Upload & Sell: Off
| p.1 #14 · Need help getting the Sekonic L358 and D800 to match... |
J.A.F. Doorhof wrote:
The "location" of the 18% gray will vary as soon as you start using different software, different curves, a different profile, a different camera etc.
The resulting rendering of 18% reflectance will vary only if you let it (or if you use software that applies processing that you can't disable.) That's not necessarily a bad thing if it's what you want. My concern is getting an accurate starting point.
I always try to explain it like this.
Forget about where 18% ends up AFTER the fact.
That works great for someone whose photography isn't dependent on accurate rendition of tones. But people who do product photography and other types of documentary style photography, that simply doesn't fly. And it's completely unacceptable for people who do art reproduction. You are thinking in terms of your own photography...but when answering someone's questions on metering I consider all types of photography.
You have a workflow that you use, in THAT workflow you want skin, materials etc. to be rendered correctly, meaning IF a material has the values 100-100-100 it should when you shoot it end up in that area (or perfect on 100-100-100 which is pretty much impossible).
Now we have an 18% reflective card and we shoot this in our COMPLETE workflow, meaning (in my case) ACR and ProPhotoRGB with a custom made profile for 5DMKIII and Leaf Credo60. Now where should the card end up. So when we forget about sRGB, aRGB and even ProPhotoRGB, just were should that card end up in the histogram when you already have everything in place ?
Over the years I've seen as many people that say 128.128.128 as people that claim differently and of course at the start I tried all methods, I've always found it weird that DURING the calibration people would have to adjust the outcome depending on the curve they were using, for me that was counteractive, in my opinion (again) when I shoot something with a fixed value it should end up where it's supposed to be.
Now when studying deeper into the material you will start to find out that there is much more to the case and in fact when you take just ONE reading of 18% gray it will never ever work, and diving even more into the matter you will very much find out (before turning mad) that it will NEVER work.
In real live let's say we have 3 anchor points metered on 50-50-50, 100-100-100 and 200-200-200
If we want to capture something 100% correct those 3 have to be spot on.
Well sorry, but that will never ever happen.
When we use for example the curve flat it will be rendered as 40-40-40, 105-105-105 and 210-210-210
However when we use the curve natural it will be rendered as 20-20-20, 120-120-120 and 240-240-240
But with the curve, silky contrast it's 34-34-34, 90-90-90 and 245-245-245
Meaning whatever profile you use the curve will be different, which is logical because a profile consists of curves.
I don't know what you're doing wrong that you can't get three relative luminances to maintain their relationship...but it works just fine for me.
The reason CIPA changed (and ISO adopted) the process of deterimining sensitivity of digital cameras, from a saturation based process to a process based on 18% reflectance, is because such a reflectance can be modeled and rendered with the same apparent brightness by any color model. In L*a*b* the card will have an L* of 50. That translates to 119 in sRGB and Adobe RGB, and around 101 in ProPhoto. As for the histogram, the RGB plots will change based on the color space because the color values are different...but the luminance plots that are based on those values, will all be in the same place.
But as I said previously, if your software is applying processing that you can't undo, then your ability to render accurately is hindered. If you use software that doesn't allow you complete control (including "no influence") over your rendering then you're at the mercy of the software maker. This is the problem people are having with Adobe Lightroom 4 and PV2012...its processing produces a different result when compared to PV2010.
To say, "...points metered 50-50-50, 100-100-100 and 200-200-200", is not really valid because RGB color space values are dependent on the color space used. So was that sRGB or Adobe RGB or ProPhoto RGB? Metering is about luminance. The luminance range between RGB values depends on the color space. Now...if you were to say that you metered "anchor points" of 18%, 36%, and 72% reflectance, and are unable to get renderings that are representative of those luminances, then I'd say you're probably using software that's forcing some processing on you...like ACR. If you were to examine the RAW file using Rawnalyze, Raw Therapee with the "Neutral" profile, or even Lightroom with PV1210 zeroed, the representative RGB values (based on your preferred color space) will be exactly where they should be. That's because such a rendering would be based on a strict transformation from the camera's color space into the color space of choice.
Such images look true to the live scene. I can take a shot of my desk, put it up on my monitor, and the rendering will appear to be a perfect match to the live view beside it. However, such images tend to not be very enticing to the eye. There's also the fact that it only really works when you have a single light source. illuminating the scene. As soon as you bring other light sources into the mix, the eye starts to pull tricks on you due to Color Constancy. In that case, there's no metering that will ever render the scene as the eye sees it.
However, you can always pick a tone, and place that one tone exactly where you want it. Then you fix the rest of the image in post...which is exactly what I imagine you do. You hold your meter by your model's face, and your calibration process results in an exposure that puts the tone of that light skin exactly where you want it. Then you fix everything else in the image to your liking. You seem to like very bright faces, which is what I'd expect from a calibration that is rendering metered tones slightly brighter. And if that's what you like...fine. Nothing wrong with that, and your images do look wonderful. But your calibration process isn't going to help the landscape photographer, or the photographer reproducing art.
If you want to apply curves and profiles to change the rendering of the image, then that's fine. But that should not be a consideration for setting exposure.
So how do we make a meter 100% accurate, well sorry that's not possible, that's also why Sekonic uses several meter readings taking into account the dynamic range of the sensor, but even than it's only a approximate setting.
So realizing this and getting a meter to do an OVERAL very good job I THINK PERSONALLY you will have to make a decision, you can meter for a 100% white patch (not paper because it contains too much in the blue channel due to whiteners) and make sure that the meter does not read more than 254.254.254, or you can meter 18% gray and place it in the center of the histogram KNOWING it's actually not correct but at least you have an anchor point.
But why must the 18% gray be in the center of the histogram? In fact, it works whereever you put it, as long as you get the exposure you're looking for. In your case, placing the metered point in the center of the histograms gives you a little bit of overexposure, which, as I said, brightens your models' faces. And this goes back to what I said in my previous post...we calibrate to a result. You calibrate for bright faces, I calibrate for accurate luminance.
In the end it all boils down to one thing...
You have accurate exposures, and I could care less if according to a formula I'm 4-5 points off, this is not visible, it however also boils down that when EVER you change workflow, curve or software you will have to do the calibration again.
On the net there are many many different opinions and all have a certain amount of truth, in the end with digital however there is no 100% accurate way of metering,
And I think that that's entirely wrong. You calibrate for accurate luminance levels. Then you can apply curves, profiles, and anything else that you want to do to your heart's desire...but your starting point should always be the same. My personal "calibration" is nothing more than an EC value applied to my gray card. I spot meter my card, add +1.3 EC, and with that I know that bright reflective whites will be near the top of saturation, but not clip. Now...why in the world would I ever what to change that calibration?? Changing it will get me either less signal (and more shadow noise,) or clipped highlights...that makes no sense!