Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: Ray on June 04, 2005, 01:55:42 am
-
The RAW converter in PS CS2 has a default automatic adjustment of exposure, shadows, brightness and contrast.
It's interesting that many images that I thought at the time were clear of the right end of the histogram, are given a compensating underexposure in ACR, like - 0.3 or -0.5.
-
Just tried it, a bit crude, doesn't seem to be SF or OOF, a bit like a very very harsh edge sharpen with blur applied to the rest.
Talking of which, I spent a while working with smart sharpen. As you pointed out in an earlier thread it needs a bigger radius than USM to achieve comparable results. I found that to keep the halo and noise levels down I had to sacrifice a lot of fine detail/resolution in the whites and darks. I tried PhotoKit and although it did the job better than smart sharpen, for portraiture it just didn't look right, the contrast between sharp and defined edge detail and unsharpened unresolved skin detail was too harsh. I sat down and compiled an action which (albeit crudely) sharpens the edges and non edges seperately with a noise reduction on the non edge areas. I find that it gives a more subtle effect of skin softening than does Photokit, it's customiseable and free!
If you're bored take a look, the first action is to replace USM 250,1,0 (my default setting for 8X12" prints from my 10D), the second is customiseable at almost every step.
I made the action specifically for portraiture, I've not tried it for landscape work. If you can tell me how to do the de-noiseing stage without using as blunt an instrument as 'despeckle', in CS, I would be grateful.
left click and 'save target as' (http://www.bphotography.co.uk/edge.atn)
-
The whole idea is to shoot a scene BEFORE the moment of interest occurs there, so that you can know your settings are appropriate.
Good L-rd.....
-
Pom,
your idea sounds equivalent to selecting a bit of underexposure compensation and then fixing the overall tonal level in the digital domain. If the highlights are blown due to sensor wells filling up, lowering ISO at the same time will not help, but at the elevated ISO setting of ISO 400 and 200 that you mention, the sensor is being "underexposed" anyway, so that is probably not the problem. Instead highlights are probably lost mainly in JPEG conversion.
So is one option to use raw with say -1 compensation, and then do batch conversion with the corresponding +1 set, and maybe an appropriate constrast level or tone curve? Cases where highlights do awry would probably then need a second custom conversion.
If you need JPEGs straight from the camera for impatient new spouses, I can see a need for an appropriate in-camera mode, but maybe such modes are already there. Perhaps a low contrast setting for the in-camera JPEG conversion, or custom tone curves (does the 20D have them?), with JPEG+RAW to allow for some later custom conversions in the tough cases. (I am still in the early stages of experimenting with low contrast in-camera JPEG versus RAW and later manipulation, via RAW+JPEG mode.)
-
Just got the 1Ds today, I see what you mean, at iso 400 a stop underexposure makes the shadow noise look as bad as iso 1600 on my 10D, WOW, at least the evaluative metering is far more accurate so I'll need to dial in less 'buffer' underexposure.
-
The difference in exposure accuracy between the 10D and 1-series is significant: it's also way better than my 1v. I can now shoot on evaluative +1/3ev for just about everything, and only really have to move off that (and shift ISO around - need a faster way of doing that) when stage lights dictate. Pom, I think the 1Ds will change the way you work.
You're telling me! I've done little testing but the 1Ds holds the highlights far better than the 10D which seems to peak much faster. The iso settings are 'true' so that my hand meter can be taken as read (the 10D seemed to be shooting at almost a stop over of sensor sensitivity, i.e. iso 200 when 100 was showing, or at least that's how it felt sometimes!) and the evaluative metering is a world apart. Being able to trust your metering and camera's exposure instead of working it out in your head the whole time to compensate for the cameras inefficiency will certainly take a lot of pressure of me when shooting weddings.
-
My experience with the 10D indicated that there was basically zero difference between the histogram indicating clipping and the RAW data being clipped, while the 1Ds & 1D-MkII have between 2/3 and 1 1/3 stop interval between histogram clipping and RAW clipping, depending on the color temp of the lighting. Daylight (5000-6000K) has the largest available headroom, and it decreases as you go above or below this range. Really low color temp stage lighting (say 2500K) decreases the clip interval to about 1/3 stop in the red channel, so that's one reason why nailing the exposure dead-on just below the clip point is critical but difficult when shooting typical concerts.
