Canon is thinking about more lenses like the RF 600mm f/11 STM and RF 800mm f/11 STM

Joules

doom
CR Pro
Jul 16, 2017
1,546
1,848
Hamburg, Germany
That is, it's actually ISO setting, when applied to the resulting image, creates visible noise. But we used the higher ISO because the image was too dark (the information was in the raw shadows with low SNR). So it's like chicken and egg problem.
The noise is not created by the ISO setting. Neither does it lower the Signal to Noise ratio. So not really chicken and egg.

Increasing the brightness of an image will however make noise more apparent in regions with poor SnR, as you say. Raising this brightness in post is not superior to doing it through ISO though if noise is the primary concern.

I'll post some pictures I took yesterday later.
 

Quarkcharmed

EOS R5
Feb 14, 2018
1,223
1,078
Australia
www.michaelborisenko.com
The noise is not created by the ISO setting. Neither does it lower the Signal to Noise ratio. So not really chicken and egg.

I never said ISO setting lowers the SNR. The higher ISO causes the camera or processing software to use lower range of the signal for producing the resulting image. Note you don't see the noise in the sensor or in the raw file. You can only see the noise in the processed image, you don't see it in the raw file. It is your interpretation of what is as a meaningful signal in the raw file.

Increasing the brightness of an image will however make noise more apparent in regions with poor SnR, as you say. Raising this brightness in post is not superior to doing it through ISO though if noise is the primary concern.

I'm not saying it's superior or inferior. My point is simply it's a bit misleading to say that higher ISO doesn't increase the visible noise. It's an observable fact that higher-ISO images have more noise.
From the practical standpoint, if the image quality is your priority, you should minimise the ISO value - in general. But there are caveats say with the R5 where ISO 400 is cleaner than ISO 320 and has greater dynamic range.
 
  • Like
Reactions: usern4cr

AlanF

Stay at home
CR Pro
Aug 16, 2012
7,776
8,545
First, I'd like to thank MichaelClark and PrivateByDesign for their carefully explained posts regarding equivalence, which I do agree with.

I'd like to give my opinions on how these 3 things compare for the same FF camera & ideal FF sensor, with the viewed "images" at the same size & brightness (where "ideal" just means 100% sensor quantum efficiency or 100% lens transmittance):
#1: Using an ideal FF 200mm f4 lens, with a 2x crop
#2: Using an ideal FF 200mm f4 lens, with an ideal 2x TC
#3: Using an ideal FF 400mm f8 lens
* All will produce the same image regarding the same angle of view, DOF, and OOF blur.
* All will capture the same rate of total photons per unit time for the image.
The differences between them are:
* #1 has the image on just 25% of the sensor area, while #2 and #3 have the image on 100% of the sensor area.
* Thus #2 and #3 have 4x the full-well capacity for the image as #1 and thus can capture 4x the total amount of light for it.
* If you stop your exposure at time T (when #1 reaches full-well for the brightest area) then #1 reaches "full exposure" while #2 and #3 are only 25% exposed but use a 4x ISO amplification to reach full exposure, and thus all three will produce the same image brightness from the same total # of photons, and thus with the same level of noise (assuming noise is mainly due to #photons captured, which I agree with).
* If you allow the exposure in #2 and #3 to continue to be four times longer (4*T) then they will reach full exposure with 4x the total photons and thus 4x less noise than #1 (but 4x longer exposure can have 4x more subject motion blur, if any).
* #1 will have less pixels displayed in the image (vs #2 and #3) and thus will have less resolution with rougher edges, but using good upsizing interpolation can start to look somewhat close to the others.
* #1 and #3 will use (typically) the same number of lens elements as far as real-world transmission loss, while #2 adds a significant # of lens elements which will slightly reduce transmission but introduce more significant contrast and resolution loss.
* Since #1 uses 25% of the sensor for the image, 75% is still recorded around it (assuming you told the camera to do so). That allows you the choice to expand in post what you can see beyond the "image" we've been discussing.

