New article - Why shoot RAW - The real answer

FM Forums \| Alternative Gear & Lenses \| Join Upload & Sell	Search Used
1 2 3 4 5 end

Archive 2010 · New article - Why shoot RAW - The real answer

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.4 #1 · New article - Why shoot RAW - The real answer

HerbChong wrote:
10 stops of dynamic range can reference maximum and minimum light levels referenced to some external value or it can merely specify how many steps there are between and arbitrary maximum and minimum. both are valid and seen in engineering literature.

Herb...

Actually, neither would be correct as stated. The last would be as the ratio between an arbitrary maximum and minimum, not the number of steps between them (a stop is a logarithmic value and cannot be used correctly as stated), the first would be as the ratio between maximum or minimum light levels referencing an external value, but as stated neither is correct since they misuse the unit in which they are stated. In all the literature I've run across (which is a fair bit, although generally the unit in question is decibels due to the industry in question, I'm in IP Networking) there's generally a certain assumption of familiarity involved. If there's a unit and an arbitrary maximum or minimum than the other maximal/minimal value is in reference to a single arbitrary value in a known unit or the unit is merely designating the separation of the two arbitrary values (IE the arbitrary maximum and minimum are 10 stops apart and thus have 10 stops of dynamic range, not there are 10 steps between the two although the latter may also be true). You cannot take two arbitrary values and designate the difference as 10 stops unless it is actually 10 stops (ie the maximum value is 2^10 greater than the minimum).

Note this is a classic case where Photographers, especially Zone System afficionados, make mistakes. 1 zone is NOT 1 stop. It can be, or it can be a multitude of other values since it is simply 1/10th of the SBR (subject brightness range). Unless you have exactly 10 stops of SBR, a zone is not a stop.

With regards to dynamic range, the maximum and minimum values used to measure it can be completely arbitrary, in fact they effectively are as the ADC implementation has a fixed maximum and minimum value so that it can correctly convert analog values between those two points to meaningful digital values.

Edited on Feb 08, 2010 at 06:19 PM · View previous versions

Feb 08, 2010 at 06:16 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.4 #2 · New article - Why shoot RAW - The real answer

Feb 08, 2010 at 06:17 PM

TonyBeach
Offline
Upload & Sell: Off

p.4 #3 · New article - Why shoot RAW - The real answer

brainiac wrote:
Correct exposure and minimum contrast in jpeg mode won't get you much less DR than a raw especially if you want to keep highlight colours real.

TonyBeach wrote:
Flat out wrong. If you utilize uni-WB you can set your exposure value as much as a full stop hotter with RAW than you can with JPEG. With JPEG the WB gain is applied to the file and is not recoverable; with RAW highlights blown by WB gain are recoverable.

mawz wrote:
Actually, that's VERY implementation-specific. Richard is correct for some implementations and you are for others.

Right, I am right as it pertains to attaining maximum DR that the sensor can deliver.

Feb 08, 2010 at 06:47 PM

brainiac
Offline
Dedicated FM
Account Locked

p.4 #4 · New article - Why shoot RAW - The real answer

Feb 08, 2010 at 07:03 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.4 #5 · New article - Why shoot RAW - The real answer

TonyBeach wrote:

brainiac wrote:
Correct exposure and minimum contrast in jpeg mode won't get you much less DR than a raw especially if you want to keep highlight colours real.

TonyBeach wrote:
Flat out wrong. If you utilize uni-WB you can set your exposure value as much as a full stop hotter with RAW than you can with JPEG. With JPEG the WB gain is applied to the file and is not recoverable; with RAW highlights blown by WB gain are recoverable.

mawz wrote:
Actually, that's VERY implementation-specific. Richard is correct for some implementations and you are for others.

Right, I am right as it pertains to attaining maximum DR that the sensor can deliver.

Actually you aren't, all uni-WB does is give you more accurate Red and Blue histograms in-camera for RAW shooting by eliminating the gain in the R and B channels, thus making it easier to get the correct exposure in RAW. It's only of any use in cameras with RGB histograms when shooting RAW and doesn't change the fact that a photographer sufficiently familiar with the performance of his camera and meter can achieve the exact same result without it.

Note that uni-WB really doesn't have any effect on the camera's actual DR, it just makes it a bit easier to achieve optimal exposure on cameras with RGB histograms and the ability to load uni-WB JPEG settings. And the optimal exposure will produce the most DR. But you do not need Uni-WB to achieve this.