-
Now that was fast, you replied while I was editing. You are right of course, I use flash for 100% of my people work, either as main light or fill, or studio setting of course. I stopped using it for landscapes as the trees kept blinking.... (sorry)
In case you're thinking that it was the flash that was overexposing, I was shooting with auto flash, fill dialed in manually at -2 stops for all the shots, the fstop was the same whether metered manually or with evaluative, I was changing the shutter speed.
-
A lot of DSLR users are purposely underexposing to preserve the highlights and correcting in RAW. However true the 'expose to the right' concept is, in the real world it still makes sense to keep most of the highlights away from the fifth section of the histogram when you can't go back and do it again. With the modern excellent charactaristics of shadow noise or the lack thereof, this isn't really a problem.
If the camera manufacturers were to develop a DSLR whose meter would be 1 stop underexposed, i.e. iso 400 was really iso 200 and the software, either in camera or as a RAW developer would bring the exposure up a stop while compressing the highlights (on a mini scale of what HDR does) , assuming that the noise control was good would that not effectively be a way to market a DSLR with less chance of losing a photo through overexposure?
Of course the documentation would have to instruct the customer to caliberate their hand meters to the camera, the noise in the shadows would have to be pretty good, and the implementation of the HDR type software would have to be carefully written, but it would be a cheaper way than installing extra low sensitivity pixels like the Fuji does.
I know that I would prefer the noise characteristics of an older generation camera if I didn't have to worry about my highlights, when I underexpose by a stop with my 10D I am essentially shooting at one higher iso as far as noise is concerned.
I know I'm rambling but if the underexposure was built into the camera's meter and 'iso' settings you could shoot using the regular modes without compensation while getting extra DR in the highlights for a more film (negative) like feel.
For a landscape shooter this might be irrelevant, for a journalist/event/wedding shooter who often have to shoot on the fly it could make a big difference.
I'm probably talking out of my bottom but if the manufacturers are having trouble getting the pixels not to hit the wall at 'x' amount of overexposure, this trick plus a HDR type software solution could work as an idiot proof solution.
Yes you can dial in underexposure to any of the shooting modes. If the default metering was underexposed then you wouldn't have to remember to dial it in, nor would you get confused when overexposing for white by a stop and finding the setting on '0'. When I shoot in Av or Tv mode it confuses the heck out of me.
-
Jonathan, I'm a wedding shooter, I cannot allow my self the luxury of checking the histogram and shooting again nor would I have time for bracketing. I have to get it right first time every time. I do the same as Jack, I have the meter in partial mode, meter of the highlights or the face in manual mode, adjust exposure to match and shoot. If I don't have time for that then I have an ambient meter around my neck when shooting and dial in a -1/2 to 1 stop from the reading dependant on the DR of the scene.
It's a pain in the neck that's all. For pleasing facial tones you need a gradual and controlled transition from the highlights to the midtones which pro neg film provides. To replicate that with digital needs careful attention both when shooting and post processing and the cameras metering doesn't usually help...
-
I found that to keep the halo and noise levels down I had to sacrifice a lot of fine detail/resolution in the whites and darks.
SNIP
If you can tell me how to do the de-noiseing stage without using as blunt an instrument as 'despeckle', in CS, I would be grateful.
Unfortunately, the CS sharpening step is hugely dependant on "how what" was used to convert the raw file and of course what camera was used and at what ISO the raw file was captured at. So what works for me may not -- actually probably won't -- work for you...
For my 1Ds2 files at ISO 100 - 400, I use RSE to convert, and generally use its sharpening set to 0 or -10, detail extraction at 0 and NR at 0.
With the above parameters, I generally set Smart Sharpen to somewhere around 100%/1.0/Gaussian/More-Accurate, the shadow 25%/25/4 and the highlight at 50%/50/4.
You obviously need to tune to your own files, but specifically the shadow radius and tonal width settings have a big effect on how the noisy areas are "smoothed" and is (not) sharpened. Higer radius = less and higher tonal width = smoother.
