Luminous Landscape Forum
Site & Board Matters => About This Site => Topic started by: amolitor on November 18, 2015, 10:24:18 am
-
I have two very short responses:
ETTL?
and
Quantization Noise
-
Perhaps someone can explain the real world implications of this concept of ISO Invariance in setting exposures. Does it not permit us to be more careful in protecting highlights from blowing at the possible expense of underexposing by 1-3 stops, knowing that we can bring up the shadows pretty cleanly? To put it differently, why do ETTR and run ANY risk of blowing highlights?
-
why do ETTR and run ANY risk of blowing highlights?
ETTR, ETTL - do not forget about rawconverter and camera profile... they may or may not wreck havoc on your left or right side ;D ... because in some converters you push or pull exposure slider after a lot of things are done to the data
-
A bit of an old story :) The first camera with that property or at least very close to was the Nikon D3X back in 2008. Canon still hasn't done it. Btw. in these forums you cannot really admit shooting a Canon ;)
-
The first camera with that property or at least very close to was the Nikon D3X back in 2008.
I 'd assume that MFDB with just one real ISO and the rest nominal ISO done by tag in raw were before 2008... no ? or may be even Sigma's FOVEON (the original) based cameras
-
I've seen this idea espoused elsewhere and, fundamentally, it's not a very good idea.
By all means, preserve the highlights you're interested in preserving, but you're better off using the analog amplifiers to whatever extent you can than you are amplifying in the digital domain, because, quantization noise. This will manifest as posterization in some cases, which can be repaired with some judgement, and will result in loss of low-contrast fine detail.
Generally, you don't lose much, and whatever you do lose was likely to be lost somewhere in the pipeline unless you're fanatically careful, but you do lose stuff and you wind up making extra work for yourself.
-
Was I supposed to think the 'pushed' version was acceptable? It looked much more noisy than the 12800 version to me.
Using higher ISO with modest underexposure would be a better bet?
Kirk
-
Was I supposed to think the 'pushed' version was acceptable? It looked much more noisy than the 12800 version to me.
Using higher ISO with modest underexposure would be a better bet?
Kirk
If anything I saw the pushed version worse than 12800. The contrast is different which make it a bit difficult to compare. But to me the 12800 looks cleaner than the 6400 pushed one.
-
We had this discussion a year ago. I posted an example (http://forum.luminous-landscape.com/index.php?topic=94724.msg780456#msg780456) showing Canon files at 5000 ISO and the other one at 500, i.e., 3 and ⅓ stops pushed. The higher-ISO one looked better.
-
In my book, ISO 640 pushed 5 stops equals ISO 20480 (640*2^5)
Am I missing something?
-
As mentioned in the article (but no harm is saying it again), the intent wasn't to espouse this approach (and yes, the low ISO / pushed image is worse), but simply to draw the concept to the attention of an audience that may not be as well informed and technical as some on this forum.
Michael
-
.. to draw the concept to the attention of an audience that may not be as well informed and technical as some on this forum...
Ha! That sounds like the Adobe's explanation for the Import redesign ;)
EDIT: Just to make shure this isn't interpreted as criticism; articles like this (back-to-basics), definitely make more sense than what Adobe did (dumbing down)
-
In my book, ISO 640 pushed 5 stops equals ISO 20480 (640*2^5)
Am I missing something?
yes, raw files.
-
As mentioned in the article (but no harm is saying it again), the intent wasn't to espouse this approach (and yes, the low ISO / pushed image is worse), but simply to draw the concept to the attention of an audience that may not be as well informed and technical as some on this forum.
Michael
I think it could be interesting to compare the result of denoising algorithms on pushed and unpushed images.
I don't know anything at all on those algorithms but* I think there could be advantages for them to have an unpushed image because more "noise details" would be available and, therefore, a more precise determination of the noise characteristics.
This could lead to a better handling of the noise.
But, again, I'm just babbling of something I know nothing about.
-
I have had a conversation with a chap who felt that he got better results with pushing and de noising. I disagreed, but they're not my pictures and it's not my call.
The "shape" of the noise is going to be different. You may or may not like the results more.
But you absolutely will unrecoverably lose certain low contrast detail. Which may or may not matter to you.
I was mainly surprised to see it on LuLa since it is essentially the opposite of ETTR.
-
I was mainly surprised to see it on LuLa since it is essentially the opposite of ETTR.
what is opposite of ETTR if we do not have enough /for whatever reason - even for the sake of experiment/ exposure (aperture and exposure time) to saturate the sensor @ lower gain /known as "ISO"/ settings ? in such situation it might be beneficial in many case based of course on a specific camera/sensor technology to use a higher gain... even when we saturate the sensor we still might do this /use higher gain still/ if clipping (or situation near clipping) does not ruin anything important vs pluses from higher gain possible positive effects on readout related noise
-
I think it could be interesting to compare the result of denoising algorithms on pushed and unpushed images.
I think it is interesting to use a raw converter with a proper raw conversion chain (that is exposure correction before WB/curves/color transform)
PS: and before demosaick too (that's obvious).
-
I was mainly surprised to see it on LuLa since it is essentially the opposite of ETTR.
Well, Michael has long been an advocate of ETTR…but this is merely an informative piece.
With my current EVF cameras I just expose for a nice JPEG, which I'm comfortable using in most cases where electronic display is the intended viewing medium. For printing or when I want to more forcefully adjust tonality I use the RAWs.
If cameras offered smart ETTR metering, and smart JPEG processing to match, I'd use that.
High-bit JPEG has been around for awhile now. Why does no camera (that I know of) offer it as an option?
-Dave-
-
I've read the article and my self I'm always trying to expose well enough, close to rock 'n roll. Why do we want to do this tests at all? In my world, just expose properly for every shoot and do a good picture. That sade, I just got my A7R II and the quality works for me. I don't push the exposure 5 stops usually. ;)
-
In my world, just expose properly
the question is not about the exposure, but rather about the gain (aka ISO) - which is not a part of exposure, but a part of your decision about exposure that you make before exposure starts and effects are applied after exposure ends...
note however that some people might also argue that for some sensors while gain is not part of exposure (aperture and exposure time = how much light you let in) it might affect how much of the said light during the exposure can ultimately be converted to photon-electrons (for example changing the gain can change sensel' well capacity and while again the gain is not part of exposure /in terms of how much light you let hit the sensor assembly surface/, it affects the the things rather during exposure process itself - but then again this mostly can be considered as a pre/post exposure effect for practical purposes)...
so to expose properly we need to know how gain works on our camera and not only that - but also if you have a particular raw converter in mind then how it handles the data, potential to screw things is enormous !