Note that 1+ stop gap in highlights doesn't even exist on some cameras. As I said, it's very implementation specific. Uni-WB only has value with some implementations and the gap between the maximum highlight value in JPEG and the maximum highlight value in RAW varies between cameras. It's only in some cameras with RGB histograms, the ability to load custom JPEG tone curves and a notable gap between the maximal JPEG and RAW highlight values that you'll see an advantage for Uni-WB.

Feb 08, 2010 at 07:07 PM

HerbChong
Offline
Dedicated FM
Upload & Sell: On

p.4 #6 · New article - Why shoot RAW - The real answer

but you can quote a signal to noise ratio between the maximum and minimum as a number of bits and that depends on the amount of noise and not the actual signal values. thus the SNR and maximum and minimum referenced to an outside standard can both be expressed in decibels or some other similar measure but mean two different things. scanners have their dynamic range specified as a number of bits of dynamic range for this reason. it means something different than the maximum and minimum brightness range they can discern and it makes sense in that context. photographers see dynamic range in one context and signal engineers see it in another.

Herb...

mawz wrote:
[You cannot take two arbitrary values and designate the difference as 10 stops unless it is actually 10 stops (ie the maximum value is 2^10 greater than the minimum).

Feb 08, 2010 at 07:32 PM

HerbChong
Offline
Dedicated FM
Upload & Sell: On

p.4 #7 · New article - Why shoot RAW - The real answer

i'd wager that shooting in bright sun or with a flash accounts for most real world shooting. that means extremely contrasty scenes are the norm and not the exception. photographers learn to control contrast, mostly by reducing it, at the same time they learn to control exposure.

Herb...

brainiac wrote:
Wayward colour balance and extremely contrasty scenes will of course make it harder to get that correct exposure, and will increase the gap between the jpeg and the raw. But there are still many cirumstances where the difference is small, and probably smaller than the stop that you estimate.

Feb 08, 2010 at 07:39 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.4 #8 · New article - Why shoot RAW - The real answer

Actually, that falls under the definition I used quite easily. Note that bits are simply another base 2 logarithmic unit in this case (used in the same way they are for ADC's) and the noise floor is being used as the minimum value since that's the minimum point where you are getting actual data.

You're suggesting a difference in definition that simply doesn't exist. What differs between photographers and Signal Engineers is the unit and thus the quantity being measured. Note that the noise floor is fundamentally the bottom limit of the DR from any imaging sensor (the Dynamic Range of a scene isn't, but that's not the DR of the camera anyways) just as the clipping point is the upper limit.

In all the cases you note, the Dynamic Range remains the difference between the maximum and minimum value of a changeable quantity expressed as a ratio (such as SNR or the contrast ratio of a display) or a logarithmic value usually in base 2 (stop, bit) or base 10 (decibel). The definition of maximum and minimum value and the quantity and unit change depending on what exactly you're measuring and how you;re using those measurements but the fundamental definition does NOT change.

HerbChong wrote:
but you can quote a signal to noise ratio between the maximum and minimum as a number of bits and that depends on the amount of noise and not the actual signal values. thus the SNR and maximum and minimum referenced to an outside standard can both be expressed in decibels or some other similar measure but mean two different things. scanners have their dynamic range specified as a number of bits of dynamic range for this reason. it means something different than the maximum and minimum brightness range they can discern and it makes sense in that context. photographers see dynamic range in one context and signal engineers see it in another.

Herb...

mawz wrote:
[You cannot take two arbitrary values and designate the difference as 10 stops unless it is actually 10 stops (ie the maximum value is 2^10 greater than the minimum).

Feb 08, 2010 at 07:55 PM

theSuede
Offline
Upload & Sell: Off

p.4 #9 · New article - Why shoot RAW - The real answer

AFAIK - which would be quite far actually, at least within the EE/DSP field - the only thing that's "negotiable" about the dynamic range definition is the definition of minimum signal quality. "Dynamic Range" in itself is a unitless ratio and application dependant in its' report nature. It can either be related to "zero", or an arbitrary fixed amount that most usually depends on the noise power or signal quantization, whichever is highest.

The "other" definition that I read between the lines in some posts shouldn't really be labeled "DR" - the correct denomination for the "number of pictorially usable brightness levels in the usable range" would be "tonal range". This is the integral sum of Signal/Noise ratios in the DR range.

The other part of the report choice is if you decide to make the value object related or measurement related. In photography, object related would be real-world Ev difference between brightest and darkest area that can contain actual (real) picture detail. Measurement related would be in raw DN or e-, and this has been subjected to electric compression effects and/or lens system DR limiting (flare and stray light).

