 5 shots with fixed illumination. On the left are the 5 fixed Illumination shots after PPing with Lightroom to make them look about the same as the 3200 ISO image.

I think this shows clearly what Guillermo did and Emil referenced...he simply didn't show the center row! Those people who initially didn't believe Thomas and Emil can see that IF you do the "experiment" the way Guillermo did, that "Yes", noise really does go down if you increase the ISO.

The next picture shows 3 crops of the pictures taken with fixed illumination (that's the way Guillermo did it). It does demonstrate what Guillermo demonstrated. Apparently, Nikon cameras sorta work the same was as Canon cameras... ;-)

The next picture shows the same 3 crops of pictures taken with variable illumination (that's the way "real" photographers would have done it).

Look closely at both sets. The series with fixed illumination doesn't look as good as the series with variable illumination. Most photographers understand this. They wonder why Guillermo, Emil, and Thomas want it done the other way...other than just to prove a point. Is there some truth that has escaped lay photographers that this trio can illuminate?

Perhaps we need to first agree on the meaning of the term "exposure". Our trio has a "sensor-centric" viewpoint...they think that "exposure" is the number of photons incident on the sensor...thus controlled by the classic duo, aperture and time, just like our grand-dads did it. Interestingly, they dismiss the light incident on the subject as being an "exposure" variable, even though it obviously affects the number of photons collected by the sensor. They also scoff at the idea that sensitivity/gain/ISO adjustments have anything to do with "exposure"...but apparently digital scaling during PPing does?

I think the two sides can be characterized as "sensor-centric" vs "camera-centric". The former group are what I call "sensor technologists"...the later group are what I call "photographers". I find that "sensor technologists" have great difficulty talking down...they basically don't do that. Instead, they talk to the "photographers" as if they were some other "sensor technologists". This approach invariably fails.

It's not totally that "photographers" can't understand what Guillermo did or the conclusions that Emil and Thomas draw from that and preach to us...much more that "photographers" don't understand why , because it seems too far removed from reality. Photographers "know" in their souls that increasing sensitivity/gain/ISO increases noise...they live with that reality daily. If there is some utility in the opposite approach that our trio takes, it isn't apparent what it is. Instead of digging in their heels and arguing incessantly, "sensor technologists" should find some different words and expressions that facilitate two-way communications...even if they are colloquial and subjective.

Bottom Line: If "sensor technologists" expect to be understood and appreciated...if they expect to be judged a valuable asset by working "photographers"...they must devise a better way of communicating their obscure thoughts.

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham

Nick in Shanghai · May 12, 2010

Don't know how you have the patience for it Chuck or the stamina.

--
Nick in Shanghai.

chuxter · May 12, 2010

Nick in Shanghai said:
Don't know how you have the patience for it Chuck or the stamina.

It took me several days...I really do have a life.

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham

Nick in Shanghai · May 12, 2010

chuxter said:
.I really do have a life.

Sure.................if banging your head against a brick wall can be called a life.

--
Nick in Shanghai.

arachnophilia · May 12, 2010

chuxter said:
It's not totally that "photographers" can't understand what Guillermo did or the conclusions that Emil and Thomas draw from that and preach to us...much more that "photographers" don't understand why , because it seems too far removed from reality. Photographers "know" in their souls that increasing sensitivity/gain/ISO increases noise...they live with that reality daily. If there is some utility in the opposite approach that our trio takes, it isn't apparent what it is. Instead of digging in their heels and arguing incessantly, "sensor technologists" should find some different words and expressions that facilitate two-way communications...even if they are colloquial and subjective.

Bottom Line: If "sensor technologists" expect to be understood and appreciated...if they expect to be judged a valuable asset by working "photographers"...they must devise a better way of communicating their obscure thoughts.

i beg to differ. they're just flat out wrong. the test is invalid, because they did it wrong. under-exposure, followed by compensation in post-production is the surest way to generate noise. it's like push-processing from the film days. if you're under-exposing an iso 100 image by four stops and then brightening it, yes, you're going to see more noise than in a properly exposed iso 1600 image. but that's only proving the above. it's comparing apples to oranges. for a proper test, you have test two properly exposed images against one another. and guess what you'll see?

every photographer knows it in their souls because it happens to be correct. and anyone who knows anything about CCD or CMOS technology would agree. but the internet is full of crackpots.

Crocodile Gena · May 13, 2010

arachnophilia said:
arachnophilia said:

chuxter wrote:

Bottom Line: If "sensor technologists" expect to be understood and appreciated...if they expect to be judged a valuable asset by working "photographers"...they must devise a better way of communicating their obscure thoughts.

