Started by uuglypher, May 17, 2016, 10:41:29 am
Quote from: uuglypher on May 20, 2016, 03:20:31 pmHi, Guillermo,The sentiments you expressed are exactly the reason that we were surprised to find variance in the extra raw-accessible dynamic range (ERADR) beyond the clipping point for the JPEG-adjusted histogram frame among camera's of the same model. That repeated tests for ERADR at different ISOs on a variety of camera models have yielded consistent results (to 1/3 stop) has reinforced reliance in the test procedure. We are not, in the manner that we test, determining the actual DR of any tested camera, simply the "extra raw-accessible dynamic range" which is then added to the ETTR exposure in an effort to utilize the maximal available dynamic range of each of our cameras. The fact remains that different cameras of the same model have yielded different numbers of 1/3 stops of ERADR. On that basis has rested the recommendation that each camera be tested for its individual allotment of ERADR. Perhaps you can explain how different amounts of determined ERADR among cameras of the same model might not, in fact, conflict with the concept of the accuracy of a single total DR determination. I've been in contact with DxO Mark but have not received helpful information in this regard nor been able to engage in any substantive discussion on the matter. It would seem most informative if someone without a "dog in this fight" would undertake similar tests for ERADR in several cameras of each of the same models and report those results. I appreciate your interest and would appreciate any further thoughts you may have.Best regards,Dave
Quote from: uuglypher on May 20, 2016, 03:20:31 pmThe fact remains that different cameras of the same model have yielded different numbers of 1/3 stops of ERADR.
Quote from: AlterEgo on May 20, 2016, 03:52:17 pmyou still did not show actual test procedure... neither links from your prev. postings in this topic showed, for example, how did you exactly test (exact OOC JPG parameters) and what was the exact camera model in question... also may be you can stop using the wording "we" , it does not add any substance.
Quote from: uuglypher on May 20, 2016, 06:19:16 pmI've been using EBTR for over a decade and it has been an interesting trip. I've been most pleased with its results when used under appropriate conditions. I've no intent to proselytize, just to explain how those of us who use the technique use it. If I tell you more than this, I'd be telling you more than I know for sure, and I'd rather not do that , considering how much of that is already going around!You'll like it or you won't. Good luck.Here y'Go!If you want to get started with Expose Beyond the Right and start taking advantage of all your camera's available dynamic range, here's how to start !:Here's the easiest way to get a handle on how much extra raw-accessible dynamic range (ERADR) your particular camera has at base ISO. On a somewhat overcast day (soft-edged shadows) find a scene with a dynamic range less than that of your sensor. In other words, a scene that produces a histogram that falls completely within your camera's histogram frame with no clipped shadows at the left (dark) end or clipped highlights at the right (bright) end.. Set the camera to its native ISO (usually ISO 100...or ISO 200 if you usually use that as the low end). Set to a specific WB (doesn't matter which one, just NOT AutoWB). Set "Creative" or "Personal Preference" Settings (brightness, contrast, Sharpness, and Saturation to "neutral". Establish a metered exposure for the scene: Take a spot meter setting from a middle gray region of the scene using either a hand-held spot-reading meter or your camera's spot meter function. Set to " manual exposure". Set to "manual focus" Attach your camera to a sturdy tripod or other solid support. Start exposing with progressively larger apertures (smaller F-Numbers) until the extreme right end of the "light pile" just barely "kisses" the right side of the histogram frame without triggering the highlight clipping warning "blinkies" (the ETTR exposure) Now, start a series of exposures by increasing each subsequent exposure by 1/3 stop of shutter speed, Add four full stops of exposure (that's twelve exposures, each 1/3 stop longer in shutter duration than the previous one). Note that the previous recommendation was for three full stops, but recent reports of a camera with more than three full stops of ERADR have suggested we ought look beyond three stops! Who knows? Your camera may be a record setter! Download the image files into your raw converter. Open the exposures in sequence and tonally normalize it by left-sliding the "Exposure" slider. As each "washed out, too bright" image is tonally normalized, you'll see the histogram slide left and