Great article.  The graphic examples were very helpful. 

Can you compare this to push and pull processing?  (maybe in another article)

Posted by .(JavaScript must be enabled to view this email address)  on  10/08  at  07:27 AM

---

“18% gray is lined up at four different ISOs… I added a dotted line that runs vertically through the chart to indicate the 18% gray (45%) chip at each ISO.”

I was going to ask about this when I saw the first chart and clearly you were thinking the same way. I am still trying to understand what the charts are showing me.

Did you line up the 18% gray as determined by the chart (before camera) or when assessing the footage (after camera)? Because if I just look at the vertical strip 18% chips you lined up vertically, they seem to get progressively lighter from top to bottom. If you lined up the chips “after camera,” should all the grays vertically be the same? Because if you’re looking at them in Photoshop, should all the 18% chips vertically stacked have the same brightness?

In the previous article, and here, you note that the strip with the tape through it is 18% gray on the chart where Alexa sees it at ISO800. So as you change the ISO, the 18% gray (before camera) moves left and right. If you lined up the chips “before camera,” should all the taped chips vertically be lined up? Because the brightness of that chip (before camera) isn’t changing?

Maybe it’s my laptop screen, or the angle of view, but the chips don’t seem to be the same on the 18% line, so I don’t understand what’s lined up and am having trouble understanding the chart.

—-
Or put another way..
You say it’s 45% grey. I put my light meter out there, and based on its internal measurements (45% grey), it says an ISO 800 camera needs a t5.6 (whatever). It is a dark scene and I want the camera to make sure I have lots of dark detail. So I ISO down to 200. I meter again and now the meter says I need to open up the lens 2 stops to t2.8.

Are the charts indicating that the change in Alexa ISO would also cause the camera interpret the same 45% grey now 2 stops darker? Or would the lens adjustments push the scale back to the left?

Posted by IEBA  on  10/11  at  11:33 AM

---

To address your first question: the 18% gray patches don’t match precisely because of how I manipulated them in Photoshop. Since LogC white clip values vary, I white-picked the brightest chip so that the brightest whites were all the same value. I didn’t want to manipulate the charts any more than that so the 18% grays don’t line up absolutely perfectly.

If I’d left the white values alone then the grays would all line up precisely but the whites would be very different values, which would also be confusing. (LogC white clip values vary depending on the ISO, because of changes in how the linear gamma is mapped.)

Still, I think the grays line up well enough to illustrate the point.

To address your second question:  If your meter said to shoot at T5.6 at ISO 800, and you changed your ISO to 200, you would have to open up the stop to T2.8. The gray chip that WAS 18% gray at ISO 800 is now two shades darker, and the gray chip that was two shades lighter than 18% is now exactly 18%.

The biggest difference is that you had seven stops of overexposure latitude at ISO 800, but at ISO 200 you now only have five because two of those stops are now distributed below 18% gray. So while the shadows will contain more detail and less noise your highlights will clip sooner.

Posted by Art Adams  on  10/11  at  03:53 PM

---

Wait, wait, I think I get what my mind has been rummaging around for a while. You keep saying dropping the ISO gives you less highlight detail (stops).

The chart had a chip with 18% gray (at ISO 800) that you marked with tape. That gray never changed “grayness” in real life. It is what it is. 18% or whatever, it is a certain amount of gray, and the luminance of that chip on the backlit chip chart never changed. It is fixed. X. And you put tape across it to help us keep track of it. I think that much I am good on.

So let’s say that X is a black man’s face. Now, whether it be highlight detail on a man’s face, or white chips a chip chart, I seem to read the chart to say there are 7 stops above X, no matter what the ISO.

So the black man’s face, at 18% gray, would always have 7 stops highlight detail above it, at every ISO, as I understand it. But you said available highlight detail goes down as ISO goes down.

If X (and the luminance of all the chips above it) doesn’t change, and we’re supposed to have less highlight detail as ISO decreases, I’d expect to count fewer chips above the chip you marked, not the same.

Posted by IEBA  on  10/11  at  10:37 PM

---

Well… I tried to lay it out in the graphics, but when you change the ISO you’re simply changing the gain on the camera. Going from 800 to 200 reduces gain (“negative gain,” although not really), so the camera becomes less noisy but also sees less highlight detail.

“If X (and the luminance of all the chips above it) doesn’t change, and we’re supposed to have less highlight detail as ISO decreases, I’d expect to count fewer chips above the chip you marked, not the same.”

That’s exactly what I’ve shown. The trick is that the chip with the tape across it is only 18% gray at ISO 800. At ISO 400 it moves one shade down and the next brightest chip becomes the new 18% gray. At 200 the chip with the tape becomes two stops darker instead of just one, and the old 18% gray is now pushed aside by the next brighter chip, which is now 18% gray.

