C500: C500 II Streaking

[Guest]

That's maybe with HDRx if that still exists, IDK.

They still have it. Requires a post workflow, and there supposedly can be motion artifacts, so people just aren’t in the habit of using it, but it can do some cool stuff. I have been meaning to mess around with it more.

Reading the manual, it seems like it is simultaneous read out, but i am not in the habit of believing anything Red says.

 

[Guest]

To me the real strengths and joys of using the C500ii have nothing to do with DR. Sure, the latitude of this camera is pretty good.

This camera does so much to move things along. It is fun and exciting. Lightweight, fast boot up, low power draw, built in ND, low noise, grab it and go, sort of ff35 motion capture goodness! Let’s just make sure the IBIS control and the streaking and things like that get corrected.

For me, i prefer the hassle of Arri, Red, and Venice, for the image quality. But, i don’t blink an eye in suggesting other cameras for specific jobs, and i sure don’t worry about lower specc’d cameras like A7’s, if that is what is called for. I love the image from the F35, but i’m not going to go around pretending it would handle high contrast situations as gracefully as the Red Gemini. And i’m not going to pretend that the Gemini has the same SOOC mono as the F35. DR and roll off matters. But so do other things. All cameras find a way to be good when used in their strengths.

 

[rexbron]

Cinematography Mailing Lists DR comparison of the FX9 and C500 MkII. Conclusion: Pretty much the same. Interesting that the FX9 needed a 1/2 stop lift to match, suggesting to me that 800 is really more like 640.

https://cinematography.net/LED-Camera/C500-2-and-FX9-exposure-range.html

 

[Airwolf]

So to update on the streaking, Canon called me Monday after much back and forth with Japan apparently, and they have determined that the streaking is not an issue, but a normal function of the sensor. They advised to shoot at 800 iso, and in increments of that. Also advised to shoot in CLog3, which basically tells me CLog2 is unusable. None of their suggestions fix the issue, as I was already shooting in CLog3 at 800 iso. So, yeah.

 

[Guest]

So to update on the streaking, Canon called me Monday after much back and forth with Japan apparently, and they have determined that the streaking is not an issue, but a normal function of the sensor. They advised to shoot at 800 iso, and in increments of that. Also advised to shoot in CLog3, which basically tells me CLog2 is unusable. None of their suggestions fix the issue, as I was already shooting in CLog3 at 800 iso. So, yeah.

What!?!

 

[Guest]

So to update on the streaking, Canon called me Monday after much back and forth with Japan apparently, and they have determined that the streaking is not an issue, but a normal function of the sensor. They advised to shoot at 800 iso, and in increments of that. Also advised to shoot in CLog3, which basically tells me CLog2 is unusable. None of their suggestions fix the issue, as I was already shooting in CLog3 at 800 iso. So, yeah.

I’ll check my test files this weekend. If i am not seeing the same thing, i’ll let you know. If my test files are different, then i’ll send you a copy so you can challenge Canon’s assumption.

 

[cpreston]

To reiterate, the streaking is CMOS smear which is common with these cameras. It occurs very high contrast images when one side of a sensor is exposed to much more light than the other and shows up as colored horizontal streaks on the same sensor lines as the exposed highlights. The C500MKII has it apparently, but it is much better than the C300MKII is similar situations.

 

[Guest]

To reiterate, the streaking is CMOS smear which is common with these cameras. It occurs very high contrast images when one side of a sensor is exposed to much more light than the other and shows up as colored horizontal streaks on the same sensor lines as the exposed highlights. The C500MKII has it apparently, but it is much better than the C300MKII is similar situations.

So, since my files don’t have the same contrast of a bright window and a black shirt, it is not likely to exhibit the same level of streaking? Or do you mean it is just more likely to be an issue in high contrast scenes since there is a higher chance of digging into the shadows for additional detail?

Well, in any case, i’ll go back to the files and look closer.

Airwolf, you may be experiencing the difference between Sony and Canon sensor technology. The C100 seemed to be using the sensor more conservatively. JCS mentioned, that it looks like Canon is extracting every last bit of usability/latitude from this sensor. If that is the case, then unlike other cameras, you don’t have to go into the depths to maximize it.