Click to expand...

arachnophilia said:
arachnophilia wrote:

i beg to differ. they're just flat out wrong. the test is invalid, because they did it wrong. under-exposure, followed by compensation in post-production is the surest way to generate noise. it's like push-processing from the film days. if you're under-exposing an iso 100 image by four stops and then brightening it, yes, you're going to see more noise than in a properly exposed iso 1600 image. but that's only proving the above. it's comparing apples to oranges. for a proper test, you have test two properly exposed images against one another. and guess what you'll see?

every photographer knows it in their souls because it happens to be correct. and anyone who knows anything about CCD or CMOS technology would agree. but the internet is full of crackpots.

Usually, when people talk about "noise", they mean the apparent noise , which is the NSR (noise-to-signal ratio): http://www.josephjamesphotography.com/equivalence/#noise

That said, most people don't really know what "noise" is. They think it's just something the sensor unwittingly "adds" to the image.

When people finally do understand what noise is (standard deviation from the mean signal for a given frequency -- see the above link), then progess can be made.

The primary source of apparent noise in an image is the light itself. High ISO images are more noisy not because the ISO is higher, but because the shutter speed is higher, which means less light falls on the sensor.

The sensor and supporting software do contribute to the apparent noise. However, their contribution is usually less at higher ISOs than at lower ISOs. But since the apparent noise at higher ISOs is intrinsically higher since less light is reaching the sensor, people incorrectly believe that the sensor "adds" more noise to the image at higher ISO settings, and fail to make the connection that it is instead the lesser amount of light getting to the sensor that causes the increase in apparent noise.

Thus, if you could take an image of a scene at a given f-ratio and shutter speed, the higher ISO setting would usually result in less apparent noise, but the higher the ISO, the more of the image would be in danger of oversaturation (blown highlights).

Hence, the "correct exposure" simply means achieving the best balance of maximizing the total amount of light that falls on the sensor vs how much of the image is oversaturated (blown) vs the necessary shutter speed to minimize motion blur and/or camera shake vs the aperture required for the desired DOF / image sharpness. In other words, purely subjective.

Is that "crackpotish" enough?

Les Berkley · May 13, 2010

Anyone take any pictures recently?

--
http://www.artisticlens.com
http://www.artisticlens.com/blog

DSPographer · May 13, 2010

I read much of the thread that you refer to ( http://forums.dpreview.com/forums/read.asp?forum=1014&message=35168630 ) and what I noticed is that after the first 5 posts almost all the following posts dealt not with how cameras work but with semantics.

I think most of the trouble could have been avoided if in the sixth post the point had simply been that using ISO to mean exposure index- as it does when using auto-exposure- is not the only meaning of the term ISO. Then that point could have been followed with a carefully worded explanation of exposure index and camera ISO gain as two separate concepts.

Just in case arguing semantics about the definition of ISO wasn't enough we then were treated to an argument about the definition of noise by two contributors that each completely understand how camera noise works but still found it necessary to argue across each other over the use of language.

I guess what I take away from that thread is that you need to be careful to watch out for disagreements that are really just over language and point it out when you get entangled in one. After that it is probably best to just let the matter drop.

chuxter · May 13, 2010

Crocodile Gena said:
Usually, when people talk about "noise", they mean the apparent noise , which is the NSR (noise-to-signal ratio): http://www.josephjamesphotography.com/equivalence/#noise

That said, most people don't really know what "noise" is. They think it's just something the sensor unwittingly "adds" to the image.

After these two weeks of angst, anyone involved with the subject thread should know what "noise" is. But your slightly condescending statement deserves:

"I know noise when I see it." ;-)

Crocodile Gena said:
When people finally do understand what noise is (standard deviation from the mean signal for a given frequency -- see the above link), then progess can be made.

That was a good link. He communicated very well, especially his analogy in the 5th paragraph. We need more clear communicating like this!

Crocodile Gena said:
The primary source of apparent noise in an image is the light itself. High ISO images are more noisy not because the ISO is higher, but because the shutter speed is higher, which means less light falls on the sensor.

BUT, the assumption that "much less light falls on the sensor" when using a higher sensitivity/gain/ISO is an arbitrary one. That single assumption is what causes this line of thinking to be "wrong". If instead, you would maintain the same light intensity on the sensor as you raised the sensitivity/gain/ISO, the result would be different. In fact, my pix showed that the higher sensitivity/gain/ISO images, when the illumination is varied to keep it constant, have more noise.

