ISO is not real In Digital Camra's

[donny1963]

, all it is, is applied GAIN to amplify the Exposure for lower lighting

I've tried to resist commenting but as others have already pointed out your understanding is not fully correct. Your statement above implies that all ISO adjustments are gain obtained by increasing voltage to the sensor to increase signal sensitivity (that's the definition of gain). The fact is that's only partially correct dependent on camera manufacturer.

There's actually three methods of adjusting ISO: native, amplified, and simulated. "Native" ISO is the ISO setting that does not require the camera to increase the voltage to the sensor to increase ISO. My K1MII has "native" ISO 100-819200. In years past "native" referred to the base ISO that the camera processor was able to handle without increasing voltage to the sensor or gain as you call it, but as processors and sensors have advanced the need to apply gain has decreased. Amplified ISO is an ISO that requires an increase in voltage to the sensor to achieve an increase in ISO. This is where you are correct in your OP statement, but this is also old technology. There might still be some manufacturers out there that do this but I'm not aware of them. Finally, Simulated ISO is when the camera uses a software algorithm to simulate even higher (or lower) ISOs. My understanding is it's applied after the shutter clicks, so if my understanding is correct then that would be another thing you were partially correct on.("it's applied AFTER THE PICTURE IS TAKEN"). The Nikon D850 for example has "native" ISO 64-25600, and software enhanced from 32-102400.

So what appears to be happening are mixed assumption partially correct and partially false. I suspect the videos you've been basing your assumptions on were either old, or the individual making them was basing their claims on older technology, and as shown above, blanket statements on ISO are not correct in today's world.

NO IT's not increasing voltage to the sensor, has nothing to do with ISO,
ISO is applied after the Image is taken.. NOT at the time the shot is taken, so voltage to the sensor has noting to do with ISO settings..
ISO is Applied Gain AFTER, I repeat AFTER the image is Taken..
the shot is taken from the sensor, to the Analog Gain, then to the AD converter, then ISO is applied as you see in the diagram..

So your wrong, and looks like YOUR your understanding is not fully correct at all..

 

[tirediron]

Well, I can understand how, based on that very simplified block-diagram how such a hypothesis could be formed, but the problem is that diagram is missing about 80% of the components that make up a digital camera, principally the power circuits; the inclusion of which would clarify the signal amplification issues causing most of your consternation. Out of curiosity, why do you believe that the [non-existent] ISO is created only in the digital amplifier and not the analogue amplifier?

 

[Overread]

Ever since Sony-made sensors in Pentax,Nikon,and Sony cameras (and in othe branded cameras) became common,and ISO invariance entered the lexicon, some of the ways we used to use ISO as an image quality control were changed profoundly.

Aye these sensors start to kick things like "expose to the right" somewhat to the edge. However I also wonder if such theories might remain long term even if just to provide a means to use the histogram and have a "target" to aim for in ideal conditions. Or has a new theory replaced it? I'd still wager the average photographer prefers to capture it as close to "right" in camera (based on given conditions and intentions for the photo). If not just to cut down on editing, but also for a more pleasing instant result.

Also I don't think either Canon or Sony sensor technology is the same as those fovorian things (spelling) that Sigma was playing around with (though which have never really gotten off the ground in a big way - or if they have only within a niche market).

 

[donny1963]

Well, I can understand how, based on that very simplified block-diagram how such a hypothesis could be formed, but the problem is that diagram is missing about 80% of the components that make up a digital camera, principally the power circuits; the inclusion of which would clarify the signal amplification issues causing most of your consternation. Out of curiosity, why do you believe that the [non-existent] ISO is created only in the digital amplifier and not the analogue amplifier?