~~~
As for the surface blur filter... For skin, try it at a very low radius, like 2 and low threshold like 5, then apply it on its own layer and dial opacity down to around 75%. ;)
-
If you need to chimp your histogram to get your exposure right then you shouldn't be shooting weddings either. How on earth did you shoot with film? Neg film does not save you for subject/ambient light ratios. Overexpose the subject and you lose your background. If you can't expose without your histogram then you don't know how to meter.
Since I changed over to auto flash and using either partial metering in manual mode or an incident meter, I chimp only every now and again and have to redo shots only very rarely.
Of course wedding photographers are prepared, in advance if there's time, for the different sets of shooting they will encounter. The PJ style shots, where a candid moment can happen anywhere in just a second is where you need to know and trust your equipment to get it right first time every time. Having the extra latitude to save you in such an occasion can help.
For example, you are shooting portraits of the bride in the shade under a tree. You take a step back and notice that the 5 year old bridesmaid has her arm round her little cousin. You have about two seconds to get the shot, maybe less, the moment they notice your camera they will freeze up. In such a case you need to change from f1.8 to f5.6, compose and shoot in about a second. Oh and they are in completely different lighting to the bride. If you need to have prepared for such a shot, have sorted out your histogram in advance, you're screwed. Those type of shots make about 40+% of my wedding work. How ever much you are ready for the 'set' scenes.....
-
Av mode with EC dialed in appropriately is the only way to go in that situation. Manual is useless when ambient fluctuates 3 stops or more moment to moment. At one of my recent concerts, the lighting fluctuated enough that my shutter speeds were ranging between 1/15 and 1/800. And getting exposure exactly right in-camera is important, because shooting at ISO 800-1600 is mandatory (no flash is ever allowed), and trying to fix underexposed shots in the RAW converter is a sure-fire recipe for disappointment. The composition also affects the correct EC amount, which can be anything from -1 2/3 for a spotlighted performer shot from the side against a black background to +1 1/3 for someone in light clothing on a white background. In a situation like that, you need to re-check camera settings every time you change compositions, and often when the lighting changes, like if it goes from a single spot to all stage lights on. If you use any one set of camera settings, you're screwed. Weddings are cake compared to concerts; concerts are usually 100% candid-type shots requiring split-second timing in a highly unstable lighting environment where 2/3 of a stop exposure can differentiate cream and crap.
Trusting settings in a concert context is somewhat like trusting jelly to stay nailed to a tree; frequent histogram verification and fine-tuning settings is the only way to ensure consistently good results. It's always a game of catch-up until you get familiar with the settings required for the various angles and lighting conditions. And even then sometimes they change.
-
Bovine excrement?
-
You're telling me! I've done little testing but the 1Ds holds the highlights far better than the 10D which seems to peak much faster. The iso settings are 'true' so that my hand meter can be taken as read (the 10D seemed to be shooting at almost a stop over of sensor sensitivity, i.e. iso 200 when 100 was showing, or at least that's how it felt sometimes!) and the evaluative metering is a world apart. Being able to trust your metering and camera's exposure instead of working it out in your head the whole time to compensate for the cameras inefficiency will certainly take a lot of pressure of me when shooting weddings.
I think you just may have put the finger on something that's been bothering me about the 20D also.
I've struggled some (but not too often) with highlights in exposures, and of course I've mainly blamed myself and not the tool. Obviously, I'm still at fault, but if the 1Ds does a better job off it, I might have some serious thinking to do, and maybe I should test that camera.
(No, wait, I can't afford a used 1Ds yet!)
-
My incident meter is still seemingly more reliable than evaluative for outdoor portraiture but although the difference was almost a stop between the readings (I focus recompose which probably doesn't help), the overexposed images were fully recoverable and when ACR'ed (-.30 exposure) were as good as the correctly exposed images. My 10D would have lost the picture with overexposed highlights on the face. I would be quite happy to use Av mode and evaluative if needed. I'm going to try spot metering from the face to see whether it's the evaluative that's overexposing or the scene just had too much DR without fill which is what I expect.
-
If I overexpose the subject the the background goes darker which I prefer to avoid.
Only if the overexposure is from flash. Otherwise the exposure balance between subject and background will remain exactly the same.