fully into the histogram frame. Do this in sequence until you find the histogram that contains a "spike" at the right end of the light pile that indicates blown highlight detail. Just a caution here... if the scene contained some specular highlights, then a very thin, one-pixel wide spike is permissible. If you see such a "specular highlight spike", go ahead and open the next frame, which should have a substantial spike of blown/clipped highlights at the right end of the light pile. The number of exposures between your brightest possible JPEG exposure (the ETTR exposure) and the first exposure with actual blown highlights( not just specular highlights) constitutes your camera's ERADR (Extra Raw-Accessible Dynamic Range).1 exposure? 1/3 stop2 exposures? 2/3 stop3 exposures? one full stop.4 exposures? one and 1/3 stops (as illustrated in the case discussed above)5 exposures? one and 2/3 stops 6 exposures? two full stops7 exposures? two and 1/3 stops8 exposures? two and 2/3 stops9 exposures? three full stops10 exposures? three and 1/3 stops11 exposures? three and 2/3 stops (not seen one of these yet...!!!!) 12 exposures? four full stops (not seen one of these yet ...!!!!!)Caveat:You should test your camera's ERADR at the ISOs you most commonly UseAs ISO increases, the camera's amount of ERADR gradually and irregularly decreases. As an example, I've a camera that, at ISO 100 has one and 1/3 Stops of ERADR. At ISO 400 it drops to one full stop, and at ISO 3200 the ERADR is 2/3 stop.How to Use that Extra Dynamic Range:So, how do you actually use that ERADR once determined? Set your exposure as for a JPEG file using either your camera's meter, or a hand-held reflective, spot-reading, or incident light meter. Enter the exposure into your camera, and, adjust that exposure so that the right end of the "light pile" of the histogram moves right to barely kiss the right side of the histogram frame juuuuuust before clipping the highlights clipping warning/the "blinkies". It doesn't matter if the right end of the light pile is a sizeable "lump" or a just a slender white line extending along the frame's base. When it just barely touches the right side of the frame. THAT is the ETTR exposure -your starting point. THEN.... add your known stops of ERADR using either slower shutter speeds or larger apertures (or a combination) and make your exposure. Note: Most who regularly use EBTR normally use 1/3 stop less than their established ERADR ,just to " play it safe" as regards possibly clipping highlights. Why "play it safe"? Because in some cameras there is some inaccuracy in tripping the "blinkies".REMEMBER: When you start using that extra dynamic range, realize that your camera's LCD display of the viewed thumbnail will, with most scenes, likely appear "washed out" with "blown highlights" (because it "thinks" it is displaying an "overexposed" JPEG file instead of a correctly exposed raw image file. So... DO NOT INTERPRET SUCH WASHED-OUT IMAGES AS "OVEREXPOSED"!!! THEY ARE CORRECTLY EXPOSED RAW IMAGE DATA FILES !!!Download the files into your computer's raw converter (Adobe Camera Raw is fine), move the "Exposure" slider to the left, and you'll see the image become tonally normalized with its full, expected tonal range and tonal spectrum! When you are new to EBTR, every time you tonally normalize a washed-out image, it's like a miracle to watch that tonally perfect image appear right before your eyes! It sort of reminds me of watching an image appear in a print in the developing tray under the dim safelight in my dad's darkroom in the 1940s."So what's the big deal about a measly Stop?It is at about this point in a class or workshop that some back-row sitter pipes up with: "C'mon, Dave, are you really telling me that just one measly stop is really worth all this trouble?"So I ask him, "What's your best, fastest lens?"Er...that'd be my Summilux-M 50mm f/1.4 ASPH. , why?"""we'll, would you mind if I took your lens and superglued it so it it couldn't open up beyond f/2 ?"or...."What if you couldn't, simply by increasing raw exposure by one stop, thereby reduce the amount of noise you might routinely capture?"or"Think of the times that limited illumination have made you wish for "just one more "measly" stop?"or,what about being able to cut your already fast shutter speed in half to permit hand-holding with a longer lens, or using a monopod instead of a tripod?Yeah, I'd say there's considerable advantage to getting every bit of extra dynamic range as you can out of that sensor that amounted to 2/3 of the cost of that new snazzy-a-sed camera body you just bought!Here we're talking about pulling out, at the very least, 2/3 stop of exposure up to two an ⅔ stops to possibly more than three stops of exposure beyond what your camera's lying JPEG-adjusted histogram suggests you can actually use! 