My opinion is that this arbitrary constraint is rooted in classical (ie, 20th century) limitations. Some current cameras have the ability to keep the illumination on the sensor constant...they do this automatically. Some photographers use this facility. Even those who lack this feature, do it manually. Pro strobe flashes have a facility to adjust their light output and pros use this to control and balance the illumination. They have no need for the lesson from the technologists that if they just raised the sensitivity/gain/ISO, they would reduce noise (assuming they didn't choose to raise the illumination level). For them, it simply doesn't work that way.

BTW, I know the ability to automatically control illumination is a "special" case, because even in 2010, digital cameras can't automatically control all sources of light. When modern photographers find this limitation, they revert to 20th century methods...they wait for the light to come to them! Or they keep walking to a place where the light is suitable.

Crocodile Gena said:
The sensor and supporting software do contribute to the apparent noise. However, their contribution is usually less at higher ISOs than at lower ISOs. But since the apparent noise at higher ISOs is intrinsically higher since less light is reaching the sensor, people incorrectly believe that the sensor "adds" more noise to the image at higher ISO settings, and fail to make the connection that it is instead the lesser amount of light getting to the sensor that causes the increase in apparent noise.

Well, I for one didn't fail to notice that connection. But I found it lacking substance. It's the kind of connection that professional scientists/amateur photographers make. To be blunt, they do this because it's counter-intuitive...they relish the astonished looks of disbelief on the faces of their lay audience. Even here, on dpr, where they can't see those faces, they can read the words of denial and engage their "victims" in debates only limited by the 150-post limit!

Crocodile Gena said:
Thus, if you could take an image of a scene at a given f-ratio and shutter speed, the higher ISO setting would usually result in less apparent noise, but the higher the ISO, the more of the image would be in danger of oversaturation (blown highlights).

The photographic demonstration by Guillermo and another in your above link by Gabor Schorr carefully sidestepped that issue, by establishing the fixed illumination vs aperture/shutter at the highest sensitivity/gain/ISO. I did the same thing, of course in reconstructing this approach. But you are correct that in actual photography, as distinct from our demonstrations, over-saturation could be an issue. To be precise, adding gain between the sensor and the ADC can't really cause "sensor" over-saturation...but it could cause over-saturation of analog gain components or truncation of the digital conversion. The effect looks the same. You may think I'm being pedantic, but this point is core to the "problem" we have...working pros don't care what part of a camera makes images look noisy or overexposed. To them a camera is a black box.

Crocodile Gena said:
Hence, the "correct exposure" simply means achieving the best balance of maximizing the total amount of light that falls on the sensor vs how much of the image is oversaturated (blown) vs the necessary shutter speed to minimize motion blur and/or camera shake vs the aperture required for the desired DOF / image sharpness. In other words, purely subjective.

Bingo! While demonstrating and explaining that "parlor trick" where noise appears to drop as sensitivity/gain/ISO is raised, is a fun exercise, it's not useful to the working pro! That entire thread should have been over on the Open Talk Forum...this thread too (I only put it here because that's where 2am_strobist's thread was).

Crocodile Gena said:
Is that "crackpotish" enough?

Your post was good, IMO. I didn't interpret arachnophilia's statement, " ...but the internet is full of crackpots... " to imply that any forum member was a crackpot...perhaps you were over-sensitive for some reason?

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham

chuxter · May 13, 2010

Les Berkley said:
Anyone take any pictures recently?

Yes, I posted mine. ;-)

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham

chuxter · May 13, 2010

DSPographer said:
I read much of the thread that you refer to ( http://forums.dpreview.com/forums/read.asp?forum=1014&message=35168630 ) and what I noticed is that after the first 5 posts almost all the following posts dealt not with how cameras work but with semantics.

Yes.

DSPographer said:
I think most of the trouble could have been avoided if in the sixth post the point had simply been that using ISO to mean exposure index- as it does when using auto-exposure- is not the only meaning of the term ISO. Then that point could have been followed with a carefully worded explanation of exposure index and camera ISO gain as two separate concepts.

Yes.

DSPographer said:
Just in case arguing semantics about the definition of ISO wasn't enough we then were treated to an argument about the definition of noise by two contributors that each completely understand how camera noise works but still found it necessary to argue across each other over the use of language.

One characteristic of many technologists is that they can only understand their own words...they simply don't have the ability to comprehend a correct sentence written with different terms. They also have very narrow definitions of terms that pro photographers use in different ways.

DSPographer said:
I guess what I take away from that thread is that you need to be careful to watch out for disagreements that are really just over language and point it out when you get entangled in one. After that it is probably best to just let the matter drop.