-
As mentioned in the article (but no harm is saying it again), the intent wasn't to espouse this approach (and yes, the low ISO / pushed image is worse), but simply to draw the concept to the attention of an audience that may not be as well informed and technical as some on this forum.
Michael
ISO 640 pushed 5 stops is 20480 as shown below. You might repeat the experiment by comparing the 5 stop push of the shot taken at ISO 640 to an exposure at 20480 or the image taken at ISO 12800 to the ISO 640 shot pushed 4 stops.
Regards,
Bill
-
ISO 640 pushed 5 stops is 20480 as shown below. You might repeat the experiment by comparing the 5 stop push of the shot taken at ISO 640 to an exposure at 20480 or the image taken at ISO 12800 to the ISO 640 shot pushed 4 stops.
Regards,
Bill
also the theory is that Sony starts may be some extra processing for raw data @ ISO25600 (and around - may be 12800 affected somewhat) under some conditions... consider histograms of darkframe shot @ ISO25600 from A7R2, but with different shot parameters (and lenses) = see the difference in green channels :
1) (http://s17.postimg.org/6gccm8pfj/alex.png)
2) (http://s18.postimg.org/pqkqrx89l/DSC09815_Sel_4119_2789_366x339.png)
-
In my book, ISO 640 pushed 5 stops equals ISO 20480 (640*2^5)
Am I missing something?
Well for my part I did not even calculate 5 stops up from 640. You are right and therefore it should be no surprise that the pushed ISO 640 looks worse than the ISO 12800. The comparison would have been better done at ISO 25600. Anyway this is old news for the more technical savvy people and still new for most people :)
-
By all means, preserve the highlights you're interested in preserving, but you're better off using the analog amplifiers to whatever extent you can than you are amplifying in the digital domain, because, quantization noise.
Quantization is a property of light, because photons. That is the reason, I do not understand what we are talking about here.
Let me give an example. I actually own a truely ISO invariant camera. It is truly ISO invariant, because it does not have an adjustable analog amplifier. The camera is an H4D-50, but other MF cameras work in exactly the same way.
The camera works in the following manner:
-it has a 16 bits ADC, meaning 2^16= 65536 levels
-interestingly, the full well (the maximum number of photons-electrons that fill the pixel) is also around 65000
-therefore the values that are recorded by the camera are something like 0 photon-electron, 1 photon-electron, 2 photons-electrons, 3 photons-electrons, ..., 65534 photons-electrons, end.
-the quantization of the ADC corresponds to the quantization of light.
Just for the record: that camera has poor low-light capabilities, for the following reasons:
-it has poor quantum efficiency, meaning your will need something like 2-3 photons to get one photon-electron
-it has high internal noise, meaning something like 8-16 photons-electrons may be added to each pixel, randomly.
But this is irrelevant to the subject of ISO invariance and quantization noise. And, if we consider that we will get 2^16 photons-electrons at ISO 100, at high ISOs, say ISO 6400, we only get 2^10. That is the problem of high ISOs: there is less light, so there are less photons. So when we are using high ISOs, we will only get values between, say, 1 photon-electron and maybe 1024 photons-electrons. The data is already very coarsely quantized in the photons domain.
So what exactly are we talking about here when we discuss quantization? Is what happens in the digital domain relevant, when the data is already quantized coarsely because of the properties of light?
-
Quantization noise means something very specific in digital signal processing.
Suppose you have light levels present in the scene that would, in the digital domain, be rendered as: 0.0, 0.25, 0.3, 0.7, 0.75, 0.9, 1.1 or something.
Since in the digital domain you only have integers, you're going to get something like 0, 0, 0, 1, 1, 1, 1 out. That difference between the analog levels and the digital ones is quantization noise.
Consider amplification. So keep things simple, let's amplify 10x. In reality, increasing exposure is done with a curve, not a constant, and 10x is unrealistic, but the ideas are the same:
Underexpose and push it in post (after the ADC): 0, 0, 0, 10, 10, 10, 10
Increase ISO instead: 0, 2, 3, 7, 7, 9, 11
By pushing in post, you're amplifying quantization noise, which shows up as posterization, and loss of detail in low contrast areas.
-
Hi,
On the other hand, the signal is inprecise. We are talking about just a few photons, that is particles of light. So assume 16 photons on average. Natural variation would be something 4 photons (standard of deviation would be 4 photons, by Poisson statistics). Now, let's assume 14 bit representation, that is 16384 values. Let's assume something like 64000 electron full well capacity and one electron per photon. So 16 photons -> 16 electrons. By and large 64000/16000 -> 4, so our 16 photons would correspond to 12 - 20 photons, natural variation taken into account. The range would result in digital numbers 3-5, and it would be decently resolved.
Es shot noise that is noise in light arriving at each pixel varies it will mask (dither) quantisation noise.
Another point may be that the A7RII is not iso invariant. It makes used of something often called "The Aptina Patent". In modern sensors a small capacitor is connected to each pixel, in order of increasing full well capacity. Having a large FWC is good for noise, and also give low ISO capability. But, a large FWC yields a low output voltage. So Aptina developed and patented a technology to connect the capacitor to the photo diode trough an extra transistor. At high ISO the transistor is closed and FWC is reduced resulting in cleaner readout.
Jim Kasson has analysed this, and found that this is happening at ISO 640 on the A7rII. So he recommends to shoot either 100 or 800 ISO. The reason he uses 800 ISO and not 640 ISO is that 100 to 800 ISO is three full EV steps. Esaier to set 800 ISO on the camera than 640 ISO.
Best regards
Erik
Best regards
Erik
Quantization noise means something very specific in digital signal processing.
Suppose you have light levels present in the scene that would, in the digital domain, be rendered as: 0.0, 0.25, 0.3, 0.7, 0.75, 0.9, 1.1 or something.
Since in the digital domain you only have integers, you're going to get something like 0, 0, 0, 1, 1, 1, 1 out. That difference between the analog levels and the digital ones is quantization noise.