It's not intended to be a full diagram of the entire works of the camera..
I didn't build the diagram, it's a pdf from the camera manufacture..
The Blue and red lines was placed by Ken, for illustration,
The diagram was created by the camera manufacture..
So if you have problems with the diagram, then take it up with Sony..

it's a diagram of the Sony Xmor-RS sensor workings..
If you don't believe me here is the link to the sony website with the diagram..

https://www.sony-semicon.co.jp/products_en/IS/sensor1/img/products/ProductBrief_IMX351_20171109.pdf

 

[Designer]

that was one of the things i stated if you have a full frame camera lets say Sony, and i got a Nikon Full frame, ..

there is really no need to shoot at way high ISO settings really because even if the picture is under exposed some you can boost it up in post, ..

1. Sensor size, per se, has nothing to do with sensitivity. Pixel size and density, yes, but not the overall size of the sensor.

2. The optimum amount of data is obtained at sensor saturation (without clipping). An underexposed data set will not produce the same image as one in which the sensor was saturated.

 

[limr]

And exactly how does all of this bickering help people take better pictures?

 

[PaulR70]

It is very common sense that all three work together, shutter speed, aperture and ISO for setting the exposure. For instance you shoot a scene and set exposure your ISO is 200, to increase the exposure by one full stop you can increase the ISO to 400 or open the aperture one full stop or decrease the shutter one full stop and you will see that each on has the effect on the metered exposure.
I agree with some of the others get out and take pics and experiment on your own to see how these settings effect exposure.

 

[tirediron]

If you don't believe me here is the link to the sony website with the diagram..

https://www.sony-semicon.co.jp/products_en/IS/sensor1/img/products/ProductBrief_IMX351_20171109.pdf

Oh, I believe the diagram is real, but you're postulating a theory and using it as supporting evidence. The problem is, your evidence is incomplete. I'm sure if you contact Sony, they will tell you that diagram is not intended to explain signal amplification. Using it to support your argument is like holding up a block diagram of an automobile which doesn't show the fuel system and stating, "This car doesn't need gasoline; see... there's no fuel tank in the diagram!" In simple terms, the evidence you're presenting has no relevance to the argument.

That said, I notice that you failed to respond to my query about why you believe that ISO is "controlled" only by the digital amplifier in that diagram and not the analogue? They seem to me (according to that diagram) to be part of the same circuit...

 

[tirediron]

And exactly how does all of this bickering help people take better pictures?

It doesn't... but it's fun!

 

[Fujidave]

Just had a read through all of this thread, and it was really great fun too. I`m now going to listen to the song Daydream Believer, a great song too

 

[Ysarex]

It is very common sense that all three work together, shutter speed, aperture and ISO for setting the exposure. For instance you shoot a scene and set exposure your ISO is 200, to increase the exposure by one full stop you can increase the ISO to 400 or open the aperture one full stop or decrease the shutter one full stop and you will see that each on has the effect on the metered exposure.
I agree with some of the others get out and take pics and experiment on your own to see how these settings effect exposure.

No, this is completely wrong. Increasing ISO from 200 to 400 will not increase exposure by one full stop under any condition.

Joe

 

[limr]

And exactly how does all of this bickering help people take better pictures?

It doesn't... but it's fun!

It's actually quite tiresome.

 

[Soocom1]

OK..
So I am going to throw in two cents worth on this.
But understand that I think the argument is a non-sequitur because digital and film are two different things doing similar functions.
But moreover, the whole ISO thing is in my opinion a fallacy argument.

like any map it is REPRESENTATIVE ONLY and not actual.
(look up map projection and distortion-distance).

The intent of the ASA, DIN, ISO or whatever acronym you want to use, is to place the sensitivity of FILM into a category of light-absorbing qualities that remain relatively consistent.

let me remind everyone who is an old film hag, that how many times were we lectured on the various different developers and their uses?
How many times were we warned not to develop certain types of B&W in certain developers because it would kill grain and contrast?
Do I need to remind said film hags that temperature of the developer and the film also plays a huge role in said development?

The ISO/ASA/DIN numbers were designed to bring consistency to the manufacturing process so it could be qualified, quantified, classified and controlled for various uses...

Ergo Soft portrait v. scientific micro photography.... Anyone remember that argument?