The subject disagreement was more than just a semantic "war", IMO. It was about whether the knowledge of technologists is relevant. Even gently implying that one of their approaches is irrelevant is like waving a red flag. Stating that they are totally wrong is like a photo-nuclear attack of the motherland. They are like bees...when one attacks, they all attack. Often they get so excited, they sting their friends. :-(

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham

Crocodile Gena · May 13, 2010

chuxter said:
chuxter said:

Crocodile Gena wrote:

Usually, when people talk about "noise", they mean the apparent noise , which is the NSR (noise-to-signal ratio): http://www.josephjamesphotography.com/equivalence/#noise

That said, most people don't really know what "noise" is. They think it's just something the sensor unwittingly "adds" to the image.

Click to expand...

chuxter said:
chuxter wrote:

After these two weeks of angst, anyone involved with the subject thread should know what "noise" is. But your slightly condescending statement deserves:

"I know noise when I see it." ;-)

Actually, no condescening attitude was implied, and I'm surprised you took it that way. Indeed, people "know noise when they see it", but if they don't know what noise is, then they can't know what causes it, how to minimize it, and what role the equipment plays in the whole affair.

chuxter said:
chuxter said:

When people finally do understand what noise is (standard deviation from the mean signal for a given frequency -- see the above link), then progess can be made.

Click to expand...

chuxter said:
That was a good link. He communicated very well, especially his analogy in the 5th paragraph. We need more clear communicating like this!

Glad you agree. Also, I want to emphasize the different between NSR and SNR. The NSR (noise-to-signal ratio) is the reciprocal of the SNR (signal-to-noise ratio: NSR = 1/ SNR). The reason for using NSR for apparent noise rather than SNR, is because a lower NSR corresponds to lower apparent noise, and a higher NSR corresponds to greater apparent noise. For the SNR, it's exactly opposite, which is a bit confusing, just as using the term "noise" to refer to "apparent noise" is confusing.

chuxter said:
chuxter said:

The primary source of apparent noise in an image is the light itself. High ISO images are more noisy not because the ISO is higher, but because the shutter speed is higher, which means less light falls on the sensor.

Click to expand...

chuxter said:
BUT, the assumption that "much less light falls on the sensor" when using a higher sensitivity/gain/ISO is an arbitrary one. That single assumption is what causes this line of thinking to be "wrong". If instead, you would maintain the same light intensity on the sensor as you raised the sensitivity/gain/ISO, the result would be different. In fact, my pix showed that the higher sensitivity/gain/ISO images, when the illumination is varied to keep it constant, have more noise.

The point is simple: the apparent noise is primarily a function of the total amount of light that falls on the sensor. The role the ISO plays is primarily as an indirect means to setting the shutter speed, and had a relatively minor effect on the apparent noise itself.

chuxter said:
chuxter said:

The sensor and supporting software do contribute to the apparent noise. However, their contribution is usually less at higher ISOs than at lower ISOs. But since the apparent noise at higher ISOs is intrinsically higher since less light is reaching the sensor, people incorrectly believe that the sensor "adds" more noise to the image at higher ISO settings, and fail to make the connection that it is instead the lesser amount of light getting to the sensor that causes the increase in apparent noise.

Click to expand...

chuxter said:
Well, I for one didn't fail to notice that connection. But I found it lacking substance. It's the kind of connection that professional scientists/amateur photographers make. To be blunt, they do this because it's counter-intuitive...they relish the astonished looks of disbelief on the faces of their lay audience. Even here, on dpr, where they can't see those faces, they can read the words of denial and engage their "victims" in debates only limited by the 150-post limit!

Quite the opposite. Once one understands what noise is, and that the primary source of image noise is the total amount of light that falls on the sensor, then everything makes perfect sense.

This opens the door for understanding what can be done at the hardware level to improve noise performance, and what cannot be done.

chuxter said:
chuxter said:

Hence, the "correct exposure" simply means achieving the best balance of maximizing the total amount of light that falls on the sensor vs how much of the image is oversaturated (blown) vs the necessary shutter speed to minimize motion blur and/or camera shake vs the aperture required for the desired DOF / image sharpness. In other words, purely subjective.

Click to expand...

chuxter said:
Bingo! While demonstrating and explaining that "parlor trick" where noise appears to drop as sensitivity/gain/ISO is raised, is a fun exercise, it's not useful to the working pro! That entire thread should have been over on the Open Talk Forum...this thread too (I only put it here because that's where 2am_strobist's thread was).

It far more than a "parlor trick" and is very useful to know for images where the histogram is smaller than the DR, and for aiding in decisions on how much noise is acceptable vs how much of the image you can afford to oversaturate.

chuxter said:
chuxter said:

Is that "crackpotish" enough?