Consider amplification. So keep things simple, let's amplify 10x. In reality, increasing exposure is done with a curve, not a constant, and 10x is unrealistic, but the ideas are the same:
Underexpose and push it in post (after the ADC): 0, 0, 0, 10, 10, 10, 10
Increase ISO instead: 0, 2, 3, 7, 7, 9, 11
By pushing in post, you're amplifying quantization noise, which shows up as posterization, and loss of detail in low contrast areas.
-
Hi,
As a rule, quantisation does not matter as it is masked (dithered) by shot noise coming from the physics of light.
This is an interesting article on some of the issues: http://www.strollswithmydog.com/how-many-bits-to-fully-encode-my-image/
Best regards
Erik
Quantization is a property of light, because photons. That is the reason, I am not I understand what we are talking about here.
Let me give an example. I actually own a truely ISO invariant camera. It is truly ISO invariant, because it does not have an adjustable analog amplifier. The camera is an H4D-50, but other MF cameras work in exactly the same way.
The camera works in the following manner:
-it has a 16 bits ADC, meaning 2^16= 65536 levels
-interestingly, the full well (the maximum number of photons-electrons that fill the pixel) is also around 65000
-therefore the values that are recorded by the camera are something like 0 photon-electron, 1 photon-electron, 2 photons-electrons, 3 photons-electrons, ..., 65534 photons-electrons, end.
-the quantization of the ADC corresponds to the quantization of light.
Just for the record: that camera has poor low-light capabilities, for the following reasons:
-it has poor quantum efficiency, meaning your will need something like 2-3 photons to get one photon-electron
-it has high internal noise, meaning something like 8-16 photons-electrons may be added to each pixel, randomly.
But this is irrelevant to the subject of ISO invariance and quantization noise. And, if we consider that we will get 2^16 photons-electrons at ISO 100, at high ISOs, say ISO 6400, we only get 2^10. That is the problem of high ISOs: there is less light, so there are less photons. So when we are using high ISOs, we will only get values between, say, 1 photon-electron and maybe 1024 photons-electrons. The data is already very coarsely quantized in the photons domain.
So what exactly are we talking about here when we discuss quantization? Is what happens in the digital domain relevant, when the data is already quantized coarsely because of the properties of light?
-
Esaier to set 800 ISO on the camera than 640 ISO.
yes, something along the lines of using a dial that selects only even ISOs (to be able to do dial it faster vs 1/3 for example) and auto ISO only using even ISO numbers...
-
Yeah, you can make quantization noise go away with dithering, and there's quite a bit of natural dither built into the system.
I don't know how it works on the Sony, but if you want to test, find yourself a very low contrast test target of some sort with a fair bit of fine detail. At some point you will see the fine detail "almost vanish" in the sense of the pattern will be vaguely visible in the noise. You'll get a sort of "modulated noise" sort of thing.
If it looks the same either way, great. In the samples I've seen (from other sensors in other cameras) the underexpose+push picture just had mush where the increase-ISO shot had the "modulated noise" appearance, at that limit.
-
Quantization noise means something very specific in digital signal processing.
I know that, thank you.
But, if memory does not fail, you are a mathematician. You should know the difference between a discrete and a continuous signal. Digital quantization only makes sense when we have a continuous signal as source.
Here we have at source a discrete signal (photons) and out digital levels are actually counting individual photons.
-
Yes, yes, I am aware that it's all discrete at the quantum level etc.
I've seen quantization noise in pictures taken with the "underexpose and push" method. It was quite close to the noise floor presented by everything else, to be sure. As you'd expect from a well-designed system. If your quantization noise is NOT on the hairy edge of visible, then you're simply wasting bits, and I am assuming that the chaps that design these things are not morons.
-
I just dug out my Oppenheim & Schafer, and I think I remember again how this goes.
The thing about quantization noise is that it doesn't actually go away, it simply adds on to whatever your dithered-in noise is. This is actually very useful if you are oversampling because it lets you trade sample rate for bit depth.
In effect, that smooth change across the sky which gets posterized can be completely reconstructed with a blur (de-noising).
In the simple model of posterization, you just get a patch of one tone separated from a patch of another tone, and a sharp line between them.
With dither noise, you get that, plus a speckled pattern of noise (which you may consider is "random spots of the darker tone in the lighter region" and "ransom spots of the lighter tone in the darker region") that gets, statistically, darker as the sky got darker. The posterization is still there, it's just hard to see, because the storm of random dots is quite dense at the "sharp line".
Then you apply a de-noiser, and the sky's lovely gradient returns, in all its glorious detail.
This doesn't work for fine detail which was less than one bit in amplitude.
The photon counting argument doesn't work, because perhaps the detail was all tucked in to 12.0 through 13.0 (well, wait, I guess maybe it does, if we assume a completely linear system that's actually counting photons? are there any sensors that actually are?
Let the ADC round it all down to 12 and the detail is simply gone, there's no getting it back.
If you let the analog amplifier do its thing, then you get that 12.0-13.0 range expanded out before the ADC gets its filthy little hands on it, and you can recover some of the detail. Yes, there will also be a lot of noise, since you're probably dealing with various kinds of noise at roughly the same amplitudes as the signals we're dealing with, but your signal will be, in some cases, present and visible. De-noising will, of course, almost certainly destroy it.
All this assumes that there IS an analog amplification step, which isn't at all clear is universally true. There's an incredible amount of speculation and blather out there, though.
It's a corner case, to be sure, but if you want to see if the system is or is not working for you, find some midtone region with very fine detail that's very low contrast, so low contrast that it's only visible as modulated noise. Test it one way, test it the other, if you can see a difference, then you can be sure the analog ISO amplifier system is bringing some value.
-
ISO Invariance is certainly something to be wished for. However, many cameras are iso-invariant when the ISO amplification overcomes the read/downstream noise of the sensor system. For example, my Canon 5D-III and 1D-IV are virtually iso-invariant at/over ISO 1600, and "usefully iso-invariant" at/over ISO 800. Testing this in a practical sense is trivial. My testing comparison is approximate based on "observing" the point where there is less that 1/3 stop improvement in visual S/N. If anyone is interested, I can show the results of my 5D-III tests.