All in all, my preference is to use is:
#3 for the best quality image if I can afford an additional lens and be willing to carry it around.
#1 would be my 2nd choice if I don't have the 400mm lens.
#2 would not be my choice, as I'd prefer #1 over it mainly for the extra 75% coverage around the image which I could use in post to better crop the final result in case the subject moved off-center enough.
Most of what you write is spot on, but a couple of points. A minor one is that photon noise varies as the square root of the number of photons. What is important to consider as well is the MTF or resolution of the sensor. For a low resolution sensor, increasing focal length is important as long as the f-number is not too far above the diffraction limit. At the other extreme, where the sensor resolution is really high, resolution is limited mainly by the diameter of the lens (entrance pupil) and not the focal length of the lens or the f-number. For example, with an extremely high resolution sensor, double the focal length of the lens with a 2xTC and you will increase the separation of two close points in the image by a factor of two but you will double the size of the diffraction disk around each and so leave the resolution unimproved. With a low resolution sensor, the distance separation doubles but the increase in size of the diffraction disk is less important as it is the size of the pixels that is more limiting.
 
  • Like
Reactions: usern4cr

Kit.

EOS 5D Mark IV
Apr 25, 2011
2,074
1,423
The SNR is lower in the shadows, and that leads to higher visible noise if the shadows are mapped to the resulting image. If they're not mapped or appear very dark and black, then there's little visible noise. If the shadows are lifted, that is mapped to mid-tones or even highlights in the resulting image, there will be more visible noise.
Human perception of the visual contrast is logarithmic, so as long as the SNR itself is not changed by a non-linear mapping, the visibility of the noise should be about the same. If the shadows are non-linearly mapped because the output media has less dynamic range than the sensor, then the noise in the shadows can be less visible, but the signal itself will be lost in the shadows, too.

That is, it's actually ISO setting, when applied to the resulting image, creates visible noise. But we used the higher ISO because the image was too dark (the information was in the raw shadows with low SNR). So it's like chicken and egg problem.
The visible noise is created by having not "enough" of (luminous) exposure in the first place. If you want to minimize the noise, the exposure should be achieved as high as practically possible and the ISO should be set for the optimal pre-ADC amplifier gain.
 

Quarkcharmed

EOS R5
Feb 14, 2018
1,223
1,078
Australia
www.michaelborisenko.com
Human perception of the visual contrast is logarithmic, so as long as the SNR itself is not changed by a non-linear mapping, the visibility of the noise should be about the same. If the shadows are non-linearly mapped because the output media has less dynamic range than the sensor, then the noise in the shadows can be less visible, but the signal itself will be lost in the shadows, too.


The visible noise is created by having not "enough" of (luminous) exposure in the first place. If you want to minimize the noise, the exposure should be achieved as high as practically possible and the ISO should be set for the optimal pre-ADC amplifier gain.

It's all true, and my point above is more about practical implications.
Generally if you're at a base ISO, you can get a cleaner image with a higher dynamic range. Setting a high ISO speed will limit you in DR and increase the noise.
Yes it's because there's less light and less information. But setting a higher ISO speed puts an upper limit on the amount of light you can capture.

In practice that means you better use base ISO plus longer exposure, than higher ISO and shorter exposure, if the image quality is important.

PS. The noise in the shadows at say ISO 100 is low and indistinguishable. When you heavily lift the shadows or push exposure slider (similar to higher ISO effect), you literally start seeing more noise. The image may become unusable, but this process quite literally creates visible noise.
 
Last edited:
  • Like
Reactions: usern4cr

usern4cr

R5
CR Pro
Sep 2, 2018
945
1,242
Kentucky, USA
Most of what you write is spot on, but a couple of points. A minor one is that photon noise varies as the square root of the number of photons. What is important to consider as well is the MTF or resolution of the sensor. For a low resolution sensor, increasing focal length is important as long as the f-number is not too far above the diffraction limit. At the other extreme, where the sensor resolution is really high, resolution is limited mainly by the diameter of the lens (entrance pupil) and not the focal length of the lens or the f-number. For example, with an extremely high resolution sensor, double the focal length of the lens with a 2xTC and you will increase the separation of two close points in the image by a factor of two but you will double the size of the diffraction disk around each and so leave the resolution unimproved. With a low resolution sensor, the distance separation doubles but the increase in size of the diffraction disk is less important as it is the size of the pixels that is more limiting.
Thanks, Alanf. I was thinking that the noise might be the square root instead of 1/x (it's been a long time since I've been to university), and your post reminds me to go back and correct what I posted.

