From Cobalt, Leica M vs world.
p.4 #1From Cobalt, Leica M vs world.
mzbe wrote:
ACR interpolates between available measured base illuminants in a profile - no user choice required (or possible)
https://www.cobalt-image.com/tutorials/the-white-balance/
Some reading on D65 vs. D50 on Cobalt's site ... so, it isn't like they aren't aware of D65's relevance ... so, why present the spectral response from D50 illuminant, instead of D65, when claiming how "inaccurate" different cameras are.
Shifting gears a bit ... so, am I to understand that basically Spectre ... is Cobalt doing for other cameras, an emulation of what Hassy is doing with HNCS (never mind the variance of D50 vs. D65 illuminants) to extract the most amount of spectral separation available to the specific camera.
This, then being different from Cobalt profiles where the intent is to emulate a series of different camera / film stock profiles.
RustyBug wrote
So, do you then have to choose between warm, cool, uber-warm profiles with their triple illuminant, or does one size fit all? Not sure I follow, here.https://www.cobalt-image.com/tutorials/the-white-balance/
Some reading on D65 vs. D50 on Cobalt's site ... so, it isn't like they aren't aware of D65's relevance ... so, why present the spectral response from D50 illuminant, instead of D65, when claiming how "inaccurate" different cameras are.
Shifting gears a bit ... so, am I to understand that basically Spectre ... is Cobalt doing for other cameras, an emulation of what Hassy is doing with HNCS (never mind the variance of D50 vs. D65 illuminants) to extract the most amount of spectral separation available to the specific camera.
This, then being different from Cobalt profiles where the intent is to emulate a series of different camera / film stock profiles.
Sep 07, 2026 at 07:26 PM
p.4 #2From Cobalt, Leica M vs world.
@RustyBug I think the misunderstanding is simpler than it may appear.
Your red-light-on-a-green-object example is correct, but it does not describe how our sensor measurement is performed. We are not photographing coloured objects illuminated by a D50 lamp. The sensor is measured wavelength by wavelength with a monochromator, so the blue response is measured directly just like the green and red responses. No part of the spectrum is missing because D50 is “too warm”.
D50 is applied afterwards as a reference illuminant when calculating that particular score. Using D65 would change the weighting and could produce somewhat different results, so it may certainly be interesting as an additional benchmark. But it would not make the measurement more complete or intrinsically more correct: D50 and D65 are simply two different standard daylight distributions.
In short: D50 does not illuminate or limit the spectral scan. It is the reference used to interpret the already measured sensor response.
Your red-light-on-a-green-object example is correct, but it does not describe how our sensor measurement is performed. We are not photographing coloured objects illuminated by a D50 lamp. The sensor is measured wavelength by wavelength with a monochromator, so the blue response is measured directly just like the green and red responses. No part of the spectrum is missing because D50 is “too warm”.
D50 is applied afterwards as a reference illuminant when calculating that particular score. Using D65 would change the weighting and could produce somewhat different results, so it may certainly be interesting as an additional benchmark. But it would not make the measurement more complete or intrinsically more correct: D50 and D65 are simply two different standard daylight distributions.
In short: D50 does not illuminate or limit the spectral scan. It is the reference used to interpret the already measured sensor response.
Sep 07, 2026 at 11:06 PM
p.4 #3From Cobalt, Leica M vs world.
Ulysseita wrote:
@RustyBug@ I think the misunderstanding is simpler than it may appear.
Your red-light-on-a-green-object example is correct, but it does not describe how our sensor measurement is performed. We are not photographing coloured objects illuminated by a D50 lamp. The sensor is measured wavelength by wavelength with a monochromator, so the blue response is measured directly just like the green and red responses. No part of the spectrum is missing because D50 is “too warm”.
D50 is applied afterwards as a reference illuminant when calculating that particular score. Using D65 would change the weighting and could produce somewhat different results, so...Show more →
Okay, so if D50 lamp is NOT the illuminant ... then what is the source illuminant from which the monochromator is measuring?
