Raw exposure bias

Started Mar 12, 2015 | Discussions
Iliah Borg Forum Pro • Posts: 29,051
Raw exposure bias
8

Many Fujifilm cameras' raw files contain the tag 0x9650, which is raw exposure bias. This tag indicates how much the raw is underexposed, and consequently, recommends the amount of brightening a raw converter may additionally apply. You can see it using an exiftool command, like this:

exiftool -iso -RawExposureBias -T <filename>

For my series, executing over a series of shots (named XE2_iso00200.raf, XE2_iso00400.raf, etc) taken at different ISO settings

exiftool -iso -RawExposureBias -T XE2_iso*.raf

resulted in

200 -0.72

400 -0.72

800 -0.72

1600 -0.72

3200 -1.26

6400 -2.26

(which, by the way, indicates that the last ISO setting where the analog gain is applied is half a stop above ISO 1600)

-- hide signature --
bowportes Veteran Member • Posts: 4,236
Re: Raw exposure bias
1

Iliah Borg wrote:

Many Fujifilm cameras' raw files contain the tag 0x9650, which is raw exposure bias. This tag indicates how much the raw is underexposed, and consequently, recommends the amount of brightening a raw converter may additionally apply. You can see it using an exiftool command, like this:

exiftool -iso -RawExposureBias -T <filename>

For my series, executing over a series of shots (named XE2_iso00200.raf, XE2_iso00400.raf, etc) taken at different ISO settings

exiftool -iso -RawExposureBias -T XE2_iso*.raf

resulted in

200 -0.72

400 -0.72

800 -0.72

1600 -0.72

3200 -1.26

6400 -2.26

(which, by the way, indicates that the last ISO setting where the analog gain is applied is half a stop above ISO 1600)

Not certain of your point. ISO 3200 requires half the light of ISO 1600 (1 stop), while ISO 6400 requires 1/4 the light of ISO 1600 (2 stops). It 's no mystery. What are you suggesting?

No analog gain is applied above ISO 1600. The data in the raw file are identical at ISO 1600 and ISO 6400.

 bowportes's gear list:bowportes's gear list
Panasonic Lumix DMC-G5 Fujifilm X-M1 Fujifilm X-T1 Fujifilm X-Pro2 Panasonic Lumix G 20mm F1.7 ASPH +13 more
OP Iliah Borg Forum Pro • Posts: 29,051
Re: Raw exposure bias
2

ISO 3200 requires half the light of ISO 1600 (1 stop), while ISO 6400 requires 1/4 the light of ISO 1600 (2 stops).

Requires for what, may I ask?

No analog gain is applied above ISO 1600. The data in the raw file are identical at ISO 1600 and ISO 6400.

Let's look at three shots, ISO 1600, ISO 3200, ISO 6400 from Imaging Resource, all taken under the controlled and rather constant light:

http://216.18.212.226/PRODS/fuji-x-e2/FULLRES/XE2hSLI01600NR1.RAF

http://216.18.212.226/PRODS/fuji-x-e2/FULLRES/XE2hSLI03200NR1.RAF

http://216.18.212.226/PRODS/fuji-x-e2/FULLRES/XE2hSLI06400NR1.RAF

and run them through

exiftool -iso -shutterspeed -aperture -RawExposureBias -T /Users/iliah/Downloads/XE2hSLI0*.RAF

Result is

1600 1/200 8.0 -0.72

3200 1/400 8.0 -1.26

6400 1/800 8.0 -2.26

Given the light is constant, and ISO 1600 being the highest with gain, one could be expecting the raw values at 1600 to be 2x larger than at 3200. But that is not so.

ISO 1600: average green for the selected gray patch = 4918.9

ISO 3200: average green for the selected gray patch = 3489.4

ISO 6400: average green for the selected gray patch = 1640.7

The difference between 3200 and 6400 is very close to 2x; that is 1 EV. The difference between 1600 and 3200 is close to 1.4x, that is ≈0.5 EV. So, the analog gain is applied up to the ISO setting of 2500 (rounded).

If one looks at true shutter speeds (non-rounded), the values are:

1600 1/194

3200 1/388

6400 1/832

That is 1/832 instead of 1/776, or 0.1 EV difference.