And I do agree that diffraction comes into play as the iris (1/f#) gets really small, and as the pixel size gets really small then you approach a limit of resolution. The basic rule of thumb that using a 2x TC or a 2x crop or just an equivalent longer lens (2x focal length & 2x f#) will give you the equivalent image still seems to hold for me since (I think) all of them will suffer this effect to a similar degree. Of course, you always have a difference between the three since one has the deleterious effects of more lens elements and one relies on the intelligence (or lack of it) in the software used to up-res the #pixels in post, and one uses a different lens of possibly different optcal quality.
 
Last edited:
  • Like
Reactions: AlanF

Lucas Tingley

Canon EOS RP
Nov 27, 2020
101
55
There's no need for an RF 400mm f/8. EF 400mm f/5.6L is relatively light and low cost, an is an excellent option especially with IBIS making its way to R series cameras. An updated RF 400mm f/5.6 IS wouldn't necessarily be much larger/heavier than a f/8 version.

An RF 500mm f/8 or an RF600mm f/8 would be much more interesting.
but the cost could be a bit more affordable, like $500
 

Joules

doom
CR Pro
Jul 16, 2017
1,546
1,848
Hamburg, Germany
I'm not saying it's superior or inferior. My point is simply it's a bit misleading to say that higher ISO doesn't increase the visible noise. It's an observable fact that higher-ISO images have more noise.
From the practical standpoint, if the image quality is your priority, you should minimise the ISO value - in general. But there are caveats say with the R5 where ISO 400 is cleaner than ISO 320 and has greater dynamic range.

As promised, I put together images to illustrate what I was talking about when saying you should not be afraid of high ISO values:


@CanonFanBoy
As your's was one of the posts pushing me to finally take some pictures on the matter, I'd just like to point out that I did just that and posted them in that thread:

 
Last edited:
Aug 11, 2019
8
9
Not dead, just mature for its intended market. If you want lenses larger than 62mm in diameter, then EOS M is not intended for you.
So dead then, because they are certainly no indication of any new lenses, an early rumour, March 2020 I think talked of a 70-300 for the system I think that would have been a terrific addition. RIP EF-m we will miss you...
 

Michael Clark

Now we see through a glass, darkly...
Apr 5, 2016
3,428
2,008
I'd have guessed it was entropy not deviation, but I thought it was unlikely Michael Clack meant entropy.



Sorry, of course. The SNR is lower in the shadows, and that leads to higher visible noise if the shadows are mapped to the resulting image. If they're not mapped or appear very dark and black, then there's little visible noise. If the shadows are lifted, that is mapped to mid-tones or even highlights in the resulting image, there will be more visible noise.

That is, it's actually ISO setting, when applied to the resulting image, creates visible noise. But we used the higher ISO because the image was too dark (the information was in the raw shadows with low SNR). So it's like chicken and egg problem.


Shot noise can be modelled very accurately using Poisson distribution.

ISO settings do *not* create noise. The noise is caused by the lack of light. If you take an image of the same scene at the same illumination level and use ISO 100 for one and ISO 1600 for the other while using the same Tv and Av that do not result in highlight clipping at ISO 1600, then amplify the ISO 100 image by four stops in post to equal the final brightness of the ISO 1600 image, the ISO 100 image will be noisier.
 
Last edited:
  • Like
Reactions: Joules

Michael Clark

Now we see through a glass, darkly...
Apr 5, 2016
3,428
2,008
I never said ISO setting lowers the SNR. The higher ISO causes the camera or processing software to use lower range of the signal for producing the resulting image. Note you don't see the noise in the sensor or in the raw file. You can only see the noise in the processed image, you don't see it in the raw file. It is your interpretation of what is as a meaningful signal in the raw file.

