Interview with Canon's Larry Thorpe on C300 - Explains a lot

[Rick Burnett]

I can't even believe this... so, let me get this straight; you choose a still image that was clearly blown up in size, distorting pixels and creating banding and whatnot, over going in to look at the film itself? You are not seriously judging the quaility of a camera by a blown out still frame used on the web! Please don't...

A little better:

View attachment 43339

​

See the highlight behind her hair? Minimal clipping, gentle rolloff. And its not that I don't want the freedom to grade the footage, being Raw doesn't necessarily help you anything with blown highlights, like I said, if the values were to high to reach within the sensor's DR then its gone, its pure white. And sometime we want pure white somewhere in the shot, be it a sky, a lamp, whatnot, but what we don't want is a clip, and the Red clip has a tiny rolloff that will make that highlight look fake, electronic even if you are able to get the clipping down to a minimum, it will look burned out, not smoothly rolling into white!

So, in that case, its not about knowing what we are doing, because sometimes we do want the exposure to go all the way. And clipping comes as an added extra that will ruin the shot, even if you grade it out, the electronic burn created by the clipping will always look fake and patchy.​

​

A litte better? Hardly. I see all the same issues in that 501x280 image. Look how soft that image is?!? At 501x280 it should be razor sharp being down converted that much. Again, if you happened to read what I wrote, you'd see I am NOT going to judge the C300 from that image. But certainly, you've not helped the case AT ALL with that screen cap. And I won't judge it from this either.

But AGAIN FOR THE THIRD TIME, you have NO GRASP of what the electronic sensor is doing. That roll-off is not caused at the sensor, on any camera. It's the post processing of what you want to do with what it recorded. AGAIN. If you want a smoother roll-off, you can fix it in post. IT'S A DIGITAL PROCESS THAT HAPPENS AFTER THE SENSOR. If you want to roll off the highlights on the Epic, you can do the SAME THING. It's just the camera processing isn't going to do it for you with the Red. That's it.

Ask yourself, why does a blue LED on the 7D clip to white? There was no red and green emitted from that LED. Where did the sensor see that? It doesn't. Instead, as the blue channel approaches clipping, the red and green channel are ARTIFICIALLY affected to make a more pleasing roll-off. There is no EXTRA data. If you want to do this in post, YOU CAN.

YOU don't want to do this. I understand that. You'd rather the camera do it for you. Again, fair enough. But, it's not a glitch, and it's not missing information.

 

[Piolet]

Let me ask you this, again, since you missed it the first time, do you think this screen grab from the C300 is appealing?

http://www.cinemaeos.usa.canon.com/...pc&utm_term=canon c300&utm_campaign=Project+A

(The lady looking to the left, scroll down)

Because I think it looks HORRIBLE. Like I said before, I don't know how that image came to be, and it's definitely from one of the films, but the softness, blockiness, banding, etc, looks terrible. Yet, I am not willing to believe that pictures represents everything that is the C300 just as much as a lot of other things I see.

Rick, that C300 photo is not a frame grab from the original footage like the true frame grab you posted from Tonaci's Red footage. It is from the XXIT film, but it looks like some webdesigner took a screenshot or something, cropped it and blew it up to fit the webpage. I saw the original footage at Paramount. This looks nothing like that.

 

[Erich Ocean]

Ask yourself, why does a blue LED on the 7D clip to white? There was no red and green emitted from that LED. Where did the sensor see that? It doesn't. Instead, as the blue channel approaches clipping, the red and green channel are ARTIFICIALLY affected to make a more pleasing roll-off. There is no EXTRA data. If you want to do this in post, YOU CAN.

Yes, but not easily, and certainly not easily for your average filmmaker.

So while what you say is true at a technical level, the fact is that Red does not allow for alternative debayer software. It's proprietary (only).

In the past RedCine did provide a way to go to RGB with no modification of the original RGB values the sensor actually recorded, which you could then process to get back the RAW sensor data in a format you could then re-debayer differently (phew!), but RED cut this path off. (I'd really like to know why, having your footage in a proprietary format seems like a really, really bad idea. People should be able to archive the RAW sensor data in an open format.)