-- hide signature --
nixda Veteran Member • Posts: 5,515
Re: Raw exposure bias

Iliah Borg wrote:

Many Fujifilm cameras' raw files contain the tag 0x9650, which is raw exposure bias. This tag indicates how much the raw is underexposed, and consequently, recommends the amount of brightening a raw converter may additionally apply. You can see it using an exiftool command, like this:

exiftool -iso -RawExposureBias -T <filename>

For my series, executing over a series of shots (named XE2_iso00200.raf, XE2_iso00400.raf, etc) taken at different ISO settings

exiftool -iso -RawExposureBias -T XE2_iso*.raf

resulted in

200 -0.72

400 -0.72

800 -0.72

1600 -0.72

3200 -1.26

6400 -2.26

(which, by the way, indicates that the last ISO setting where the analog gain is applied is half a stop above ISO 1600)

There is some useful info in all these tags. However, I cannot get that information with the command you specified. I am working with X-E1 files, but I also can't access the information using the X-E2 file from the gentleman who reported underexposure. All I get is a "-". Is there a tag group name that needs to be specified? Which version of exiftool are you using?

 nixda's gear list:nixda's gear list
Fujifilm X-E1 Fujifilm XF 14mm F2.8 R Fujifilm XF 18-55mm F2.8-4 R LM OIS Fujifilm XF 55-200mm F3.5-4.8 R LM OIS Fujifilm XF 27mm F2.8 +1 more
OP Iliah Borg Forum Pro • Posts: 29,051
Re: Raw exposure bias

Is there a tag group name that needs to be specified?

I copied the command from my terminal window

Which version of exiftool are you using?

Version is

ibmacbook:~ iliah$ exiftool -ver

9.87

Can it be some space missing, or maybe a part of the filename interpreted as a parameter?

The command without filename is

exiftool -iso -RawExposureBias -T

There needs to be a space between -T and the filename

-- hide signature --
nixda Veteran Member • Posts: 5,515
Re: Raw exposure bias
1

Iliah Borg wrote:

Is there a tag group name that needs to be specified?

I copied the command from my terminal window

Which version of exiftool are you using?

Version is

ibmacbook:~ iliah$ exiftool -ver

9.87

OK I got it now. I was using version 9.82. Latest is 9.89. Silly me; making sure one is up to date is the first thing I usually recommend when something like this happens...

 nixda's gear list:nixda's gear list
Fujifilm X-E1 Fujifilm XF 14mm F2.8 R Fujifilm XF 18-55mm F2.8-4 R LM OIS Fujifilm XF 55-200mm F3.5-4.8 R LM OIS Fujifilm XF 27mm F2.8 +1 more
OP Iliah Borg Forum Pro • Posts: 29,051
Re: Raw exposure bias
1

I was using version 9.82

Right, this tag appeared in exiftool 9.86.

-- hide signature --
nixda Veteran Member • Posts: 5,515
What is the meaning of raw exposure bias?

Iliah Borg wrote:

Many Fujifilm cameras' raw files contain the tag 0x9650, which is raw exposure bias. This tag indicates how much the raw is underexposed, and consequently, recommends the amount of brightening a raw converter may additionally apply. You can see it using an exiftool command, like this:

exiftool -iso -RawExposureBias -T <filename>

For my series, executing over a series of shots (named XE2_iso00200.raf, XE2_iso00400.raf, etc) taken at different ISO settings

exiftool -iso -RawExposureBias -T XE2_iso*.raf

resulted in

200 -0.72

400 -0.72

800 -0.72

1600 -0.72

3200 -1.26

6400 -2.26

(which, by the way, indicates that the last ISO setting where the analog gain is applied is half a stop above ISO 1600)

What is the meaning of raw exposure bias?

I am currently trying to get my head around the various exposure-related entries in the EXIF data and how they relate to the typical APEX quantities. I am having a very hard time to make full sense of it all. Does raw exposure bias factor into any of that? For example, does it affect how the camera determines "Brightness Value", "Light Value", or any other quantity?

 nixda's gear list:nixda's gear list
Fujifilm X-E1 Fujifilm XF 14mm F2.8 R Fujifilm XF 18-55mm F2.8-4 R LM OIS Fujifilm XF 55-200mm F3.5-4.8 R LM OIS Fujifilm XF 27mm F2.8 +1 more
OP Iliah Borg Forum Pro • Posts: 29,051
Re: What is the meaning of raw exposure bias?

