Watching the C300 MarkII "features" video, I was struck by the characterization of the new 15 stop DR as supporting greater "shadow protection". I've always felt the weakness in the C300 has been in how it handles highlights (clipping rather than rolling them off nicely).

I dug up dynamic range charts for the original c300 as well as Alexa and can see that the new camera improves highlight DR by 1 stop and shadow DR by 2 stops compared to the C300. The Alexa, on the other hand has an additional stop of range in the highlights (at native iso) compared to the mark 2, but two stops less in the shadows. Additionally, alexa's DR seems adjustable based on ISO...meaning you can provide additional highlight protection by simply underexposing and increasing gain, and vice versa. Canon's chart however indicates a consistent DR distribution (6.3 stops above 18%grey, 8.7 below) throughout the higher ISOs.

I guess my question is this (and it's something I've never quite understood about how we look at ISO and DR on these cameras) -- can we simply underexpose our grey card by a stop to get an additional stop of highlight protection (and thus...mimic Alexa), and is there any penalty to this given our new 10/12bit color space, and apparently improved S/N ratio?

Also, Can anyone explain the discrepancy between the ISO/DR chart which shows 6.3 stops of highlight range, and the C-Log2 graph that appears to show 8+ stops of range above 18%?

I'm all ears.

I guess my question is this (and it's something I've never quite understood about how we look at ISO and DR on these cameras) -- can we simply underexpose our grey card by a stop to get an additional stop of highlight protection (and thus...mimic Alexa), and is there any penalty to this given our new 10/12bit color space, and apparently improved S/N ratio?
Firstly, you have to think about whether you're talking about the raw off the sensor signal or a coded camera.

It's easier to think first of the signal straight off the sensor.

Imagine the camera is pointing at a DR chart, with a very wide range - let's say 20 stops. Now let's say we adjust the iris such that the middle 14 stops are clearly resolved, the darkest three effectively black, the lightest three effectively white. It's the relationship between f stop and the overall brightness that therefore defines the ISO.

Now imagine closing the iris one stop. You'll still get 14 stops defined - but they'll shift along such that the darkest four are now pure black, the lightest two pure white. Compared to the previous example, that's simply equivalent to doubling the lighting level and stopping down one stop - so exactly the same ISO. This is consequently the native ISO for the sensor - and can only take one value.

Well, almost. It assumes the sensor itself is the weakest link, but it's conceivable that the weakest link is the A-D convertor and the analogue electronics. And this is the case in the Alexa, owing to it's relatively large photosites and the inherent high DR of the sensor itself. It gets round it by having two read mechanisms - one biased to read the lowlights, the other the highlights, and combining the two to keep the full range of the sensor itself. Hence it's excellent results in this respect.

In most cameras, the A-D element should be capable of getting most of what the sensor can deliver, so normally we don't need to worry.

If you think about what an ISO increase actually means, then the physics of what the sensor itself is doing can't change. That's the significance of what a camera having a single native ISO means. Changing ISO means (for the example above) stopping down two stops for a two stop increase in ISO, and getting the same result for the same lighting level.

But at the sensor level, the two darkest blocks that gave detail in the original case now won't, end of story. The ISO increase is being achieved by a remapping of the sensor output levels to video levels - and in the simplest possible case that's effectively the same as a gain increase, and increased noise. Practically, it may be more complicated than that - varying the remapping with level. And that's why it becomes difficult to directly answer your questions, there become too many variables. But by and large if you increase ISO you pay a penalty - but it's possible to trade what that penalty is.

Hi David.

Thanks for the explanation. And I totally get what you're saying. I think my difficulty is in understanding how two sensors that are probably remarkably similar under the hood can place the mid-tone at such differing levels. In a perfect world, the 18% grey would sit right in the middle, with equal amounts of range above and below it. Yet the Alexa seems skewed to protect highlights, and the C300 biased toward protecting the shadows. It seems a bit arbitrary to me. Hopefully at NAB we'll glean some more information about how canon is pulling these 15 stops off the sensor, and how much we can dive into the shadows to save the highlights.