The ISO today has a similar role (sensitivity to light numbers) but are typically rated at levels and based on a completely different mechanical aspect that film. So the ISO numbers are designed to bring the photographer into a position of having similar results with digital using a similar numbering system to that of film, but the two are two totally different mechanical processes that are not comparable.
They just do the same thing differently.
Think V type engine to a wankle.

 

[TCampbell]

There's a mix of some "right" and some "wrong" info in the thread.

ISO & Gain

ISO is basically the label used because photographers are familiar with it. Digital sensors don't have a concept of ISO ... the camera has a concept of "gain" and "offset" (we'll ignore offset ... that'll just create more confusion). The camera controls which allow adjustment of "ISO" setting are really just a photographer-friendly user-interface to control the more technical aspects of the camera (which doesn't really use ISO -- but it will ultimately correlate your ISO selection to some application of gain. I'm not being specific about *how* gain is applied because it will vary based on camera make & model... but one way or another, it will ultimately apply gain.)

Digital sensors do not have adjustable sensitivity. At a low level, photons are collected and converted to electron voltage. These accumulated volts are capped based on the full-well-depth of the sensor (greater "full well depth" results in better "dynamic range" because you have a greater range of possible electron volts.)

When the exposure is complete, various types of gain may be applied. There may be analog gain (prior to analog-to-digital-conversion or ADC). The ADC will convert electron volts to ADUs (Analog Digital Units) (sometimes also called a DN or "digital number"). This is not necessarily a one-to-one mapping (and usually is not). After ADU conversion, a digital gain can be applied (essentially multiplying the digital values by a scaling factor). The result of the combination of any analog gain (if any) combined with digital gain (if any) are combined and this total "gain" (which ultimately results in digital values) can be stored as a file that has had "ISO" applied to it (so ISO is really just a label used to express the application of gain and is never applied until after the exposure has completed).

I'll hint at another area ... which is that often some application of gain is required. I mentioned that the ADC from electron volts is usually not a one-to-one mapping. ADUs are integer values -- not fractions. So heres' a problem...

Suppose we have four pixels and they have electron volt levels of .2 e-, .4 e-, .7 e-, and 1.4 e-. In a "one to one" mapping of electron volts to ADUs, these would correspond to the ADU values of 0, 0, 1, and 1. In reality we know that .4 is double .2 and 1.4 is double .7. Yet both .2 and .4 map to 0 ADU and both .7 and 1.4 map to 1 ADU. So we lose tonality. If we apply some gain (suppose we multiply everything by 10) then we can distinguish between each one. Sometimes a little gain is a good thing -- it's not all evil. (but having ADU's map to 0 is evil ... and that's where "offset" comes in. I'm leaving offset out of the conversation except to allude that the thing does exist and is a requirement in digital cameras.)

There's more than one kind of "gain". As an example... my old Canon 60Da will apply only analog gain at very low ISO values (e.g. 200 & 400). It applies incrementally more gain as you increase the ISO... until somewhere around ISO 800. At this point it reduces the amount of incremental analog gain and starts shifting to digital gain. At higher ISO's the incremental gain is mostly just digital. So depending on the "ISO" setting, you may get no gain, some analog gain, or some analog gain AND some digital gain. Some cameras mostly do digital gain with very little analog gain. Specifically what the camera does will depend on the make & model.

So it is accurate to say camera sensors do not have ISO. (ISO is a feature of the user-interface). But that ISO does map to some amount of gain (however the camera decides to apply it). A key difference is that with film, a change of film type can result in a change in sensitivity. One film type really may be more or less sensitive than another film type. With a digital camera, your sensor is whatever it is ... and remains that way (adjusting ISO doesn't change anything in the sensor.)


Noise

Increasing gain (regardless of analog or digital) DOES result in noise being more noticeable (if when shooting RAW you do not see a relationship in noticeable noise corresponding to an increase in "ISO" then your camera is "cooking" its RAW files. This is not unusual. Most cameras seem to "cook" the RAW data. Some are pretty aggressive about it... some are a bit more subtle.)

How would you know?