The trouble with jpg's as I see it (I DON'T usually do 4k clicks per day. I've topped 2k once when I had to do both the lead, artist portraits, costume portraits and last dress rehearsal in one day) is that most cameras just can't move the endpoints of the curve far enough apart to fit my taste. Decrease "contrast" or use a custom, less steep tone curve, and all that happens is that you flatten the midrange. You don't extend your used range anything at all.

For minimal PP, jpg is just fine if you don't mind getting less colour accuracy/resolution than the camera really allows. One thing to remember IF you know that you are going to PP or even just resize the final results is to use the absolute minimum of NR possible in the camera. Noise serves as a non-random dither, and can extend your usable tonal range by quite a bit. NR tends to posterize results quite badly, and this is NOT recoverable in post. If you keep noise in the jpg though, that can up to double your tone resolution in the highlights...

Feb 08, 2010 at 08:57 PM

brainiac
Offline
Dedicated FM
Account Locked

p.4 #10 · New article - Why shoot RAW - The real answer

theSuede wrote:
The trouble with jpg's as I see it (I DON'T usually do 4k clicks per day. I've topped 2k once when I had to do both the lead, artist portraits, costume portraits and last dress rehearsal in one day) is that most cameras just can't move the endpoints of the curve far enough apart to fit my taste. Decrease "contrast" or use a custom, less steep tone curve, and all that happens is that you flatten the midrange. You don't extend your used range anything at all.

...unless you build your own profile with Picture Style Editor, in which case you can curve it how you want, and produce the compressed shadows/highlights and amplified midrange that you seek. In JPEGs. And there's room in the camera for 3 variations of that.

Feb 08, 2010 at 11:31 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.4 #11 · New article - Why shoot RAW - The real answer

theSuede wrote:
AFAIK - which would be quite far actually, at least within the EE/DSP field - the only thing that's "negotiable" about the dynamic range definition is the definition of minimum signal quality. "Dynamic Range" in itself is a unitless ratio and application dependant in its' report nature. It can either be related to "zero", or an arbitrary fixed amount that most usually depends on the noise power or signal quantization, whichever is highest.

Yeah, I'm sortof misusing Unit in my above descriptions, where I really should be using scale (ie decibel scale rather than decibel unit). There is however some variation in how you define the maximum point, particularly with how WB is handled on Bayer sensors, you can pick the maximum value where you get good chroma data or the maximum value where you gte valid luminance data but may get false colour. It clearly doesn't vary as much as the possible minimum values though.

Note I'm also coming at this from an EE background (I'm an Electronics Engineering Technician and currently doing my B. E Eng) although my background is communications systems, primarily IP.

The "other" definition that I read between the lines in some posts shouldn't really be labeled "DR" - the correct denomination for the "number of pictorially usable brightness levels in the usable range" would be "tonal range". This is the integral sum of Signal/Noise ratios in the DR range.

The other part of the report choice is if you decide to make the value object related or measurement related. In photography, object related would be real-world Ev difference between brightest and darkest area that can contain actual (real) picture detail. Measurement related would be in raw DN or e-, and this has been subjected to electric compression effects and/or lens system DR limiting (flare and stray light).

The trouble with jpg's as I see it (I DON'T usually do 4k clicks per day. I've topped 2k once when I had to do both the lead, artist portraits, costume portraits and last dress rehearsal in one day) is that most cameras just can't move the endpoints of the curve far enough apart to fit my taste. Decrease "contrast" or use a custom, less steep tone curve, and all that happens is that you flatten the midrange. You don't extend your used range anything at all.

For minimal PP, jpg is just fine if you don't mind getting less colour accuracy/resolution than the camera really allows. One thing to remember IF you know that you are going to PP or even just resize the final results is to use the absolute minimum of NR possible in the camera. Noise serves as a non-random dither, and can extend your usable tonal range by quite a bit. NR tends to posterize results quite badly, and this is NOT recoverable in post. If you keep noise in the jpg though, that can up to double your tone resolution in the highlights...

Feb 08, 2010 at 11:32 PM

Bifurcator
Offline
Dedicated FM
Upload & Sell: On

p.4 #12 · New article - Why shoot RAW - The real answer

theSuede wrote:
The "other" definition that I read between the lines in some posts shouldn't really be labeled "DR" - the correct denomination for the "number of pictorially usable brightness levels in the usable range" would be "tonal range". This is the integral sum of Signal/Noise ratios in the DR range.

Hopefully you meant "tonal resolution". The range is still between 0% and 100%. The resolution of that fixed range is dependent on the kinds of units used and the number of bits of tonal resolution. One is a fidelity issue and the other is actually an issue of range.