What is the meaning of raw exposure bias?

It is an instruction for a raw converter to apply certain amount of default (silent, behind the scenes) brightness compensation.

You can shoot a scene at different ISO settings (including intermediate ones) and fixed white balance and see the progression. Next, to see if it depends on white balance, change it and take couple of shots at ISO 3200.

-- hide signature --
nixda Veteran Member • Posts: 5,515
Re: What is the meaning of raw exposure bias?

Iliah Borg wrote:

What is the meaning of raw exposure bias?

It is an instruction for a raw converter to apply certain amount of default (silent, behind the scenes) brightness compensation.

If the camera specifies that there should be a brightness compensation applied to an image, does it mean that the metering is somehow off, or is it a calibration issue of the analog-to-digital converter, or something else? In other words, why is it necessary to have such a bias?

Also, what does a value of -0.72 mean? Should the brightness be reduced or increased by 0.72 stops?

Finally, is the raw exposure bias indicating that ISO 800 isn't really ISO 800 but a different value (when comparing ISOs between different camera makers), which gets at the nation that Fuji tends to overstate its ISO numbers, compared to some other mainstream camera makers?

 nixda's gear list:nixda's gear list
Fujifilm X-E1 Fujifilm XF 14mm F2.8 R Fujifilm XF 18-55mm F2.8-4 R LM OIS Fujifilm XF 55-200mm F3.5-4.8 R LM OIS Fujifilm XF 27mm F2.8 +1 more
OP Iliah Borg Forum Pro • Posts: 29,051
Re: What is the meaning of raw exposure bias?
1

If the camera specifies that there should be a brightness compensation applied to an image, does it mean that the metering is somehow off

It only means that the compensation is to be applied to comply with some standard. In practice it is the compensation between the exposure meter calibration point for the given ISO setting and the SOS (Standard Output Sensitivity) ISO recommendation (Fujifilm are using SOS in their current cameras).

Also, what does a value of -0.72 mean? Should the brightness be reduced or increased by 0.72 stops?

Increased by 0.72 stops. Compare to baseline exposure compensation in DNG documentation.

Finally, is the raw exposure bias indicating that ISO 800 isn't really ISO 800 but a different value (when comparing ISOs between different camera makers), which gets at the nation that Fuji tends to overstate its ISO numbers, compared to some other mainstream camera makers?

This tag indicates that normal ISO overstating is slightly more than 2/3 of a stop, and more for high ISO. That is done in order to protect highlights and allow film simulations to work.

-- hide signature --
nixda Veteran Member • Posts: 5,515
Re: What is the meaning of raw exposure bias?

Iliah Borg wrote:

If the camera specifies that there should be a brightness compensation applied to an image, does it mean that the metering is somehow off

It only means that the compensation is to be applied to comply with some standard. In practice it is the compensation between the exposure meter calibration point for the given ISO setting and the SOS (Standard Output Sensitivity) ISO recommendation (Fujifilm are using SOS in their current cameras).

Also, what does a value of -0.72 mean? Should the brightness be reduced or increased by 0.72 stops?

Increased by 0.72 stops. Compare to baseline exposure compensation in DNG documentation.

Finally, is the raw exposure bias indicating that ISO 800 isn't really ISO 800 but a different value (when comparing ISOs between different camera makers), which gets at the nation that Fuji tends to overstate its ISO numbers, compared to some other mainstream camera makers?

This tag indicates that normal ISO overstating is slightly more than 2/3 of a stop, and more for high ISO. That is done in order to protect highlights and allow film simulations to work.

Fantastic! Thank you so much!

One final question (hopefully): do you have a formula, or a reference, or a link to where I can find how raw exposure bias is related to any APEX standard formulas?

E.g.,

Bv = log2 (A²/T/Sx/N) (Brightness value)

Bv: brightness value; A: f-number; T: exposure time; Sx: ISO; N: constant that establishes the relationship between the ASA arithmetic film speed ´┐╝and the ASA speed value

Is the 'N' related to raw exposure bias?

Or:

A²/T = B*Sx/K (exposure equation)

K: reflected-light meter calibration constant

Is the 'K' related to raw exposure bias?

