cgardner
Offline
• • • • •
Upload & Sell: Off
|
 p.1 #2 · Mark III / 580EX bug? | |
The flash metering is different from ambient. Not sure about the Mark III but on other bodies there are two options for flash metering: evaluative and average (i.e. center weighted). Evaluative is the default for flash.
I shot with manual flash using power / distance for many years and had to re-learn flash on Canon's terms, and that required understanding how the camera metering worked to get optimal, predictable results. Whenever I get new gear I do some baseline testing to figure out how it differs from what I'd used before. One of the first things I do is put the camera into the most automatic mode like P and just go around and shoot stuff seeing how well the camera handles AF and exposure. I then repeat the process in Av mode, controlling DOF creatively and letting the camera handle exposure. Then finally I shoot manually to see if I can out-guess and out-perform the camera metering. So when I first got my 580ex flashes I put it on the camera in P mode with flash in ETTL mode at FEC=0. One of the first things I tried is what you did; point the camera out a open doorway. Much to my surprise the camera did a very good job of exposing the outdoor ambient lit doorway and the flash lit interior wall. Hmmm... I thought, this is one wickedly smart camera...
I've done quite a bit of testing with Canon flash, including shooting the same mixed ambient / fill-flash scene in all 18 possible metering combinations on my 20D with three different crops to evaluate how each worked and how they compare. You can see the results here: LINK. My conclusion? Of all the possible metering methods Evaluative ambient + Evaluative flash offers the fastest path to correct exposure IF you used this workflow:
1) Compose and shoot letting evaluative take its best guess
2) Evaluate the results with the expectation that the camera will only get it right about 50% of the time
3) Using the tools the camera provides (over-exposure warning and histogram) evaluate the test shot
4) Dial in the needed correction factors (EC and FEC)
What you are doing by spot metering, using FEC etc. is trying to out-guess the camera without really understanding how it is interpreting the scene. The above approach starts from the camera's baseline. That baseline, for better or worse, will be consistent in similar shooting situations. By starting from a baseline its easier to recognize whether or not the camera will perform well in various situations.
It is also very important to understand the difference between a shot which looks correctly exposed PERCEPTUALLY and one which is correctly exposed TECHNICALLY in the sense that detail is preserved in the textured highlights.
This shot is ambient only in flat light adjusted to retain texture in the white towel:
http://super.nova.org/TP/DR_FlatLight.jpg
This is that same target, adjusted to retain texture in the white towel in back light, ambient only:
http://super.nova.org/TP/DR_Backlight.jpg
That file is correctly exposed in the technical sense of retaining highlight detail but I'm sure 99 out of 100 people seeing to would say it is under-exposed because the middle-tones look darker than the same scene would look by eye. That's a result of using a camera which can record about a 7 stop range of detail to record a scene with about 10 stops of contrast: when the highlights are correctly exposed everything else gets recorded darker than seen by eye. That of course is why we use fill-flash, to reduce the dynamic range of the scene, at least to the extent the flash can reach, to match the range of the sensor. Remember how I said evaluative is wickedly smart? This is what happened when I just reached up and turned on the 580ex flash in the bracket above the camera, which was 10 feet away from the foreground target in high speed FP mode:
http://super.nova.org/TP/DR_FlashFill.jpg
The evaluative metering was able to match the exposure of the foreground to the background. The ambient lit portion of the white towel remained correctly exposed and the portion in the shadows was brought up to match it technically in the sense of exposing it the same. What is interesting here is that technical / perceptual "knife" cuts both ways. While the the exposure is now a good technical match of ambient and flash exposure, as evidenced by looking at the histogram of the card area...
http://super.nova.org/TP/DR_FillHisto.jpg
... perceptually it may appear the foreground is overexposed because in person looking at that scene our eyes would react to the brighter background and cause it to look darker. Not as dark as the lame sensor of the camera recorded it with ambient only, and not quite as flash at FEC=0. So to make it look more realistic perceptually I would want to dial in .3 to .6 stops of minus FEC.