Click to expand...

chuxter said:
Your post was good, IMO. I didn't interpret arachnophilia's statement, " ...but the internet is full of crackpots... " to imply that any forum member was a crackpot...perhaps you were over-sensitive for some reason?

I was a bit condescending with that comment, however.

thomasnb · May 13, 2010

Firstly, thank you, Charles, for a useful set of images which shows that even with a D300 (which has a very flat read noise/ISO curve) the advice Emil and myself were giving - to use maximum exposure compatible with your pictorial needs and then maximum ISO compatible with that exposure, is indeed good advice.

chuxter said:
Look closely at both sets. The series with fixed illumination doesn't look as good as the series with variable illumination. Most photographers understand this. They wonder why Guillermo, Emil, and Thomas want it done the other way...other than just to prove a point. Is there some truth that has escaped lay photographers that this trio can illuminate?

Simply because you have onece again put up a straw man so you can knock it down. Guillermo, Emil or myself have never suggested 'doing it the other way'. We suggest maximum exposure compatible with your pictorial aims, maximum exposure means as much light as you can get (if you have control over the light), biggest aperture, longest shutter speed, indeed, the most light you can get on the sensor without having less DoF or more motion blur than you want. Which bit of that do you not understand?

chuxter said:
Perhaps we need to first agree on the meaning of the term "exposure". Our trio has a "sensor-centric" viewpoint...they think that "exposure" is the number of photons incident on the sensor...thus controlled by the classic duo, aperture and time, just like our grand-dads did it.

Only because that is the meaning of exposure, not just according to Gillermo, Emil or myself but also according to every optical or photographic textbook textbook and organisations such as the ISO, for whom meanings are rather important. We just go with what the term means instead of what it doesn't mean.

chuxter said:
Interestingly, they dismiss the light incident on the subject as being an "exposure" variable, even though it obviously affects the number of photons collected by the sensor.

That is simply a lie. None of us has dismissed the light incident on the subject. From my first post in the past thread, here:
http://forums.dpreview.com/forums/read.asp?forum=1014&message=35179767

chuxter said:
chuxter said:

'Exposure' means the amount of light (light intensity times shutter time) at the sensor. The only things that adjust exposure are the subject lighting, the aperture and the shutter speed.

Click to expand...

So, please retract your lie.

chuxter said:
They also scoff at the idea that sensitivity/gain/ISO adjustments have anything to do with "exposure"...but apparently digital scaling during PPing does?

I refer you to my quote above. So please retract that lie too.

chuxter said:
I think the two sides can be characterized as "sensor-centric" vs "camera-centric". The former group are what I call "sensor technologists"...the later group are what I call "photographers". I find that "sensor technologists" have great difficulty talking down...they basically don't do that. Instead, they talk to the "photographers" as if they were some other "sensor technologists". This approach invariably fails.

What you actually mean is that you completely and humiliatingly lost the argument, so you thought you'd start a new thread, tell a few lies about what others said in that thread in the hope that most readers wouldn't be bothered to track back to that thread and find out.

chuxter said:
It's not totally that "photographers" can't understand what Guillermo did or the conclusions that Emil and Thomas draw from that and preach to us...much more that "photographers" don't understand why , because it seems too far removed from reality. Photographers "know" in their souls that increasing sensitivity/gain/ISO increases noise...they live with that reality daily.

The may 'know' that, but it is wrong, as your own experiment shows. It is not increasing gain/ISO (the ISO control does not change the sensitivity of the sensor, although there is an argument that it changes the sensitivity of the system, by lowering the input referred noise floor) which increases noise, it is reducing the exposure. Which is why I say, as I have said all the way through this sorry debate, if you want to minimise visible noise, maximise the exposure.

chuxter said:
If there is some utility in the opposite approach that our trio takes, it isn't apparent what it is. Instead of digging in their heels and arguing incessantly, "sensor technologists" should find some different words and expressions that facilitate two-way communications...even if they are colloquial and subjective.

Or, you could just stop lying about what we said.

chuxter said:
Bottom Line: If "sensor technologists" expect to be understood and appreciated...if they expect to be judged a valuable asset by working "photographers"...they must devise a better way of communicating their obscure thoughts.