From AI:
The Light Color of the Illuminant:
Because a monochromator's sole purpose is to isolate individual colors out of a mixed source, the illuminant light color must be raw, continuous "White" light containing all wavelengths simultaneously.If you feed a monochromator a single-color light source (like a green laser), it won't have any other wavelengths to separate. To ensure it can measure across the full ultraviolet, visible, and infrared spectrums, instruments utilize specialized, high-intensity laboratory lamps:
For the Visible Spectrum (380nm – 780nm): The illuminant is a Tungsten-Halogen lamp. Natively, its light color looks like a warm, bright yellowish-white (similar to a premium studio hot light or standard architectural bulb). It provides a completely continuous, gap-free rainbow of wavelengths across the entire visible spectrum.
Is this ^ correct?
This, then meaning your are essentially measuring the BFA influence, similar to DXO.
Then, assuming that this has been a misunderstanding that an incomplete illuminant (D50 vs. D65) was being used ... back to my earlier question of how accurately does your correction achieve what Hassy has achieved where I get what I shot. Meaning, that if I'm set at 6500K (D65), and the illumination of my scene is 4300K reflecting off a neutral subject, that it registers as 4300K ... thereby revealing the TRUE color of the light (vs. a WB adjustment, that emulates 4300K as the neutral point).
I ask that because I do prefer the color separation that I get with my Hassy (vs. M10-R for instance). I'd like to understand "how close" can I get to that with my other cameras. I mostly shoot my (non-event) camera's at 6500K to get the REAL color of the light ... i.e. not the same as shooting for a WB adjusted color. As a technical matter, there are times that I WANT my light to render the color that the LIGHT is, with the subject merely being the canvas for the light color (previous image being such an example). That is the inverse of a WB adjusted correction.
I realize that this isn't how many folks approach things, but for me, 6500K is the anchor point of neutrality.
I understand that "Daylight" = 10-2, etc. as the temps move across the day, but also as your orientation faces N-E-S-W, via AI=AR, the color of your illuminant is in constant variation, unless you are in a controlled environment. For that reason, I want to SEE what the colors are (i.e. 4300K = 4300K, etc.), then make my decisions if I want to retain / reveal that native illuminant (or with the understanding of its native spectrum), or make an adjustment to the WB in the case of revealing the native color of the object vs. the native color of the light.
I can't have it both ways, simultaneously ... except when my illuminant IS 6500K (which is a very infrequent, not consistently reproducible in ambient settings). So, my preference is optioned to choose 6500K so I can understand what the native color of the light is in that moment. From that, I can assess more. Bottom line (for me), I want 4300K = 4300K, etc. If I continue to shoot at 6500K, how closely would a Spectre profiled, non-Hassy body render to the the Hassy HNCS rendering. So, in that regard, I suppose I'm using HNCS as a pseudo-reference point.
@RustyBug@ I think the misunderstanding is simpler than it may appear.
Your red-light-on-a-green-object example is correct, but it does not describe how our sensor measurement is performed. We are not photographing coloured objects illuminated by a D50 lamp. The sensor is measured wavelength by wavelength with a monochromator, so the blue response is measured directly just like the green and red responses. No part of the spectrum is missing because D50 is “too warm”.
D50 is applied afterwards as a reference illuminant when calculating that particular score. Using D65 would change the weighting and could produce somewhat different results, so...Show more →
Okay, so if D50 lamp is NOT the illuminant ... then what is the source illuminant from which the monochromator is measuring?
From AI:
The Light Color of the Illuminant:
Because a monochromator's sole purpose is to isolate individual colors out of a mixed source, the illuminant light color must be raw, continuous "White" light containing all wavelengths simultaneously.If you feed a monochromator a single-color light source (like a green laser), it won't have any other wavelengths to separate. To ensure it can measure across the full ultraviolet, visible, and infrared spectrums, instruments utilize specialized, high-intensity laboratory lamps:
For the Visible Spectrum (380nm – 780nm): The illuminant is a Tungsten-Halogen lamp. Natively, its light color looks like a warm, bright yellowish-white (similar to a premium studio hot light or standard architectural bulb). It provides a completely continuous, gap-free rainbow of wavelengths across the entire visible spectrum.
Is this ^ correct?
This, then meaning your are essentially measuring the BFA influence, similar to DXO.