I know when i went from Canon to Sony, i was over exposing the Sony, and then once i got used to Sony, i was under exposing Canon. Perhaps this is also a case of the way fs5 prioritizes shadow to highlight ratios, vs the way Canon does. As i don’t imagine the dynamic range of the Fs5 and C500ii being drastically different. However, if you are used to going by eye, there is a transitional period between Canon and Sony exposure ratios, before the eye is retrained.

I’ll send you one of my files, that way you can see if my file has the same issues. If not, you can go back to Canon and ask them to reconsider.

 

[cpreston]

So, since my files don’t have the same contrast of a bright window and a black shirt, it is not likely to exhibit the same level of streaking? Or do you mean it is just more likely to be an issue in high contrast scenes since there is a higher chance of digging into the shadows for additional detail?

On the C300 II, you could see this streaking pop up in the shadows even on a proper grade of the footage, especially if the ISO was raised very high.

 

[Barry G]

To reiterate, the streaking is CMOS smear which is common with these cameras. It occurs very high contrast images when one side of a sensor is exposed to much more light than the other and shows up as colored horizontal streaks on the same sensor lines as the exposed highlights. The C500MKII has it apparently, but it is much better than the C300MKII is similar situations.

If I remember right the streaks that Airwolf were seeing are mostly related to underexposure, and not the classic CMOS smear you describe. If you look at a number of the over/under tests that are floating around the web, you typically see them start to appear around 4.5 stops under, which is an exposure level that is generally considered unusable on any currently marketed camera.

Re Airwolf's comment...from my experience with Clog2 on this camera it is very usable, much more usable than it was on the C300II. I've tried to reproduce Airwolf's experience on a number of occasions. The only way I could get it to occur was to raise "noise floor" detail to around 40-50 ire on the scopes (this was mirrored in his examples) ...this isn't really how Clog2 (or any log) is designed to be used. It is beyond the limit of this camera (and pretty much any other camera I know of). It may not be pretty, but I don't know of another camera thats pretty that far south.

The best practice for Clog2 has always been to rate it at 800 and then offset the exposure upwards about 1/2 stop-1 stop from the meter/chart. Do the opposite, and it's at your own peril. I've found that with the C500II it functions quite well at 800ISO without an exposure offset, but I still wouldn't be underexposing because that's just ignoring what the sensor is capable of, or how the gamma is designed to be used.

FWIW, CML just published its latitude comparison of the C500II and FX9. The results might be a little surprising to some. (they usually are, as CML actually uses standardized exposure charts and a strict methodology). In their test, they found that the FX9 is actually about a half stop slower than its ISO rating, which, once corrected for, has a strange effect on all that rumored extra highlight DR that get talked about) Anyway, Geoff concluded that the two cameras have nearly identical latitude, with nearly identical highlight stops, and the C500 perhaps having a tiny bit more in the shadows, but he called that too close to tell. I dropped the two tests into a timeline so that they can be viewed together and he's pretty spot on in his assessment. The cameras clip (highlights and skin tones) at the same points, with the FX9 losing more color as the exposure increases above clipping. In the undexposure portion, the FX9 shifts green by -4 whereas the C500 stays neutral all the way to -6. (although neither is "good" past -2 according to him). I don't see a significant difference between them in the usable range. They both develop a certain amount of mottled / streakiness around 4 stops under, and maybe the FX9 is a little worse, but it doesn't matter much, as neither are "good" at that point.

I'm sure there will be some disagreement on this one, as his tests don't necessarily square with some of the other comparisons out there (the .5 stop exposure issue being responsible for most of that). Based on his standard of "good" versus "usable" they both have 5.5 stops of latitude 3.5 over and 2 under, and I didn't see anything in the test that suggests otherwise --as Geoff says the shadows are a little bit open to interpretation, but the highlight data is very clear.

 

[Guest]

C-log in my experience, is already digging down into the shadows. For people that expose by eye, and coming from other picture styles, or other camera brands, this may cause heavy under exposure.

In this screen shot A is clog, and B is neutral picture style on the 1DC. Otherwise same exact settings taken seconds apart.

It can be easy to react to clog by eye the same way one might be used to exposing from other picture styles or other brands. It is another reason to get used to scopes and meters when switching between cameras.