Bottom line, you're misrepresenting what others say in a pathetic attempt to have another bite at an argument that you lost completely.
--
thomas

thomasnb · May 13, 2010

arachnophilia said:
arachnophilia said:

It's not totally that "photographers" can't understand what Guillermo did or the conclusions that Emil and Thomas draw from that and preach to us...much more that "photographers" don't understand why , because it seems too far removed from reality. Photographers "know" in their souls that increasing sensitivity/gain/ISO increases noise...they live with that reality daily. If there is some utility in the opposite approach that our trio takes, it isn't apparent what it is. Instead of digging in their heels and arguing incessantly, "sensor technologists" should find some different words and expressions that facilitate two-way communications...even if they are colloquial and subjective.

Bottom Line: If "sensor technologists" expect to be understood and appreciated...if they expect to be judged a valuable asset by working "photographers"...they must devise a better way of communicating their obscure thoughts.

Click to expand...

i beg to differ. they're just flat out wrong.

Yes it is, but none of us said what he said we did. Chuxter was lying about that.

arachnophilia said:
the test is invalid, because they did it wrong. under-exposure, followed by compensation in post-production is the surest way to generate noise.

That's right. That is why you need both to ensure that you choose the maximum exposure available to you (it will usually be limited by your depth of field and motion blur requirements) and, at the same time, ensure that the image is not 'under exposed' - 'over exposed' (in vernacular terms - of course a misuse of the term 'exposure') is far better, so long as you don't clip the highlights. That means choosing the highest ISO that will not clip the highlights to go with your selected exposure.

arachnophilia said:
it's like push-processing from the film days. if you're under-exposing an iso 100 image by four stops and then brightening it, yes, you're going to see more noise than in a properly exposed iso 1600 image. but that's only proving the above. it's comparing apples to oranges. for a proper test, you have test two properly exposed images against one another. and guess what you'll see?

It's not really the same as film days. The noise you see in a digital image is directly the effect of the random arrival of individual photons (or at least the major part) whereas the grain you see in a film is to do with the structure of the emulsion. The real problem with push processing in film isn't that it produces more 'noise', but rather that it increases contrast to levels that rapidly become unusable. The problem with 'push processing' in digital is that in many digital cameras the late part of the read chain produces a lot of noise. Digital pushing amplifies that noise, whereas ISO adjustment applies the 'push' before the source of that noise.

arachnophilia said:
every photographer knows it in their souls because it happens to be correct. and anyone who knows anything about CCD or CMOS technology would agree. but the internet is full of crackpots.

Like the ones who lie about what others say because they can't bear to lose an argument.
--
thomas

chuxter · May 13, 2010

Crocodile Gena said:
Indeed, people "know noise when they see it", but if they don't know what noise is, then they can't know what causes it, how to minimize it, and what role the equipment plays in the whole affair.

"People" just keep the ISO at the base value. Sure, sometimes there is not enough light to take a picture at base ISO, but that doesn't mean that "people" don't know what to do. They either add more light and minimize noise or boost ISO to get more noise, but at least a picture of some sort. There is no secret that the sensor technologists can pass on to the "people" that will help them make a better choice in this instance!

Crocodile Gena said:
...I want to emphasize the different between NSR and SNR. The NSR (noise-to-signal ratio) is the reciprocal of the SNR (signal-to-noise ratio: NSR = 1/ SNR). The reason for using NSR for apparent noise rather than SNR, is because a lower NSR corresponds to lower apparent noise, and a higher NSR corresponds to greater apparent noise. For the SNR, it's exactly opposite, which is a bit confusing, just as using the term "noise" to refer to "apparent noise" is confusing.

Yes, it's confusing. But consider that there are other camera metrics that are also "up-side-down". Example: f/# goes down when the aperture goes up. Photographers have come to terms with "up-side-down" metrics and should have no problem with SNR. But I don't object to NSR.

Crocodile Gena said:
Crocodile Gena said:

Crocodile Gena said:

The primary source of apparent noise in an image is the light itself. High ISO images are more noisy not because the ISO is higher, but because the shutter speed is higher, which means less light falls on the sensor.

Click to expand...

Click to expand...

Crocodile Gena said:

BUT, the assumption that "much less light falls on the sensor" when using a higher ISO is an arbitrary one. That single assumption is what causes this line of thinking to be "wrong". If instead, you would maintain the same light intensity on the sensor as you raised the ISO, the result would be different. In fact, my pix showed that the higher ISO images, when the illumination is varied to keep it constant, have more noise.

Click to expand...

The point is simple: the apparent noise is primarily a function of the total amount of light that falls on the sensor. The role the ISO plays is primarily as an indirect means to setting the shutter speed, and had a relatively minor effect on the apparent noise itself.