-
Here's a much better idea that can be easily implemented on any RAW-capable digital camera right now:
Digital Exposure And Metering Strategies (http://www.visual-vacations.com/Photography/exposure_metering_strategies.htm)
-
I'm a wedding shooter,
As a totally OT aside, have you tried the new "surface blur" filter in CS2 yet?
,
Jack
-
If you can tell me how to do the de-noiseing stage without using as blunt an instrument as 'despeckle', in CS, I would be grateful.
Neat Image or Noise Ninja...
-
I would love to have your shoulder strength for two 1 series bodies. You shoot a lot of concert stuff don't you? With lights flashing on and off all over the place how can you shoot that way, don't you need to trust your settings to get it right?
-
So is one option to use raw with say -1 compensation, and then do batch conversion with the corresponding +1 set, and maybe an appropriate constrast level or tone curve? Cases where highlights do awry would probably then need a second custom conversion.
That is precisely what I do, the problem is that calming the highlights down is not that precise in ACR, I use the minus exposure plus brightness and a tweak of the curves. For portraiture I use a standard Highlights=2 contrast =3 in the shadow/highlights tool as I find that it gives a more film like response.
Jonathan, can you not use Av with spot metering on the face? Would that not keep the exposures more even as you don't have to worry about the ambient or spotlighting?
I use an incident meter indoors and outdoors when the light is stable-ish. If the light is changing I can usually see it happening and re-meter. I bought a 1Ds today and am going to experiment with using the spot meter in manual mode to meter off faces which should be more accurate then doing the same thing with partial metering which is all the 10D has.
I found in Av mode that the shutter speeds can jump so far between shot to shot that I can't really trust that I'm shooting at handholdable speeds. My auto flash does not have hi sync so using Av and juggling the ISO to keep the shutter speeds between 1/60 and 1/200 is a pain in the neck.
I'm usually shooting max iso 400 and RAW of course, a stop underexposed doesn't really bother me at all so when I meter off the face, I keep that setting in manual and underexpose a bit. That way I preserve the highlights, keep better tonality in the face and am usually within a stop of the perfect exposure. I check regularly of course.
-
BJL,
I recall your mentioning this before. Are you implying there's a little man inside the camera that thinks, 'bugger, he's got the wrong exposure again. I'll go easy on the amplification'?
I haven't noticed that high ISO's have a tendency to clip less with overexposure, in RAW mode. But maybe you are right.
Mainly I am implying that at higher ISO setting, the amount of light reaching the photsites is less (higher exposure index, meaning a shorter exposure time for the same f-stop). Higher ISO settings entail a higher amplification factor in the charge-to-voltage conversion pre-amplifier, used to convert electron counts from photosites into input voltages for the A/D converter.
As far as I know (Jonathan might know in more detail), highlight clipping can have at least three causes, or which only the first is naturally avoided by using higher exposure index ("ISO")
1) over-full electron wells
2) maxed out pre-amplifier or A/D converter
3) at JPEG conversion
Every doubling of ISO above the minimum halves the electron count produced at wells, but then every stop of + compensation has the opposite effect. The electron count from the wells is then amplified (on-chip pre-amplifier with Canon CMOS) before going into the A/D converter, so that, with no exposure compensation, any ISO setting will produce the same signal strength going from the pre-amplifier into the A/D convertor. So it is possibe that at this stage, the extra amplification used at higher ISO causes pre-amplifier "clipping", or exceeds the maximum input level that the A/D converter can handle.
Finaly, if you use "straight" conversion from RAW to JPEG, whether in-camera or on-computer, the highest output levels might get clipped or compressed. JPEG tends to allow for only about 2 1/2 stops from mid-tone to "level 255", whereas th ndstruy standard for DLSR sensor base ISO speed measurement is placing mid-tones at 18% and the full well capacity at 170%, a bit over three stops of headroom.
If so, lowering the contrast setting (applied during JPEG conversion) could salvage something.
-
Just got the 1Ds today, I see what you mean, at iso 400 a stop underexposure makes the shadow noise look as bad as iso 1600 on my 10D, WOW, at least the evaluative metering is far more accurate so I'll need to dial in less 'buffer' underexposure.