'Tain't chicken feed, McGee! "One measly stop of exposure? I wouldn't give that up, as we said back in the day, "...for love nor money"'Why don't I just compile a list of camera brands and models and state the tested ERADR of each? Because three camera's of the same make and model can have three different amounts of ERADR at a given ISO! Also, I've received some unpleasant responses from camera owners who used the ERADR amount that I had found in their camera model and they overexposed some important "irreplaceable" images. (they just hadn't caught on that they had to test their own camera!) So, I'm sorry...but you just gotta test each individual camera!As the spirit moves and time permits, I'm happy to help anyone test their own camera. or to examine their ERADR series of exposures. If appropriately prepared, I'll report back the ERADR determined. Just contact me by personal message.So far, I've encountered two camera models represented by more than one camera. Two cameras of one model and four of another...and so far no two cameras of the same model have had the same allotment of ERADR. The two-camera pair were ⅔ stop apart, and the four camera quartet had a range difference of one and ⅔ stops. Go test some cameras. Find out for yourself! I'll be glad to help evaluate your ERADR exposures as time permits. I'm glad to help anyone pursuing this interesting phenomenon (and to discover a way to assure that you are using your camera's maximal available dynamic range).Give it a go! Some wil try it, like it, and never look back! For others it'll be too much trouble. I've been using EBTR for twelve years when conditions were appropriate. Give it a go! You'll never know if you'll like it or not until you actually try it. But by this time I've realized that there will always be some sideline sitters who'll carp about why it won't work, yet never try it themselves. It has been ever thus!Best regards, and let me know how you do.Dave
Quote from: AlterEgo on May 20, 2016, 09:12:54 pmDave, a lot of words - so what were the cameras and then you probably can show us couple of raw files as a proof so that we can see how everything was identical yet the 2 different cameras of the same model/make/etc did so differently in the identical conditions (identical scene & framing, light, OOC JPG parameters, camera settings, lens, firmware version, etc, etc - everything is the same) are producing identical JPGs (OOC JPG histogram-wise) yet with a lot (>> 1/3 EV) different saturation in raw channels ... that will be (the only) really interesting point in your posting, sorry... so far you did not produce any proof at all.as for everything else, thank you, but I am using rawdigger ( www.rawdigger.com ) to find how my camera's spot metering is calibrated and settings I outlined above do serve me well (as verified by rawdigger again)... so do other people and Guillermo Luijk most probably likes to use his own tool (histogrammar)
Quote from: Guillermo Luijk on May 20, 2016, 07:55:11 pmHi Dave, some of us have already walked through the paths you are following time ago. What you call ERADR is not DR but highlight headroom between the camera JPEG and the RAW file, and will strongly depend on camera processing. This includes adjusted WB, output color profile, contrast/saturation settings,... Also the kind of light will have an influence becase the spectral components will impact in the RAW channel distribution itself.More logical than measuring "JPEG vs RAW" headroom, is to measure the "Metering vs RAW" headroom, i.e. know how many stops you can add to the exposure meter 0.0 point before having actual RAW clipping. Still the scene lighting will have a strong impact here.E.g. RAW file metered over a gray uniform surface under tungsten light and exposed accordingly:RAW histograms:Canon 350D:Canon 5D:Even with that reddish light the G channel captures photons the most, having around 3,5 stops of RAW headroom vs camera metering (apparently a bit more in the Canon 350D, but remember we are exposing in 1/3EV steps so there is room for a slight deviation).Regards
Quote from: uuglypher on May 23, 2016, 01:34:04 amYe Gads...Take off for the weekend and return to find everyone's been busy refuting much of what I may have long held as carefully considered reliable opinion, if not actually irrefutable fact!Guillermo states:" ....some of us have already walked through the paths you are following time ago. What you call ERADR is not DR but highlight headroom between the camera JPEG and the RAW file, and will strongly depend on camera processing. This includes adjusted WB, output color profile, contrast/saturation settings,... Also the kind of light will have an influence because the spectral components will impact in the RAW channel distribution itself."