As stated, there's only one accepted definition for "DR" or "HDR". Although it is fairly common for photographers to misuse HDR when they really mean tone compression, tone remapping, or etc. But since at least one of the processes does include the use of an actual HDR image in interim I suppose it's forgivable.

WB is of no consideration in either definition of range and resolution. Neither is white point and black point. All are arbitrary values within the physical limitations of the file format weather user specified, preset, or algorithmically defined.

Feb 09, 2010 at 12:39 AM

veroman
Offline
Dedicated FM
Upload & Sell: On

p.4 #13 · New article - Why shoot RAW - The real answer

The article is thin, simplistic and a waste of reading time. It's also old hat .....

Feb 09, 2010 at 12:48 AM

cogitech
Offline
Dedicated FM
Upload & Sell: On

p.4 #14 · New article - Why shoot RAW - The real answer

I shoot RAW...always will.

Not good enough to shoot JPEG.

I'm glad there is RAW.

(Is this Haiku?)

Feb 09, 2010 at 01:07 AM

Daniel Heineck
Offline
Upload & Sell: Off

p.4 #15 · New article - Why shoot RAW - The real answer

Dynamic Range: The ratio between the maximum recordable signal and the noise floor. For a CMOS camera, we should be reporting this in volts/volts (photodiode), and in coulombs/coulombs for a CCD camera.

Bit Depth: how finely we want to divide the dynamic range.

Tone Mapping: how we interpret the relationship between input and output.

Using anything much more specific than that is likely to get you in trouble, as we've seen here.

I have my MSEE, which is to say I don't know jack, seriously.

Feb 09, 2010 at 01:21 AM

Bifurcator
Offline
Dedicated FM
Upload & Sell: On

p.4 #16 · New article - Why shoot RAW - The real answer

Daniel Heineck wrote:
Dynamic Range: The ratio between the maximum recordable signal and the noise floor. For a CMOS camera, we should be reporting this in volts/volts (photodiode), and in coulombs/coulombs for a CCD camera.

Bit Depth: how finely we want to divide the dynamic range.

Tone Mapping: how we interpret the relationship between input and output.

Using anything much more specific than that is likely to get you in trouble, as we've seen here.

I have my MSEE, which is to say I don't know jack, seriously.

Works for me.

Me is MSIT(CS). And you're right... Just enough to be dangerous.

Feb 09, 2010 at 02:33 AM

theSuede
Offline
Upload & Sell: Off

p.4 #17 · New article - Why shoot RAW - The real answer

There are as you say quite a few possible ramifications - and they're mostly "not important" unless you're comparing two separate measurement devices. Then you have to make sure that measurements from both compared subjects are measured/calculated the same way... And preferably in a way that's relevant to the use of what you're measuring.

Brainy - last time I tried that I was still at least an Ev short of what a separate raw-converter could provide. That was a few PSE revisions ago with the 1D3, things might have changed... I'll check again. I'm also interested in how accurately I can get the HSL correction to work "in camera" (if at all), so I'll spend some time with that pretty soon.

Bifurcator - Now that's STILL one more definition - is this "measurement resolution related" or "signal quality related"? :-)
...But I actually meant "range" - as in "signal quality related minimum step count". This resolution has very little to do with how many bits/steps you quantize your measurement with (as long as you use a quantization step small enough - 12 bits or more). This value wouldn't change if you used 16, 18, 20 or 24bit ADC's... It is defined by the amount of minimum signal change steps you can fit in between the lower (noise floor) and upper (clipping) boundary with respect to signal noise present at signal level. These "steps" get wider and wider the stronger the signal gets - as minimum noise is always the square root of the signal. Gamma correction fortunately coincides reasonably well with this - otherwise jpg's in 8 bits would be pretty useless...

Feb 09, 2010 at 04:02 PM

TonyBeach
Offline
Upload & Sell: Off

p.4 #18 · New article - Why shoot RAW - The real answer

brainiac wrote:
TonyBeach wrote:

brainiac wrote:
Correct exposure and minimum contrast in jpeg mode won't get you much less DR than a raw especially if you want to keep highlight colours real.

TonyBeach wrote:
Flat out wrong. If you utilize uni-WB you can set your exposure value as much as a full stop hotter with RAW than you can with JPEG. With JPEG the WB gain is applied to the file and is not recoverable; with RAW highlights blown by WB gain are recoverable.

mawz wrote:
Actually, that's VERY implementation-specific. Richard is correct for some implementations and you are for others.

Right, I am right as it pertains to attaining maximum DR that the sensor can deliver.