For most stuff, I shoot Av mode, evaluative metering, +1/3 EC. For most outdoor/sunlit stuff that'll get you pretty close; I shot an outdoor all-day music festival last Saturday (>3300 frames, a new 1-day personal record!) with those settings on the 1Ds and 1D-MkII for about 80% of the daytime stuff. After dark I had to use a variety of settings for different shots, anywhere from -2 to +2/3, depending on the elements in the shot and the stage lighting.
-
I recently did a wedding in Thailand; shooting everything from indoor, outdoor bright directional sun, heavy shade in same sun, afternoon with low directional sun and finally late evening and fireworks, and hardly shifted off Av mode + evaluative the whole day (except for fireworks, obviously). No blown highlights, lots of noise-free detail in the shadows which was easily lifted out from the high contrast shots. I found this a more challenging metering problem than most I've done in the UK, where you're pretty much assured of a nice tupperware-grey sky to light everything pretty evenly and flatly! I also used a 550EX dialled down half a stop, and the equipment (though heavy) worked beautifully together.
The 1DsMkII is a fantastic camera - the best I have owned by a long mark.
-
Not really when the overexposure is on the face losing skin detail. The straight evaluative exposure blew the white shirt of my subject completely which is why I decided to go for an incident reading.
With a -1 stop, i.e. the incident reading, the picture was perfect straight out of the camera. Apart from which if I overexpose the subject the the background goes darker which I prefer to avoid.
I've set my '1Ds' custom settings in ACR as: exposure -1.5, brightness +150. It gives the same result as the standard setting (0.0,+50) but with better highlight control. I also straightened out the highlight end of the curve in ACR for slightly less contrast in the highlights while maintaining good overall contrast at a standard setting of +45.
I usually have to batch a lot of photos at a time, though nothing like you seem to be doing thank G-d, Finding a set of values that I can batch all the photos to so that they need only minor tweaking of exposure afterwards is very important to me.
BTW I'm having difficulty finding a level of USM that works for portraiture on the 1Ds. With the 10D I was using 230,1,0. With these new pics I haven't made up my mind yet, they need more sharpening but they also have worse grain. For custom enlargements I use my own edge sharpening action but a standard setting for the proofs I haven't found yet.
-
From a practical standpoint I think it would be easier to dial in any compensations directly myself -- at both ends of the process -- rather than have the meter calibrated so it read some value under and a raw processor compensate for it. Also, often the difference between a blown or clipped channel and a full DR exposure is just 1/3 stop...
In the end I think it makes a lot more logical sense to learn your system and develop a standardized workflow like Jonathan's article suggests.
For landscape I am usually in manual mode and double-checking the histogram. For action, I am lazier and put the cam in Av and dial in -1/3 or -2/3 exposure comp (depending on the camera I'm using), then deal with it later in raw or post.
Bottom line is I prefer to be the one making the decisions....
Cheers,
Jack
-
Jonathan, I'm a wedding shooter, I cannot allow my self the luxury of checking the histogram and shooting again nor would I have time for bracketing.
That's bovine excrement. I shoot weddings as well, and have no trouble finding time to point the camera down the aisle before the wedding party starts walking it to double-check the histogram. The whole idea is to shoot a scene BEFORE the moment of interest occurs there, so that you can know your settings are appropriate. This is not hard to do, all you need is knowledge of the flow of events so that you know what's going to be happening and where before it happens. If you're not already accomplishing that, you have no business shooting weddings. It's one of several reasons I always shoot the rehearsal; I have a chance to shoot the venue beforehand and familiarize myself with the lighting, see the exact sequence of events and figure out where I want to be, and deal with any restrictions there may be regarding flash usage, movement restrictions, etc. before the actual event. And one can usually get some decent candids as well.
-
I chimp a lot because I try to get every shot as close to ideal as possible without blowing highlights, so I get anal about 1/3-stop adjustments that most photographers wouldn't worry about. But I know when to chimp and when to just shoot. If I'm off a little bit, it usually doesn't matter, because the difference usually isn't something that would be visible in a print without doing a side-by-side comparison with a loupe.