This is useful when shooting high ISO scenes. If I am shooting a scene that may require 1600 to 6400 ISO (say a dance performance) using my preferred exposure, I will park my camera at ISO 1600 (or even ISO 800) and shoot. This will give some highlight headroom and avoid blown highlights that might happen if I shot ISO 6400. The (for me very) minor downside is that chimping images on the camera screen will look dark.
-
The photon counting argument doesn't work, because perhaps the detail was all tucked in to 12.0 through 13.0 (well, wait, I guess maybe it does, if we assume a completely linear system that's actually counting photons? are there any sensors that actually are?)
We have roughly the same number of photons as we have levels on the camera I was talking about. On modern cameras, we are actually counting a few (that is: less than 10) photons on the low light levels of the picture at high isos. Basically, photon shot noise (which is quantization noise for photons) is about the same level as ADC quantization noise, so you can't really ignore one and say you are only interested in the other.
-
...Why do we want to do this tests at all? In my world, just expose properly for every shoot and do a good picture. ...
I think one should always try to expose properly - i.e. set exposure time and aperture so as to have sufficient (but not too much) light falling onto the sensor.
The problem is what to do when conditions and artistic ambitions exclude "proper exposure". Perhaps you want significant DOF and not too much motion blur in a dark scene. Then your image sensor is going to be under-exposed.
You might still like to take the picture, but knowing how to deal with it can be a good thing. One solution is to increase the in-camera ISO setting. This works well with my Canon camera up to at least ISO1600. Another solution is to increase the "brightness" slider in your raw developer.
-h
-
I wonder: If this sensor is really ISO-invariant, wouldn't it then be better if the ISO set in the camera would only be registered as a setting in the RAW file, just like the white balance?
It seems that having the ISO setting applied to the RAW files only results in possible loss of highlight information, with no gains in return.
-
I wonder: If this sensor is really ISO-invariant, wouldn't it then be better if the ISO set in the camera would only be registered as a setting in the RAW file, just like the white balance?
It seems that having the ISO setting applied to the RAW files only results in possible loss of highlight information, with no gains in return.
A similar method has been suggested before by Michael for ETTR that a tag was registered and brightness on the LCD was adjusted accordingly. This would also be required with your suggested approach.
-
I wonder: If this sensor is really ISO-invariant, wouldn't it then be better if the ISO set in the camera would only be registered as a setting in the RAW file, just like the white balance?
this was already done for a long time ... for example some (old) MFDBs and old Sigmas
-
ISO Invariance is certainly something to be wished for.
actually no... for as long as we have the same DR @ base ISO I'd prefer less (as much less as possible) then 1 stop DR drop with the every next full stop in gain increase. - I think you 'd prefer that too upon thinking a little bit... that means that if my situation allows or forces me to expose less (no sensel saturation) then I actually wish to trade unused sensel capacity for lesser readout-related noise...
-
As long as there is noise that can be overcome with some kind of analog amplification step, ISO invariance isn't a very good idea.
When there is no such noise, you might as well amplify in the digital domain.
-
A7R2, tricks with (unswitchable) NR @ 1sec, 4sec, etc = http://blog.lexa.ru/2015/11/20/sony_a7r_ii_temnovoy_shum_v_zavisimosti_ot_vyderzhki_i_chuvstvitelnosti.html
use translate google com (this blog is of rawdigger/fastrawviewer code developer, who works in team with IB).
-
I think some of you guys have missed the point. (Not naming anyone, because I'm so polite. ;) )
ISO invariance is mainly useful for those who shoot in manual mode, selecting shutter speed and aperture, visible in the viewfinder, and taking note of the exposure scale, also visible in the OVF.
It is often described, by fanatical purists, that ETTR applies only to base ISO, where full-well capacity can be achieved.
In relation to this point of view, when it is obvious that the combination of chosen shutter speed and aperture will result in a higher-than-base ISO, the ISO-Invariance camera allows one to ignore the time-consuming procedures of determining the correct ISO for a 'pseudo' ETTR exposure.
If the variance is huge, like 5 stops, then there might be some advantage in raising ISO. If we're talking about one, or two, or even three stops, then the best practice, with a 'so-called' ISO-Invariant camera, is simply to underexpose, in accordance with choice of shutter speed and aperture at base ISO, or even higher than base ISO, if that's what the camera is already set at.
It's quicker and it might enable one to get the shot, rather than miss it as a result of fussing around with regard to ETTR in relation to one's chosen ISO setting. It removes the fastidious and time-consuming concern about ETTR in relation to a chosen ISO
I always shoot in manual mode. I'm in charge. ;)
-
It's quicker and it might enable one to get the shot, rather than miss it as a result of fussing around with regard to ETTR in relation to one's chosen ISO setting. It removes the fastidious and time-consuming concern about ETTR in relation to a chosen ISO
when you have a decent drop in noise in deep shadows with ISO variant camera then you simply put your camera in matrix metering and auto ISO and then use your M mode... much better than fixed ISO with M mode with ISO invariant camera in real life when you don't have time to patiently meter the scene... so __IF__ DR @ base ISO is the same, ISO variant camera simply better - period
-
ISO Invariance is certainly something to be wished for. However, many cameras are iso-invariant when the ISO amplification overcomes the read/downstream noise of the sensor system. For example, my Canon 5D-III and 1D-IV are virtually iso-invariant at/over ISO 1600, and "usefully iso-invariant" at/over ISO 800. Testing this in a practical sense is trivial. My testing comparison is approximate based on "observing" the point where there is less that 1/3 stop improvement in visual S/N. If anyone is interested, I can show the results of my 5D-III tests.
I disagree. ISO invariance is not desirable, having a large dynamic range sensor is.
The problem with Canon sensors is not that they are not ISO invariant, but the fact they have a relatively poor dynamic range at base ISO.
Should Canon sensors begin at ISO100 having the same DR as Sony sensors, anyone would prefer the Canon behaviour (holding nearly the same DR at ISO200 or even ISO400 as at ISO100).
Regards
-
The problem with Canon sensors is not that they are not ISO invariant, but the fact they have a relatively poor dynamic range at base ISO.
Should Canon sensors begin at ISO100 having the same DR as Sony sensors, anyone would prefer the Canon behaviour (holding nearly the same DR at ISO200 or even ISO400 as at ISO100).
Regards
exactly as I noted above (if DR @ base ISO...)