I have the feeling that all of this is connected to apparent inconsistencies in the EXIF information of my X-E1 RAF files (and potentially those from other Fuji cameras).

 nixda's gear list:nixda's gear list
Fujifilm X-E1 Fujifilm XF 14mm F2.8 R Fujifilm XF 18-55mm F2.8-4 R LM OIS Fujifilm XF 55-200mm F3.5-4.8 R LM OIS Fujifilm XF 27mm F2.8 +1 more
nkbj
nkbj Regular Member • Posts: 263
Re: What is the meaning of raw exposure bias?
1

Iliah Borg wrote:

What is the meaning of raw exposure bias?

It is an instruction for a raw converter to apply certain amount of default (silent, behind the scenes) brightness compensation.

You can shoot a scene at different ISO settings (including intermediate ones) and fixed white balance and see the progression. Next, to see if it depends on white balance, change it and take couple of shots at ISO 3200.

You can also see the progression of values if you change the DR settings:

DR100: -0.72

DR200: -1.72

DR400: -2.72

It seems that the FUJIFILM cameras are simply lowering exposure to protect highlights with these settings.

-- hide signature --

Regards,
Niels Kristian Bech Jensen

OP Iliah Borg Forum Pro • Posts: 29,051
Re: What is the meaning of raw exposure bias?

nkbj wrote:

Iliah Borg wrote:

What is the meaning of raw exposure bias?

It is an instruction for a raw converter to apply certain amount of default (silent, behind the scenes) brightness compensation.

You can shoot a scene at different ISO settings (including intermediate ones) and fixed white balance and see the progression. Next, to see if it depends on white balance, change it and take couple of shots at ISO 3200.

You can also see the progression of values if you change the DR settings:

DR100: -0.72

DR200: -1.72

DR400: -2.72

It seems that the FUJIFILM cameras are simply lowering exposure to protect highlights with these settings.

Not that simple. It is the instruction for "main" exposure, in case the DR is achieved through 2 exposures or two sets of pixels.

If you could upload those three raw files, we can look into them and see what is happening in the particular camera.

-- hide signature --
OP Iliah Borg Forum Pro • Posts: 29,051
Re: What is the meaning of raw exposure bias?

nixda wrote:

Iliah Borg wrote:

If the camera specifies that there should be a brightness compensation applied to an image, does it mean that the metering is somehow off

It only means that the compensation is to be applied to comply with some standard. In practice it is the compensation between the exposure meter calibration point for the given ISO setting and the SOS (Standard Output Sensitivity) ISO recommendation (Fujifilm are using SOS in their current cameras).

Also, what does a value of -0.72 mean? Should the brightness be reduced or increased by 0.72 stops?

Increased by 0.72 stops. Compare to baseline exposure compensation in DNG documentation.

Finally, is the raw exposure bias indicating that ISO 800 isn't really ISO 800 but a different value (when comparing ISOs between different camera makers), which gets at the nation that Fuji tends to overstate its ISO numbers, compared to some other mainstream camera makers?

This tag indicates that normal ISO overstating is slightly more than 2/3 of a stop, and more for high ISO. That is done in order to protect highlights and allow film simulations to work.

One final question (hopefully): do you have a formula, or a reference, or a link to where I can find how raw exposure bias is related to any APEX standard formulas?

APEX is modified in EXIF standard, the constant for speed scaling is different, please use http://www.exif.org/Exif2-2.PDF

Other than that, right - it is 'K' that is different.

-- hide signature --
bowportes Veteran Member • Posts: 4,236
Re: Raw exposure bias
1

Iliah Borg wrote:

The difference between 3200 and 6400 is very close to 2x; that is 1 EV. The difference between 1600 and 3200 is close to 1.4x, that is ≈0.5 EV. So, the analog gain is applied up to the ISO setting of 2500 (rounded).

If one looks at true shutter speeds (non-rounded), the values are:

1600 1/194

3200 1/388

6400 1/832

That is 1/832 instead of 1/776, or 0.1 EV difference.

-- hide signature --

The value shifts from -.72 at ISO 2000 to -.83 at ISO 2500. So the analog gain ends somewhere just under 2500 ISO.

Your information here was very informative.  THANKS!!

