The subject of the "JPEG+Raw" option arose in another thread. Rather than hijack that thread to continue this secondarily introduced topic, it seemed altogether meet and proper to start a new thread for discussion thereupon.


THE FALLACY OF THE "JPEG+ RAW" OPTION

In recommending use of the "JPEG + raw" option offered by many cameras, you are urging emasculation of the major part of the advantages provided by properly exposed raw image data over a normally exposed JPEG file.

For a raw image data capture to deliver the vastly improved image data quality of which it is capable, it must be exposed by using at least some additional dynamic range beyond that available to expose the JPEG file. Sure, a raw file captured with the same exposure as a JPEG file will have enough "headroom" with which to recover some blown highlight data. The problem is that one camera may have an extra 1/3 stop of raw-accessible DR, while another camera of the same brand and model may have as much as two and 2/3 stops to perhaps even three full stops or more of that extra raw-accessible DR. It is in testing one's camera for, and then using that extra allotment of raw-accessible DR that the great advantages of capturing raw image data are fully realized.

It is not convenient (nor, evidently, sufficiently profitable) for the camera manufacturers to disclose their failure to determine and then report on each camera's allotment of that extra DR beyond the amount necessary to meet the needs of an 8-bit JPEG file. They thus incur no need for an in-camera raw histogram adjusted to the DR available for raw exposure.

It is no problem to assure that a given sensor will offer a specific dynamic range of exposure. The manufacturer evidently cannot, however, predictably assure how much extra dynamic range exists in each sensor! This problem of "process variance"/"performance variance" has been a bug-a-boo of the silicon-based semiconductor device production industry since Shockley invented the transistor back in the late 1940s. It includes our cameras' light sensors...and it continues today....it's simply a fact of life with silicon-based semiconductor devices.

So, in restricting exposure of a raw capture to that necessary for a JPEG image file, the "JPEG + raw" option relegates the captured raw data to the same data quality (sound:noise ratio) as that of its paired JPEG partner. The raw image data so exposed thus becomes functionally an 8 bit-depth image file delivered in a 16 bit depth environment.

Depending on the amount of unused raw-accessible DR the Raw data file will, in most cameras, suffer a 50% to 75% (or more) loss of its potential tonal and color spectra. that's an incredible waste considering that the sensor accounts for 1/2 to 2/3 of the price you paid for the camera body!

The question may be asked: "Will one only gain an advantage in raw by using an exposure that will blow the highlights in the JPEG file?

The short answer is an unqualified "YES !"

Why?

Because: If one shoots JPEG + raw and exposes enough to the right to gain benefit from the more properly exposed raw image data, the JPEG file will, indeed, be irretrievably overexposed. The raw image data, however, will show its advantage in being able to be completely tonally normalized in the raw converter ( like Adobe Camera Raw) with no blown highlight detail as long as the exposure did not exceed use of that particular camera's extra raw-accessible dynamic range (which must be determined for each individual camera by the new owner). The camera manufacturers don't do this for the purchaser. It may be a reasonable assumption that the per-unit cost of doing this would result in an unacceptable increase in the price of each camera. The camera manufacturers are silent on this matter.

Thus, the "JPEG+raw" option is a sop offered to the uninformed photographer who wants to feel as if he were in the vaunted vanguard using the highly touted "raw capture".

In summary, no favor is done by advising use of the of the "JPEG+raw" option without simultaneously informing the advisees of the cost they incur in severely reduced raw image data quality resulting from limiting its exposure to that suitable only for a JPEG image file.

Until such time as the camera manufacturers provide an accurate in-camera raw histogram, the fallacy of the "JPEG+raw" option will persist.

Regards,
Dave

You make a very good point. While saving a raw file along with a jpeg is better than not saving a raw file at all (such as folks who haven't learned to process raw files but hope to one of these days), it has its limitations: you have to decide which one to expose for, and those decisions may well render one or the other less than optimal.

I gave up saving jpegs long ago for that reason. My raw files are exposed NOT for what they look like SOOC but moreso for what I need to harvest from them. So the resulting companion jpeg may be too bright or too dark, depending on what I am shooting and what I mean to do. An ETTR file is gonna look too bright, but one that is exposed to cover a too-wide dynamic range where highlights are likely to be blown may look too dark. But those are choices I make based on what I know about my camera and the scene itself.

Thanks for reminding everyone of this.

(It is one of several reasons why it makes no sense to participate in the occasional challenge on the Main section asking people to post raws and companion jpegs from the same capture to compare quality).