There are 4 variables in "exposure": Illumination, Aperture, Time, and Gain. They all are independent variables. It's possible to fix any 3 and vary the 4th to get perfect "exposure". BUT other IQ aspects may suffer. As long as only "exposure" is reviewed, then any of these 4 variables can be the controlling variable. You simply can't say that " The role the ISO plays is primarily as an indirect means to setting the shutter speed. "...well, you can say that, but it's not true any more than saying that " The role ISO plays is primarily an indirect means to setting the aperture. "

Crocodile Gena said:
Crocodile Gena said:

Crocodile Gena said:

The sensor and supporting software do contribute to the apparent noise. However, their contribution is usually less at higher ISOs than at lower ISOs. But since the apparent noise at higher ISOs is intrinsically higher since less light is reaching the sensor, people incorrectly believe that the sensor "adds" more noise to the image at higher ISO settings, and fail to make the connection that it is instead the lesser amount of light getting to the sensor that causes the increase in apparent noise.

Click to expand...

Click to expand...

Your statement that " ...since the apparent noise at higher ISOs is intrinsically higher since less light is reaching the sensor... " is only true if you first make that arbitrary assumption that the A, T, & I is fixed. If you let any other variable change in opposition as the ISO increases, the amount of light reaching the sensor is constant.

Crocodile Gena said:
This opens the door for understanding what can be done at the hardware level to improve noise performance, and what cannot be done.

There are simply NO pro photographers on this Forum that give a whit " ...what can be done at the hardware level to improve noise performance... ". They want technologists to do that work for them! And it's better that it's done silently in the background. I'll agree that it's an interesting subject, but it needs to be on a different Forum. This forum is for Pros to discuss how to price their services/products and ask/answer technical questions on a level and manner that they can understand.

I started this thread to:

Show that they made an arbitrary choice to fix all other exposure parameters, other than ISO, then claimed a demonstration of that "proved" that "noise goes down as ISO goes up".
Show that if the demonstration had allowed two parameters to vary in opposite directions, the noise would have trended in the opposite direction.
Show that they "cheated" by not highlighting that the demonstrator had processed the images afterward to "correct the exposure".
Open a discussion (that you and I are having) about the technologists being incapable of writing in a way that lay "persons" can understand what they are saying...perhaps uncover a better way to communicate between levels?

Crocodile Gena said:
It far more than a "parlor trick" and is very useful to know for images where the histogram is smaller than the DR, and for aiding in decisions on how much noise is acceptable vs how much of the image you can afford to oversaturate.

Nah! It's a "parlor trick" intended to wow the natives. Trust me...there is nothing there that a working stiff needs to know. ;-)

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham

thomasnb · May 13, 2010

chuxter said:
Even gently implying that one of their approaches is irrelevant is like waving a red flag. Stating that they are totally wrong is like a photo-nuclear attack of the motherland. They are like bees...when one attacks, they all attack. Often they get so excited, they sting their friends. :-(

I think you have that completely the wrong way round. If you actually look hat the thread in question it goes like this.

OP asks why 'when i over or under expose by + or -1EV there is noticeable noise in the image'
Four posts are made in which it emerges that the OP has been in auto ISO mode. Two posters give incorrect accounts of why this occurs in auto ISO mode.
I make a post explaining in some deatil the sources of noise and why the OP is seeing an increase in noise both for over and under exposure. I finish with the (correct) advice 'If you're really concerned to minimise noise, the technique is simple. Always set the maximum exposure (largest aperture and shutter speed) that you can.' I also correct the two mistakes that have been made.
Will Rushmore enters the discussion responding to my correction. After a short (two post) semantic discussion, we agree
Dan DeLion responds to my correction, saying that the increase in noise is due to 'Over or under exposure results in moving the average pixel exposure up or down the sensors characteristic curve. To correct for this, the image's curves need to be readjusted. Noise, which is always there, is accented by the curve readjustments.' Since the sensor has a linear response, this is incorrect. I say so, there follows a 22 post sub-thread where dan insists that sensors are non-linear. Emil comes in at post 13 (hardly a quick response for a bee) giving hard evidence of the linearity of sensors. By this time the only argument Dan has is playground insults.
Simultaneously, there is another sub-thread going on. This one starts with me offering to rephrase anything I haven't explained clearly. This subthread goes on for 12 posts, mostly myself and Will Rushmore, where I have to continually restate what i said as Will says I said things i didn't. nonetheless the discussion remains civil.
AKfotogeek comes in, saying how useful my posts have been to him. This compliment seems to be too much for some to bear, because:
whoisjohngault comes in, saying he doesn't understand it at all. So I have another go at explaing.
Your good self comes in, starting with a sententious little exhortation to me to be open minded, and ending with a phrase tantamount to calling me a religious nut. Nonetheless, i take you seriously and post two considered an polite replies. There follows a five post discussion where you very quickly descend to derision and insults (which seems to be your usual style of argument) and also accuse me of misquoting and misunderstanding Emil
Emil enters the discussion to say that I have understood and interpreted him correctly. there follows a bizzare exchange in which Emil seems to change in your eyes from being your preferred expert to another religious nut. In the meantime, Emil has taken apart your line of argument completely.
Thereafter Emil and I get into a fairly long discussion with djjoofa, who starts off saying that the reduction in visible noise is due solely to digital tone adjustment. Finally he decides that what he had meant all along was that it was due to the combination of gain change and digital tone adjustment, which is what I'd said in the third post I'd made in the thread.