You can generate BIAS frames by leaving the lens cap on the camera (or body cap) and shoot a series of very fast exposures (the shortest possible exposure your camera will allow. Many CCD & CMOS sensors used for astro-imaging actually have a setting to generate BIAS files. A BIAS file is basically a zero-duration image -- meaning it "powers up" the sensor and then immediately performs a "read-out" without actually doing an exposure. On a DSLR, you can't dial in a zero-duration exposure. But you can set the shutter speed to something very short (e.g. 1/4000th sec) and take 50 shots at base ISO with the lens cap on. Those are basically "BIAS" frames. An analysis of these frames would reveal about how much electron voltage is ever-present on the chip just by virtue of the fact that it was "powered up" and then "read out" without ever actually taking a photo.

You can also generate "dark frames". Dark frames have a duration. They are basically normal exposures.... except with the lens-cap still on.

Even though you know that no actual "light" (photons) were collected, the pixel values in these exposures wont actually be zeros... they'l be some non-zero value. Also... they all collected the same amount of light (none) and on the sensor with the same amount of power applied to power-up the sensor. You would think they would all have the SAME non-zero value. But they wont. Each pixel might be about the same... but subtle variations from pixel to pixel.

Any amount of real photons collected represent "signal". The goal of these tests are to not receive any signal.

You would collect all this data and then calculate the statistical mean & standard deviation. It's the "deviation" from pixel to pixel that represents your camera's "noise" (and it is ever-present in every exposure ... even exposures where you don't see it. There is no such thing as an exposure that doesn't have noise.) There are several reasons why it's impossible to eliminate all noise (though it it is possible to reduce it). Keep in mind that one source of noise comes from quantum effects and there's absolutely no getting rid of that -- it is the nature of how the universe works. But things like read-noise and dark current tend to be high on the list of things that contribute to noise (dark current is basically heat ... in the electromagnetic spectrum everything is a "photon" and your sensor is basically a photon counter. If the sensor gets warm, that heat is basically infrared light.. and those are photons. They're not photons from the scene you were trying to photograph, but they're still photons. This is why many dedicated astrophotography imaging sensors have cooling units to keep the chip cold. Two of my imaging cameras have cooling systems that can drop the chip temperature to about 40°C below whatever the ambient temperature is. (If it's 80°F / 26°C, I can drop the chip to 8°F / 13°C to try to fight back the noise from heat.)

When you "amplify" the data (regardless of how you do it) you amplify everything... the stuff you do want (signal) and the stuff you don't want (noise). In other words if we multiply every value by 10x... then the "noise" gets 10x stronger and the "signal" gets 10x stronger (the ratio is maintained). But the differences between the darkest pixels and lightest pixels is increased and now the noise starts to become noticeable. This will always happen and again... if you don't see this happening then either you've done something wrong OR your camera is cooking the RAWs.

The cameras only real controls are aperture size and exposure duration. ISO (or any other label you want to use to describe applied gain) is always applied only after the shot is completed (hence it isn't a "real" part of exposure ... it's more of a post-processing step.)

I did see in the video the example that you could just apply the exposure increase in post-processing software. There are a lot of variations in how to do this. While you could apply the gain linearly ... you could also use a non-linear process in an effort to protect highlights from being blown.

In other words, we can use simple math idea to express the idea of increasing the exposure ... but the actual implementation might not necessarily be a linear process. Or another way to think about it is what the CAMERA does to increase gain might not be the same algorithm that your COMPUTER uses to to increase gain.

The bottom line is that the camera exposes a user-interface to the photographer and the user-interface includes terms like "ISO". But at a technical level, the chip itself doesn't have a concept of "ISO" ... ISO is some process (and I say "some" process because the exact process can vary from camera to camera) of applying gain to the data after the shot is completed.



So all of this is whatever it is.... all the being right or wrong won't actually change how the camera really works. Knowing the causes of noise and what ultimately elevates the noise to the point of being noticeable (and knowing how to post-process it with common tools) can help control it to the point where it's really no big deal. I don't actually spend too much time worrying about noise in my images since there are some good techniques to handle it.