Quote from: uuglypher on May 23, 2016, 01:34:04 amAnother thoughtful and knowledgeable photographer of reasonable note, Jeff Schewe, has published another classic example of accidental, inadvertent EBTR in "Un-debunking ETTR"http://schewephoto.com/ETTR/In fact, Jeff's example of accidental EBTR has undoubtedly sparked the interest of more than a few to pursue testing their own their own cameras for "whatever it is" beyond the right of the JPEG clipping spot that does a superlative job of standing in for extra raw-accessible dynamic range when, to our dismay, we hear that extra raw-accessible dynamic range being proclaimed not to exist.
Quote from: bjanes on May 23, 2016, 07:26:57 amIMHO, Guillermo is exactly correct. Differences in highlight headroom between cameras of the same make and model derive not from variations in the sensor itself, but from variations in the in camera processing by the JPEG engine settings that Guillermo mentions and/or from differences in raw converters. For example ACR/LR have a BaselneExposure tag (see page 38 of the DNG specification 22.214.171.124) that adjusts the exposure setting of the converter and also a BaselineExposureOffset (page 71) that further adjusts exposure when a DNG profile is in use. ACR/LR also are image adaptive and apply automatic highlight recovery with PV2012. For these reasons, ACR/LR are not good tools to evaluate highlight headroom, and use of Rawdigger or a similar program is essential for this purpose.Jeff's points are well taken, but if you look at the raw file of Jeff's waterfall image with Rawdigger, you will see that the ACR histogram is misleading and that the image is not so severely overexposed as the ACR histogram would lead one to believe and "highlight recovery" is less impressive than the example would leave one to believe. If highlight colors are clipped, there is really no way to recover them exactly and intelligent guesses are necessary. ACR/LR tend to render received highlights as neutral gray/white and this often works well, since many highlights are indeed neutral. However, in many cases the highlights may not be neutral. See Guillermo's DCRaw section on white balance and look at Figure 9. ACR recovers the blown skin highlights as neutral, but DCRaw offers an option to look at colors adjacent to the blown highlights, and deduce that the blown colors of the face are skin tones.Since modern sensors such as found in the Nikon D810 and other recent cameras have such a high dynamic range, ETTR is less valuable than previously. With this camera, it is better to have slight under-exposure of the highlights than risk the data loss that blown highlights impose. Regards,Bill
Quote from: BartvanderWolf on May 21, 2016, 06:27:36 amI agree. My experience with multiple camera bodies does not indicate large variations, on the contrary. It's more likely to cause variations by illumination or exposure metering not being constant.As for determining the Raw headroom to achieve ETTR, a tool like Rawdigger, or even Fast Raw Viewer, is indispensable. Rawdigger will let you see exactly what the maximum DN is per color plane, although some cameras digitally amplify (some) channels to facilitate lossy compression with minimal impact on image quality. Subsequent WhiteBalancing could affect the compromises one makes regarding acceptable clipping. FRV will allow to determine an Auto Exposure correction based on the amount of clipping one can tolerate, if any (see attached settings). With an exposure bracketed series, only the 'under-exposed' images get a positive exposure correction, others stay at zero because some data is already clipped. Even in the case of one or two channels being clipped in Raw, modern Rawconverters do a reasonable job of reconstruction the missing channel data whit their Highlight recovery algorithms.There is also another potential issue with ETTR, and that has to do with some Rawconverters that introduce color twists when pulling the ETTR exposure of a low contrast scene to a more normal level during the conversion.Cheers,Bart
Quote from: Lundberg02 on May 21, 2016, 06:12:45 pmSemiconductor mfgs deal with process variability by assigning different model numbers to the units in different sigmas of the output lots.
