Perhaps. To quote another photographer, “You say yes, I say no.”
Either that or they believe, perhaps correctly, that ISO invariance is not something people do well to think about.
For those who don’t need it, it is not overrated. It is irrelevant.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
ISO invariance - do you use it?
- Thread starter IPTAK
- Start date
The Ghost of Caravaggio
Senior Member
- Messages
- 1,006
- Solutions
- 1
- Reaction score
- 747
Sometimes.
With the FUJIFILM X-Pro2 ISO invariance is useful when I’m in a dynamic situation and the lighting changes. For example, a wedding reception or other event where I’m doing candid work and subjects are in different lighting conditions.
With the camera in manual exposure mode, I set the shutter speed fast enough to minimize resolution loss from camera shake and, or subject motion. Likewise. I pick a lens aperture to achieve a conservative DOF. I try to use a similar subject to camera distances (for useful DOF). I don’t think about them again unless I change lenses or people are dancing, etc.
In bright light (which is uncommon) my camera ISO setting is 200. In low light I use a camera ISO setting of 800. The X-Pro 2 has dual conversion gain. There are two, separate ISOless regions – ISO 100-636 and 800 - 25,600.
I use the X-Pro 2 optical view finder so I can focus, compose and decide when to press the shutter button in ambient light. I can also see what’s happening outside the frame line-estimates and choose my composition accordingly. A minor benefit is increased battery life.
Camera operation is simplified to proper focusing. I use the optical view finder and AF-S with the multi-mode focus region. The AF focus region is fixed in the frame center. A half shutter-button press initiates AF. Focus is confirmed when the focus region box turns green a small green dot appears in the OVF and a the sound of a faint audio beep. The DOF is set such that the focus and recompose method is practical.
This allows me to concentrate on composition and timing. I eliminate the light meter and histogram displays in the OVF. I don’t have to think about switching between automated focusing modes/options when situations change.
I don’t review images when I’m working. People who used/use film regularly will understand.
I use raw files.
During post production rendering, image brightness is optimized for each selected photo.
I am well-aware using all the camera’s automation features (and raw files) can produce excellent results as well. In some ways full automation is more complicated. It takes me time to learn how all the automation features work together. I have to trust I won’t loose a photograph because I miss set an automation parameter or the people who designed the automation algorithms didn’t anticipate my needs. Perhaps this means FUJIFILM’s automation is poorly implemented. If your camera’s automation would achieve a 100% success rate in these circumstances, then congratulations.
Lastly, I enjoy operating this camera in a way that is quite similar to how I used a film rangefinder. The main difference is I use AF-S instead of turning the lens ring. In MF mode (using the focus ring manually), the X-Pro 2 has a small digital-image display in the lower corner to check focus. The display can be set to display just the focus region with a simulated split-image rangefinder or with focus peaking. This display is too small for me, so I use AF-S.
--
_____________________
“…the mathematical rules of probability theory are not merely rules for calculating frequencies of random variables; they are also the unique consistent rules for conducting inference (i.e., plausible reasoning)”
E.T Jaynes, Probability Theory: The Logic of Science
With the FUJIFILM X-Pro2 ISO invariance is useful when I’m in a dynamic situation and the lighting changes. For example, a wedding reception or other event where I’m doing candid work and subjects are in different lighting conditions.
With the camera in manual exposure mode, I set the shutter speed fast enough to minimize resolution loss from camera shake and, or subject motion. Likewise. I pick a lens aperture to achieve a conservative DOF. I try to use a similar subject to camera distances (for useful DOF). I don’t think about them again unless I change lenses or people are dancing, etc.
In bright light (which is uncommon) my camera ISO setting is 200. In low light I use a camera ISO setting of 800. The X-Pro 2 has dual conversion gain. There are two, separate ISOless regions – ISO 100-636 and 800 - 25,600.
I use the X-Pro 2 optical view finder so I can focus, compose and decide when to press the shutter button in ambient light. I can also see what’s happening outside the frame line-estimates and choose my composition accordingly. A minor benefit is increased battery life.
Camera operation is simplified to proper focusing. I use the optical view finder and AF-S with the multi-mode focus region. The AF focus region is fixed in the frame center. A half shutter-button press initiates AF. Focus is confirmed when the focus region box turns green a small green dot appears in the OVF and a the sound of a faint audio beep. The DOF is set such that the focus and recompose method is practical.