-
when you have a decent drop in noise in deep shadows with ISO variant camera then you simply put your camera in matrix metering and auto ISO and then use your M mode... much better than fixed ISO with M mode with ISO invariant camera in real life when you don't have time to patiently meter the scene... so __IF__ DR @ base ISO is the same, ISO variant camera simply better - period
Wouldn't that procedure result in the risk of blown highlights?
-
Wouldn't that procedure result in the risk of blown highlights?
where was the last time you saw a camera overexposing with it's automatic metering ? on the contrary - they underexpose all the time ... so matrix metering is the safe bet.
-
Saying that ISO invariance is better than ISO variance is like saying that a 2.8 lens is better than a 4.0 lens. We all know that some 4.0 lenses are better than some 2.8 lenses.
But everything else equal, ISO invariance is preferable above ISO variance because it offers the opportunity to adjust the ISO after the shot has been taken.
-
But everything else equal, ISO invariance is preferable above ISO variance because it offers the opportunity to adjust the ISO after the shot has been taken.
you can't change the readout noise after the shot /I mean when you have raw file already/ (which is the only point in gain change for ISO variant sensors before exposure starts)... so if you on purpose (whatever it is) or forced (by the subject - for example it is moving) undersaturate the sensor you are better off with an option to trade the unused sensel capacity for reduced readout noise...
and no you are not "changing ISO" after the shot - you simply multiply the data (with potentially lower S/N, baked in that data forever, than was possible)...
so any ISO variant sensor with equal DR @ base ISO to your ISO invariant sensor is logically always better = the only /and real/ opposite case is when ISO variant sensor (like Canon's) are not offering same DR @ low gains -> so there you beat Canon's there... but it is not because they are ISO variant... their ISO variantness is also a result of the real reason why they have lower DR @ low gains.
-
Please note that I said "everything else equal", which includes noise at any ISO setting.
-
Please note that I said "everything else equal", which includes noise at any ISO setting.
you can't have that by defintion... because the point of ISO variant sensor is exactly in the lower readout noise :-) ... the DR drops by less than a stop when you go stop higher in gain just exactly because readout noise drops... and with ISO invariant sensor the noise does not drop - otherwise it is no longer invariant.
-
I'm a theatre-photographer and for me Iso-invariance is a blessing, but I don't push it further then maximal 3 stops.
Most of the theatre-photography you do in manual because you cannot trust your meters anymore because of the vastly black backgrounds and the changing intensity of the lighting.
So if I shoot manually close to normal ranges there is always the danger of blowing out the highlights. I measure once at very close distance an actors face and the turn my exposure value two steps down. I push camera monitor almost up to maximal and then I photograph the performance without any risk of overexposure.
Later in LR I will make the necessary corrections.
I works very well for me and I use this method for over a year now.
Working with a nikon D4s
Ton
-
I'm a theater photographer and for me Iso-invariance is a blessing, but I don't push it further then maximal 3 stops.
Most of the theatre photography you do in manual because you cannot trust your meters anymore because of the vastly black backgrounds and the changing intensity of the lighting.
So if I shoot manually close to normal ranges there is always the danger of blowing out the highlights. I measure once at very close distance an actors face and the turn my exposure value two steps down. I push camera monitor almost up to maximal and then I photograph the performance without any risk of overexposure.
Later in LR I will make the necessary corrections.
I works very wel for me and I use this method for over a year now.
Ton
you simply miss the point that if you think that matrix metering will overexpose (they don't unless you have ancient camera with like a dozen segments - or you are saving specular highlights) then you can simply lock your gain (ISO) on ISO variant camera with the same effect as you have now ;D ... just think a little bit - you are not losing anything with ISO variant camera because at any fixed ISO it is has either equal (at base) or less (further up from base) noise (readout) vs ISO invariant one and you can use ISO variant camera as ISO invariant but not vice versa... try to move your mind from Canon's sensors for a while.
-
No, you can't. If you use an ISO variant camera at ISO 400 and later on, you discover that you should have used ISO 800 (with the same exposure), you're out of luck: you'll never get the quality that you would have gotten if you used ISO 800 in the first place.
With an ISO invariant camera, you are able to develop the photo as ISO 800 and get the same quality.
-
With an ISO invariant camera, you are able to develop the photo as ISO 800 and get the same quality.
you totally miss the point that with ISO variant camera I already have better readout noise @ ISO400 than you with ISO invariant camera by definition (of variantness) so if I need to push by a stop in raw converter I am in a better situation than you ;D... you start with the worse readout noise, I start with better - we both push by a stop... and if I am paranoid about blowing something then I can just start @ base ISO where we by definition have the same situation... so I have a choice - you don't
-
By all means, preserve the highlights you're interested in preserving, but you're better off using the analog amplifiers to whatever extent you can than you are amplifying in the digital domain, because, quantization noise.
I see that you have qualified this comment later, but let me describe what I see as the trade-offs, and a sweet spot between "constant ISO speed; fix the levels in post" and "pseudo-ETTR by increasing the ISO speed". (*)
At low levels of analogue gain, noise from the ADC ("quantization noise") might be a significant part of all shadow noise, so that increased analogue gain helps. But beyond some point (maybe around ISO 800 with many recent non-Canon sensors?), the noise in the analogue amplified signal is well above the ADC's noise contribution so that further analogue gain does not really help, and to much of it can hurt, for example by causing clipping of some highlights that one's metering overlooked. I refer to highlights that did not overfill the photosites, but whose signal is then amplified into clipping.
The practical conclusion for me is that I will normally set the ISO speed to get the desired levels in the standard JPEG conversion, but in some low light situations where "proper exposure" requires an elevated ISO speed, but there is also a risk of amplifier clipping from highlights, the safe strategy with an "ISO invariant" sensor is to err on the side of a bit of underexpose. Thus, I no longer see much point to the "high ISO speed pseudo-ETTR" strategy of raising the ISO speed with careful metering with the goal of placing the brightest highlights at raw levels just a hair below maximum. (There is still a place for choosing exposure time and sensor exposure levels to maximize photon counts without overfilling too many photosites.)
(*) I put that "pseudo" in "pseudo-ETTR" because increasing the ISO speed with the same exposure time etc. does not increase exposure, as in the amount of light gathered by the sensor. Some might call this "purist" or "extremist"; I just call it using words correctly.
-
Sony a7RII ISO Invariance, should you underexpose, or use high ISO?