 

[Barry G]

C-log in my experience, is already digging down into the shadows. For people that expose by eye, and coming from other picture styles, or other camera brands, this may cause heavy under exposure.

It can be easy to react to clog by eye the same way one might be used to exposing from other picture styles or other brands. It is another reason to get used to scopes and meters when switching between cameras.

Thanks, James, for re-iterating something that was said very early in this thread (and is just right). I find with the C500II that if I'm "eyeballing it" it's much safer to be eyeballing with the wideDR LUT (or user LUT) staring me in the face...that gives you a sense where your exposure really is. When setting exposure to a log monitor image, my eye automatically just goes to the waveform, because I know from experience those raised log shadows are just gonna mess with my game . And the new LUT button on the side of the camera making it super easy to just toggle between gammas and the WF reflecting the log readings regardless of what LUT on is showing, this camera has plenty of tools to help you get exposure right. That said, I do wish there was a way to bring up the WF as a full screen overlay on the camera monitor...that tiny thing is really hard to read, and I have found myself bumping the clip point more than a few times.

 

[cpreston]

It does have the false color, which might be useful for a quick on/off to see overexposed and underexposed areas. I don't think you can change the levels, though.

 

[Guest]

Thanks, James, for re-iterating something that was said very early in this thread (and is just right). I find with the C500II that if I'm "eyeballing it" it's much safer to be eyeballing with the wideDR LUT (or user LUT) staring me in the face...that gives you a sense where your exposure really is. When setting exposure to a log monitor image, my eye automatically just goes to the waveform, because I know from experience those raised log shadows are just gonna mess with my game . And the new LUT button on the side of the camera making it super easy to just toggle between gammas and the WF reflecting the log readings regardless of what LUT on is showing, this camera has plenty of tools to help you get exposure right. That said, I do wish there was a way to bring up the WF as a full screen overlay on the camera monitor...that tiny thing is really hard to read, and I have found myself bumping the clip point more than a few times.

I am not as familiar with Wide DR. It is my assumption, that wide DR and S-Cinetone profiles have less detail preserved in the shadows than traditional picture styles and luts.

I still can’t quite judge the image in Wide DR, so, as Cpreston mentioned, i tap the false colours on/off, and sometimes check the log look to see what it going on beneath the roll offs, and get a better idea what i might have in the raw later. But overall, quite pleased with wide dr as a standard lut/profile.

 

[Barry G]

I am not as familiar with Wide DR. It is my assumption, that wide DR and S-Cinetone profiles have less detail preserved in the shadows than traditional picture styles and luts.

If by traditional you mean various standard, rec.709 style LUTs, comparatively wide dr preserves more shadow, although less than a log gamma. When Lybezki was describing his custom LUT for revenant, he was describing a wide dr type LUT, although his was probably a bit wider — the primary purpose of these LUTs is to let you see into the shadows more, and they often roll off the highlights a little less as well. Not sure where cinetone sits in there.

Back on the streakiness. I hooked the camera up to a larger monitor today, and spun the dials looking for evidence of streaking.. it arrived in an chart image that was 6-7 stops underexposed, at an ISO of 25,600 in the deepest gloss black chart patch), that had been raised by the gain to about 30 ire. There was no tonality below this point, and thus the streaks.

While I was at it, I noted a few other things. Canon has redone the noise reduction section of the camera (compared to the c300ii, at least), and added a temporal filter along with the spatial filter. I found that a setting of 1 for each takes the edge off this cameras 800 iso noise, seemingly non-destructively. At higher ISO’s, increasing the temporal noise to its max (3), does a nice job of quieting some of the worst noise, while leaving the general noise structure intact for post denoising. At this setting, the noise floor “streaking” seems less noticeable. Increased spatial NR, eats detail and just makes the noise lumpy.. so I’d avoid it. Along with trying to raise the noise floor from the dead.