Anyway, at least in the context of Red, highlight handling is something that is not fully controllable in post; you can only do what Red's own debayer software supports. Many people think that both Arri, Canon (and to a lesser extent, Sony) do better in this area when it comes to highlights and skin tones. YMMV.

 

[Rick Burnett]

Rick, that C300 photo is not a frame grab from the original footage like the true frame grab you posted from Tonaci's Red footage. It is from the XXIT film, but it looks like some webdesigner took a screenshot or something, cropped it and blew it up to fit the webpage. I saw the original footage at Paramount. This looks nothing like that.

Believe me, I could post other examples that really show the C300 shining. The only reason for posting that was that the origination and/or processing that happens to a lot of things you see online is unknown. If I were Canon though, I'd be replacing that image IMMEDIATELY as it looks horrible. Compared to the painstaking detail they've provided so far, some of their image selections in their documentation online have definitely been less than stellar (boy on bike). But, with all things, and the point again, how can you know? Until it is your eyes shooting a scene with a camera, how that camera responds is just short of a guess. Likewise, pushing that footage in post and looking into the shadows, only then can you see what kind of range you have, or, how the slopes of the curves are affecting what you are capturing. One person says "Hey, there isn't any detail in the shadow" and then they find out, oh, the blacks are crushed, so if I just dial back a few settings, then "Wow, yeah, this camera is totally amazing!!!"


There are SO MANY variables to what is happening to the footage that almost ALL comparisons totally suffer from subjective bias. The technical discussions try to get around that as much as possible to give some sort of 'boundaries' for where a camera performs. But even beyond that, for people planning on the future of their productions, when they can't get their hands on the camera, all they have is the information that is presented to them. If a camera gets 13 stops of DR, that's it. It doesn't magically have more because you want to believe it does. If your scene has more than 13 stops, it will either clip on one end, or it will get lost in the noise at the other. Period. (And even then, how much does that bottom stop sit above the noise? What does the noise look like? Is it acceptable). SNR doesn't tell you what the noise looks like. And I'll say noise is DEFINITELY a VERY subjective subject. So there are things that CANNOT be answered easily with technical discussions either, but for now, that is the starting point for people evaluating if they are going to move to this camera. Everyone has a different set of criteria for what is important to them, and price is definitely one of those things.

I want the camera. But right now, with our production company, *I* cannot justify the price/performance for our audience. Others can. In truth, there is nothing wrong with that. It's just a part of business.

 

[Rick Burnett]

Yes, but not easily, and certainly not easily for your average filmmaker.

So while what you say is true at a technical level, the fact is that Red does not allow for alternative debayer software. It's proprietary (only).

In the past RedCine did provide a way to go to RGB with no modification of the original RGB values the sensor actually recorded, which you could then process to get back the RAW sensor data in a format you could then re-debayer differently (phew!), but RED cut this path off. (I'd really like to know why, having your footage in a proprietary format seems like a really, really bad idea. People should be able to archive the RAW sensor data in an open format.)

Anyway, at least in the context of Red, highlight handling is something that is not fully controllable in post; you can only do what Red's own debayer software supports. Many people think that both Arri, Canon (and to a lesser extent, Sony) do better in this area when it comes to highlights and skin tones. YMMV.

De-Bayering does not do highlight handling, it's just the process of spreading the pixel color information into the adjacent different colored pixels (interpolation) through a variety of weighting methods based on the algorithm. This does not destroy information either. It's actually creating information where none exists. Which is what the Canon sensor is not doing. The Canon sensor is skipping the benefits of debayer by sacrificing resolution to simplify the process. Of course, at 1080p with a 4K sensor, that extra resolution would be lost anyways, so it doesn't matter (and makes sense to me why they did what they did). Neither of these things have anything to do with the highlight handling.

But two things; 1) if you are spending the money you are spending on an Alexa or Epic, I don't think you are the 'average' filmmaker, but even if you are 2) this is why you hire someone to color your film for you. This is why people specialize in JUST THAT because it's an art, and one that clearly is needed since EVERY movie credits I have watched includes such a person.