You don't see anything in a raw file until it has been processed. What you see then is only one of a near infinite number of possible interpretations legitimately derived from the information in the raw file. There's no such thing as an unprocessed or unedited raw file displayed on a screen. If you're not selecting how the raw data is being converted to a viewable image, then whoever wrote the default development profile has decided it for you.

I'm not saying it's superior or inferior. My point is simply it's a bit misleading to say that higher ISO doesn't increase the visible noise. It's an observable fact that higher-ISO images have more noise.
From the practical standpoint, if the image quality is your priority, you should minimise the ISO value - in general. But there are caveats say with the R5 where ISO 400 is cleaner than ISO 320 and has greater dynamic range.

To maximize image quality, one should maximize the amount of light falling on the sensor until just before the highlights begin to clip. If it isn't possible to do that at ISO 100 (because, for example, the subject is moving and the light is dim), then the best image quality for a given amount of light entering the camera is the *highest* ISO that doesn't allow clipping! You'll get better image quality shooting at ISO 3200, f/2.8, 1/1000 in the high school stadium in which I shoot often than you will using ISO 100, f/2.8, 1/1000, and pushing the entire image five stops in post.
 
  • Like
Reactions: Joules

Michael Clark

Now we see through a glass, darkly...
Apr 5, 2016
3,428
2,008
So dead then, because they are certainly no indication of any new lenses, an early rumour, March 2020 I think talked of a 70-300 for the system I think that would have been a terrific addition. RIP EF-m we will miss you...

The cameras and lenses in the EOS M system are still selling rather nicely, thank you very much. They've all but replaced Rebels/xx00D models everywhere but North America and Western Europe. They're not for gear hounds who need a new, improved lens twice every year. They're for folks who want a small, light, and affordable camera (not cameras, camera) and a lens or three that will last them for several years without any thought of "upgrading" every time another hot rumor hits the inter-webs.
 

usern4cr

R5
CR Pro
Sep 2, 2018
945
1,242
Kentucky, USA
Shot noise can be modelled very accurately using Poisson distribution.

ISO settings do *not* create noise. The noise is caused by the lack of light. If you take an image of the same scene at the same illumination level and use ISO 100 for one and ISO 1600 for the other while using the same Tv and Av that do not result in highlight clipping at ISO 1600, then amplify the ISO 100 image by four stops in post to equal the final brightness of the ISO 1600 image, the ISO 100 image will be noisier.
*EDIT* - I'm trying to get clarification on this issue (thanks Joules), so what I've mentioned below may not be correct (well, that wouldn't be the first time :ROFLMAO: ! )

I'm not following you here. Using the same speed and f# for both (which does not cause highlight clipping) will capture the same #photons for both images. If you shoot RAW, the ISO while viewing later (or in post) will show the ISO 100 picture darker, but adding +4 EV in post will get the same brightness. For non-clipped images in RAW, changing in post ISO(if possible) and EV offset have the same effect and they're just software values.

Now, if you shoot jpg then the encoding of the image in camera memory will be a cause of difference due to the ISO 100 encoding with 4 fewer bits (hence a darker image). But that's not caused by the ISO setting itself, but by improper encoding range of the jpg image which drops the low 4 bits of the image instead of saving it.

So the ISO is not inherently causing noise. Jpg (but not RAW) encoding is causing noise.
If I'm not understanding something, please "illuminate" me! ;)
 
Last edited:

Joules

doom
CR Pro
Jul 16, 2017
1,546
1,848
Hamburg, Germany
If you shoot RAW, the ISO while viewing later (or in post) will show the ISO 100 picture darker, but adding +4 EV in post will get the same brightness. For non-clipped images in RAW, changing in post ISO(if possible) and EV offset have the same effect and they're just software values.
[...]
So the ISO is not inherently causing noise. Jpg (but not RAW) encoding is causing noise.
If I'm not understanding something, please "illuminate" me! ;)
As far as I can tell, all the talk of noise and such here was concerning RAW, not JPEG. Lossy compression introduces many artifacts that have nothing to do with the noise being talked about here.

