Uuglypher wrote:
[uquote=kymarto]Loss of 50 to 75% or more of tonal or color spectra? That sounds like a gross exaggeration. Please tell us how you arrived at that figure.
Quote:
Hi, Toby,
Thanks for joining in.
Given that the brightest available stop of exposure contains half (50% , if not more...) of its bit-depth's contribution of tonal and chromatic spectrum (depending on how much of the lower reaches of the DR was used...) -and that most cameras (about 70% of those tested) have AT LEAST one full stop of ERADR (if not more)... then for each stop of available exposure sacrificed , there is loss of 50% or more of its bit-depth's contribution. If two full stops of ERADR are sacrificed, the the loss is 75%. If two and 2/3 stops of ERADR are sacrificed, the loss is beyond 75%. And if the choice was also to expose to the left, well, you do the math.
Thus, if I were to fully expose my raw captures according to the JPEG histogram, Each image would have testified to sacrifice: one and 2/3 stops worth of tonal and chromatic spectra (somewhere between 50% and 75%) provided by their 14 bit-depth. To my mind, I'd rather not waste that much of the potential contribution of the sensor of a camera, 1/2 to 2/3 of the price of which supposedly went to the cost of the sensor.
This is all based on the twice-as-much /half-as-much relationship between full stops upon which we base so much of our exposure theory and practice.
The 2004 white paper by Bruce Fraser ""RAW Capture, Linear Gamma, and Exposure" is quite illuminating.
http://www.adobe.com/digitalimag/pdfs/linear_gamma.pdfDave
Hi, Toby, br Thanks for joining in. br br Given ... (
show quote)
Hi Dave,
OK, that sounds reasonable. I certainly agree that if you're shooting jpg+raw and exposing to keep the jpgs from clipping you are losing a big chunk of what the raws can do. However I must in good conscience point out a couple of things.
The first is that this is all a bit logarithmic: just as +3dB doubles the sound power but really doesn't seem like a large jump in volume to the ear, so too in the image: underexposing by a full stop and then raising the values to look like what it would have been had you exposed one stop more really doesn't look that much different in normal viewing--often even two stops of underexposure don't really seem to affect the final image that much, even though you are losing then your 75% of tonal values, as it were. Certainly on very close viewing the differences will be noticeable, but generally speaking what is there will by far outweigh what is not--just as the difference between what we can see when we light a single candle in a dark space (as compared to no candle) is MUCH greater that the difference between one candle and two. This is certainly not a reason not to use the full dynamic range of the sensor, but that brings me to my second point.
Here we must consider cost/benefit questions. The benefits of using the full dynamic range are obvious (though perhaps not as great in reality as the numbers might seem to indicate, as per the above)--but what are the costs?
Here we have to take a step into the real world and consider situations. Under controlled conditions, where time is not critical, I fully agree that ERADR is absolutely recommendable. BUT... but...In the field, with changing conditions with time-critical subjects, one has to balance the possibility of blowing highlights past ERADR and losing the acceptable exposure, against keeping a bit of headroom for the inevitable changes in maximum brightness in the scene that will inevitably occur.
Working professionally in audio, I might point out that standard practice in setting the zero reference level for recording is either -12 or -14dB, even sometimes -20dB. Heavens! We are losing something like 97% of our dynamic range by the way you count, and yet no one will ever notice the difference unless they listen to absolute silence in an anechoic chamber. And even if there was a noticeable difference, one would have to balance that against possible peak distortion if one were to raise the zero level, one instance of which would ruin a recording.
I think you see my point. ERADR is not a panacea, and can actually be more than counterproductive in many normal shooting situations. Even if one shoots for jpg exposure in a combined jpg+raw situation, the raw file will contain substantially more information, which can go into making a much better image that what is available in the jpg in any case, with the added insurance of some headroom if necessary. As good as with ERADR? Clearly not, but one must be aware of the difference, and weigh which is the better compromise in any given situation.