My answer: Best option is neither. Use as low an ISO as possible coupled with the use of a tripod, a wide aperture, and a long exposure.
When that is not possible, then use a higher ISO and “expose to the right”.
Never underexpose, unless to accommodate highlights.
Why?
Camera pixels have a characteristic called the “Full Well Capacity”, i.e. the maximum number of electrons the pixel can hold without saturating or “blooming” (spilling electrons out to adjacent pixels). See
http://www.clarkvision.com/articles/digital.sensor.performance.summary/
for an excellent explanation.
I do not know exactly how the analog-to-digital conversion (ADC) is implemented in the Sony A7RII, but I have designed sensor interfaces in CCD cameras and I suspect the general principles are the same. That is, the analog voltage from each pixel, that is proportional to the number of electrons in that pixel after exposure, is fed to an “Analog Front End (AFE)”. The first function in the AFE is a “Sample and Hold”. That is followed by a Variable Gain Amplifier, which is followed by an Analog-to-Digital-Converter. The gain (amount of amplification) of the variable Gain Amplifier corresponds to the ISO setting. For a weak signal (low light) it amplifies the signal a lot to match the voltage range of the ADC. For a strong signal (strong light), the gain is set to a low value (low ISO) so that the ADC is not overloaded.
What will happen if you deliberately underexpose?
The gain of the Variable Gain Amplifier will be set to a low value, and the voltage going to the ADC will be very low.
For a 5-stop underexposure, you will only be using 1/32 of the range of the ADC.
In other words, you will only be using 9 bits of a 14-bit A to D converter, and you will be multiplying the Quantizing Noise by a factor of 32.
Compare Fig #2 withe Fig #4 in Michael’s article. To my eyes this extra noise in Fig #4 is very obvious.
-
For a 5-stop underexposure, you will only be using 1/32 of the range of the ADC.
In other words, you will only be using 9 bits of a 14-bit A to D converter,
I can't figure out why so many people have adopted the phrase "dynamic range" to describe "low noise shadow recovery". It seems to me that more often than not, one desires to recover shadows after making a relatively low dynamic range exposure.
When ever I encounter the phrase dynamic range I think of dynamic range and so I find the recent discussions about "dynamic range" confusing.
It seems simpler to call shadow recovery, "shadow recovery" so that people can focus on and discuss the differences in the ADC systems.
-
When that is not possible, then use a higher ISO and “expose to the right”.
Never underexpose, unless to accommodate highlights.
Why?
. . . Camera pixels have a characteristic called the “Full Well Capacity”,
. . .
For a 5-stop underexposure, you will only be using 1/32 of the range of the ADC.
In other words, you will only be using 9 bits of a 14-bit A to D converter, and you will be multiplying the Quantizing Noise by a factor of 32.
Compare Fig #2 withe Fig #4 in Michael’s article. To my eyes this extra noise in Fig #4 is very obvious.
I disagree about "ETTR at high ISO" or the reasons stated above: once the ISO speed is much beyond the "unity gain" level where noise from the sensor is amplified above the nose level forth ADC, further amplification just increases the risk of blowing highlights by amplified clipping, while adding nothing to the image quality in the rest of the image, because shot noise and sensor noise are dominant over "ADC noise".
Yes fives stop of underexposure leaves a 14-bit ADC using only the bottom 9 bits of its range – but those are enough to cover all the actual information from the sensor anyway in situations like Michael's example where you would have to push the ISO more than 5 stops above minimum, and specifically more than fives stops above "unity gain" level. Because then all the photosites are underexposed by more than 5 stops (getting less than 1/32 of their full well capacity) so the signal from the sensor has less than 9 stop of DR: any extra amplification just adds less significant bits of pure noise at the bottom of the useful signal, with only at most the 9 most significant bits containing any useful information: this does not improve signal quality, and if taken too far, increases the risk of clipping.
As has been pointed out, Michael's comparison photos are off by one stop: ISO 640 pushed by 5 stops should be compared to ISO 204800 (not 102400) to be comparing equal exposure level, meaning equal amount of light reaching the photosites, as with equal exposure time and aperture.
-
ISO 640 pushed by 5 stops should be compared to...
It was my impression that the example of ISO 640 pushed by 5 stops was meant to be compared to the "properly" exposed ISO 640 example. The recently popularized notion of "invariance" seems related to a practice of leaving the ISO alone while shooting with a priority for both shutter and aperture regardless of resulting exposure.
I inferred that the higher ISO shot was included as an after thought, or as a casual comparison.
-
Photographers do this a lot. They think that if one quantity is bigger than another, then the other one kind of goes away.
We see the same fallacy in sharpness discussions.
-
I disagree. ISO invariance is not desirable, having a large dynamic range sensor is.
The problem with Canon sensors is not that they are not ISO invariant, but the fact they have a relatively poor dynamic range at base ISO.
Should Canon sensors begin at ISO100 having the same DR as Sony sensors, anyone would prefer the Canon behaviour (holding nearly the same DR at ISO200 or even ISO400 as at ISO100).
Regards
In a practical sense, "Iso Invariant" means that, after the sensel charge is read, there is no further significant noise contribution to the pixel. You can alter the effect of ISO amplification and see no additional noise. This allows the maximum dynamic range to be achieved --- remember that DR is a signal to noise relationship. ISO Invariance implies maximum dynamic range for that sensor. Of course, you want a sensor that has the lowest possible sensel read noise and ISO invariance together.
The Canon (like my 5D-III) is "Iso Variant" because, after the pixel is read (which is very low noise on Canon sensors - among the best available), additional downstream noise is created in the sensor processing system. The implication is that some amplification of the pixel values can be done in such away that it raises the signal above the down-stream added noise. In a simplistic sense, if the pixel voltage is one volt and the down-stream noise is 2 volts, you can, for example, amplify the pixel above the noise to 3 volts by setting an ISO value. This means that the actual signal can be detected cleanly since it well above the noise. Unfortunately overall, this also means that the base ISO DR of the Canon is not very good - but turns out to be very good ISO 1600 and higher.
-
Photographers do this a lot. They think that if one quantity is bigger than another, then the other one kind of goes away.
We see the same fallacy in sharpness discussions.
