Question about dynamic range: RAW-JPEG

It has been my assumption, and I underline assumption, that the dynamic range of a RAW file is about one stop greater than that of a JPEG. As such, in a situation in which the scene's dynamic range exceeds the camera's capability, a JPEG file correctly exposed to record highlight detail will still lose roughly one stop of shadow detail compared to a similarly exposed RAW file. Am I wrong about this; is there some algorithmic linear factor that makes RAW's one stop advantage only applicable to highlight information?

I ask this, because whenever I read, "if you get the exposure right in camera, you do not need RAW," I feel the need to point out that on one end or the other of the tonal spectrum, you are going to lose about a stop's worth of information compared with RAW, if, once again, the setting's dynamic range extends beyond that of the camera's sensor. However, this may be an erroneous assumption.

Any clarification would be great.
Thanks

I don't know about an extra stop but more range within stops is the way i understand it

I too would like to know if the above is fact....

sjones wrote in post #5895532
It has been my assumption, and I underline assumption, that the dynamic range of a RAW file is about one stop greater than that of a JPEG. As such, in a situation in which the scene's dynamic range exceeds the camera's capability, a JPEG file correctly exposed to record highlight detail will still lose roughly one stop of shadow detail compared to a similarly exposed RAW file. Am I wrong about this; is there some algorithmic linear factor that makes RAW's one stop advantage only applicable to highlight information?

I ask this, because whenever I read, "if you get the exposure right in camera, you do not need RAW," I feel the need to point out that on one end or the other of the tonal spectrum, you are going to lose about a stop's worth of information compared with RAW, if, once again, the setting's dynamic range extends beyond that of the camera's sensor. However, this may be an erroneous assumption.

Any clarification would be great.
Thanks

Your assumption is incorrect. There is no difference in dynamic range ... the same data in the RAW file is what is used to create the JPG which then gets reduced to a compressed 8-bit image that has far less data, but still the same dynamic range. The rest of the bits go to bit heaven where they get recycled. And, in either case, it should be obvious that nothing can be captured beyond the range of the sensor -- otherwise it would not be beyond the range of the sensor.

The advantages of RAW are that you have all of the captured data to edit the image and that can be a huge advantage in situations such as trying to salvage some blown highlights in ACR or Lightroom where you need all of the image data that you can get. The camera doesn't try to recover blown highlights because it requires a human in the loop to make judgment calls concerning the editing process. If the exposure is a "perfect" (whatever that means) capture and doesn't "need" any post processing, I still do it anyway. There is nothing to say that you can't just do a straight conversion to jpg without any bit twiddling. I've never had a perfect shot that couldn't be made more perfect. But, sooner or later, you have to kill the Photoshopper because the customer is waiting for his/her pictures.

rooeey wrote in post #5896078
I don't know about an extra stop but more range within stops is the way i understand it

I too would like to know if the above is fact....

That would be more "resolution" within a stop ... a stop represents a given range of luminance and you can't expand the range "within" it without expanding the range itself ... and then it would not be a "stop" any longer. The luminance quantization levels represent one type of "noise" in an image that shows up at the darkest regions as loss of detail as subtle changes in light levels get blocked together into a single luminance level.

bill boehme wrote in post #5896256
Your assumption is incorrect. There is no difference in dynamic range ... the same data in the RAW file is what is used to create the JPG which then gets reduced to a compressed 8-bit image that has far less data, but still the same dynamic range. The rest of the bits go to bit heaven where they get recycled. And, in either case, it should be obvious that nothing can be captured beyond the range of the sensor -- otherwise it would not be beyond the range of the sensor.

The advantages of RAW are that you have all of the captured data to edit the image and that can be a huge advantage in situations such as trying to salvage some blown highlights in ACR or Lightroom where you need all of the image data that you can get. The camera doesn't try to recover blown highlights because it requires a human in the loop to make judgment calls concerning the editing process. If the exposure is a "perfect" (whatever that means) capture and doesn't "need" any post processing, I still do it anyway. There is nothing to say that you can't just do a straight conversion to jpg without any bit twiddling. I've never had a perfect shot that couldn't be made more perfect. But, sooner or later, you have to kill the Photoshopper because the customer is waiting for his/her pictures.