I think people have subscribed to the 'one-man-band' philosophy as if that were the norm in high end filmmaking, but it isn't. I'm still waiting to see a crew list of ONE person at the cinema.

 

[Jadotch]

I think people have subscribed to the 'one-man-band' philosophy as if that were the norm in high end filmmaking, but it isn't. I'm still waiting to see a crew list of ONE person at the cinema.

It does not happen. Although on El Mariachi, Robert Rodriguez takes up half the credits, another quarter are taken up by his fictitious names. I know it is the exception that proves the rule.... Still it is amazing what that man can do.

 

[Rick Burnett]

It does not happen. Although on El Mariachi, Robert Rodriguez takes up half the credits, another quarter are taken up by his fictitious names. I know it is the exception that proves the rule.... Still it is amazing what that man can do.

It is a good movie I own it on DVD. Of course, didn't a whole bunch of time and money get spent to make it ready for cinema? After it got picked up?

 

[Nuno Dias]

But AGAIN FOR THE THIRD TIME, you have NO GRASP of what the electronic sensor is doing. That roll-off is not caused at the sensor, on any camera. It's the post processing of what you want to do with what it recorded. AGAIN. If you want a smoother roll-off, you can fix it in post. IT'S A DIGITAL PROCESS THAT HAPPENS AFTER THE SENSOR. If you want to roll off the highlights on the Epic, you can do the SAME THING. It's just the camera processing isn't going to do it for you with the Red. That's it.

If a camera gets 13 stops of DR, that's it. It doesn't magically have more because you want to believe it does. If your scene has more than 13 stops, it will either clip on one end, or it will get lost in the noise at the other. Period.

Isn't that contradictory?

See you you're saying. First you say, 'if you want a smoother roll off, you can fix it in post, its a digital process that happens after the sensor'. This would have any basis on reality if the camera was able to record every single exposure of any given situation, from a dark shadow all the way to a bright could in the sky,let's say 27 stops! But the camera has only 13!

How will you fix this in post?

From stop 13 over, you're white, fully exposed! In the Red sensor, by stop 14, you're clipping!

 

[Erich Ocean]

De-Bayering does not do highlight handling, it's just the process of spreading the pixel color information into the adjacent different colored pixels (interpolation) through a variety of weighting methods based on the algorithm. This does not destroy information either.

That's what's so annoying about the Red! It does destroy information. Suppose I shoot 5K FF, and I debayer to 5K RGB, so there is a 1:1 relationship between sensor "pixels" and image pixels. Meaning, each image pixel is covered by exactly one R, G, or B sensor "pixel"; the rest are interpolated. Now here's what's annoying: the values that are in the original sensor are also modified.

If they weren't modified, you could just drop the RGB values that didn't correspond to original "sensor pixels" and you'd get back the original RAW data. You cannot do this with Red today. This is because Red's demosaic software does modify the original sensor "pixel" values; it changes them to some other value.

Now, it didn't used to. There was a flag you could pass to RedCine on the command line to not modify them (that flag is now gone).

It's also my understanding the color of the missing sensor values is subject to highlight handling (e.g. boosting red and blue towards white when green clips), and that this is done during the de-mosaic process, but theres' no point arguing over it.

 

[Jadotch]

It is a good movie I own it on DVD. Of course, didn't a whole bunch of time and money get spent to make it ready for cinema? After it got picked up?

Yeap. A lot of work did go into it after the fact. I think they redid most of the sound professionally. (I honestly do not know what else was done.) I just had to take up your challenge with the best answer that I could think of.

PS I think there is a lot of myth about the one man band because of Robert Rodriguez as well. He does a lot himself, but by no means all of it.

PSS Blair Witch Project (2 man band) is another, but it looks crappy, but they used that to their advantage.

 

[Piolet]

Isn't that contradictory?

See you you're saying. First you say, 'if you want a smoother roll off, you can fix it in post, its a digital process that happens after the sensor'. This would have any basis on reality if the camera was able to record every single exposure of any given situation, from a dark shadow all the way to a bright could in the sky,let's say 27 stops! But the camera has only 13!