Then, assuming that this has been a misunderstanding that an incomplete illuminant (D50 vs. D65) was being used ... back to my earlier question of how accurately does your correction achieve what Hassy has achieved where I get what I shot. Meaning, that if I'm set at 6500K (D65), and the illumination of my scene is 4300K reflecting off a neutral subject, that it registers as 4300K ... thereby revealing the TRUE color of the light (vs. a WB adjustment, that emulates 4300K as the neutral point).
I ask that because I do prefer the color separation that I get with my Hassy (vs. M10-R for instance). I'd like to understand "how close" can I get to that with my other cameras. I mostly shoot my (non-event) camera's at 6500K to get the REAL color of the light ... i.e. not the same as shooting for a WB adjusted color. As a technical matter, there are times that I WANT my light to render the color that the LIGHT is, with the subject merely being the canvas for the light color (previous image being such an example). That is the inverse of a WB adjusted correction.
I realize that this isn't how many folks approach things, but for me, 6500K is the anchor point of neutrality.
I understand that "Daylight" = 10-2, etc. as the temps move across the day, but also as your orientation faces N-E-S-W, via AI=AR, the color of your illuminant is in constant variation, unless you are in a controlled environment. For that reason, I want to SEE what the colors are (i.e. 4300K = 4300K, etc.), then make my decisions if I want to retain / reveal that native illuminant (or with the understanding of its native spectrum), or make an adjustment to the WB in the case of revealing the native color of the object vs. the native color of the light.
I can't have it both ways, simultaneously ... except when my illuminant IS 6500K (which is a very infrequent, not consistently reproducible in ambient settings). So, my preference is optioned to choose 6500K so I can understand what the native color of the light is in that moment. From that, I can assess more. Bottom line (for me), I want 4300K = 4300K, etc. If I continue to shoot at 6500K, how closely would a Spectre profiled, non-Hassy body render to the the Hassy HNCS rendering. So, in that regard, I suppose I'm using HNCS as a pseudo-reference point.
Sep 08, 2026 at 06:00 AM
p.4 #4From Cobalt, Leica M vs world.
@RustyBug
I think part of the confusion here comes from mixing the normal conversational use of words like “color”, “white light” and “the color of an object” with the much stricter technical vocabulary of colorimetry. This happens very easily in this field because we use the same words for rather different concepts.
Your question, “What color is the object?”, is actually a very good example.
In normal conversation it is a perfectly clear and meaningful question.
In technical colorimetry, however, the question is fundamentally incomplete. An object does not possess “color” as an intrinsic physical property in the same way that it possesses mass, dimensions, or a spectral reflectance.
What the object physically has is a spectral reflectance function (R(lambda)) — or, depending on the material, spectral transmittance, absorption, etc.
Color is the perceptual result. In other words, color exists in the observer, not inside the object.
The CIE definition itself describes perceived color as a characteristic of visual perception, and explicitly states that it depends not only on the spectral distribution of the stimulus, but also on its size, shape, surround, the observer’s state of adaptation, and the viewing situation.
Your photograph is actually an excellent illustration of this.
Let us assume, for simplicity, that both faces are made of perfectly homogeneous material, have identical spectral reflectance, and behave as Lambertian diffusers.
The two surfaces can therefore have exactly the same intrinsic spectral reflectance (R(lambda)).
But if they receive two different incident spectral distributions (E_1(lambda)) and (E_2(lambda)), the light leaving them will also have two different spectral distributions.
For an ideal Lambertian surface, approximately:
L(lambda) propto E(lambda)R(lambda)
So we have:
L_1(lambda) propto E_1(lambda)R(lambda)
and
L_2(lambda) propto E_2(lambda)R(lambda)
Same object reflectance. Different light reaching the observer. And therefore potentially different perceived colors.
That is why asking for the “real color” of the object has no unique technical answer without specifying an illuminant and viewing conditions.
Human vision then adds another layer of complexity: chromatic adaptation, adaptation luminance, surround, local contrast, photopic/mesopic/scotopic state, field of view, simultaneous color induction, prior visual context, and so on.
Our visual system is remarkably good at estimating surface properties despite changes in illumination — what we loosely call color constancy — but that does not mean that the physical spectrum reaching our eyes remains unchanged.