Thanks for the response, and yes, you are right, it makes sense that since the JPEG file is simply compressing the same data collected by RAW, the dynamic range should be the same. Yet, then the question centers on if the process of compression can potentially sacrifice a stop or so of data. That is, and I realize this might be getting into semantics, but if the highlights are blown, they are just that, blown. RAW cannot fabricate this lost data, it must have been captured at time of exposure, so it would appear that the loss of such highlight detail could be due to the compression process, not necessarily because of overexposure.

But OK, let's just say that RAW's advantage is that in some cases, it can recover seemingly blown highlights. So my question now is, if the image in JPEG was exposed in such a way as to capture these highlights, would it come at any expense to the shadow detail in terms of increased noise or lost information compared with shooting in RAW?

No, by visible inspection, a jpeg and a RAW would appear to be very similar if no in-camera manipulation was carried out on the jpeg. However, the RAW has more latent potential, compared to the jpeg, where it's strictly "wot U see is wot U got", like it or not.

If you like, it can be cooked to develop more subtle flavours without burning?

I do so like the mental image of "bit heaven". Where all the cropped pixels go to live in CMOS bliss, where voltage never fades and noise is banished.

Richard

Sure, because in order to preserve the highlights in jpg you have to give less exposure. This means darker, blocked up shadows. Moreover, because in 8 bits so few tonal levels carry the image data, any attempt to lighten the shadows may run into problems of posterization and noise.

Sjones said, "if the image in JPEG was exposed in such a way as to capture these highlights,..." The only way this can be done is by reducing exposure, so by visual inspection the jpg would be darker.

Yes, but the same exposure for the highlight detail using RAW would give you the same darker image. To argue, as the poster is trying to, that "Yes, but I'm going to treat the jpeg exposure differently just to prove my point" is to admit he really knows that the RAW files contains loads more information. So I still say, if everything is kept the same, there would be no "visible" difference at first glance.

Richard

Lowner wrote in post #5896724
Yes, but the same exposure for the highlight detail using RAW would give you the same darker image. To argue, as the poster is trying to, that "Yes, but I'm going to treat the jpeg exposure differently just to prove my point" is to admit he really knows that the RAW files contains loads more information. So I still say, if everything is kept the same, there would be no "visible" difference at first glance.

Richard

That the RAW file contains considerably more information than does JPEG is axiomatic, no one is debating this. In the past, very recent past, my argument has been that shooting JPEG to correctly capture highlight data would result in lost shadow detail compared with a RAW shot similarly exposed to capture highlight data, if, of course, the dynamic range of the scene photographed exceeded that of the camera's sensors.

However, after hearing so many photographers adamantly state, "if you expose correctly in camera, you don't need RAW," I figured, well, there just might be something to this. Therefore, I wanted to confirm my previous assumptions, which differs from trying to prove a point.

This is the way that I had previously envisaged the comparison. You have a 12-centimeter horizontal line, with each centimeter marked off and representing one stop. Then, you have a RAW ruler that is, say 8 centimeters, and a JPEG ruler that is 7 centimeters. Do you see where I am going with this? Yet, when dealing with digital and such, I felt compelled to check if this was a flawed analogy, if the issue was, in fact, more complicated than this.

Frankly, I would shoot RAW regardless, because in the days of digital, I converted everything to monochrome, and to have done so in JPEG would have been more destructive. Because I enjoy post processing, and that this is the hobby, time and patience is not an issue.

RAW does give you more dynamic range IMHO, although you may have to work to achieve it. The default JPEG processing in-camera actually clips some of the highlight and shadow detail (i.e. the top and bottom of the range) and this can never be recovered. In RAW, you can use this detail (although processing the file with everything on default does not do so).

Take a look at a typical review on dpreview:
40D review

They achieve 9.1 stops of dynamic range from an ISO100 JPEG, but nearly 11 stops from RAW. They also make the point that the default Adobe Camera RAW conversion delivers less dynamic range than JPEG from the camera - you need to play with the settings to get the 11 stops.

gcogger wrote in post #5897478
... They achieve 9.1 stops of dynamic range from an ISO100 JPEG, but nearly 11 stops from RAW. They also make the point that the default Adobe Camera RAW conversion delivers less dynamic range than JPEG from the camera - you need to play with the settings to get the 11 stops.