That depends entirely on the scene you are shooting. I accept that the histogram is not a good guide to what I vaguely called "correct exposure" in my original comment. I think you are attempting to provide a definition of what that phrase means. Wayward colour balance and extremely contrasty scenes will of course make it harder to get that correct exposure, and will increase the gap between the jpeg and the raw. But there are still many cirumstances where the difference is small, and probably smaller than the stop that you estimate.

I did say "as much as a stop." Yes, often it is more like 1/3 of a stop. The biggest differences occur when the gain to the red channel (along with even fairly conservative in-camera image settings) causes that channel to blow.

Feb 09, 2010 at 07:21 PM

TonyBeach
Offline
Upload & Sell: Off

p.4 #19 · New article - Why shoot RAW - The real answer

mawz wrote:
TonyBeach wrote:

brainiac wrote:
Correct exposure and minimum contrast in jpeg mode won't get you much less DR than a raw especially if you want to keep highlight colours real.

TonyBeach wrote:
Flat out wrong. If you utilize uni-WB you can set your exposure value as much as a full stop hotter with RAW than you can with JPEG. With JPEG the WB gain is applied to the file and is not recoverable; with RAW highlights blown by WB gain are recoverable.

mawz wrote:
Actually, that's VERY implementation-specific. Richard is correct for some implementations and you are for others.

Right, I am right as it pertains to attaining maximum DR that the sensor can deliver.

Actually you aren't, all uni-WB does is give you more accurate Red and Blue histograms in-camera for RAW shooting by eliminating the gain in the R and B channels, thus making it easier to get the correct exposure in RAW. It's only of any use in cameras with RGB histograms when shooting RAW and doesn't change the fact that a photographer sufficiently familiar with the performance of his camera and meter can achieve the exact same result without it.

No again, RAW converters can do funny things when applying WB gain; using uni-WB eliminates that problem. Also, if you use a CC30M filter to tame the green channel, you gain another 2/3 of a stop of DR in many situations, and you are making your life even more difficult trying to keep WB accurate in-camera doing that.

Note that 1+ stop gap in highlights doesn't even exist on some cameras. As I said, it's very implementation specific.

It works with Nikon's more recent cameras, and I also see it with my Sony A850.

Feb 09, 2010 at 07:28 PM

mawz
Offline
Dedicated FM
Upload & Sell: Off

p.4 #20 · New article - Why shoot RAW - The real answer

TonyBeach wrote:
mawz wrote:

Actually you aren't, all uni-WB does is give you more accurate Red and Blue histograms in-camera for RAW shooting by eliminating the gain in the R and B channels, thus making it easier to get the correct exposure in RAW. It's only of any use in cameras with RGB histograms when shooting RAW and doesn't change the fact that a photographer sufficiently familiar with the performance of his camera and meter can achieve the exact same result without it.

No again, RAW converters can do funny things when applying WB gain; using uni-WB eliminates that problem. Also, if you use a CC30M filter to tame the green channel, you gain another 2/3 of a stop of DR in many situations, and you are making your life even more difficult trying to keep WB accurate in-camera doing that.

Note that 1+ stop gap in highlights doesn't even exist on some cameras. As I said, it's very implementation specific.

It works with Nikon's more recent cameras, and I also see it with my Sony A850.

Umm, no. Uni-WB does absolutely nothing to affect how RAW converters handle WB gain and may seriously screw with the default conversions if the RAW converter is using the camera's WB settings by default. All it does is give you a histogram which more closely resembles the actual RAW file when shooting in RAW on a camera which can implement it. That's it. For anyone who is is truly competent with other metering methods Uni-WB won't give them anything (but might make their life easier if they like chimping the Histogram for exposure). Since WB in the RAW file is merely metadata, the RAW file from a Uni-WB shot and a standard WB shot using the same exposure settings will produce the same final result.

Using a CC30M filter to tame the green channel will have an actual effect in the RAW converter, but at the cost of causing WB problems since you are mucking with the Bayer demosaicing algorithm's basic assumptions. The green channel is stronger for a reason, because the human eye is far more sensitive to green. It's not a very good idea to use a CC30M if you want anything resembling accurate colour.

And try Uni-WB on an Oly E-3 sometime. if you can find 1/10 of a stop between RAW and the Netral JPEG's that is. Oly does wonders with their JPEG algorithms, but not so well with the DR of their sensors.

Feb 10, 2010 at 03:23 AM

1 2 3 4 5 end
FM Forums \| Alternative Gear & Lenses \| Join Upload & Sell 1 2 3 4 5 end