Based upon this information, I wonder whether the high ISO NR of skin doesn't begin at precisely the point where -.72 begins to decline.

 bowportes's gear list:bowportes's gear list
Panasonic Lumix DMC-G5 Fujifilm X-M1 Fujifilm X-T1 Fujifilm X-Pro2 Panasonic Lumix G 20mm F1.7 ASPH +13 more
OP Iliah Borg Forum Pro • Posts: 29,051
Re: Raw exposure bias

bowportes wrote:

Iliah Borg wrote:

The difference between 3200 and 6400 is very close to 2x; that is 1 EV. The difference between 1600 and 3200 is close to 1.4x, that is ≈0.5 EV. So, the analog gain is applied up to the ISO setting of 2500 (rounded).

If one looks at true shutter speeds (non-rounded), the values are:

1600 1/194

3200 1/388

6400 1/832

That is 1/832 instead of 1/776, or 0.1 EV difference.

The value shifts from -.72 at ISO 2000 to -.83 at ISO 2500.

Right, seems like up to but not including 2500. Without rounding the value is close to 2200.

-- hide signature --
bowportes Veteran Member • Posts: 4,236
Re: Raw exposure bias
1

Iliah Borg wrote:

bowportes wrote:

Iliah Borg wrote:

The difference between 3200 and 6400 is very close to 2x; that is 1 EV. The difference between 1600 and 3200 is close to 1.4x, that is ≈0.5 EV. So, the analog gain is applied up to the ISO setting of 2500 (rounded).

If one looks at true shutter speeds (non-rounded), the values are:

1600 1/194

3200 1/388

6400 1/832

That is 1/832 instead of 1/776, or 0.1 EV difference.

The value shifts from -.72 at ISO 2000 to -.83 at ISO 2500.

Right, seems like up to but not including 2500. Without rounding the value is close to 2200.

So my questions to you are practical ones:

1. Some have called the Fuji X cameras "ISOless." To what extent, if any, would you say this is true, and what difference does it make for image quality?

2. What practical difference does the Analog signal amplification (as compared to non-analog amplification) make for image quality?  I'm not asking a general or theoretical question here, but am wondering whether anyone has examined the "analog" difference for Fuji cameras specifically.  I know that analog amplification is reflected in the raw data, but beyond that, does it make a visible difference in the resulting images?

3. Relatedly, is underexposing a shot by two stops at 400 ISO and then pushing it by 2 stops (to 1600 ISO) via in-camera reprocessing essentially equivalent to underexposing a shot by 2 stops at 1600 ISO and then pushing it back to 6400 ISO?  I'm curious because the former would lose the analog amplification in the raw data while the raw files in the latter case would be essentially the same.  (Let's assume for this question that analog amplification ends at 1600.)  Again, this question reduces down to, What is the advantage of analog amplification of the raw data for Fuji cameras in particular?

4. How much would I really lose in the final image if I shot an image at 400 ISO underexposing by three stops as compared to shooting the image at 3200 ISO and EC-0 in the first place?

I'm becoming redundant. Help me to understand what practical difference some of this makes.

 bowportes's gear list:bowportes's gear list
Panasonic Lumix DMC-G5 Fujifilm X-M1 Fujifilm X-T1 Fujifilm X-Pro2 Panasonic Lumix G 20mm F1.7 ASPH +13 more
OP Iliah Borg Forum Pro • Posts: 29,051
Re: Raw exposure bias

1. Some have called the Fuji X cameras "ISOless." To what extent, if any, would you say this is true, and what difference does it make for image quality?

If you can take several shots of a fine detailed subject, well-focused, at fixed exposure and using more or less stable light, varying ISO only, it can be checked.

2. What practical difference does the Analog signal amplification (as compared to non-analog amplification) make for image quality?

For numeric analysis, a series of shots of an even surface, out of focus, are useful. A monitor screen in a darkened room at 1/15 is usually a good target.  For the first shot the highest "analogue" ISO setting, spot meter and add 2 EV to set the aperture. Other shots are taken at the same shutter speed and aperture, only changing the ISO setting.

For visual analysis, a white feather is a good target; procedure is the same. It is good to have both series.

-- hide signature --
nkbj
nkbj Regular Member • Posts: 263
Re: What is the meaning of raw exposure bias?

Iliah Borg wrote:

If you could upload those three raw files, we can look into them and see what is happening in the particular camera.

@Iliah: I have sent you a PM with a link to the files.

-- hide signature --

Regards,
Niels Kristian Bech Jensen

Keyboard shortcuts:
FForum MMy threads