Indeed, but one should also avoid the opposite mistake of pursuing perfection in one respect (like minimizing quantization noise) at the cost of ignoring faults in other respects (the greater imperfections of higher levels of analog gain, like amplifier nonlinearity and clipping). I have perhaps given the wrong impression that I support the dogma that "more that unity gain is pointless", so let me clarify:
At some ISO speed setting (seemingly in the range 400 to 800 with many good current sensors) the analog gain brings the noise level entering the ADC up to the noise level then caused by quantization in the ADC: I believe that this is what some call "unity gain". Contrary to the above dogma, I see that it can still help to amplify a few stops beyond that, so that the noise floor in the incoming analog signal is at about two or four times the quantization noise, but beyond about that, quantization noise is insignificant, and reducing it further with further analog gain is of no practical value. Indeed, digital gain in raw conversion becomes a better approach: it is perfectly linear, and one can back-off from clipping – which I suppose is then more accurately called "arithmetic overflow".
To quantify: describing the ADC quantization noise as of level 1 (a 1 bit error?), if the noise in the input signal is amplified to be two stops higher it is at level 4. Then the combined effect in the deepest shadows is the RMS combination sqrt(1^2 + 4^2) = 4.12, so raising the deep shadow noise floor and lowering the total engineering dynamic range by 0.04 stops compared to what one would get by avoiding ADC quantization noise effects entirely. This I would call measurable but in practice negligible.
Raising the ISO speed by one more stop, to three stops over unity gain, the noise floor in the amplified analog signal is at eight times the quantization noise floor, and the quantization noise raises that total noise from 8 to 8.06, a change of 0.01 stops, so a tiny 0.03 stops better than the previous "two stops over unity gain" case. If the cost of this 0.03 stop improvement in the deep shadows is the risk of clipping the highlights one stop earlier . . .
P. S. In all but the deepest shadows, photon shot noise overwhelms these other noise sources anyway.
TL;DR Pardon all the numbers: my point is that these numbers are negligible, and we should not obsess about pushing up the analog ISO speed gain in pursuit of perfect ETTR histogram placement in low light situations! Good modern sensors often let us use any ISO speed in a safe range that avoids the problems at either extreme, so we can think more about composition instead of worrying about absolutely perfect light metering and settings choices.
-
...
At some ISO speed setting (seemingly in the range 400 to 800 with many good current sensors) the analog gain brings the noise level entering the ADC up to the noise level then caused by quantization in the ADC: I believe that this is what some call "unity gain". Contrary to the above dogma, I see that it can still help to amplify a few stops beyond that, so that the noise floor in the incoming analog signal is at about two or four times the quantization noise, but beyond about that, quantization noise is insignificant, and reducing it further with further analog gain is of no practical value. Indeed, digital gain in raw conversion becomes a better approach: it is perfectly linear, and one can back-off from clipping – which I suppose is then more accurately called "arithmetic overflow".
...
TL;DR Pardon all the numbers: my point is that these numbers are negligible, and we should not obsess about pushing up the analog ISO speed gain in pursuit of perfect ETTR histogram placement in low light situations! Good modern sensors often let us use any ISO speed in a safe range that avoids the problems at either extreme, so we can think more about composition instead of worrying about absolutely perfect light metering and settings choices.
My interpretation of "unity gain" (from simple reading) :
Unity gain happens at some value of the camera ISO setting when 1 electron produced in a pixel results in 1 ADU at the ADC output, that is the conversion factor is now equal to "1". I suspect that higher than unity gain, in general, does not result in significant improvement ... other than to suppress pattern noise. Apparently, this unity gain is the preferred operating point for astronomical photographers.
I interpret this to apply to my Canon 5D-III - at the point where the sensor system becomes virtually "iso invariant" at ISO 1600 when simple digital scaling in ACR is "identical" to in-camera ISO scaling.
-
There seems to be a major point which has been overlooked in this thread. The Sony A7R2 is not an ISO-Invariant camera, according to DXOMark tests, especially with regard to DR.
For example, at ISO 100 (actually ISO 74) the DR of the A7R2 is 13.9 EV, as measured by DXO. If one underexposes by 5 stops at ISO 100, instead of increasing ISO to 3200 for a pseudo-ETTR shot, then the DR of the resulting underexposed shot will be 5 EV lower (ie. 13.9 - 5 = 8.9 V).
However, the DXOmark Dynamic Range measurements at ISO 3200 show a DR of 11.01 EV. That means that DR is only 2.89 stops down instead of the expected 5 stops down that would result if the camera were truly ISO-Invariant.
It seems clear from these results that choosing to raise ISO instead of underexposing will always produce a better dynamic range with the Sony A7R2. However, the gain in DR varies according to the starting ISO. For example, underexposing 5 stops at ISO 3200, instead of raising ISO to 102,400, will result in a DR which is worse by only 0.4 EV, approximately.
http://www.dxomark.com/Cameras/Compare/Side-by-side/Nikon-D810-versus-Sony-A7R-II___963_1035
-
However, the gain in DR varies according to the starting ISO. For example, underexposing 5 stops at ISO 3200, instead of raising ISO to 102,400, will result in a DR which is worse by only 0.4 EV, approximately.
http://www.dxomark.com/Cameras/Compare/Side-by-side/Nikon-D810-versus-Sony-A7R-II___963_1035
Thank you Ray. That has been my understanding from DXO and from Jim Kasson's blogs. The a7rII is ISO invariant in steps.
Dave
-
Thank you Ray. That has been my understanding from DXO and from Jim Kasson's blogs. The a7rII is ISO invariant in steps.
Dave
Yes. Having checked the graph again, I would describe the A7R2 as being ISO-Invariant only for the 3 stops between ISO 3200 and ISO 25,600, and the one stop between ISO 51,200 and 102,400. For all other full-stop steps there is at least 1/3rd of an EV advantage in DR by increasing ISO just one stop instead of underexposing, and sometimes as much as a 0.5 EV advantage.
-
My takeaway: unless I'm working slowly and/or methodically with a tripod I can ignore all this stuff. In practice I've been able to get more than good results with the ISO cranked up, provided I expose enough for a pic that already looks good SOOC.
-Dave-
-
My takeaway: unless I'm working slowly and/or methodically with a tripod I can ignore all this stuff. In practice I've been able to get more than good results with the ISO cranked up, provided I expose enough for a pic that already looks good SOOC.