How will you fix this in post?

From stop 13 over, you're white, fully exposed! In the Red sensor, by stop 14, you're clipping!

I'm not sure if I understand Rick, either. To me, the sensor does matter. Seems like GIGO theory: garbage in, garbage out.

If one sensor records more information than another, then post can't make the second one equivalent to the first.

 

[BobbyMurcerFan]

Rick,

I just want to reiterate what I said in post 211, that the line between RAW and image processing is not nearly as clearly as we'd like to think. There is significant image processing that all Reds do in-camera. RAW is not a "raw" stream off the sensor. Some processes, like sensor noise denoising, are certainly done in-camera. Others, it's not so clear. Essentially, RAW is image information arranged exactly how the pixels of the sensor are arranged; that's it.

Whatever Red does or does not want to do in-camera is totally up to them. They could boots the blue channel, lower the green and apply some extra noise reduction, etc. As long as the recorded image information is in a 5K Bayer pattern, it's RAW. Same idea goes for ArriRAW. And similarly for Sony's F65, which from what I've heard takes image processing to a whole new level in terms of extent.

So there is simply no standard as to what you can or can't do to an image and still be allowed to call it RAW, except of course that it has to be arranged as the sensor is.

As for debayering, I believe Red encrypted .R3D b/c Cineform pissed them off by coming up w/ alternate algorithms that they felt were superior. Nonetheless, it does seem that debayering itself should not affect highlight handling much at all, but then again who knows.

My personal feeling is that the reason why DSLR's look so good is because they are using image processors that are a generation or two ahead of most video camera's. How accurately one can reproduce what a DIGIC IV does in Premiere, Avid MC or Resolve is beyond me. But I imagine that there is a lot of subtlety going on, including things like scene detection.

Is there some extra mojo that the Alexa has? Probably so. Does it mean that the Epic is so off that hobbit faces need to be clown painted? No way. Are people probably misusing their Red Ones and Epics to get inferior results? Sure.

I'm totally open to the idea that the Epic needs special care when shooting skin tones. But I'm also open to this being blown way out of proportion. I've heard from more than a few people that both cameras can be seamlessly intercut.

 

[Rick Burnett]

Isn't that contradictory?

See you you're saying. First you say, 'if you want a smoother roll off, you can fix it in post, its a digital process that happens after the sensor'. This would have any basis on reality if the camera was able to record every single exposure of any given situation, from a dark shadow all the way to a bright could in the sky,let's say 27 stops! But the camera has only 13!

How will you fix this in post?

From stop 13 over, you're white, fully exposed! In the Red sensor, by stop 14, you're clipping!

It's not contradictory at all. You are implying that the highlights on the C300 have more information in them, and I'm telling you that's not the case of why they look like they do. From what I gather, you are assuming that the Epic is just clipping out and it's throwing away data that the C300 is keeping. The example I gave, the 7D, has (to me) very nice clipping. But, what it shows is NOT what is in the scene. If I film a blue LED and it clips to white, there was no red/green element for the camera to capture. THAT is what I am explaining. THAT is the color science that went into making the clipping look more appealing, but, it was a post process algorithm that had nothing to do with the sensor. As that channel approached a full clip in the blue channel, on the 7D, a curve was applied to the red/green values relative to what the blue channel was doing, at the top end only. I lost dynamic range in the red/green channels because the HL protection methodology wanted to make it NOT clip to a single channel.

Once you understand this, you understand how to fix this in post if you need it. The same is true with complex colors that have different intensities across multiple colors. One of those colors is going to saturate first NOT all of them most of the time. So, you clamp down as ONE of the colors reaches over-exposed as to prevent the color shifts if one channel blows out before the others. Basically, ratio speaking, again, to make a more gradual and pleasing blow out, they all change in that protection zone to get a better look.

I'm not saying the process is absolute simple, because ALL three channels are independent, all three can clip on their own. But, it's not a clipping of the sensor problem, it's a clamping problem on how to handle that blowout.