The ISO setting in the camera is different from raising the image brightness digitally in post. ISO is amplification of the electrical signal before it is converted to digital. What that means is that the noise in the image is also amplified, but any noise that is added through the electronics in the remaining steps leading up to a digital image is not amplified. If you do all your brightness increase in software, even the noise introduced by the electronics involved in the ADC get's 'amplified' (digitially). Especially with older sensors (anything older than the M6 II, 90D, 1DX III, R6 or R5), this will lead to an image looking siginficantly worse if shot at a low ISO and raised in post rather than just shot with a high ISO in the first place.

Feel free to run the experiment yourself, if image quality at high ISO is a topic of interest to you. I have also done exactly that a little while back and tried to share the results and my understanding of the theory behind it in this post here:

 
  • Like
Reactions: Michael Clark

usern4cr

R5
CR Pro
Sep 2, 2018
945
1,242
Kentucky, USA
As far as I can tell, all the talk of noise and such here was concerning RAW, not JPEG. Lossy compression introduces many artifacts that have nothing to do with the noise being talked about here.

The ISO setting in the camera is different from raising the image brightness digitally in post. ISO is amplification of the electrical signal before it is converted to digital. What that means is that the noise in the image is also amplified, but any noise that is added through the electronics in the remaining steps leading up to a digital image is not amplified. If you do all your brightness increase in software, even the noise introduced by the electronics involved in the ADC get's 'amplified' (digitially). Especially with older sensors (anything older than the M6 II, 90D, 1DX III, R6 or R5), this will lead to an image looking siginficantly worse if shot at a low ISO and raised in post rather than just shot with a high ISO in the first place.

Feel free to run the experiment yourself, if image quality at high ISO is a topic of interest to you. I have also done exactly that a little while back and tried to share the results and my understanding of the theory behind it in this post here:

OK, I wasn't aware that each ISO level had a different amount of amplification of the pixel signal before ADC. I've seen graphs of IQ (SNR or DR, I don't recall) that show a clear relation to ISO but with a single big "improved jog" at ISO 400 (relative to lower ISOs). - I thought that was because at ISO 400 (and above) it increased the amplification before ADC, which would mean (to me) that it doesn't change amplification before ADC at all the other ISOs. Can you explain what this jog is, if indeed each ISO level has its own different amplification before ADC conversion?
 

usern4cr

R5
CR Pro
Sep 2, 2018
945
1,242
Kentucky, USA
As far as I can tell, all the talk of noise and such here was concerning RAW, not JPEG. Lossy compression introduces many artifacts that have nothing to do with the noise being talked about here.

The ISO setting in the camera is different from raising the image brightness digitally in post. ISO is amplification of the electrical signal before it is converted to digital. What that means is that the noise in the image is also amplified, but any noise that is added through the electronics in the remaining steps leading up to a digital image is not amplified. If you do all your brightness increase in software, even the noise introduced by the electronics involved in the ADC get's 'amplified' (digitially). Especially with older sensors (anything older than the M6 II, 90D, 1DX III, R6 or R5), this will lead to an image looking siginficantly worse if shot at a low ISO and raised in post rather than just shot with a high ISO in the first place.

Feel free to run the experiment yourself, if image quality at high ISO is a topic of interest to you. I have also done exactly that a little while back and tried to share the results and my understanding of the theory behind it in this post here:

OK - I've read (& re-read) your detailed "Equivalency - now with pictures" post. It didn't mention whether images were stored as raw or not, but I'll assume they were stored as uncompressed raw (as I think you implied in your above reply). Of course, higher ISO values will require that the final image has more amplification applied to the #photons sensed. But there are many steps to be done before you see the image, which are: ADC conversion, storage in raw format to the in-camera file, de-bayer-interpolation, and export to the user for either post processing/printing/or viewing in EVF. The ISO amplification could conceivably be done before any of these 4 steps. Your article doesn't mention whether this amplification is before the ADC, or sometime after it. You mentioned in your above reply that the amplification is done before ADC - I don't doubt you (as I am no expert here), but is that something that is mentioned in some technical document for the latest Canon sensors? I'd be interested in the details, and what would explain the unexpectedly large increase in IQ when the ISO is increased from below IS0 400 to ISO 400.
 
<-- start Taboola -->