This allows me to concentrate on composition and timing. I eliminate the light meter and histogram displays in the OVF. I don’t have to think about switching between automated focusing modes/options when situations change.
I don’t review images when I’m working. People who used/use film regularly will understand.
I use raw files.
During post production rendering, image brightness is optimized for each selected photo.
I am well-aware using all the camera’s automation features (and raw files) can produce excellent results as well. In some ways full automation is more complicated. It takes me time to learn how all the automation features work together. I have to trust I won’t loose a photograph because I miss set an automation parameter or the people who designed the automation algorithms didn’t anticipate my needs. Perhaps this means FUJIFILM’s automation is poorly implemented. If your camera’s automation would achieve a 100% success rate in these circumstances, then congratulations.
Lastly, I enjoy operating this camera in a way that is quite similar to how I used a film rangefinder. The main difference is I use AF-S instead of turning the lens ring. In MF mode (using the focus ring manually), the X-Pro 2 has a small digital-image display in the lower corner to check focus. The display can be set to display just the focus region with a simulated split-image rangefinder or with focus peaking. This display is too small for me, so I use AF-S.
--
_____________________
“…the mathematical rules of probability theory are not merely rules for calculating frequencies of random variables; they are also the unique consistent rules for conducting inference (i.e., plausible reasoning)”
E.T Jaynes, Probability Theory: The Logic of Science
Last edited:
Quarkcharmed
Senior Member
Users shouldn't have to be thinking about it. Camera software should provide appropriate shooting modes and settings that fully utilise the camera capabilities.
ThrillaMozilla
Veteran Member
Sad, isn't it? I would be happy if I could be confident that the camera would not blow major highlights.
SrMi
Veteran Member
Highlight clipping is dependent on exposure and ISO. ISO invariance is unrelated to highlight clipping.
Last edited:
SrMi
Veteran Member
I only have a Q3, which improves by about .5 stops with DCG (@ISO 400). X2D and R5 improve more at the DCG point, making the difference more visible.
Here are excerpts at ISO 1600, 400, and 200, with the same exposure, lifted in post to equalize brightness. Everyone can test their camera and decide whether to pay attention to DCG.
ISO 1600
ISO 400
ISO 200
SrMi
Veteran Member
ISO standard is irrelevant when shooting raw.
For some people, anything but P mode is an "occult process." Other photographers want to maximize the technical quality of their images when they have the opportunity. That requires knowledge of their tools
Quarkcharmed
Senior Member
It depends on what you mean by 'relation'. It's very related indeed, the main reason to use ISO invariance is to avoid/reduce highlight clipping.
--
https://www.instagram.com/quarkcharmed/
https://500px.com/quarkcharmed
Last edited:
SrMi
Veteran Member
I can agree with that. The other use is to avoid constant and unnecessary fiddling with manual ISO, but the former is more relevant.
Yes, I know. It is unfortunate that “ISO” is used in so many different contexts. It causes confusion.
Hmmm, I don’t use P much, but I also don’t think I can actually leverage the undocumented and subject to change internals of the camera to take better pictures in M. Instead I go by what I can see while I use the documented features of the camera. Your mileage may vary.
Maybe the cameras already do that. I am very skeptical of all this speculative stuff about invariance, etc. But I don’t object to people enjoying themselves any way they see fit with their gear.
SrMi
Veteran Member
To the best of my knowledge and experience, dual conversion gain is part of the hardware and cannot be changed for a camera model. Some manufacturers do not document it, but testing and measurement reveal that information. Leica specifies the Dual Basis ISO ranges in the manual for SL3.
The knowledge is available, and it can lead to technically better images. I do not understand why it should not be used.
SrMi
Veteran Member
Why speculative? The invariance ranges have been documented by measurements and can be verified with experiments.
ThrillaMozilla
Veteran Member
It's not speculative. The behavior of the cameras is fairly well tested and understood. It's just a little error-prone to use the information correctly, especially in the field.
Last edited:
Quarkcharmed
Senior Member
How is it speculative? It's based on measurements, tests and practice. It's measured on photonstophotos, I tested it myself and it works for me.