Also, depending on the sensitivity of the B/R/G channels they do not blow out at the same time either. This is why with a very flat light source, you end up seeing one color first always when the blowout happens. HOWEVER, that said, you can fix that. You just apply a curve to the response to flatten out the response.

And I think you don't understand what these camera tests are showing you. If you have 27 stops, the Canon cannot see more than 12+ either. And they are not present in the highlights as well. When someone says a camera shows 12 stops, it's 12 stops. That's it. If you look at most of the tests, you will see the crushed blacks and rolled off HLs in those 12 stops. The protection INCLUDES those stops as well. Just look at how the stops are not linearly displayed vertically. This may help you understand what those numbers mean.

It's the same reason people were having trouble at first trying to expose shots on the AF100 with regards to the 7D. If you exposed the skin at the same IRE, the content above and below that was TOTALLY different. Those spots where the skin was at, while looking IDENTICAL between the two in terms of luminance, were not the whole story. The curves in each camera were NO WHERE NEAR the same.

This is why it is hard to compare compares with different curves. How the image maps, which is almost always no linearly, is different. Further, the amount of headroom and tailroom above and below that same IRE measurement are different. Some cameras give extra to the highlights, some give more to the shadows. BUT, that doesn't mean you can't get the same results on both. YOU can push that midpoint up and down in post. So, if you want more HL room, you can change your exposure to compensate for that. (granted the noise matters on your camera, and will be a limit). This is what I had to do with the AF100 to get the shadows where I wanted them.

 

[BobbyMurcerFan]

I went to the C300 event at Paramount and saw it's footage projected on the big screen. Compared to an Epic, those images looked noisey, blocky and yes blown out. There was simply no comparision.

 

[Rick Burnett]

I'm not sure if I understand Rick, either. To me, the sensor does matter. Seems like GIGO theory: garbage in, garbage out.

If one sensor records more information than another, then post can't make the second one equivalent to the first.

If two sensors can measure let's say 13 stops of light, they both are recording the same amount of information. Granted, DR tests have some subjective questions about the bottom end (ie, how much a stop is out of the noise floor, and really, what that noise looks like).

But, if two sensors clearly see 13 stops, then you can be pretty sure they are going to be able to capture the same amount of luminance range. If the Epic does 13.5, it would stand to reason that it captures more information that the C300.

It's not the sensor that handles the highlights. The sensor is an array of photodiodes basically. What that sensor sees is in terms of how much capacity the photodiode can hold. Here is a simplified circuit for one:

http://en.wikipedia.org/wiki/Active_pixel_sensor

The sensor itself is pretty basic in terms of what is there. Of course, the row/column and supporting circuitry including the A/D converters add a lot of complexity, but from a highlight perspective, this doesn't matter. There isn't anything in the sensor that is dealing with this. If you have enough room in your photodiode, then that is what is important.

Don't confuse this with other aspects of the sensor, like active area, microlens on top, color filter array, quality of the amplifiers, etc. These ARE all important to the image. VERY important. But when it comes to hilights, you are talking about being so far from the noise floor, and with sensors having 12+ stops of light in DR, this isn't the problem. The problem is in the post processing and how it handles ALL the channels with regards to 1, 2 or even 3 of them approaching clipping. Canon has a great algorithm, no doubt. It might even be patented for all we know. But, it's still in the digital domain that this is happening and that means you can do this with other sensor information in post. How much you can do this will depend on the quality of the codec and the curves, no doubt, but with EPIC and it's bit depth, I wouldn't be worried.

Every conversation I've seen on DVXUser where a user posted something with horrible highlights, another user was able to help them fix it. It's all in understanding what is going on.

Again, I will repeat because I think it is necessary, I would rather have this in camera as well, myself, personally, but what I am saying is where the problem is coming from is not the sensor, it's the digital processing, and that can be at least compensated for in all cases I have seen.

 

[BobbyMurcerFan]

Isn't that contradictory?

See you you're saying. First you say, 'if you want a smoother roll off, you can fix it in post, its a digital process that happens after the sensor'. This would have any basis on reality if the camera was able to record every single exposure of any given situation, from a dark shadow all the way to a bright could in the sky,let's say 27 stops! But the camera has only 13!