-Dave-
Haven't you got that the wrong way round, Dave? If you are able to work slowly and methodically with a tripod, then you can ignore all this stuff because you have the time to get ETTR exposures (or pseudo-ETTRs) right.
If your subject is stationary, and/or lighting good, and/or desired aperture wide, you can use base ISO for maximum DR, or bracket exposure for merging to HDR if necessary. In such circumstances, the ISO-Invariance qualities of the camera become irrelevant.
-
Thank you Ray. That has been my understanding from DXO and from Jim Kasson's blogs. The a7rII is ISO invariant in steps.
That is correct. As I understand things, ISO invariance (ISO-Less) occurs when the input referred read noise (expressed in electrons) holds constant as the ISO is raised. As the chart shown below taken from Bill Claff's web site demonstrates, the Sony shows 3 discrete decrements in read noise as the ISO is raised. This is likely due to the Aptina Aptina DR-pix (http://www.photonstophotos.net/Aptina/DR-Pix_WhitePaper.pdf) technology that is said to be implemented in this sensor.
The 7RMii demonstrates only minor changes in the input read noise from ISO 640 to 25600 and is nearly ISO-less. This is the range that Michael discussed in his post. I do not think DR changes of less than 0.5 EV are really significant.
The Nikon D7200 is virtually ISO-less from ISO 400 to 5100. The Canon 5DS R does not appear to be ISO-less at any ISO, but it's readnoise is competitive above ISO 1600. The Nikon D800e falls between the two extremes.
Regards,
bill
-
Haven't you got that the wrong way round, Dave? If you are able to work slowly and methodically with a tripod, then you can ignore all this stuff because you have the time to get ETTR exposures (or pseudo-ETTRs) right.
Unless for some reason I've chosen a specific ISO value I just let it float in camera, caring not a whit whether the gaining up—when it happens—is happening in the analog domain, the digital domain or some combo of the two. That's what I mean by ignoring "all this stuff."
-Dave-
-
Unless for some reason I've chosen a specific ISO value I just let it float in camera, caring not a whit whether the gaining up—when it happens—is happening in the analog domain, the digital domain or some combo of the two. That's what I mean by ignoring "all this stuff."
-Dave-
I see! It's been mentioned before that matrix metering with automatic ISO rarely overexposes. I must admit I haven't tried this since I began using the 'effectively' ISO-less Nikon cameras. I've assumed that an automatic ISO setting would tend to overexpose the very bright areas in a scene if those areas cover only a small proportion of the scene, since the automatic metering produces an average reading which is not necessarily ideal for any particular part of the scene.
My usual procedure is to separate focusing from metering, allocating the AF-On button to focusing, using a single focusing square, and allocating the half-pressed shutter to the exposure reading in relation to the position of the focusing square.
This set-up allows me to move the focusing square around the scene and watch how the exposure reading changes on the metering scale visible in the viewfinder. If I consider a small portion of bright, cloudy sky within the scene might be overexposed, and I want to capture the maximum detail in that small portion of the scene, I simply increase the shutter speed with my thumb on the wheel until the meter in the viewfinder looks right when the focusing square is positioned over that brightest part of the sky. I then recompose the scene and take the shot.
-
That is correct. As I understand things, ISO invariance (ISO-Less) occurs when the input referred read noise (expressed in electrons) holds constant as the ISO is raised. As the chart shown below taken from Bill Claff's web site demonstrates, the Sony shows 3 discrete decrements in read noise as the ISO is raised. This is likely due to the Aptina Aptina DR-pix (http://www.photonstophotos.net/Aptina/DR-Pix_WhitePaper.pdf) technology that is said to be implemented in this sensor. .....
One thing to consider how "constant" is the input-referred noise. One way is to think it is "truly constant". The other is to put bounds of the noise change. Bill Claff personally chooses to infer "ISO invariance" is when the noise varies by 1/3 or less EV for higher "measured ISO".
-
One thing to consider how "constant" is the input-referred noise. One way is to think it is "truly constant". The other is to put bounds of the noise change. Bill Claff personally chooses to infer "ISO invariance" is when the noise varies by 1/3 or less EV for higher "measured ISO".
I haven't seen Bill's post on that subject, but it is in essential agreement with what I said. There is not much difference between <0.5 and <=0.3.
Bill Janes
-
I haven't seen Bill's post on that subject, but it is in essential agreement with what I said. There is not much difference between <0.5 and <=0.3.
Bill Janes
Bill,
As I understand, and have observed, a difference of 0.5 EV in dynamic range is noticeable and therefore significant at a pixel-peeping level, although generally not particularly significant.
If one underexposes by 2 stops at ISO 100, with the A7R2, instead of raising ISO to 400, then according to the DXOMark tests, one will have sacrificed a full stop (or EV) of DR, which is definitely significant.
However, adopting the same procedure with the Nikon D810, underexposing by 2 stops at ISO 100 instead of raising ISO to 400, one sacrifices approximately 1/4th of an EV worth of DR, which is definitely insignificant.
-
you totally miss the point that with ISO variant camera I already have better readout noise @ ISO400 than you with ISO invariant camera by definition (of variantness) so if I need to push by a stop in raw converter I am in a better situation than you ;D... you start with the worse readout noise, I start with better - we both push by a stop... and if I am paranoid about blowing something then I can just start @ base ISO where we by definition have the same situation... so I have a choice - you don't
I stand corrected. I did not know that an ISO variant camera by definition has the same readout noise at base ISO as an ISO invariant camera.
I thought ISO invariance just meant that shooting at increased ISO delivers the same results as shooting at base ISO and then pushing the image an equivalent number of stops. (using the same exposure settings ofcourse)
-
I thought ISO invariance just meant that shooting at increased ISO delivers the same results as shooting at base ISO and then pushing the image an equivalent number of stops. (using the same exposure settings ofcourse)
you can have that either with ISO-by-tag (the true ISO-less camera, no analog anb/or digital gain - just firmware writes a tag telling raw converters to push behind the scene) or with constant readout related noise when gain is applied (of any kind - analog or whatever)... now because we consider theoretical sensors where the only difference is what happens with increase in ISO (gain) we shall assume that readout noise is the same at "base" ISO...
if we will start to consider real cameras then things get tainted because Canon sensors are not using the same technology as Sony sensors for example and that makes readout related noise not equal.