John Sheehy
Forum Pro
Nobody else can tell you whether it is better to accept a little bit more read noise or clip highlights. That is not a scientific or technical question; this is a subjective one which must be decided by you, the photographer, but based on technical facts, to understand the trade-offs.
IMO, the ISO-invariant bandwagon is premature in many cases, and most cameras or ISO ranges of specific cameras that are deemed "ISO-invariant" are not really so, unless your compositions are "mid-to-high key", lacking in darker tonal ranges. Almost all digital cameras use increasing analog gain going up all or most of the ISO range, and it's only a matter of how bright you want to make the darker areas of the scenes, which determines if there is a visible noise penalty from using an ISO setting lower than standard for your level of exposure.
Many people assume that ISO-invariance can be determined from graphs of noise or DR vs ISO setting, but the reality is that graphs can only clearly show "negatives", not "positives", for ISO invariance. The graphs at DxOMark and P2P are based on monolithic standard deviations of noise, not the actual spatial character of noise, which has as much or even more influence on the visibility and inconvenience of noise than modest standard deviation differences.
Imagine that you had an empty glass jar and containers of red, green, and blue sand. Imagine that you poured red for a while, then green for a while, then blue for a while, etc. By the time you filled the jar, you would have distinct bands of color which could be seen from quite a distance. If you go as far as you can from the jar, where you can still clearly see the separate colors, and someone else picked up the jar, put a lid on it and shook it vigorously, you would see all the color disappear and the jar would look grey to you. Yet, the standard deviation of the sand did not change, nor did any histogram you may have had of the grains of sand. Both P2P and DxOMark data is given in such a way that it is indistinguishable from and assumed to be the "shook up" version of the sand art, ignoring the "poured" or correlated version, and sensor read noise is actually a little bit like the poured sand.
SrMi
Veteran Member
The trade-off between loss of relevant detail in highlight vs more noise should be clear: more noise is always better than loss of detail. Noise can be handled quite well with recent tools. Lost detail may be recoverable either with post-processor reconstruction using unclipped channels (which can lead to discoloration) or using content-aware fill to add artificial detail in the white holes.
It depends on the amount of light, not the kind of image. Suppose the highest ISO needed without clipping highlights is in the ISO invariant range. In that case, you can use ISO invariance to reduce ISO (not exposure) and be safe from accidental highlight clipping without data loss.
Check out P2P's shadow improvement charts, and you will see that many modern cameras do not improve noise when raising ISO after the DCG point. Experiments have confirmed this. If you have an experiment that goes against it, please share it.
Experiments have confirmed the data measurements on P2P.
Quarkcharmed
Senior Member
It's a very good analogy with the jar and coloured sand, but in terms of our invariance problem, is raising the ISO setting equivalent to shaking the jar? If it's thermal noise, for example, I think it will be amplified regardless. Only regular patterns produced during ADC conversion will be reduced.
With my Canon R5, I don't see any significant difference when setting ISO 2-3 stops under, e.g. ISO 800-1600 instead of 6400-12800. There's high frequency banding (thin regular stripes) but it's there at high ISOs regardless of analog or digital pushing.
Also in the R5 (and other Canons) they only use analog amplification at whole ISO stops (100, 200, 400, 800, 1600...) and ISO is fully digital in between.
So even if there was banding, it'd make zero difference if you used, say, ISO 800 instead of 1200, but again you would be saving 2/3 stops of highlights.
ThrillaMozilla
Veteran Member
That doesn't seem to be true. The saturation data suggest otherwise.
From "Input-referred Read Noise", https://www.photonstophotos.net/Charts/RN_e.htm
I thought it might be true for the M6II, but not that either.
Last edited:
Quarkcharmed
Senior Member
It's normalised as far a I understand, so I'm not sure how you can infer that from the saturation data?
You can see triplets in the read noise, also in the PDR and shadow improvement charts.
Actually Bill Claff himself pointed it out a while ago
ISO on the R5
FYI Is this related to the person who posted some weeks back under the handle "vegetable"? I presume that this is a trolling post. Umm..what? Trolling? I thought the video was helpful. The video isn’t related to the issue. That's all you needed to say and not accuse me of trolling. It was...
Similar threads
Keyboard shortcuts
- f
- Forum