Quote from: TonyW on May 21, 2016, 12:12:10 pmHi DaveInteresting points but I guess that your audience may be novice users and the intent is to show that there is more headroom shooting raw than JPEG and proving the point that the camera LCD histogram display and the blinkies are based on and only reliable indicators for JPEG camera images? Of course I may be misunderstanding.In other words a 'correctly' exposed JPEG will be an underexposed raw, or perhaps more correctly an image with less than optimal exposure.Exposure Beyond The Right (EBTR) is not a term that I care for as it implies that you can pass the far right 'wall' of any histogram which of course is/may be correct for most/all in camera histograms for raw but not necessarily the case for all raw editors due to the potential for behind the scenes activity to present an 'acceptable' on screen renderingFor instance Adobe are among those that apply a baseline correction for specific cameras and this coupled by other background changes with process versions and rendering can lead a user to believe that they may be overexposing when in fact the opposite is true - possibly up to 2 EV.Others such as RawTherapee seem to show the image close to that captured by the sensor i.e. without much in the way of brightening the imageMost camera exposure meters (and separates) are likely to be calibrated to around 12-14% and I would expect that the you may have a maximum of around +3 stops before clipping with many cameras. So once you find this point you can meter your important highlight areas and apply the correction without being concerned with the camera histogram display I also agree with others that large variations with cameras is likely to be variations in illumination, metering or even (rare I think) shutter speed or aperture variations
Quote from: uuglypher on May 23, 2016, 09:01:15 amOf course, The technique I am using may not be disclosing "Raw headroom" ...but if not, what is that stuff acting as if it really were useful dynamic range? That's a totally serious question, by the way...
Quote from: uuglypher on May 23, 2016, 09:26:02 amYour concern regarding the raw converter used may well have merit. My experience over the years has been limited to Adobe Camera Raw in its various iterations, except for having started with some plug-in raw converters prior to the advent of ACR. ACR was a breath of fresh air hat has continued to improve phenomenally as time has gone on.
Quote from: uuglypher on May 23, 2016, 09:26:02 amMy experience over the years has been limited to Adobe Camera Raw in its various iterations, except for having started with some plug-in raw converters prior to the advent of ACR. ACR was a breath of fresh air hat has continued to improve phenomenally as time has gone on.
Quote from: uuglypher on May 23, 2016, 09:26:02 am....Your concern regarding the raw converter used may well have merit. My experience over the years has been limited to Adobe Camera Raw in its various iterations, except for having started with some plug-in raw converters prior to the advent of ACR. ACR was a breath of fresh air hat has continued to improve phenomenally as time has gone on. Cheers,Dave
Quote from: uuglypher on May 23, 2016, 08:55:44 amHi, Bill,I appreciate your comments and and certainly am aware of the pervasive concern that perhaps sacrificing some DR may be preferred to loosing some detail to clipped highlights in the normalized EBTR exposure. The saving grace of EBTR is that the described method of determining the ERADR --the extra raw-accessible dynamic range - if practiced with reasonable care at the chosen ISO - is a guar-an-day-um-tee that highlights will not be blown. It is the common mantra that clearly identifies those who have not actually tried to determine ERADR and practice EBTR.Here is another example of EBTR:
Page created in 0.048 seconds with 15 queries.