cgardner
Offline
• • • • •
Upload & Sell: Off
|
 p.3 #12 · Technique for shooting brides dresses? | |
With respect to Canon ETTL evaluative metering there's a big difference between what is technically correct exposure (i.e. blow no highlights) and what looks correctly exposed perceptually.
Because of the short range of the sensor, a nominally exposed (i.e. no blown highlights anywhere) file will look grossly underexposed perceputualy in contrasty lighting conditions such as backlight from the sun or when flash is bounced and there are unfilled shadows.
Regardless of how flash is metered it will obey the laws of physics no? So what's actually happening with average metering is that average, weighted towards firing more flash by the predominantly dark background, will actually blow some of the highlight detail when adjusted to make the overall shot look well exposed perceptionally. Perceptually we will accept blown highlights if we don't expect to see detail there, such as the flat part of a dress. Its only when stuff like the lace and beadwork disappears that overexposure will be noticed perceptually. So in technical terms many average metered or manual flash shots, adjusted perceptually are actually over-exposed in some highlights.
Once you grasp the fact that ETTL evaluative will not blow ANY highlights it changes the way you evaluate the scene and apply FEC. When I first started using ETTL, after 30+ years with manual and auto-thyristor Vivitars I made FEC decisions the same way I had based on what I thought the perceptual average was.
Auto thyristor averaging flashes aren't foolproof or automatic either, it just takes a much greater change in the scene to shift the average. For example, if using a 285HV in auto mode at the f/8 setting the exposure might be perfect for a H&S shot of the groom, but when the bride was added the additional light reflecting from the dress would be enough to affect the auto sensor on the flash. Seeing more light it would output less flash and instead of shooting at f/8 you'd open the lens a bit more, otherwise the shot with the bride would be underexposed.
Because weddings create constantly changing scene contrast between a viewfinder filled with mostly white in one shot and a viewfinder filled with mostly black void (open space) the the next no automatic exposure system is 100% reliable and foolproof. That's why most wedding shooters use M mode. Not because its simpler, but because it is more predictable in situations where there's only one opportunity to get the shot.
I'd been using one and two flashes for so long in manual mode that when I looked at a scene I didn't measure distance in feet mentally, I just looked and thought "Hmmm... she's f/7 away" and adjusted the aperture without much thought. With two flashes you simply move the off camera light in and out as the shooting distance changes to retain the same ratio - quite easy if your off camera light is on a wheeled IV stand like mine is. Its a technique I learned working for Zucker in the 1970s. He introduced and popularized the idea of using dual flash for "candid" wedding reception coverage.
If you manual flash consistently for about a week and pay attention to the f/stops and distances you'll find the process will also become instinctive: you'll measure distance in f/stops too.
Like everyone else I was really pissed-off at spending $800 on a pair of 580ex flashes to replace the trusty Vivitars only to get worse results with ETTL-II metering. And like any good gear-head I was determined to find a way to make it work.
Once I figured out how ETTL evaluative keyed off the highlights and erred on the side of not blowing any it shifted my mental baseline. I realized that as I say above the same laws of physics applied to power and distance. I also noticed that ETTL-II shots while looking underexposed overall, rarely had any significant image areas with any blow highlights where as my perceputally based manual exposures did because I was basing the exposure decision on the mid-tone detail. Then it finally dawned on me that ETTL-II was design to make FEC=0 a starting baseline of "OK I didn't blow any highlights, but now you figure out how to handle the scene perceptually".
That was a eye opener in two ways: I understood how evaluative metering worked and realized that in many cases in the past I did blow highlights to make the exposure look normal (as by eye) on the playback in the midtones and shadows perceptually. I was still "exposing for the shadows" perceptually, but just not noticing or discounting the blown highlights.
Consider that if you were to shoot the same scene first in M, then in ETTL evaluative, then ETTL average you could make them look exactly the same perceptually and the actual flash power used would be identical. But if adjusted perceptually for the midtone detail they would all probably have identically blown highlights. Hmmm... You begin to realize the problem isn't the flash, but rather the fact the sensor can't record the scene correctly due to the fact it has a longer range.
Fact is if you take any shot and expose it nominally for no clipping anywhere it will start to look darker than normal in the midtones and shadows whenever the range of the scene in f/stops exceeeds the DR of the sensor. There's a simple way to compensate for that perceptually - move the middle slider in levels to the right or adjust the middle in curves. It doesn't increase the range of the capture but it will lighten the middle and 3/4 tones fooling the eye into thinking more detail has been captured.
With a bit more experimentation I found that the over-exposure warning showed me exactly when optimal exposure of the highlights was occuring. Optimally exposing the highlights also means that the full range of the sensor is being used. That is when I began to realize ETTL-II actuall understood digital exposure better than I did.
Its also important to realize that I've used dual flash since 1972 and rarely take a single flash shot indoors if I can avoid it, even if only to park the off camera light behind for back/rim light. As a result I never worried about shadow noise with digital because I could simply add fill as needed to pull up the shadow detail above the noise threshold of the sensor. Conversely when I did use single flash or ambient without fill it was easier to see, by comparison, how the limits of the sensor affected perception of "correct" exposure.
Using two flashes, which I did from the start with the Canon system, was quite simple in ETTL mode. Set the desired A:B ratio then dial in FEC until the highlights were correctly exposed by monitoring the OEW. Sure the amount if FEC varied, but as noted above the same variation had occured when using auto thyristor averaging too. I accept it as a fact of life with auto metering.
With a single flash I found that if I let ETTL guide me into exposing the highlights correctly the camera would sometimes capture images which looked overexposed perceptually but which contained the fullest possible range of actual detail from the scene from highlight down to the noise threshold of the sensor. It requires a bit more work in Photoshop to pull that detail in the midtones up to the point it looks normal perceptionally and similar to a perceptually adjusted "average" metered shot, but from a techical point of view the net result is better.
An interesting side effect of using E-TTL and letting it "police" my exposure based on the brightest highlights in the frame was realizing that in many cases I had not been composing my flash shots very well. Accepting that some blown highlights are OK to get exposure of the midtones right leads to not worrying too much if something in the foreground got blown out. But knowing that ETTL-II evaluative would key the whole exposure off that foreground distraction caused me to compose more carefully so what I wanted correctly exposed was closest to the flash, and everything I wanted correctly exposed was the same distance -- the advice suggested earlier as the path to more predictable ETTL exposures.
So rather than getting frustrated with ETTL and dismissing its a worthless I discovered how to make it work: 1) Always use two flashes when possible, and 2) expect to make some adjustments of the mid-tones in Photoshop to compensate for the lame range of the sensor.
In the days before digital when shooting transparencies the only option available was #1, which is why in the good old days the barrier between being a hobbiest and a pro wasn't the camera gear but rather the fact the pro usually had $20,000 invested in lighting equipment.
The best investments digital shooters can make today is buying a second flash and taking the time to learn how to how to use it with ETTL ratios and M power settings. Once the second flash is in play ETTL becomes more practical and reliable because its no longer necessary to base exposure on perception of the mid-tones. In all cases the highlights can be exposed nominally and the full range of the sensor exploited. Like many things the reasons why it works do not become apparent until its used.
|
actually I laugh when I think back to how bad it was, luckily RAW photos are pretty good when it comes to WB correction