I believe Gale Tattersall is presenting some sample footage at the Canon Theatre, so hopefully some of the questions will be answered by that.

Barry
____________________________________________________

If I want my c300mkii to look like an Alexa, what's a guy to do?

I agree that it seems a bit arbitrary.

If you want more detail in the shadows, expose higher. You'll get less noise in the lower levels at the risk of clipping highlights. But maybe there are no highlights in the scene.

If you want more highlight latitude, expose lower. You get more noise in the shadows that cover up low level details, but maybe there is nothing dark in the scene.

And yes, a neutral shoot puts the gray card in the middle, but where did you place the gray card? Are there multiple characters, one near the key, another with only the fill, and a third farther away from the light? If so, where does the gray card go?

To me, this comes down to artistic intent. The one technical consideration might be that the C300ii has its colors calibrated at one level and the Alexa's colors are calibrated a stop lower. Then again, how much can this matter, given that your character might move from silhouette to the sunlight through the window.

I've recently studied the ACES workflow in which you shoot a gray card and bring the footage into ACES with that card a mid-gray and color balanced - for normal scenes. But not all scenes are "normal". One might intentionally apply a color cast (like when shooting 6500K at sunset) and one might want the character's face to be very dark or porcelain white. (Marlene Dietrich generally requested to be shot a stop brighter than other characters.)

So aside from color calibration, I think it's fairly arbitrary. And I don't think that a difference of one stop will push the calibrations too far out of whack, though in my mind's eye, I can see dark faces going a bit magenta and bright faces going a bit yellow.

Also, Can anyone explain the discrepancy between the ISO/DR chart which shows 6.3 stops of highlight range, and the C-Log2 graph that appears to show 8+ stops of range above 18%?

Simply bug.

From what I could glean on the nab floor this morning... C-log 2 is lifting the midtones and shadows significantly, compressing the highlights nicely. Exactly what I want from a camera, and more in line with Alexa and s-log3, and exactly the opposite of c-log which pushes the midtones down.. LCD is much improved... Accurate color and a lot sharper. Gobs of detail. No visible noise on the monitors. Much cleaner looking than the c300. Clog 2 still seems to have relatively good color ungraded, albeit a lot flatter and lighter than the old camera.

Got to spend a little time with the C300 Mark II at EVS yesterday (nice store BTW, plus rentals and nice stages right there). The camera they had there was most likely one of the same prototypes we saw at NAB, but without the crowds there was an opportunity to dig in and examine the image. A couple of takeaways:

The dynamic range is really phenomenal, you can push the exposure up 3-4 stops above normal and have really usable data in the highlights, (the difference with the original C300 in the skintones as you increase exposure is most remarkable), or equally, you can protect the highlights with a lot of range left in the shadows. Also on the low light side, at ISO 25,600, while the noise is still quite obvious, the sharpness and dynamic range didn't seem to suffer. (for perspective, the room was lit by a couple of Kino 4-bangers set about 5 feet from the targets, I was shooting with both ND wheels engaged for 10 stops of ND). At base ISO, there was absolutely no moire visible on the resolution chart (not that that was ever much of an issue on the C300, but I think the new camera is improved in this manner). The new handle is much improved, a solid as a rock with lots of mounting points. (and...the redesigned LCD/controller unit now sits straight when mounted on the front of the handle --woo-hoo).

The only negative was the pattern noise I reported at NAB that's visible at all ISO's in the deepest tones. Gale Tattersall was quoted at NAB as saying this camera has more like 18 stops of DR...and I'd agree, but I think at least one of those stops will go away once Canon buttons up those shadows. I think that 15 stops will be an easy target for this camera to hit. It remains to be seen what the how the new codec will play into this, but at least on the monitor, there is very little noise visible in the image at most ISOs above 20 ire or so...The increased highlight range should allow you to "expose to the right" to get an extremely clean signal, even in very low light situations.