This is also why the idea that “full spectrum white light reveals the inherent color of the object” is problematic technically.
There is no unique physical spectrum called “white”.
D50 can look white.
D65 can look white.
Illuminant A can appear essentially white after sufficient chromatic adaptation.
Yet their spectral power distributions are very different.
Now, regarding the monochromator: the AI answer you quoted is essentially correct in the practical sense, but its explanation is slightly misleading.
A monochromator does not technically require a source that is “white” in the perceptual sense. It requires a broadband source providing sufficient radiant energy over the wavelength range we want to measure.
For visible/NIR work, quartz-tungsten-halogen sources are very commonly used because they provide a smooth continuous spectrum. Xenon, deuterium and other sources can be used depending on the wavelength range and instrument design.
The important point is this:
**the source feeding our monochromator is NOT D50.**
And its CCT does not define the illuminant under which the camera is being evaluated.
The broadband lamp is basically an energy source.
The monochromator takes that energy and isolates a very narrow wavelength interval.
We expose the sensor at, say, 400 nm, measure its response, then 410 nm, 420 nm, etc., across the visible spectrum.
What we obtain are the actual spectral sensitivity functions of the sensor:
S_R(lambda),quad S_G(lambda),quad S_B(lambda)
The radiant output at every wavelength is independently measured/calibrated, so the response can be normalized to the known amount of incident radiant energy.
The original spectral shape or CCT of the broadband lamp therefore does not become part of the “color” measurement. It matters instrumentally — available power, stability, signal-to-noise ratio, etc. — but it is not a colorimetric reference illuminant.
Quartz-tungsten-halogen sources are indeed widely used for exactly this reason: they have a smooth continuous spectrum through the visible and NIR and are very stable.
This distinction is important because once we have measured (S_R(lambda)), (S_G(lambda)), and (S_B(lambda)), we do not need to physically illuminate the camera with D50, D65 or A to determine how it responds under those illuminants.
We can calculate it.
For any surface reflectance (R(lambda)) and any illuminant (E(lambda)), the response of camera channel (i) is essentially:
C_i=int E(lambda)R(lambda)S_i(lambda),dlambda
Change (E(lambda)) from D50 to D65 to A and we can calculate the corresponding camera response from the same measured spectral sensitivities.
So D50 is not a limitation imposed during measurement. It is a reference condition used later when evaluating the colorimetric behaviour.
This also answers the question about whether Hasselblad, Leica, etc. “benchmark their BFA toward D65, D50 or A”.
I cannot speak for their internal design process, but fundamentally a Bayer CFA does not have to be thought of as being physically “D50” or “D65”. It has three spectral sensitivity functions. Once those functions are known, their colorimetric properties can be evaluated under any illuminant.
Why, then, do we report our tests using D50? Because D50 is not an arbitrary choice we made.
It is one of the central reference conditions of international color management, graphic arts and photographic reproduction.
The CIE now formally defines D50 as a standard illuminant, alongside A and D65, and specifically notes its extensive use in graphic arts and photography.
More importantly for our field, the ICC Profile Connection Space is D50-based. Colorimetric data entering an ICC workflow is chromatically adapted to the D50 PCS reference condition.
Print evaluation environments and standardized graphic-arts viewing conditions are also based around D50.
One small technical clarification: CIELAB mathematically can be calculated relative to different reference whites, so it is not correct in an absolute mathematical sense to say that “Lab can only be D50”. But in an ICC/graphic-arts workflow, when we talk about PCS Lab and standardized reproduction colorimetry, D50 is the reference condition.
Could we run exactly the same analysis under D65?
Absolutely. And perhaps it would actually be interesting to publish that comparison.
But D50 and D65 are relatively close daylight illuminants, so I would expect the general ranking and conclusions to change relatively little, although individual errors would of course move.
Illuminant A would be a much more severe change and would be interesting for a different reason, because its spectral distribution is dramatically weighted toward longer wavelengths compared with either D50 or D65.
But none of these tests would tell us the “true color” of the object.
They would tell us how accurately a particular sensor represents a set of spectral reflectances under a defined illuminant and a defined observer model.
And that distinction is basically what colorimetry is all about.