 

[Guest]

If by traditional you mean...

no

 

[modernhuman]

Good points. Very aware of camera DR vs display DR. The Alexa has cleaner DR. This is because of better analog components and better digital converters. Similar to how high end digital audio converters have lower noise floor and thus more dynamic range. An Alexa costs orders of magnitude more than any of these cameras. So components are flat out better. Alexas will continue to be best for image quality, until people shut up about resolution, low light and hi frame rates. Arri understand cinema and how cinema DPs work, but they have also surpassed film with DR. So it makes sense why he is making the choices he does, given his natural light technique, and a camera with the most natural contrast possible. This camera has allowed him to veer away from a more traditional contrast curve like the one Deakins uses. No knocking deakins, btw. One of my favorites. But there’s no question that Lubezki is doing stuff they would never have tried with film. Also, I totally understand market segments. But I’m interested solely in cinema and cinematic story telling. Not ENG. Thanks for your insights Barry.

 

[Guest]

Alexas will continue to be best for image quality, until people shut up about resolution, low light and hi frame rates.

I hate to beat a dead horse. But I do love talking about the same thing over and over again.

Doesn't the Alexa employ dual gain simultaneous read out to get that dynamic range? The pixels don't hurt, but as far as being about 1.5 stops better than a 6K camera without the Arri dual gain patent... doesn't that mean Arri is doing it with a bit of a trick? A good trick, mind you, but it isn't just because of 2K or better pixels, it is also because of their patent. Similar to Red's internal raw. Other companies might be able to do it, but they don't. but it is better for Arri marketing to say "better pixels". While there is truth to it, there is also a dual gain real time HDR thing going on, as well. if the Venice could use that, I think it would surpass the old 2K Alexa in DR.

Arri has an advantage in that there employees are film makers, and I believe they have higher film making aesthetics to the engineers at other companies, and so their end result looks and feels more like what film makers want. And they can charge a lot for the hard work they put in to their cameras, but I wouldn't say 2K alone is the reason for the camera's dominance today. I'd say it is likely more their patent, or an unwillingness on Japanese companies to kill the price bench marks that Arri is keeping afloat (tinfoil hats on). IF an Arri package coasts $100K, then Canikony can charge a little less for whatever they put out. As seen with the Black Magic race to the bottom, they can't keep zcam or other companies from competing on those margins. perfect cameras for the compulsive upgrader. Each release is "never before seen specs in this price bracket". But I believe that Canikony can't get past Arri in DR, due to patent control over the more straight forward and effective way to get dual scan read out for HDR images.

So, I don't believe it is because someone on a forum said that they prefer 4K, that camera companies have been stuck behind Arri for the last decade.

 

[modernhuman]

Totally with you about the dual gain patent. But Alexa has bigger pixels, quite a bit bigger than most. The physics of that can’t be ignored. The noise goes down with bigger pixels. So that’s why I keep going back to the resolution race. The irony is most film cinematographers try and soften their images, with filters, vintage lenses, diffusion, flares. So 2k really is fine for most film. But it’s all these pixel peepers and pc gamer types who want this ultra sharp stuff. I think Sony was on the right path with the 12 megapixel A7Sii (big pixels=clean shadows=clean gain), but now it’s too late and 6 and 8k are the new marketing buzz words. S35 sensors should be around 3k and full frame around 4.5k for perfect debayer. No need for more.

 

[Guest]

Totally with you about the dual gain patent. But Alexa has bigger pixels, quite a bit bigger than most. The physics of that can’t be ignored. The noise goes down with bigger pixels. So that’s why I keep going back to the resolution race. The irony is most film cinematographers try and soften their images, with filters, vintage lenses, diffusion, flares. So 2k really is fine for most film. But it’s all these pixel peepers and pc gamer types who want this ultra sharp stuff. I think Sony was on the right path with the 12 megapixel A7Sii (big pixels=clean shadows=clean gain), but now it’s too late and 6 and 8k are the new marketing buzz words. S35 sensors should be around 3k and full frame around 4.5k for perfect debayer. No need for more.

But I see less noise in the Red Gemini than the Arri Alexa Mini at iso1600.

The race war is slightly overblown, as tech has surpassed some of the previous limitations.


The way I see it, is that 4K was inevitable. So all the HD and 2K devotees just slowed down the inevitable giving 4K and 8K an excuse to not be as good. But now that we have 8K, people are beginning to understand that DR is important. There was no way around it. Companies were going to go with the easy marketing wording. 2K slowed down image quality in the long run.