Somebody is fooling themselves if they think that they are capturing that wide of a dynamic range. At the lowest stop, there would only be 8 discrete luminance levels to convey any detail in the image. That small number of levels is insufficient to be of any practical use.

bill boehme wrote in post #5900022
Somebody is fooling themselves if they think that they are capturing that wide of a dynamic range. At the lowest stop, there would only be 8 discrete luminance levels to convey any detail in the image. That small number of levels is insufficient to be of any practical use.

The most widely used definition of the DR floor seems to be "when the S/N is such that detail is swamped by noise," but I've never seen a standard value for that S/N. Bill, you're an engineer. Doesn't ISO or some other international body maintain photographic standards?

tzalman wrote in post #5901354
The most widely used definition of the DR floor seems to be "when the S/N is such that detail is swamped by noise," but I've never seen a standard value for that S/N. Bill, you're an engineer. Doesn't ISO or some other international body maintain photographic standards?

There are a very large number of things that influence the dynamic range of camera sensors. At the top end, there is the obvious maximum amount of charge that the sensors can handle without causing problems like "bloom" and a lot of factors influence where this point is set. At the other end, there are many more things that influence where the minimum level is set. Various noise types are probably the main factors ... and there are more types of noise that are involved than one would imagine.

Although the sensor itself is analog, its conversion to digital introduces some additional types of noise because of quantization round off and the discrete number of levels available for a given range of brightness, but normally not a limiting factor in DSLR sensors because other factors have already decided the limits. However, ignoring any other more restrictive limiting factors, quantization does create a physical limit on the dynamic range of a sensor for any given quantization level.

Our eyes and film react to light in an approximately logarithmic fashion, but digital sensors respond linearly to brightness. We see each change in exposure value (or f stop) as equal increments, but the actual change in light per f stop is a factor of two (either doubling or halving). As a result, when the analog signal is digitized, the brightest f stop range which is a factor of two change in brightness uses half of the available quantization levels. By the time that the brightness has changed by six f stops, there are very few discrete levels available to represent the levels of detail within that range. Here is a chart that I made showing quantization levels for 12 bit and 14 bit cameras where I arbitrarily show Zone IX being chosen as the brightest exposure zone and an 18% reflectance gray card would then, hopefully, fall in Zone V.

Quantization Levels

Exposure Zone........12 Bit Quantization........14 Bit Quantization
Zone IX...............2048...............8192
Zone VIII...............1024...............4096
Zone VII...............512...............2048
Zone VI...............256...............1024
Zone V...............128...............512
Zone IV...............64...............256
Zone III...............32...............128
Zone II...............16...............64
Zone I...............8...............32

In actuality, the brightest exposure Zone for the 12 bit sensor would make more sense to be Zone VIII in order to help keep the darker areas of the image from being too badly blocked up. I would guess that when there are fewer than 32 bits to express details in an exposure zone that image detail would be next to useless.

The chart above helps to explain part of the reason that the 40D produces a cleaner image than the XTi.

Regarding ISO standards, I once briefly reviewed one of them that was mostly applicable to film, but I do not think that what it said is really pertinent to the question which I interpret to ask whether ISO 100 in digital is the same s ISO 100 in film with respect to getting the right exposure. This would be an apples and oranges question that does not have a simple straightforward answer, but to simplify things, I would say that the answer is "yes".

Bill Boehme Rather than "Quantization" don't you think "Contrast Ratio" is more accurate a term here or is there a distinction I am missing?

Also, if image detail is measured by contrast, and contrast is the dynamic difference in tonal values within an image, isn't there a way to shift some of that “detail-as-tonal range” from the highlight areas into the shadow using Photoshop's shadow/highlight plug-in or curves. When this difference is less than 1:256 or 2.5% a gap will show in the histogram, but with the 12 bit 1:2048 or 14 bit 1: 8192 images no real detail is lost or created since even less detail can be used than can be either displayed or printed.

Monitors can only show about 8 stops at 1:300 or so.
Print can only hold about 5 stops that’s only 1:32 contrast ratio

Only If you shoot towards zone 9 for the highlight, that you can always drag the shadows down into zone 2 if the dynamic range of the photographic scene is less than 7 or 8 stops?

In short-if we capture a scene that has 5-6 stops can we not stretch the larger tonal areas from the highlights to the shadow end while maintaining actual detail and yet giving us a 4 or 3 stop increase in the dynamic range of the final image? This would not, seem to work with exposures that took the initial 9 stop the camera is capable of, for the same reasons I mentioned here.

Is my thinking correct on this or is it way off?