I think part of the confusion here comes from mixing the normal conversational use of words like “color”, “white light” and “the color of an object” with the much stricter technical vocabulary of colorimetry. This happens very easily in this field because we use the same words for rather different concepts.
Your question, “What color is the object?”, is actually a very good example.
In normal conversation it is a perfectly clear and meaningful question.
In technical colorimetry, however, the question is fundamentally incomplete. An object does not possess “color” as an intrinsic physical property in the same way that it possesses mass, dimensions, or a spectral reflectance.
What the object physically has is a spectral reflectance function (R(lambda)) — or, depending on the material, spectral transmittance, absorption, etc.
Color is the perceptual result. In other words, color exists in the observer, not inside the object.
The CIE definition itself describes perceived color as a characteristic of visual perception, and explicitly states that it depends not only on the spectral distribution of the stimulus, but also on its size, shape, surround, the observer’s state of adaptation, and the viewing situation.
Your photograph is actually an excellent illustration of this.
Let us assume, for simplicity, that both faces are made of perfectly homogeneous material, have identical spectral reflectance, and behave as Lambertian diffusers.
The two surfaces can therefore have exactly the same intrinsic spectral reflectance (R(lambda)).
But if they receive two different incident spectral distributions (E_1(lambda)) and (E_2(lambda)), the light leaving them will also have two different spectral distributions.
For an ideal Lambertian surface, approximately:
L(lambda) propto E(lambda)R(lambda)
So we have:
L_1(lambda) propto E_1(lambda)R(lambda)
and
L_2(lambda) propto E_2(lambda)R(lambda)
Same object reflectance. Different light reaching the observer. And therefore potentially different perceived colors.
That is why asking for the “real color” of the object has no unique technical answer without specifying an illuminant and viewing conditions.
Human vision then adds another layer of complexity: chromatic adaptation, adaptation luminance, surround, local contrast, photopic/mesopic/scotopic state, field of view, simultaneous color induction, prior visual context, and so on.
Our visual system is remarkably good at estimating surface properties despite changes in illumination — what we loosely call color constancy — but that does not mean that the physical spectrum reaching our eyes remains unchanged.
This is also why the idea that “full spectrum white light reveals the inherent color of the object” is problematic technically.
There is no unique physical spectrum called “white”.
D50 can look white.
D65 can look white.
Illuminant A can appear essentially white after sufficient chromatic adaptation.
Yet their spectral power distributions are very different.
Now, regarding the monochromator: the AI answer you quoted is essentially correct in the practical sense, but its explanation is slightly misleading.
A monochromator does not technically require a source that is “white” in the perceptual sense. It requires a broadband source providing sufficient radiant energy over the wavelength range we want to measure.
For visible/NIR work, quartz-tungsten-halogen sources are very commonly used because they provide a smooth continuous spectrum. Xenon, deuterium and other sources can be used depending on the wavelength range and instrument design.
The important point is this:
**the source feeding our monochromator is NOT D50.**
And its CCT does not define the illuminant under which the camera is being evaluated.
The broadband lamp is basically an energy source.
The monochromator takes that energy and isolates a very narrow wavelength interval.
We expose the sensor at, say, 400 nm, measure its response, then 410 nm, 420 nm, etc., across the visible spectrum.
What we obtain are the actual spectral sensitivity functions of the sensor:
S_R(lambda),quad S_G(lambda),quad S_B(lambda)
The radiant output at every wavelength is independently measured/calibrated, so the response can be normalized to the known amount of incident radiant energy.
The original spectral shape or CCT of the broadband lamp therefore does not become part of the “color” measurement. It matters instrumentally — available power, stability, signal-to-noise ratio, etc. — but it is not a colorimetric reference illuminant.
Quartz-tungsten-halogen sources are indeed widely used for exactly this reason: they have a smooth continuous spectrum through the visible and NIR and are very stable.
This distinction is important because once we have measured (S_R(lambda)), (S_G(lambda)), and (S_B(lambda)), we do not need to physically illuminate the camera with D50, D65 or A to determine how it responds under those illuminants.
We can calculate it.