 

[Ysarex]

OK..
So I am going to throw in two cents worth on this.
But understand that I think the argument is a non-sequitur because digital and film are two different things doing similar functions.
But moreover, the whole ISO thing is in my opinion a fallacy argument.

like any map it is REPRESENTATIVE ONLY and not actual.
(look up map projection and distortion-distance).

The intent of the ASA, DIN, ISO or whatever acronym you want to use, is to place the sensitivity of FILM into a category of light-absorbing qualities that remain relatively consistent.

let me remind everyone who is an old film hag, that how many times were we lectured on the various different developers and their uses?
How many times were we warned not to develop certain types of B&W in certain developers because it would kill grain and contrast?
Do I need to remind said film hags that temperature of the developer and the film also plays a huge role in said development?

The ISO/ASA/DIN numbers were designed to bring consistency to the manufacturing process so it could be qualified, quantified, classified and controlled for various uses...

Ergo Soft portrait v. scientific micro photography.... Anyone remember that argument?

The ISO today has a similar role (sensitivity to light numbers) but are typically rated at levels and based on a completely different mechanical aspect that film. So the ISO numbers are designed to bring the photographer into a position of having similar results with digital using a similar numbering system to that of film, but the two are two totally different mechanical processes that are not comparable.

Right, they are not comparable.

Film: ISO 5800:1987 https://www.iso.org/obp/ui/#iso:std:iso:5800:ed-2:v1:en
Digital: ISO 12232:2019 https://www.iso.org/obp/ui/#iso:std:iso:12232:ed-3:v1:en

It is possible to read them. Problem is ISO needs (appropriately) to be funded and so the standards are for sale at about $120.00 a pop. Lot of money to spend to settle an troll argument.

In this case however there's a backdoor of sorts. ISO did the right thing in establishing the standards and worked closely with the industry. In the case of digital they relied on CIPA which in effect wrote the standard for them. CIPA fortunately publishes all of their technical documents on the web for free and here's the document the ISO standard is based on: http://www.cipa.jp/std/documents/e/DC-004_EN.pdf

I already posted that earlier as proof for donny that he's clueless but hey, it's a technical paper.

Confounding the likes of Tony Northrup and Ken Wheeler the digital ISO standard was unable to bring the industry into unified agreement and so it in fact elaborates variations in methodology to determine ISO. Derrel earlier mentioned DX0. This is informative: ISO sensitivity | DxOMark In that document DX0 explains how they treat ISO. What they do is noted in the standard but not embraced by any camera manufacturer. In other words all of the camera manufacturers use a different method to determine ISO values than does DX0.

Recently brought up again by Tony Clickbait this is a commonly reoccurring topic that's spreading misinformation all over Youtube (there isn't enough yet). The camera manufactures tell us what they're doing. Look in the EXIF data of your image and you'll see either SOS or REI noted as the method used to determine ISO. Nikon and Canon both use REI while Fuji leads the group using SOS (you can find both in that CIPA doc).

Nobody is lying or trying to deceive us. All of the camera manufactures are in close compliance doing what they tell us they're doing.

And ISO is not applied gain. The standard leaves implementation entirely up to the camera manufacturer. Gain is a problematic word to us in describing ISO implementation because it carries connotations from audio engineering that are not entirely appropriate. While I'm writing this TimC posted a long response and provided some explanation of gain/ISO. He did a good job. "Gain is a measure of the ability of a two-port circuit (often an amplifier) to increase the power or amplitude of a signal from the input to the output port by adding energy converted from some power supply to the signal." And TimC noted that what ISO does to the sensor signal is accomplished in multiple ways. Electronic gain is one method, digital scaling which is also used really is something else. My hesitation in just accepting the term gain is because it implies for a lot of people (audio connotation) an increase in sensitivity. And as TimC very correctly pointed out, "Digital sensors do not have adjustable sensitivity."

Back a few pages here TimT got it right and I'll quote: "ISO on a camera is simply a calibration of the middle grey tone to an RGB co-ordinate."

Joe

They just do the same thing differently.
Think V type engine to a wankle.