C500: C500 II an interesting little wrap up.

[robotfist]

To me the C300 III’s curve looks a bit more gradual descending from the highlights into the midtones. But maybe I’m just drunk.

 

[Guest]

To me the C300 III’s curve looks a bit more gradual descending from the highlights into the midtones. But maybe I’m just drunk.

The aspect ratio of the charts are a little different. Gives the C3003 that lean.

Red: C3003, Blue:C5002
In this case, it would appear the the C500ii had slighlty more noise, brighter mids, and softer highlight roll off. And the C3003 with cleaner shadows, “richer” mid tones, and sharper highlight roll off.

You can see the 3 and the 6 are still darker on the C3003, but then the 1 and 2 are the same levels. Means the C3003 curve is actually less gradual, except in the shadows.

 

[Barry G]

Yes, I appreciate the cinema5D imatest because it's repeatable and since the same testers are putting together a database of cameras, at least you can draw inferences about relative performance.


...But if you look at the noise floor of the C300III, isn't there a bit of fuzz rising up to the top of that 13th stop? We can see the shape of the 14th+ stops but they're overlapping with a lot of noise. In contrast, the c500ii has a super clean noise floor.



As for indicating color performance at the extremes of DR, can't we glean some insight into that by the latitude over and underexposure tests that cinema5D does? Progressively changing the exposure in each direction and then correcting it and seeing how it looks?

I think "from the beginning" ie -- ~5 years ago when C5d started running these tests -- they clouded things up with a lot of editorializing...in particular their use of the term "usable Dynamic Range" coupled with their reliance on xyla/imatest as their sole testing method. Imatest doesn't tell you anything about usability (their reasoning for using XF-AVC instead of raw in the current tests tells you as much... Xf-AVC gives them a better score so they are using that, but XF-AVC is in no way more "usable" than Canon's RAW light.) Back when they were taking "guff" from the peanut gallery, it was because they were applying this "usable" metric to compare cameras and coming up with some fascinating deviances from what the cameras were capable of (or not capable of).

Today I think they do a lot better job...they've added latitude tests which I think give us a better metric of "usability" and are doing a more consistent job of exposing the xyla than they used to.

As for what you're seeing in the C3003/c5002 waveforms, I'm seeing the opposite. The 13th step of the C500 has more noise (thus looks more raised) and the noise floor is "all noise", where as the C300 has much smoother steps from 8-13 and then appears to be registering more signal variation within the noise floor, with less overall noise there. I guess its ok for two of us to look at the same thing and see it differently, though.

 

[robotfist]

The aspect ratio of the charts are a little different. Gives the C3003 that lean.

Red: C3003, Blue:C5002
In this case, it would appear the the C500ii had slighlty more noise, brighter mids, and softer highlight roll off. And the C3003 with cleaner shadows, “richer” mid tones, and sharper highlight roll off.

You can see the 3 and the 6 are still darker on the C3003, but then the 1 and 2 are the same levels. Means the C3003 curve is actually less gradual, except in the shadows.

Good call with the overlay! Looking at that, the highlights look extremely close to me. I do wish that in both cameras the drop into the midtones was as gradual a curve as it is from the midtones into the shadows.

I feel like they should perform this same test with an Alexa to offer a baseline for which to interpret all their other tests.

 

[ahalpert]

I feel like they should perform this same test with an Alexa to offer a baseline for which to interpret all their other tests.

They did at some point and rated the Alexa at 14 stops. I can't find their original test post but they reference it in this article: https://www.cinema5d.com/the-cinema5d-camera-lab-is-back-dynamic-range-tests/

Of course, people are also happy with the Alexa for allocating more of those stops above middle gray and having the more gradual highlight roll-off, which the DR number doesn't tell us. But still.

So, in essence, their testing finds the new Canons to have close to a stop less DR. (We can argue about the C300III having usable shadow stops or not, but at any rate they're in the shadows and not the highlights).