How will you fix this in post?

From stop 13 over, you're white, fully exposed! In the Red sensor, by stop 14, you're clipping!

Are you really incapable of composing a shot w/ 13 stops? 99% of the time, extra stops are going to be in the extreme shadows or specular highlights anyway. Guess what, no one is looking there anyway.

It's like wanting volume control that goes to 11 when you moving it past 8 already hurts your ears.

 

[Rick Burnett]

Rick,

I just want to reiterate what I said in post 211, that the line between RAW and image processing is not nearly as clearly as we'd like to think. There is significant image processing that all Reds do in-camera. RAW is not a "raw" stream off the sensor. Some processes, like sensor noise denoising, are certainly done in-camera. Others, it's not so clear. Essentially, RAW is image information arranged exactly how the pixels of the sensor are arranged; that's it.

Whatever Red does or does not want to do in-camera is totally up to them. They could boots the blue channel, lower the green and apply some extra noise reduction, etc. As long as the recorded image information is in a 5K Bayer pattern, it's RAW. Same idea goes for ArriRAW. And similarly for Sony's F65, which from what I've heard takes image processing to a whole new level in terms of extent.

So there is simply no standard as to what you can or can't do to an image and still be allowed to call it RAW, except of course that it has to be arranged as the sensor is.

As for debayering, I believe Red encrypted .R3D b/c Cineform pissed them off by coming up w/ alternate algorithms that they felt were superior. Nonetheless, it does seem that debayering itself should not affect highlight handling much at all, but then again who knows.

My personal feeling is that the reason why DSLR's look so good is because they are using image processors that are a generation or two ahead of most video camera's. How accurately one can reproduce what a DIGIC IV does in Premiere, Avid MC or Resolve is beyond me. But I imagine that there is a lot of subtlety going on, including things like scene detection.

Is there some extra mojo that the Alexa has? Probably so. Does it mean that the Epic is so off that hobbit faces need to be clown painted? No way. Are people probably misusing their Red Ones and Epics to get inferior results? Sure.

I'm totally open to the idea that the Epic needs special care when shooting skin tones. But I'm also open to this being blown way out of proportion. I've heard from more than a few people that both cameras can be seamlessly intercut.

Absolutely. RAW is not a defined standard, it's what any manufacturer decides to call RAW. And what you get from one cameras RAW is going to be different from another.

There is a TON of information pertaining to sensors I have not talked about, because as you point out, there is a lot of different magic that can happen. Look at the quantity of settings in the AF100 or FS100! The sharpening algorithms, etc. There is a lot of different tools going on in real time in the camera to modify that image to make it more appealing.

I've listened to people talk about Sony and their noise reduction circuitry in the F3 and FS100, these are all post process techniques, after acquisition, but more than almost always likely before compression into the codec. Have you ever done long exposure on a CMOS sensor for pictures through a telescope? Looked at hot pixels that have faster response curves than other pixels on the sensor? ALL of these things make a difference. ALL have to be accounted for depending on what your shooting requirements are.

That said, the question becomes how good of a source can you get out of the camera that you can do post processing OUTSIDE the camera. Sharpening for instance, why bake it in? You can sharpen in post if you want. When making sharpening circuits for a digital camera, you've got to consider power / speed / cost. ALL three of those things matter. Power for battery, speed to make sure you can modify a frame in the time it takes before you need to process another one, and cost makes sense obviously. Just because something is implemented in hardware doesn't make it better. It will make it faster, no doubt, but there are some things better left to post. I think neat video does an AMAZING job reducing noise in a scene. How does that compare to the noise reduction in camera? Certainly better than what is built in to the 7D.

With a camera like the Epic, the question becomes how good is the material coming out?

Some may say that the codec in the C300 is no issue, but that is clearly not always the case. I've seen people post comparisons and showing issues where that codec breaks down. Same thing with AVCHD. Sure, I've used it with doing special effects work, and I can tell you right now, the macro blocking of H.264 is a pain in the neck if you start applying ANY curves to the image. I am not saying that what the C300 is inferior or superior. I'm just saying compression has it's price. The less you take away information, and that is what is happening, the more you have to work with if you need it.