In your hypothetical grab-shot situation I'd be prepared because A, I'd have already done some test shots in the lighting not under the tree and know what the appropriate settings were, and B, because I shoot weddings and events in general with both my 1-series bodies on my person. If I'm in a situation where there's two significantly different lighing conditions adjacent to each other, I'll have one body set for each lighting condition, and if I see a grab-shot opportunity, I can drop one body, grab the other, and get the shot while you're still fiddling around changing settings. Or I'll have worked out some camera settings that will work well for both lighting conditions; Av mode can be handy for that sort of thing. Don't asssume that meticulous preparation and successfully capturing grab-shots are mutually exclusive, the truth is that the first is often key to the second.
-
your idea sounds equivalent to selecting a bit of underexposure compensation and then fixing the overall tonal level in the digital domain. If the highlights are blown due to sensor wells filling up, lowering ISO at the same time will not help, but at the elevated ISO setting of ISO 400 and 200 that you mention, the sensor is being "underexposed" anyway, so that is probably not the problem. Instead highlights are probably lost mainly in JPEG conversion.
BJL,
I recall your mentioning this before. Are you implying there's a little man inside the camera that thinks, 'bugger, he's got the wrong exposure again. I'll go easy on the amplification'?
I haven't noticed that high ISO's have a tendency to clip less with overexposure, in RAW mode. But maybe you are right. :D
-
Jonathan, can you not use Av with spot metering on the face? Would that not keep the exposures more even as you don't have to worry about the ambient or spotlighting?
Heh. Not when you're shooting racially mixed groups, and especially not when faces are off-center or you're shooting wide shots. I'm in California, where whites will be a minority in 10 years or so if current demographic trends continue. Black, Hispanic, Filipino, and Asian people are common, as well as every possible combination of the above. One of the bands I shot yesterday had 2 white girls, one of whom was a blonde with a very fair complexion, and two black girls, one of whom had very dark skin. Spot metering would have made exposure vary nearly two stops, depending on which girl I was shooting. So I used Evaluative with +1/3 EC and things came out pretty consistent.
Also, that idea is totally useless when trying to get wide shots of the whole band and stage setup, as it is fairly unlikely you'll be able to spot meter on a face when the people are moving around too fast to keep a focus point on their face. If there's even one located where you'd need it.
Manual mode is not an option when stage lighting fluctuates 2-4 stops every few seconds, and spot metering is only useful when shooting manual is practical.
I'm usually shooting max iso 400 and RAW of course, a stop underexposed doesn't really bother me at all
Get out of the habit of shooting a stop under with the 1Ds; if you need more than +1 adjustment in ACR then image quality will be negatively impacted, especially when shooting above ISO 200. At ISO 800, you want to be within 1/3 stop of ideal or you'll run into noise issues and occasionally banding. Ideally, you want the ACR exposure slider between 0 and -.4 for best results.
-
I do a lot of events with the sort of lighting variety Jonathan mantioned, with the added fun that action is important (often dance events and night parades, when action moves from stage / TV lighting through to nighttime streets in HK). Weddings I do rarely, and I'm very happy I don't have to make a living from them. I've experienced the exposure problems and shot failures you both mention (Jon and Pom), and remain thankful that the client rarely has the sense of quality and criticality that the photographer does. That was with film and the 10D, before I got the 1DsMkII.
The difference in exposure accuracy between the 10D and 1-series is significant: it's also way better than my 1v. I can now shoot on evaluative +1/3ev for just about everything, and only really have to move off that (and shift ISO around - need a faster way of doing that) when stage lights dictate. Pom, I think the 1Ds will change the way you work.
Anyone remember the manual days when exposure for weddings was 'dress highlights +3' to get it in Z8 or 'face +1' to get it in Z6?! I actually often still use the latter, plus it's ethnic variants, for portraits.
-
I think you just may have put the finger on something that's been bothering me about the 20D also.
I've struggled some (but not too often) with highlights in exposures, and of course I've mainly blamed myself and not the tool.
Actually, the 1-series metering accuracy is one of the reasons I have not yet abandoned my 1D2 for the smaller and lighter 20D...
-
the overexposed images were fully recoverable and when ACR'ed (-.30 exposure) were as good as the correctly exposed images.
-.30 exposure adjustment in ACR means you nailed the exposure dead on with the 1Ds in the vast majority of cases. As long as you haven't clipped non-specular highlights, "overexposure" is a good thing.