For any surface reflectance (R(lambda)) and any illuminant (E(lambda)), the response of camera channel (i) is essentially:
C_i=int E(lambda)R(lambda)S_i(lambda),dlambda
Change (E(lambda)) from D50 to D65 to A and we can calculate the corresponding camera response from the same measured spectral sensitivities.
So D50 is not a limitation imposed during measurement. It is a reference condition used later when evaluating the colorimetric behaviour.
This also answers the question about whether Hasselblad, Leica, etc. “benchmark their BFA toward D65, D50 or A”.
I cannot speak for their internal design process, but fundamentally a Bayer CFA does not have to be thought of as being physically “D50” or “D65”. It has three spectral sensitivity functions. Once those functions are known, their colorimetric properties can be evaluated under any illuminant.
Why, then, do we report our tests using D50? Because D50 is not an arbitrary choice we made.
It is one of the central reference conditions of international color management, graphic arts and photographic reproduction.
The CIE now formally defines D50 as a standard illuminant, alongside A and D65, and specifically notes its extensive use in graphic arts and photography.
More importantly for our field, the ICC Profile Connection Space is D50-based. Colorimetric data entering an ICC workflow is chromatically adapted to the D50 PCS reference condition.
Print evaluation environments and standardized graphic-arts viewing conditions are also based around D50.
One small technical clarification: CIELAB mathematically can be calculated relative to different reference whites, so it is not correct in an absolute mathematical sense to say that “Lab can only be D50”. But in an ICC/graphic-arts workflow, when we talk about PCS Lab and standardized reproduction colorimetry, D50 is the reference condition.
Could we run exactly the same analysis under D65?
Absolutely. And perhaps it would actually be interesting to publish that comparison.
But D50 and D65 are relatively close daylight illuminants, so I would expect the general ranking and conclusions to change relatively little, although individual errors would of course move.
Illuminant A would be a much more severe change and would be interesting for a different reason, because its spectral distribution is dramatically weighted toward longer wavelengths compared with either D50 or D65.
But none of these tests would tell us the “true color” of the object.
They would tell us how accurately a particular sensor represents a set of spectral reflectances under a defined illuminant and a defined observer model.
And that distinction is basically what colorimetry is all about.
Sep 08, 2026 at 06:48 AM
|
p.4 #5From Cobalt, Leica M vs world.
Thanks, I think you are understanding my question(s) much better, now.
Yes, I understand the D50 is a standard ... and (artificial) lighting used in viewing applications is inherently closer to D50 standard, than D65 (hard to find 6500K bulbs, 5000K-5500K much more prevalent).
I guess I would differ about your statement of proximity for D50 vs. D65 being relatively close. Imo a 1500K shift isn't trivial. But, hey, I guess that's just me.
Thanks again for the input and entertaining my on the subject. And, I truly appreciate your recognition of my posted image as an illustration to the matter.
Yes, I understand the D50 is a standard ... and (artificial) lighting used in viewing applications is inherently closer to D50 standard, than D65 (hard to find 6500K bulbs, 5000K-5500K much more prevalent).
I guess I would differ about your statement of proximity for D50 vs. D65 being relatively close. Imo a 1500K shift isn't trivial. But, hey, I guess that's just me.
Thanks again for the input and entertaining my
on the subject. And, I truly appreciate your recognition of my posted image as an illustration to the matter. Sep 08, 2026 at 07:07 AM
p.4 #6From Cobalt, Leica M vs world.
Raamiel wrote:
@dalegaspi@
I agree with some of your points, but I’d like to clarify a few aspects more precisely.
To better understand Adobe profiling, it helps to keep the ideal architecture in mind.
A DNG profile (`.dcp`) is designed to be modular. It can contain characterization data for one illuminant, or for two or three illuminants. In its simplest form it can contain only a single matrix, but a complete profile—such as Cobalt profiles, and also the various Adobe Standard profiles—is composed of:
ColorMatrix
ForwardMatrix
HueSatMap
LookTable
ToneCurve
The first three are illuminant-specific, while `LookTable` and `ToneCurve` are global. They come after the colorimetric stage.
The typical embedded profile found inside...Show more →
thank you for responding, i appreciate the exhaustive explanation.
one final question.. am i correct about this statement?