The important thing to note about that exercise is that it started by getting the ambient exposure correct TECHNICALLY in the sense of not clipping the highlights of the towel. In that test shot in Av mode at f/2.8 it was necessary to dial in minus 1-2/3 stops of EC to get the highlights correctly exposed. I do that with a very simple and method. First I use the towel as a target. It has texture and is the perfect proxy for textured highlights in a portrait subject. Then I just use the over-exposure warning in the playback to tell me when the towel is clipping. Yes that based on the tiny JPG thumbnail, but by comparison with RAW file images of the towel with various levels of exposure I know that when the exposure is set so the towel is 1/3 stop below clipping in the camera OEW the RAW file exposure is optimal; not on the bleeding edge of clipping but with a margin of error and contrast which ensures the sparkle of specular highlights and the subtle details surrounding them are preserved.
Getting the ambient exposure correct in the technical sense is the key to making fill flash work in evaluative mode. If I had adjusted the ambient exposure perceptually based on how I thought the mid-tones in the foreground should look (i.e. conventional wisdom) when flash was added the resulting shot would have been grossly under-exposed. Why? Because the camera metering is designed to deal with the TECHNICAL REALITY not what looks right perceptually. But as the test also demonstrates you can't disengage the brain when putting the camera and flash in auto mode because even when the camera does get the fill flash perfect in the technical sense, it still might not look normal perceptually.
How Evaluative Works:
What ETTL-II evaluative does is compare the ambient metering at the point where AE is locked (i.e. half press of the shutter) with the pre-flash return, which occurs between the time the shutter is fully pressed and the shutter opens to fully reveal the sensor so the flash can be fired.
Reflected ambient light isn't affected by distance. If you shoot a white sofa which is 10ft away and 10ft from a white wall in flat ambient light the sofa and wall will be the same tone of white. Flash falls off about 2 stops each time the distance doubles, so about 4x more pre-flash will reflect off the couch than the wall. Evaluative compares the ambient and pre-flash reflections over the many metering zones (35 on my 20D, 63 on yours if memory serves) and can logically deduce from the fact there was no change zone-to-zone between the center and edges with the ambient reading, but the pre-flash return was 4x brighter in the center relative to the edges that there must be something closer and /or more reflective in the center.
By comparing relative ambient and pre-flash levels zone-by-zone, evaluative metering can construct a virtual radar-like 3D map of the scene making an educated guess which zone contains significant highlight detail which should be preserved. What it seems to do amazing well is discount small specular reflections off reflective objects which total confuse average metering.
With any flash, regardless of how power is regulated, exposure is only correct at one distance from the flash. So while it is possible to use something in the foreground to frame a subject further back in an ambient only shot, that will not work in a flash shot because either the foreground will be either blown out if exposure is based on the middle ground using FEL, or the middle ground subject will be under-exposed if with auto metering which will react to the near object. The solution to that dilemma is to learn how to compose shots for flash. This even affect posing in portraits. If you use a pose where a bare shoulder is closer to the flash than the face it will be physically impossible to expose both of them correctly. One of the reason the "classic" poses became classics is because they look natural and at the same time keep the front of the face as close or closer to the key light than any other body part so the face can be correctly exposed and all other body parts rendered darker and less distracting. When flash shots are composed to keep what is important closest to the source of the flash the odds of ETTL metering getting the exposure correct with FEC = 0 increase dramatically
Because evaluative flash metering is based on a zone-to-zone comparison I find it works better in conjunction with evaluative ambient metering. It is also important that the ambient exposure is locked on exactly the same scene content as the when the shutter is fully pressed. Tv mode and focus / recompose are not good choices when using flash because they will cause the ambient and flash exposures to be based on different scenes. What I suggest when shooting in Av mode is moving AF lock to the * button an using this workflow:
1) Lock AF with *
2) Recompose
3) Half-press shutter to lock evaluative ambient meter, then fully press to fire pre-flash and take the shot.
That will allow the camera to meter the ambient in flash in the same place with the highest degree of artificial intelligence. You paid $4,000 for your camera because it can do everything automatically, no? So give it a chance to do its thing BEFORE you start to second guess it. Try the baseline approach of test shot, evaluate, correct. If you work with evaluative from a consistent baseline of EC =0 and FEC =0 you will quickly come to understand when the camera metering gets exposure correct from that baseline and when it is fooled. It is a machine so it will be fooled consistently and with a bit of experience from the "zero" baseline you will learn to anticipate how it will be fooled and be able to adjust the baseline. For example when using diffusers on my flashes I start with a baseline FEC of + 1-1/3 stops because I've found on average that's what it takes to correctly expose the highlights.
Chuck
|