18% gray moves around depending on the ISO. That’s what the ISO is really doing: arbitrarily placing 18% gray on the dynamic range scale, based on where ISO is set.

And when you move 18% gray in either direction on the scale, you lose one stop from one side and allocate it to the other. That’s why ISO 400 loses one stop from highlights and adds one to the shadows: you’ve just moved where the camera places 18% gray. The brightness seen by the sensor is exactly the same, but you’ve changed how the camera responds to it. You’d open up one stop and shoot.

The graphics above really do explain it fairly well. I’m not sure what else I can tell you.

Posted by Art Adams  on  10/11  at  11:32 PM

---

The dotted line shows where 18% gray is. The chip with the tape on it is only there to give you a reference and see how 18% gray at ISO 800 moves around. That chip is only middle gray at ISO 800.

The dotted line shows where 18% falls at different ISO’s, and that’s what you really want to watch.

Posted by Art Adams  on  10/11  at  11:34 PM

---

me:
>>If X (and the luminance of all the chips above it)
>>doesn’t change, and we’re supposed to have less
>>highlight detail as ISO decreases, I’d expect to
>>count fewer chips above the chip you marked,
>>not the same.

Art:
> That’s exactly what I’ve shown. The trick is that
> the chip with the tape across it is only 18% gray
> at ISO 800. At ISO 400 it moves one shade down and
> the next brightest chip becomes the new 18% gray.

Here’s the crux. The chart chip with the tape never changed luminance. When you say it “moved down” you’re measuring after capture, by assessing the LogC files, which you yourself said “[LogC] it’s not meant to be viewed on a Rec 709 display. LogC is for storage only, not for viewing.”

If, as you said, “I set that first chip just to the point of clipping at ISO 800… well, no matter what ISO I set the camera to, that chip was always just clipping—and without changing the stop!”

Then line up all the chip charts so that the clipping white is in the same spot (since it is the same exact luminance value going into the camera where the camera clipped) then line up all those white chips, but the dotted line through that, and let’s see how the grays differ as they go down the chart.

I did this here:
http://ieba.wordpress.com/2010/10/13/alexa
And there, I agreed with you. You are right, the Alexa does have fewer stops at ISO 200 than 800, but it’s fewer TOTAL stops. Middle gray moves down because the difference between each step is larger. It moves down because, as you decrease ISO and have fewer total steps, you get from black to your middle gray quicker. This says to me that, at lower ISOs, there’s LESS shadow resolution, not more.

At least that’s my interpretation.

Posted by IEBA  on  10/13  at  11:36 AM

---

Hi IEBA-

You are misinterpreting my data, and you seem to be hung up on where the tape is placed on the chart. That was only meant to show where 18% gray started off, at ISO 800; it is not 18% gray at every ISO. That’s the point of the article.

As I mentioned in the article, the chart that you’ve manipulated was tweaked roughly in Photoshop and was never intended to be used as any sort of exact latitude reference, but only as a rough conceptual guide. All the values that you’ve noted on the chips are values created in Photoshop as a result of white balancing the white chips and making them appear white—something that doesn’t happen in LogC as white clips at different values depending on the ISO—and don’t reflect any real data or values.

I wish I could spend more time explaining this to you… but I can’t. It’s probably best if you get your hands on an Alexa and perform your own tests. You can do this with a RED as well.

Posted by Art Adams  on  10/13  at  12:14 PM

---

Hi Art, your Alexa’s articles are great and really helpful.

I have this queston regardng metering:

If I wanna use my spot lightmeter with LogC… should I first stop a 18% gray at 40% -as Arri states- on the waveform (or until the Alexa’s false color turns “Green”)... then meter it with my spotmeter and compare both readings? If I do that I will get 1 stop of difference between both… does it mean that I should use my lightmeter with a “one stop down” calibration?

Bottomline: Should I calibrate my lightmeter to read the 18% gray as the LogC curve sees it (one stop down)? or should I use it normally (spotmeter reading it at 50%)? (entering same ISO/fps values on both as usual).

Thanks!
Luis

Posted by .(JavaScript must be enabled to view this email address)  on  03/11  at  05:53 PM

---

Dear Art ,

would you say that the camera’s negative or positive gains controller at the sensor output act accordingly to the relative iso’s setting?

Posted by .(JavaScript must be enabled to view this email address)  on  06/01  at  05:45 AM

---

Name:

Email:

Location:

URL:

Smileys

Remember my personal information

Notify me of follow-up comments?

Submit the word you see below: