Back To Listings RSS Print

Arri Alexa’s Dynamic Range: It’s All in How You Use It

Always push exposure to the limits. But learn where the limits are first!

By Art Adams | September 28, 2010

Knowing that the Arri Alexa has 15 or so stops of dynamic range is nice, but it's hardly useful information unless you know how those stops are arranged at different ISO settings. Fortunately I've done the hard work for you.

Or rather I should say that Adam Wilt and I did the hard work, as we both traveled to Chater Camera in Berkeley recently with a DSC Labs DX-1 17-stop dynamic range chart intent on discovering Alexa's hidden secrets. (The biggest secret was "Why is this Alexa just sitting here?" as we'd managed to luck upon the only day during a multiweek period when it was actually available for testing.)

Dynamic range charts are a lot of fun if you want to play games like "My camera's dynamic range is bigger than yours!" but they don't tell us how to exploit dynamic range in our work. My goal in these tests is to find how that dynamic range is distributed, which is much more useful to me as a cinematographer.

The only solid exposure reference that an HD camera offers is the point where the sensor clips, and the DSC 17-stop chart is designed with that in mind: by exposing the chart such that the first chip just hits white clip we can count how many chips are visible until they blend into the black background. As each chip is one stop, counting the visible chips to the right of white (don't count the first one for this test) gives us a number for the camera's full dynamic range in lens stops.

When using a camera "film style" we don't set the exposure based on clipping, but with an eye to exposing the image properly as a whole. My preference is to use the Zone System and expose certain objects in the scene so that they reproduce with a specific tonality. That can only be done by knowing how a camera or film stock responds to different reflective values of light and using that knowledge in concert with a spot meter. The Zone System hinges around 18% gray, and as there isn't a high dynamic range chart in existence that is set up to measure exposure in relationship to that middle tone, I set out to make one.

I set the Alexa to record ProRes 4444 to SxS cards in "legal" mode. I set the Alexa's ISO at Arri's recommended "sweet spot" of 800 and, while watching a waveform monitor, I set the exposure on the DSC Chart so that the first chip just flattened out into clipping. Then I looked at the waveform monitor and determined which step fell closest to 18% gray (generally 42% to 45% on a waveform), and marked that chip by putting a thin strip of black tape vertically through it. The tape creates a "notch" in the waveform trace and allows us to watch what happens to that middle gray reference as we change the camera's ISO settings.

Adam and I only tested four ISO settings-1600, 800, 400 and 200-because the chart is calibrated in one stop increments. (We tried a Stouffer wedge chart with 1/3 stop increments on a prior occasion and beyond a certain point the steps became too fine to count in the shadows.) We shot the chart at each ISO setting in both LogC and Rec 709 mode, and then did an exposure test simply to measure overall dynamic range. This resulted in two sets of results: one that showed how many total stops were visible below white clip on the chart, and another that showed how those stops were distributed above and below 18% gray on the waveform monitor. The first tells us how much range the camera has overall at a specific ISO, and the other tells us how much exposure latitude we have before we hit white clip or the noise floor (black).

Turn the page to get started...

Page 1 of 4 pages 1 2 3 4 Next »

Editor's Choice
PVC Exclusive
From our Sponsors

Share This

Back To Listings RSS Print

Get articles like this in your inbox: Sign Up

Comments

Adrian Jebef: | September, 30, 2010

I have been using the ALEXA on a number of commercial spots recently and have been undeniably impressed with the camera’s dynamic range and smoothness of image.  I performed by own ASA / Gain test just a few weeks back in an attempt to find out how much DR separation may be lost when dialing the camera down from 800 to 400 or even 200.  I’m still trying my best to make sense of my results as the differences between these ASA settings were hardly noticeable.  You and Adam did a great job with this DR chart as I can pick out some subtleties in the -3 to -7 stop range between the 800/400/200 settings. I’ve noticed that the -3 stop begins to be pulled slightly lower on the waveform (by minus a couple of IREs) and -4, -5 etc. follow suit.  In contrast, 18% Grey seems to hold consistently around 45 IRE no matter what ASA you set the camera to.

This is little to no news, which in turn is good news.  Setting the ALEXA to ASA 200 to shoot a mid-day exterior will allow you to toss the ND’s and not worry about losing any DR in the process.  This camera is more than impressive; it’s quite ridiculous really.

Oh by the way, I’m a firm believer that a rec709 gamma has no use in today’s digital cinema cameras.  To that end I am only referring and commenting on the ALEXA’s Log C capture setting.

And as a last minute comment to your test, perhaps you would have seen the ALEXA map highlights into 100/105/110 on the scale at ASA 800/400/200 if you had another +chip or two on the chart…

Keep up the good work.

Rami Katzav: | September, 30, 2010

Thanks so much for this helpful test, and especially for the summery table.

Nick Shaw: | September, 30, 2010

Very interesting article.  Thanks.

One comment though. You say you recorded ProRes in “legal” mode. My understanding from the ALEXA I tested, and also the manual on ARRI’s site, is that ProRes is always “legal”. The “extended range” option affects only the REC OUT SDI.

Not a big deal at all but worth noting, as it means there is no “bigger bucket” available in ProRes. If people are switching to “extended range” and expecting to see the same in ProRes in post as they see on a waveform monitor on set, that is not correct.

In my opinion you would not want to put extended range into a ProRes file anyway. Many apps (After Effects, Nuke - although there is a work around, Smoke on Mac, to name a few I have tested) clip anything outside “legal” when they read ProRes files.

Or am I wrong?

IEBA: | September, 30, 2010

I find technical geekery at this level fascinating, but I may be perhaps missing a fundamental point… if the “base” ISO of the Alexa is 800, and you notch the reference for middle grey based on that setting, and you adjust ISO up and down (and I would assume, some other setting so they whole chart doesn’t similarly go up and down the scope) then why does the Notched reference reliably go up or down with the ISO setting? Shouldn’t the middle grey stay in the middle?

Also, how is it that the Alexa resolves more white steps at ISO 1600 than at a lower ISO? I look even at the base ISO 800 and see several unused black steps (crushed) and the last white step (in LogC) leaves ample room for more bright steps above it.

Art Adams: | October, 01, 2010

Ieba- that notched grey doesn’t stay in the middle was the point of this test. Remember that 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!

What that tells me is that the chip sees what the chips sees, period. You can manipulate how that is interpreted by changing the ISO setting, which simply moves middle gray around to different values. You still have the same amount of latitude available, it’s just distributed differently.

So you can have eight stops of overexposure latitude at ISO 1600
if you can tolerate the noise. You can have five stops of overexposure latitude at ISO 200 if you don’t have a lot of really bright highlights.

This is important to know, because everyone knows that boosting gain adds noise and lowering gain decreases noise, but do they know that lowering gain reduces overexposure latitude, or that raising gain increases it?

When it comes to sensors, you don’t get anything for free. smile

Art Adams: | October, 01, 2010

Nick- that’s something I don’t know for sure either. I do know that when I look back at all the ProRes that I’ve shot on Alexa so far, even in extended mode, I’ve never had a value above 100%.

It is important to note, though, that Alexa’s monitor out never goes beyond 100% while the record out does.

Art Adams: | October, 01, 2010

Hi Adrian- I’m very glad you enjoyed the article, but I do have to take issue with what you said about shooting at ISO 200 and throwing away your ND filters. You can do that… but there’s a price.

The price is that highlights will hard clip at five stops above 18% gray. If you’re shooting under low contrast conditions then that maybe acceptable. Otherwise you’ll give yourself an extra stop of overexposure latitude by shooting at ISO 400, and another stop by shooting at 800.

There’s an interesting possibility that I believe I’ve seen when shooting RED footage, but haven’t confirmed through testing: when the ISO is reduced, and there are fewer steps between grey and white clip, in theory each step should be more contrasty.

At ISO 800 there are seven one-stop steps between middle gray and white clip, but at ISO 200 there are only five. Theoretically those five steps are each covering more ground between gray and white than if there are seven steps, so each step should cover more tonalities. I’m not sure what the practical limitations are, but highlights could be a bit “crunchier” at lower ISO’s.

Here’s another example: if you expose caucasian flesh tone at, say, one stop brighter than 18% gray on your spot meter, that value should be brighter at ISO 200 than at ISO 800 because middle gray is 1/5 of the way to white at ISO 200, while it’s 1/7 of the way to white at ISO 800.

I think I’ve seen this in RED at lower ISO’s. I just shot a green screen piece yesterday on a RED MX and I rated it at 400 just to reduce noise and give the blue channel a healthy exposure. At that ISO I only got 3.5 stops of overexposure before white clip (instead of 4.5 at 800) and I think some of the highlights got a little compressed. (We were on a stage so I fixed that with the lighting.)

Something to think about…

IEBA: | October, 01, 2010

Color me more confused.

Set the camera to an ISO of 800.
I can set my manual lens to X for good exposure.
Adjust the camera ISO to 200.
And I _don’t_ have to adjust the lens?

This breaks every basic photography rule I know.

Then, if I read this right…

Setting a lower ISO for cleaner images means I actually have less latitude. (fewer stops till clipping) but gaining up (where I would presume that I would be more likely to clip whites) actually means the opposite: that I have more stops before white clips.

Do I have this right?

Art Adams: | October, 01, 2010

I probably didn’t explain this well.

When I did the Alexa test, I set the white chip to just clip at ISO 800.

At ISO 400, the chip still “just clipped” but 18% gray became one stop darker.

At ISO 200, the chip still “just clipped” but 18% gray became two stops darker.

In each case the exposure didn’t change, only how the camera interpreted it. And middle gray moved around a bit, which in the real world of non-test-chart-applications has profound effects on the image. smile

In your case, when you changed from ISO 800 to ISO 200, you would have to open up two stops in order to keep things looking somewhat “normal” between the exposures.

In my case, when watching a test chart, I just watched my middle gray “notched” reference move up and down on the waveform. If I’d wanted to keep that “notched” reference at middle gray I would have had to change exposure, but that wasn’t really the point of the test.

This is where you ask… THEN WHAT THE HELL WAS THE POINT OF THE TEST??? smile

To me, knowing a camera has X stops of latitude tells me nothing unless I know how they are laid out. I’m an old film guy, and I want to know how many stops above 18% grey before I lose detail in highlights and how many stops below 18% grey before I lose detail in the shadows. By setting up an 18% grey reference at ISO 800 I was able to see that.

And then, by changing the ISO, I was able to see the penalties imposed when changing the gain around for higher or lower ISO’s. The distribution of stops above and below 18% gray changed, which goes to show that changing ISO’s is not a free lunch.

By watching the “notched’ reference I could then see what was going on. At ISO 400 I was pushing all the values one stop down toward black; at ISO 1600 I was pushing all the values one stop up.

Art Adams: | October, 01, 2010

And, above, where I say that 18% grey became one stop darker at ISO 400, and then two stops darker at ISO 200, I meant to say that the “notched” reference changed in that way. It was no longer 18% grey-it had drifted, and a new chip became 18% grey.

If I can be more confusing please let me know. smile Maybe I should take another pass at this article.

IEBA: | October, 01, 2010

I actually GET the point of the test, and knowing how a camera interprets light at different ISO settings is critical. I mean, your work here is kin to what we find in the ASC manual.

What I DON’T get is the changing ISO doesn’t change the _exposure_ just how different greys are interpreted (which seems absolutely wrong for a “digital” cinema camera, but I won’t get into THAT)

If this were FILM, and you filmed a chip chart, developed it, one lighted it, and scoped it to find your 18% grey… and then changed ISO setting, you’d either crush your whites, crush your blacks, or you’d HAVE to adjust the lens to compensate.

I don’t understand how swapping out the camera for a digital one means that changing the ISO now DOESN’T affect the amount of light the camera uses to get the exposure.

i.e. if you Set your basis at ISO 800, then stop down two stops to 200, I would fully expect:
A - the chip chart to “peak” at 60-70 ire on the waveform.
B - the “curve” of the steps to be identical because it’s DIGITAL and if digital can’t maintain a response curve while changing ISO, then I consider that a BAD thing.

And because the Alexa obviously can’t maintain the same response curve as you change ISO (something I expect every digital still and cinema camera to do), your testing and results are CRITICAL to actually using the Alexa for real moviemaking.

For instance, I’d LOVE for you to combine the premise of this article and your 5D latitude test (http://provideocoalition.com/index.php/aadams/story/canon_5d_how_much_dynamic_range_does_it_have_really/) and see how changing the ISO in the 5D affects the latitude, response curve, exposure, and gain structure.

I’d expect that the latitude and response curve would remain the SAME at different ISO settings, the exposure of the lens WOULD have to be adjusted for different ISO settings and the gain structure to be similar to what you found here, fuzzier scope traces with higher ISOs.

But then I would have thought the same for Alexa (being a replacement for a “film"making camera) and you clearly busted that myth.

Art Adams: | October, 01, 2010

I’ll try to address your points one by one:

>>>If this were FILM, and you filmed a chip chart, developed it, one lighted it, and scoped it to find your 18% grey… and then changed ISO setting, you’d either crush your whites, crush your blacks, or you’d HAVE to adjust the lens to compensate.<<<

That’s correct, and that would be the case here as well if I were basing all this on 18% grey. But I’m not “printing” to compensate for where the new grey would fall, but instead showing you where the new middle grey would fall on the exposure scale, and how that relates to the value that was grey at the original ISO.

The only difference between what I’m doing, and what you’re describing with film, is that I’m not “printing” my original 18% grey reference so that it’s always grey as I move the ISO around. Instead I’m “printing” white to always be white. If you moved the ISO of a film stock around and printed it so that D-max is always the same you’d get exactly the same results.

And if you look at the waveforms you can clearly see that whites and blacks ARE getting crushed. For example, there are seven stops between middle grey and white at ISO 800, but only five at ISO 200. That results in highlights that have less contrast and separation and are “crushed.”

>>>I don’t understand how swapping out the camera for a digital one means that changing the ISO now DOESN’T affect the amount of light the camera uses to get the exposure.<<<

It does. It works exactly the same way. Although, and this is an important point, it’s NOT like changing film’s processing. Pushing and pulling film are completely different processes. This is what would happen if you shot film of an 18% grey card and then changed the ISO settings without changing development, and then printed so that white was always consistent.

>>>i.e. if you Set your basis at ISO 800, then stop down two stops to 200, I would fully expect:
A - the chip chart to “peak” at 60-70 ire on the waveform. <<<

Actually you’ve got that reversed: going from ISO 800 to 200 would require opening up two stops if you were trying to keep the middle grey value on the chart consistent. 200 is less sensitive than 800, so it requires more light.

That’s the thing, though. I didn’t stop down. I just changed the gains in the camera. The values were remapped so that a step on the chart that’s two stops brighter than the original grey is now the new middle grey. If I were printing film based on D-max I’d get exactly the same response.

>>>B - the “curve” of the steps to be identical because it’s DIGITAL and if digital can’t maintain a response curve while changing ISO, then I consider that a BAD thing.<<<

“Digital” is not magic, and a silicon sensor is an analog device. The digital part comes into play when the analog voltages from the photosites on the sensor are turned into numbers in the DSP.

The curve of the steps are basically identical in LogC. They can’t be identical in Rec 709 because Rec 709 is a very confined space in which all that data has to be crammed. If you have five stops of overexposure latitude versus seven then the shoulder (knee) of the curve is going to have to change to hold all of that. RED is the same way, as is any camera if you could change the gain settings enough.

>>>And because the Alexa obviously can’t maintain the same response curve as you change ISO (something I expect every digital still and cinema camera to do)<<<

That’s a misconception on your part. Every camera made functions this way. You don’t normally see it because you don’t have so many ISO options. I could get the same results on any camera that has -3, 0, +3, +6, +9, etc. gain settings.

And this works exactly the same way on the RED. The MX sensor gives about 4.5 stops of overexposure latitude at ISO 800, but I tend to rate it at 400 as I like smoother tones with less noise. At 400 I get one less stop of overexposure latitude for a grand total of 3.5 stops, and the highlights are a little crushed because I now only have 3.5 stops between middle gray and white instead of 4.5 stops.

This isn’t really a demo of how the Alexa is different from other cameras, other than its incredible ability to handle overexposed highlights. Instead it’s a demo that shows how EVERY camera handles changes in gain.

>>>For instance, I’d LOVE for you to combine the premise of this article and your 5D latitude test (http://provideocoalition.com/index.php/aadams/story/canon_5d_how_much_dynamic_range_does_it_have_really/) and see how changing the ISO in the 5D affects the latitude, response curve, exposure, and gain structure.<<<

The number of stops of latitude over/under 18% grey will change, and the noise will change, in exactly the same way you see in this test.

>>>I’d expect that the latitude and response curve would remain the SAME at different ISO settings, the exposure of the lens WOULD have to be adjusted for different ISO settings and the gain structure to be similar to what you found here, fuzzier scope traces with higher ISOs.<<<

Why would you expect that? You wouldn’t expect that in film. If you overexpose film by one stop and print it down you’re pushing the grain down at the expense of losing highlight latitude, because your new middle grey is one stop closer to D-max. If you underexpose film you’re pulling the shadows, and the grain, up, and you’ll have better highlight handling because your new middle grey will be one stop closer to D-min.

>>>But then I would have thought the same for Alexa (being a replacement for a “film"making camera) and you clearly busted that myth.<<<

Given that, based on your own examples, it responds exactly the same way film would under the same test, I don’t understand why you’d assume that.

Digital cinema’s greatest weakness has been the inability of electronic cameras to adequately handle highlight detail. Now that there’s a camera on the market that can handle seven or eight stops of overexposure latitude, it seems that last barrier is ready to fall.

The only thing special about Alexa is its enhanced dynamic range. In every other respect it basically responds as any other camera would. And if I did this same tests with a film stock I’d get the same results as far as where middle grey would fall on the chart and how that affects over- and under-exposure latitude.

IEBA: | October, 01, 2010

Okay, changing gears here, I overlooked a key point in the article… and I see that the charts do indeed reflect “lower” overall light entering the camera as you set the ISO lower.
= = = = = = = = = = =

Which brings up a new question…
If you drop the ISO 2 stops from 800 to 200, why not then open the iris two stops to compensate?

You stated: “I haven’t changed the T-stop on the lens or the lighting on the chart.” and the white that was just under clipping at ISO 800 is now 83 IRE (ISO 200, LogC).

At ISO 200, you state: “We’ve now lost another stop of overexposure latitude [from ISO 400], and our notched reference is down two stops from where it started at ISO 800.”

Okay, but what would happen if you opened up the iris and put the white chip at clipping, just like you did at ISO 800? Wouldn’t that bring up the notch reference two stops and “give back” those two stops of overexposure latitude?

I’m unclear how underexposing the chart by two stops (it tops out at 83 ire, not 100 ire) means you “lost” overexposure latitude. The way I interpret whats happening is that, the latitude is still there, you just need to open the iris back up to bring the top chip to clipping and THEN assess the chip chart.

By decreasing the ISO, and not opening the lens, you are making the scale darker and pushing more of the scale below 50%... As indicated by the now -10 steps at the bottom, as opposed to only +5, topping out at 83 ire.

I ask this specifically because, on the first page of the article, you state: “We shot the chart at each ISO setting ... results: ... showed how many total stops were visible below white clip on the chart, and another that showed how those stops were distributed above and below 18% gray on the waveform monitor. The first tells us how much range the camera has overall at a specific ISO, and the other tells us how much exposure latitude we have before we hit white clip or the noise floor (black).”

But if you never move white back up to clipping, either through aperture adjustment or brightness of the scale itself, then how can you determine how much exposure you have before white clip when your whitest white is at 83 ire?

Art Adams: | October, 01, 2010

>>>Okay, but what would happen if you opened up the iris and put the white chip at clipping, just like you did at ISO 800? Wouldn’t that bring up the notch reference two stops and “give back” those two stops of overexposure latitude?<<<

No, because the sensor is already clipping.

Think of it this way: the sensor only has as much range as it has, period. Based on these tests it looks like it has around 14-15 stops of latitude. At ISO 800, you’ve got seven stops of range above 18% grey and seven-eight stops below 18% grey.

Changing the ISO doesn’t add range, it just redistributes it: at ISO 200, the camera now only has five stops of range above 18%-the NEW 18%, because the old 18% grey has been pushed down by lowering the gain-and the rest of the stops fall below 18%. The sensor is still clipping. Normally you’d open up two stops to bring 18% grey back up to “normal” but in this case I didn’t, because I wanted to show what happened to the original ISO 800 18% grey value.

By opening up the aperture two stops all that would happen is that I’d put the notched reference grey back at 45% and two additional steps on the left side of the chart would be clipped, instead of just the first chip.

So you’ll still have 14-15 stops of dynamic range, but only five will be over 18% grey at ISO 200.

What I don’t think you’re understanding is that the first step of the chart in every LogC waveform is clipped, even though it isn’t at 100%. If you opened up the aperture further, every other step on the chart will rise up and flatten out next to the first chip.

>>>I’m unclear how underexposing the chart by two stops (it tops out at 83 ire, not 100 ire) means you “lost” overexposure latitude.<<<

In LogC, at ISO 200, the maximum level the signal will reach will be about 85%. Opening up the stop won’t change that as every other chip will clip when it reaches 85%. In LogC gamma the ceiling level at which everything will clip drops with the ISO. It can do this because it’s not meant to be viewed on a Rec 709 display. LogC is for storage only, not for viewing.

I’m going to have a further discussion with Arri to learn more about LogC so I’ll find out more then. The only ISO that allows a signal to reach 100% is 1600.

In Rec 709 gamma the max will be at 100%, but that’s a different gamma. LogC can do what it wants as it’s meant for storage, not for viewing. Rec 709’s white, by definition, is always at 100%.

>>>The way I interpret whats happening is that, the latitude is still there, you just need to open the iris back up to bring the top chip to clipping and THEN assess the chip chart.<<<

Why? At ISO 200 you wouldn’t restore dynamic range, you’re just causing two more chips on the left of the chart to clip.

>>>But if you never move white back up to clipping, either through aperture adjustment or brightness of the scale itself, then how can you determine how much exposure you have before white clip when your whitest white is at 83 ire?<<<

Because, as I’ve explained, the MAXIMUM WHITE VALUE changes in LogC. As the ISO drops, the MAXIMUM WHITE VALUE lowers. At ISO 200 it’s 85%. At 1600 it’s 100%. At 800 it’s 95%. At 400 it’s 905. Why is that? I don’t know. That’s just the way LogC works.

Sony S-Log is similar, as you can never push the signal over 104% no matter how hard you try.

But just because the top value on the scale doesn’t go beyond 83-85% DOESN’T MEAN THAT IT’S NOT CLIPPED. If you cranked the stop open further at ISO 200 you could clip the entire chart, but all the chips would clip at 83-85% and not 100%. For some reason, 85% is the maximum level you can reach in S-Log at ISO 200.

If you go back and look at the Rec 709 waveforms for every example you will see that they ALWAYS reach 100%, regardless of the ISO, because that’s what the Rec 709 standard calls for. LogC is not a Rec 709-compliant gamma and can do whatever it wants to as you have to grade it for Rec 709 later. LogC is very efficient at storing information for grading, it is not meant to be viewed directly.

I’ll emphasize this one point one more time:

IN EVERY WAVEFORM SHOWN IN THIS ARTICLE THE FIRST CHIP AT THE LEFT OF THE CHART -IS- CLIPPING IN EVERY SINGLE CASE. Clipping is defined by opening up the exposure until the first chip hits a hard ceiling, flattens out, and won’t move beyond that level. The point at which that happens varies by ISO in LogC. Why? Because. I don’t know yet. In the Rec 709 waveforms, though, clipping is always clearly at 100%, regardless of the ISO.

Nick Shaw: | October, 01, 2010

>>I’m going to have a further discussion with Arri to learn more about LogC so I’ll find out more then. The only ISO that allows a signal to reach 100% is 1600.

If you go to www.arridigital.com you can download the ALEXA LUT package, which includes a very informative html document with plots of the curves. The very first graph there is a plot of the sensor linear signal against LogC, which makes it very clear why only ISO 1600 reaches 100%.

Art, you may want to ask ARRI if you can reproduce that graph in this article.

IEBA: | October, 01, 2010

I think this may be the basis of my confusion. Applying traditional television measuring gear (waveform) to new technologies that act differently.

“Clipping is defined by opening up the exposure until the first chip hits a hard ceiling, flattens out, and won’t move beyond that level. The point at which that happens varies by ISO in LogC. Why? Because. I don’t know yet.”

Coming from my broadcast TV background, it’s like saying, you have to shoot at 18db of gain in order to get 100ire whites. If you knock it down to 9db or 0db of gain for less grain / noise, then your whites top out at 83 ire and your midpoint has dropped two stops, and the entire scale is now smooshed in the greys.

Nobody’d buy that camera.

We expect to adjust sensitivity and have the same response measured on the scopes.

As to the Rec 709, while it looks quite pleasing, I agree with you that a lot of stops are smooshed at the ends of the S-curve. Also, you clearly discount it:
- “it was created at a time when cameras could deliver only about five stops of dynamic range—so that’s all Rec 709 is designed to hold.
The trick has become cramming 10+ stops of dynamic range into a small five stop ‘bit bucket’ in such a way that it still looks good.”

So I wasn’t really considering it as a useful measuring or storage tool.

So my takeaway is this:
we have LogC which maintains a fairly linear response, but isn’t properly measurable using traditional broadcast waveform scopes.
And we have Rec 709 which crams 10+ stops into a 5-stop bit bucket through a strong S-curve. Which doesn’t help the footage.

Yes, the article is quite enlightening. Makes me want to go and try this with my HDV camcorder connected to a waveform monitor (though I lack the scale) and see what happens at different gain settings. I’m still interested in how a 5D / 7D would compare.

Nick Shaw: | October, 02, 2010

>>Coming from my broadcast TV background, it’s like saying, you have to shoot at 18db of gain in order to get 100ire whites. If you knock it down to 9db or 0db of gain for less grain / noise, then your whites top out at 83 ire and your midpoint has dropped two stops, and the entire scale is now smooshed in the greys.

>>Nobody’d buy that camera.

But LogC doesn’t work that way. LogC, like the original Cineon curve it is designed to replicate, is a way of holding image data much higher that the nominal scene white. How much higher on a digital sensor is dependent on ISO. Video cameras only go up to the scene white level with only the extra 9% IRE as headroom above it.

The ‘wasted’ headroom at lower ISOs in LogC is not seen once a viewing LUT has been applied. The extra headroom in a LogC image is essentially there only to be ‘dug into’ in grading - a slight over-simplification, but basically true.

Art Adams: | October, 02, 2010

>>>Coming from my broadcast TV background, it’s like saying, you have to shoot at 18db of gain in order to get 100ire whites.<<<

A great example of this is the old SDX-900. If you ran it in -3db gain you could never get a highlight that would hit 100%. The signal topped out in the high 90’s. If you had zebras set at 100% you’d run into some surprises, as you’d never see zebras kick in.

>>>we have LogC which maintains a fairly linear response, but isn’t properly measurable using traditional broadcast waveform scopes.<<<

Incorrect. We can measure it fine, it just won’t look good on a monitor.

As I’ve said numerous times now: LogC is not meant to be viewed directly, so it can do whatever it wants to do in the camera. What’s important is that you can grade it in post (something done more in the film world than the broadcast world) and create extraordinary looks that couldn’t be done in camera.

And although it won’t look right on a monitor it is very easy to see what’s going on using a waveform monitor.

>>>And we have Rec 709 which crams 10+ stops into a 5-stop bit bucket through a strong S-curve. Which doesn’t help the footage.<<<

Sure it does. It looks fine. Most broadcast cameras should be so lucky as to have that much dynamic range.

There’s nothing new about the Alexa other than its dual gain architecture. It otherwise acts very predictably compared to existing cameras. It’s old technology used in new ways.

ericescobar: | October, 05, 2010

Thanks so much for doing this Art. This is such great info, very thorough and smart.

Quentin Brown: | October, 05, 2010

Thank you very much for this in depth article Art. It’s really great to be able to go into working with the Alexa with all these tests and data done and laid out so clearly.
All of the above is as expected based on my experiences with the RED and changing ISO’s and dynamic range. Great to have the specifics for the Alexa though.

I too was a little confused looking at the Log C charts through the different ISO’s, on the surface they look like a measurement of the changing exposed stops on a fixed scale and that in the lower ISO’s there is room at the top of the scale for more stops if the lens was opened up more.
Knowing that you did open up more and established that those highest chips were at a new lower clipping point really helps. Perhaps you should put this info into the article somewhere, I was wondering about it as I read through.

>>IN EVERY WAVEFORM SHOWN IN THIS ARTICLE THE FIRST CHIP AT THE LEFT OF THE CHART -IS- CLIPPING IN EVERY SINGLE CASE. Clipping is defined by opening up the exposure until the first chip hits a hard ceiling, flattens out, and won’t move beyond that level. The point at which that happens varies by ISO in LogC. Why? Because. I don’t know yet. In the Rec 709 waveforms, though, clipping is always clearly at 100%, regardless of the ISO.<<

Joshua Jackson: | October, 27, 2011

Ok.  But…

In reality, middle grey never “moves” around at all.  It technically exists between chip number 2 and 3 if the densities of the test chart is scaled to be 1 stop between each chip. 
For instance, if the 0 chip is scaled to 100% and it has a typical starting density of .05, then a chip at 1/2 stop between chips 2 and 3 would give a density of .80 (which at a base density of .05, gives .75 density above 100%).  Which equals… 18 % transmission.  The true 18 % that is occuring in the SCENE, regardless of exposure.

Please correct me if I’m wrong, but it sounds like this is where most of the confusion is bourne in the conversation.

Joshua Jackson: | October, 27, 2011

It’s this “middle grey” terminology that caused the most confusion it seems.
Ha!  Get it?
Ok, I desist.  But if people can can differentiate this from a “correct exposure” test, I think the point of this test (to find where to place that “correct exposure” via acquisition speed for the sake of latitude) will make more since.

Joshua Jackson: | October, 27, 2011

The “Zone System” cannot be properly recreated with transmission wedges or printed scales, as 100% transmission/reflected would be reached quickly a little past zone VII.  What Art is doing is artificially recreating this zone system with a transmission strip to predict what those +100% speculars would do for the Alexa.

Art Adams: | October, 28, 2011

How can the Zone System not be properly recreated using printed scales when that’s exactly how Ansel Adams developed it in the first place?

I have this argument regularly with video engineers who believe that the only proper 18% gray reference is a material that passes or reflects exactly 18% of the light it is lit with. I come at this from a film perspective, where we create our own densities and think of middle gray as being something very different.

When I aim my spot meter at something and set the resulting exposure on a film camera, whatever I aimed the meter at will be captured at a density that will reproduce 18% gray. It doesn’t matter how much light is reflected off of it or how much light it is transmitting—the result will be the same on the negative: middle gray.

When I aim a video camera at a wedge chart and place the 18% chip at 40 ire, I will reproduce that particular wedge as middle gray. I can change the exposure and make any other wedge on that chart appear to be 18% gray as well. Just because there’s one wedge that transmits 18% of the lit that hits it doesn’t mean that I can’t expose any of the other wedges to look exactly the same. I can, very easily.

The argument that 100% transmission would be reached very quickly just past Zone VII is an outdated argument that assumes that modern cameras follow the Rec/ITU 709 spec to the letter, which means that there’s only about 2.5 stops of headroom above middle gray. There are currently no cameras out there that do this, which is why I say that they are “Rec 709 compliant” instead of being true Rec 709 cameras. They all play games with gamma in order to cram a lot of dynamic range (in Alexa’s case, 14 stops) into Rec 709’s 5 to 6 stop container).

The Zone VII argument implies a strict Rec 709 curve, which is an option in some modern cameras but is never, ever used. (Compare the Rec 709 and Cine 4 gammas in a Sony EX3 or F3, for example, and you’ll see why.)

And the argument that this isn’t a “correct” exposure test is null as well. Any test that results in knowledge as to how a camera captures and processes light is valid, and as I know that this particular scale is calibrated in exact one stop increments I can learn quite a lot about any camera regardless of which wedge lands on 18% gray.

Joshua Jackson: | October, 28, 2011

I believe I didn’t communicate very clearly.
I actually very much so agree!  In matter of fact, it is this difference between the engineers plotting exact real world reflection/transmission values and the photographer scaling the settings for the exposure/densities they desire for the scene that causes this debate.
Stops, contrast and density are all rigid measurements relative to a given scene, exposure, eye, etc.
It’s the “relative” part engineers overlook and the “rigid measurements” that return as blank stares from most artists.
Unfortunately, it seems that this line blurs from time to time resulting in questions like “why are you changing the ISO but not the lens?”  Which obviously overlooks the point of this test.
I believe this test is VERY much so valid and essential.  It is a fantastic method for testing the dynamic range of the camera and the latitude it offers under various speeds. 
Lastly, for the Zone System.  If you’ve got 11 steps of 10 stops, you’ve got the tools for a Zone System.  I think the reason engineers are baffled by this is that when they go about thinking how to create a chart with real-world INTENSITIES (would’ve preferred italics over caps!) they do the math and find Zones VIII-X range from 144% to 578%.  To get this via transmission/reflectance (with real world Intensities) they’d have to normalize the values and end up capping 100% anyway thus defeating the purpose AND overlooking the point of the Zone SYSTEM.

Joshua Jackson: | October, 28, 2011

Sorry, not sure why SYSTEM is caps in that last sentence…

C47man: | July, 22, 2014

On the off chance someone still reads these, I was hoping for a bit of clarification here. I understand this test shows that reducing ISO on the Alexa without changing T/Stop, Shutter, or any other exposure parameter will result in an image with equivalent brightness in Rec709 but a lower brightness in LogC. This means the whites are clipping lower and lower on the IRE scale as you decrease ISO. You essentially trade away overexposure latitude in exchange for neater blacks and less noise.

So, to utilize this on set, does this mean I should set the Alexa to ISO 400 but leave my light meter rated to ISO 800 and make no adjustments to my exposure value? This would result in me underexposing my image, which feels counter intuitive. I’m just a bit confused on how to read the results of this excellent test.

Art Adams: | July, 23, 2014

You don’t really need to worry about the exact value at which whites clip as we’re looking only at LogC here, which is a storage medium only: it’s not meant to be looked at directly. White and black levels will be set in the grade.

I shoot Alexa at 400 all the time with my meter set at 400. If you view the image through the Rec 709 LUT you won’t see any difference in white clip level.

Please login or register to comment