With Spectre, the thing is you think about a linear response (sorry couldn't think of a better term) that allows you to make uniformity across camera sensors...which also means that the Color science of manufacturers are essentially out the window...you can't have "Canon colors" when you are using this in conjunction with other cameras since you can't have both. the "S curve" is gone. Also, remember when i said "accurate" is not necessarily "pleasing?" yeah, that applies here now. I guess you can say that Cobalt is "less opinionated"
@dalegaspi@
I agree with some of your points, but I’d like to clarify a few aspects more precisely.
To better understand Adobe profiling, it helps to keep the ideal architecture in mind.
A DNG profile (`.dcp`) is designed to be modular. It can contain characterization data for one illuminant, or for two or three illuminants. In its simplest form it can contain only a single matrix, but a complete profile—such as Cobalt profiles, and also the various Adobe Standard profiles—is composed of:
ColorMatrix
ForwardMatrix
HueSatMap
LookTable
ToneCurve
The first three are illuminant-specific, while `LookTable` and `ToneCurve` are global. They come after the colorimetric stage.
The typical embedded profile found inside...Show more →
thank you for responding, i appreciate the exhaustive explanation.
one final question.. am i correct about this statement?
With Spectre, the thing is you think about a linear response (sorry couldn't think of a better term) that allows you to make uniformity across camera sensors...which also means that the Color science of manufacturers are essentially out the window...you can't have "Canon colors" when you are using this in conjunction with other cameras since you can't have both. the "S curve" is gone. Also, remember when i said "accurate" is not necessarily "pleasing?" yeah, that applies here now. I guess you can say that Cobalt is "less opinionated"
Sep 08, 2026 at 09:08 AM
p.4 #7From Cobalt, Leica M vs world.
So, I've been going through the reports for the various cameras ... never minding the D50 vs. D65 aspect ... the reports indicate how far off Adobe Standard is from the D50 reference (and how much correction Spectre presents). Two questions:
1) Do I recall that Adobe change their "Standard" profile(s) in recent years. Curious to whether or not there was a difference in the Fadgi scores between Adobe's different profile(s) and profile versions?
2). If these reports are using Adobe profiles, and not OEM profiles ... curious what the Fadgi scores are like for OEM profiles for a given camera. For instance, how does the M10R with Adobe Standard profile compare to M10R with Leica native profile for the M10R (etc. for other Leica, Sony, et al).
Any chance you have OEM on OEM profile results, or just Adobe Standard on multiple brands / models?
Will be curious also to see when you have SL3-P (and S1R II) data, as those are likely to be "long term" cameras for me.
Also, as I was watching one of the video's ... it was advocating the difference between Adobe Standard and Cobalt HDR, but while the HDR made things "punchier", it also struck me that the Vivid or Landscape profiles of Adobe were not being compared with, either. I guess just a lot of variables for profiles ... which I totally get, akin to profiles in film stock.
I also noticed that it seemed like there was ONE specific red swatch that ALL models were unable to get a good fadgi correction on. Just some things I notice, as I continue to try and better understand what your profiles might do for me.
1) Do I recall that Adobe change their "Standard" profile(s) in recent years. Curious to whether or not there was a difference in the Fadgi scores between Adobe's different profile(s) and profile versions?
2). If these reports are using Adobe profiles, and not OEM profiles ... curious what the Fadgi scores are like for OEM profiles for a given camera. For instance, how does the M10R with Adobe Standard profile compare to M10R with Leica native profile for the M10R (etc. for other Leica, Sony, et al).
Any chance you have OEM on OEM profile results, or just Adobe Standard on multiple brands / models?
Will be curious also to see when you have SL3-P (and S1R II) data, as those are likely to be "long term" cameras for me.
Also, as I was watching one of the video's ... it was advocating the difference between Adobe Standard and Cobalt HDR, but while the HDR made things "punchier", it also struck me that the Vivid or Landscape profiles of Adobe were not being compared with, either. I guess just a lot of variables for profiles ... which I totally get, akin to profiles in film stock.
I also noticed that it seemed like there was ONE specific red swatch that ALL models were unable to get a good fadgi correction on. Just some things I notice, as I continue to try and better understand what your profiles might do for me.
Sep 11, 2026 at 09:18 PM