There is a conceptual problem expressed above. Each camera brand, model or sample does not have some variation in this "extra raw-accessible DR". Each model, given the exact same configuration and the exact same scene, will show the exact same resulting DR. There is no unmanageable variation in the silicon of the sensor! Sample variations are probably less that 1/4th fstop, and generally about 1/10th of an fstop. A given camera cannot be measured and then proclaimed to have 1/2 fstop or whatever amount of extra DR compared to the JPEG histogram. Different images will cause the amount seen to change, and different camera JPEG configuration (such as brightness or saturation) will also change it. It is not a fixed thing individual to each camera, it is specific to how a given raw sensor data set is processed. The data set changes with the scene and the processing changes with the JPEG configuration.

The idea that by shooting in RAW+JPEG and exposing for either one causes the other to be less than optimal is essentially true, and it certainly is detrimental to expose for the JPEG when fully intending to use the RAW data. But there are some easy, and some not so easy, ways to deal with that if one either thinks it is important or just wants to experiment.

The obvious is UniWB, which will simply discard all usefulness of the JPEG other than producing an accurate histogram. Not much point in then shooting RAW+JPEG though.

But in the exact same line of thinking, if the actual extra DR available (imagined, measured, or otherwise) is expected to be 2/3rd of an fstop: that can be adjusted very precisely in the JPEG configuration using the "brightness" slider. At least with a higher end Nikon camera that means going to the "Picture controls" menu, selecting whichever is being used (such as Standard, Neutral, Vivid, etc) and setting the brightness to a lower value. Lower contrast and lower saturation might also help, but the most significant parameter is brightness.

If the Picture control is set to -2/3rd of a stop lower, and exposure is made to get a proper looking JPEG image, the effect will be to increase the RAW data exposure by that same 2/3rd of a stop. Both the RAW and the JPEG data sets will be "correct". Problem solved, and shooting RAW+JPEG works fine.

The real problem is determining what should actually be configured, because it will change every time the scene being photographed is significantly different. That just makes it less than perfect, but it absolutely will reduce the error. If one finds that virtually all of the images being shot have some fixed "extra DR", then setting the Picture control to reduce brightness of the JPEG by that amount will virtually always be correct. (Meaning if you shoot very similar types of images most of the time it will work very well most of the time. That is probably true for many, though not most, photographers.) The risk is that if it is set for adding 3/4 fstop of exposure, on the rare occasion when it needs only be 1/2 fstop the RAW data is going to be clipped by that 1/4 of a stop. Unlike not setting it at all and needing to boost brightness when processing RAW data by 3/4 for some and rarely only by 1/4, the clipped data images cannot be corrected in post processing.

thanks, Dave.
You seem to have researched this topic quite extensively and you put forth some compelling arguments. Are we to assume then, if we are shooting RAW, we should be overexposing on purpose in order to gain the full DR that the sensor is capable of? Would this not require that every shot would have to be extensively manipulated to gain acceptable results? Would bracketing a composition not achieve the same result with three or more exposures? Does you technique give us the ability of having enhanced DR without the hassle of multiple shots?
Erich

thanks, Dave.
You seem to have researched this topic quite extensively and you put forth some compelling arguments. Are we to assume then, if we are shooting RAW, we should be overexposing on purpose in order to gain the full DR that the sensor is capable of? Would this not require that every shot would have to be extensively manipulated to gain acceptable results? Would bracketing a composition not achieve the same result with three or more exposures? Does you technique give us the ability of having enhanced DR without the hassle of multiple shots?
Erich

Great thread Dave, although much of it is on the cusp of what I can understand. With regard to jpeg + raw, and referencing the quotes below, are you saying the RAW data is reduced by virtue of the actual jpeg+raw setting, or, is it that by relying on the histogram you are only exposing for the DR range of the jpeg instead of the full range of a raw capture. Also, what is ERADR?

Quote:
So, in restricting exposure of a raw capture to that necessary for a JPEG image file, the "JPEG + raw" option relegates the captured raw data to the same data quality (sound:noise ratio) as that of its paired JPEG partner. The raw image data so exposed thus becomes functionally an 8 bit-depth image file delivered in a 16 bit depth environment.

Depending on the amount of unused raw-accessible DR the Raw data file will, in most cameras, suffer a 50% to 75% (or more) loss of its potential tonal and color spectra. that's an incredible waste considering that the sensor accounts for 1/2 to 2/3 of the price you paid for the camera body!
Later…
We must realize that every exposure results in capture of a mass of "raw" data. When the photographer elects to receive a JPEG file, that 8 bit-depth file is extracted from the massive captured data, processed according to the manufacturer's algorithms to result in a presumably acceptable image straight out of the camera (SOOC). The rest of the data is discarded.

When the delivery option "JPEG + raw" is chosen, the extracted JPEG file, as well as a file of raw image data are presented to the photographer...unfortunately with the raw image data having been exposed only as if it were a jpeg file. : end quotes