That's Red's philosophy. Love it or hate it, they STILL give you more than you are going to get with the C300. Is all that data usable? That's a GOOD question. But you get to decide that. I am NOT buying a Scarlet-X. I don't need all that extra freedom because the cost for it isn't worth it for me. I'm not saying it isn't valuable, but for what I do, I think the extra cost for media/drives/conversion is overkill.

Why do you think these comparisons are so hard for people to talk about?

- different algorithms for color handling/curves in camera
- different algorithms for image enhancement in camera
- different algorithms for compression in camera
- different algorithms for color handling/curves in post
- different algorithms for image enhancement in post
- different algorithms for compression in post

and ALL of these wrought with artistic, non-technical, decisions as well. THAT is why it is hard to compare a lot of these things pertaining to color/DR/SNR.

At least rolling shutter is simple and easy to measure!

 

[Piolet]

Are you really incapable of composing a shot w/ 13 stops?

You didn't have to word it like that. These little jabs is what gets both sides riled up.

The point about the sensor is if one sensor records more information than another, say 14 stops versus 10 stops, then the highlight characteristics will look different on the 14 stop sensor. By highlight characteristic I mean the tonal response as it goes up to the limit of the sensor. So given the same processing, whether in camera or in post, the image from the 14 stop originated material will not look the same as the 10 stop originated material.

 

[Rick Burnett]

See, why do you have to word it like that? These passive aggressive little jabs is what turns a thread about Canon interviews into 28 pages of who knows what. I'm not excusing Nuno's outbursts either, but comments like this aren't helping.

The point about the sensor is if one sensor records more information than another, say 14 stops versus 10 stops, then the highlight characteristics will look different on the 14 stop sensor. By highlight characteristic I mean the tonal response as it goes up to the limit of the sensor. So given the same processing, whether in camera or in post, the image from the 14 stop originated material will not look the same as the 10 stop originated material.

Yes and no. I only say that because if you don't care about the shadows, you could just expose for the highlights and get the same response. Where those 14 and 10 stops are applied can be different in how they are lined up. Remember that the response curves in cameras are not linear as well so they can be compressed and expanded BUT you are correct in that 2 stops of information is coming from some combination of the top, bottom or both.

But, the yes part, pertaining to the earlier question, if the epic is 13.5 and the C300 13.5, then they should be able to obtain the same shadow and highlight detail provided the noise signature is acceptable in both. They maybe the same noise magnitude distribution wise, but, that doesn't mean both are desirable to the same level. Not taking into account any destructive processing like the codec.

 

[indiawilds]

I went to the C300 event at Paramount and saw it's footage projected on the big screen. Compared to an Epic, those images looked noisey, blocky and yes blown out. There was simply no comparision.

I value your opinion since you have seen the actual screening. Was it in the main theater?
I read the main theater was better than the projections in the other two screens.

I saw this post by Vincent Laforet. quoting him here: "Two of the most exciting things about the Canon C300 are the sharpness of the image, and also the "organic" or granular structure of the noise pattern the sensor produces" One can see the detailed blog post here: http://blog.vincentlaforet.com/2011/11/09/canon-c300-organic-noise/

He uses the word "organic" for noise. I have seen lot of complains about Epic being noisy. Need to dug up where all I had read it. Philip Bloom says about Scarlet "As it’s the same chip as the Epic i doubt it will be as good in low light as the Canon C300 but it will be upgradable to the new Dragon chip in 2012." He has an Epic and says Scarlet has the same sensor as Epic, so he doubts that it will be as good in low light as Canon C300. The link is here: http://philipbloom.net/2011/11/04/newcameras/

From my perspective, the good thing about Epic is the high resolution and the footage (especially in wildlife) becomes kind of future proof for use lateron.

Frankly speaking, I am looking for an objective discussion to know the relative merits of the C300, I hope this thread can stick to that without arousing passionate responses.

Cheers,
Sabyasachi