So, the history of the thread seems to be mainly me responding, initially by myself, to 5 stingers, of which you were the most angry and excited. So, who is it that's like a swarm of bees?

As for the open minded thing, you signed out with 'As long as Phil allows me to continue, I will argue against your screwy ideas whenever I notice that you guys are misinforming people...'. I observed at the time how open minded that was of you. Now it seems you've discovered that we were right all along. ROTFLMAO.

--
thomas

chuxter · May 13, 2010

thomasnb said:
Firstly, thank you, Charles, for a useful set of images which shows that even with a D300 (which has a very flat read noise/ISO curve) the advice Emil and myself were giving - to use maximum exposure compatible with your pictorial needs and then maximum ISO compatible with that exposure, is indeed good advice.

Thomas, I'm NOT going to respond to your questions, statements, & accusations. In the thread started by 2am_strobist, you posted 55 times, debating with Will Rushmore, WhoIsJohnGault, DanDeLion, Joofa, HarryThePhot, and others, including me. 55/150 is 37%! You don't know how to listen to others (when your mouth is open, your ears don't work). You often interpret benign remarks as insults. You try to dominate any conversation. I hope everyone ignores you. I am.

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham

thomasnb · May 13, 2010

...yes I am lying about that thread. I'm just going to keep on lying and misrepresenting what he said and when he says I have, I'll ignore it, making it look like his fault. You are pathetic.

chuxter said:
thomasnb said:

Firstly, thank you, Charles, for a useful set of images which shows that even with a D300 (which has a very flat read noise/ISO curve) the advice Emil and myself were giving - to use maximum exposure compatible with your pictorial needs and then maximum ISO compatible with that exposure, is indeed good advice.

Click to expand...

Thomas, I'm NOT going to respond to your questions, statements, & accusations. In the thread started by 2am_strobist, you posted 55 times, debating with Will Rushmore, WhoIsJohnGault, DanDeLion, Joofa, HarryThePhot, and others, including me. 55/150 is 37%! You don't know how to listen to others (when your mouth is open, your ears don't work). You often interpret benign remarks as insults. You try to dominate any conversation. I hope everyone ignores you. I am.

Edit: BTW, your own contribution to this thread is so far 8/19 = 42%, mostly lies about what I and others said. So what was that you just said about trying to dominate any conversation'?
--
thomas

ejmartin · May 13, 2010

It's really not so hard. The point of the little demonstrations in the other thread is to demonstrate why you raise the ISO when the light is low, rather than push a lower ISO in the raw conversion stage. It's because at the higher ISO's, the camera-generated contribution to noise is less relative to the given light signal. If it were otherwise, then one would be better off shooting at the lowest ISO always, and push-processing in raw conversion. But it's not, so (big surprise) you're usually better off raising the ISO when the light is low.

A side point to the little demonstrations is that the benefit to raising the ISO diminishes to a negligible amount at some (camera-dependent) ISO; beyond this point there is little difference between using any higher ISO setting, vs using that ISO and pushing in raw conversion to the desired output brightness. And though there is little difference wrt noise, use of the lower ISO leaves more highlight headroom should you want it.

For Guillermo's 350D, that point of diminishing returns is somewhere between ISO 800 and 1600; for the D300, it's considerably lower. If you behave yourself, I'll tell you how to determine this max useful ISO for any given camera without having to conduct a whole bunch of tests.

--
emil
--

http://theory.uchicago.edu/~ejm/pix/20d/

Nerd Wars...Part 1

Forum Pro

Forum Pro

Leading Member

Forum Pro

Leading Member

Veteran Member

Senior Member

Senior Member

Senior Member

Forum Pro

Forum Pro

Forum Pro

Senior Member

Forum Enthusiast

Forum Enthusiast

Forum Pro

Forum Enthusiast

Forum Pro

Forum Enthusiast

Veteran Member

Similar threads

Keyboard shortcuts