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

Status
Not open for further replies.
indiawilds said:
I don’t agree with one of the moderators saying that the C300 will only be of interest for people who has a ASC attached to their names. If you think so, then please wait and watch.
Click to expand...
That's not what I said. I said "if you don't have ASC after your name, then you're not in the target market for this camera." Meaning, they aimed and designed (and priced) this camera towards the needs of the Hollywood cinematographer. That doesn't mean that other people won't find it interesting, of course -- I mean, the Red One was designed for the same market, and thousands of non-ASC, non-Hollywood shooters bought it too. And I'm sure that many outside the target demographic will buy the C300 (or at least rent it).
 
Yep, I'm leaning that way, but I'm hoping I come to my senses and realize that my 5DII with a VAF-5DII filter will be enough. Thank god there's a 3 month cooling period that this cam is unavailable to buy. After the C300's release, I've also imposed a 6 month waiting period of reading all the reports and tests on the forums. If I can find my login and password, I'm going to give Red forum a shot of talking me out of it. I'm sure they'll do an admirable job. :-)
 
GH2user said:
It's a great codec. BUT IT'S FROM A $3000 CAMCORDER.
Click to expand...
No, it isn't. It's from a $34,000 broadcast camcorder. That codec was first used in the Sony PDW700 and is now used in their PDW800, among others.

Canon stuck the same codec in a $7,500 camcorder. Then they further put it into a $3,000 camcorder. And I have it in a $2700 NanoFlash too.

So is your logic that Sony should now drop the price of the PDW800 to $3,000 because Canon put the same codec in a $3,000 camcorder?

And who cares really? It's not that good of a codec anyway. It's still a 1995-era, 8-bit, long-GoP MPEG2. It's not in the same league as AVC-Intra or HDCAM-SR or ProRes or CineForm or RedCode Raw or any other truly modern codec. Yes it's accepted by the BBC for broadcast, I'm not knocking that, but -- it's hardly cutting-edge technology and certainly not anything to get all worked up over.

And it is not, in any way, a determining factor in what the camcorder should be priced at. The same codec is used in the $3,000 XF100, the $7,500 XF305, the $20,000 C300, the $28,000 PMW500 and the $34,000 PDW800.

I wouldn't even pay $3500 for the C300 simply for the fact of a lack of over-cranking alone.
Click to expand...
Then you have clearly decided that you are not in the target market for this product. This is good news: it means you no longer have to waste your time, energy, or breath on this product, you can move along.
 
Nuno Dias said:
That's fine, but I know that when you have a clipped image, no matter what you try to do in post, you won't get rid of it, because the sensor wasn't able to record any information on those highlights, so you can't make up for something that is not there. Unless, of course, you replace the entire sky (or whatever it is that is clipped) but I surely hope that that's not what they did in the Canon shorts...

When I first posted frames from the C300, I specifically choose shots that had hard highlights in them, either the sun or production 'lights', meaning, that if the camera is easy to clip, in those scenes they would most certainly clip, and they didn't

But I'm not saying this is a fact, just what I saw from those shorts that's all.
Click to expand...

I'm sorry, but your understanding of why digital CMOS sensors clip has NOTHING to do with Ansel Adams or his zone system, period. Further, you couldn't be further from the issue if you tried.

Let me ask you this, why is it that out of the box, the 7D, 5D, T3I, 60D all handle highlight clipping better than even the AF100 or FS100, or in my opinion, even the F3? I'll give you a hint, it's not the sensor.

When film reaches overexposure, the photons hitting the film have more of a chance of NOT hitting a silver halide grain. Therefore, the highlights roll off because the sensitivity of the film goes down as you approach completely exposed.

With a CMOS sensor, any of them, you don't get this top end. The photons can continue to affect the charge of the photosite until it is saturated. The difference between the sensors, if we hold dynamic range to be the same, and sensitivity, is the clamping algorithms used to affect what happens when an overexposed condition occurs. Not that there isn't technology with multiple transistors that can affect the sensitivity curves, those are going to create a totally different non-linearity as well, but I see no difference in those in terms of highlight handling.

If you take BOTH the AF100 and FS100 and dial the saturation down until you match the performance of say the 7D, which I have, the highlight handling is AMAZINGLY improved.

Now granted, I will say, I prefer that the camera do this and I not have to worry about it in post, to say you cannot fix it in post is wrong. What you have to understand is what is happening in the camera. First, you have to look at how fast each of your channels overexposes, then, you have to take away some of your dynamic range to handle what happens when you get close to an overexposure. As most of us with struggled with the AF100 and FS100 can attest that once you understand what is going on, fixing it in post is definitely possible.

But let's take it a step further, have you studied the overexposure between say the FS100 and the a Canon sensor like the 7D? I've shot blue LEDs before to try and understand the science that Canon was using. On the FS100, as expected, it clipped to an ugly blue. On the 7D it clipped to white. But, the truth is, there was no white on that LED. So from a scientific level, the FS100 is being more accurate. Not that it is a good thing, but it helps to understand what you need to do to fix.

As the blue channel is reaching fully exposed, in the color science after the sensor, you know that you need to ramp up the red and green channels as well to handle the clip gracefully. You do this by sacrificing the DR of the red and green channels at this point as the blue approaches a clip.

And to that, it's all curves. That's it. If you want to do this in post you can.

Again, I absolutely agree that I think Canon has some GREAT HL clipping algorithms, but from what I have seen and my research, it's not the sensor. AND it can be saved in post if you know what you are doing. I've seen MANY people here even show their examples of fixing ugly highlights on the AF100 specifically.

Now pertaining to the detail in the shadows, it's all about DR. I've seen PLENTY of material posted with even the Epic that blew me away detail in the shadows. Even the early 'Leo' video with the epic shooting in near dark was great. Not that I don't think the shadow handling of most cameras today are good, including what little I have seen of the C300.

Now, a better argument would have been the sensitivity of the C300 being what it appears seems to truly be more sensitive than a lot of other cameras, and in that, with natural lighting, you have more of a chance to get more shadow detail, but, it doesn't preclude the other cameras from getting that shadow detail, it's just the shadows would need to compensate with a bit of light from the sensitivity difference of the camera. Given the C300 has 13 stops then it should be very much on par with the Epic which I've read at 13.5, once graded properly, you would be able to get the same range (if Canon is being conservative with their 13+ and I'll give them the benefit of the doubt).

And I will say this, because this was a HUGE learning lesson for me. When the Nikon D90 came out, I watched EVERY video I could find online and was TOTALLY amazed at how great the video looked. Then I bought the camera. Then I discovered the worst rolling shutter I have ever seen, a huge lack of exposure setting because it couldn't be set manually, and worst of all, a skipped line on the sensor like every 7th line or something like that. I even went back and watched the videos and realized, given they were compressed for the web, it had hidden the line problem. At the time NO one had talked about the exposure control issue, and I didn't realize how bad the rolling shutter was because no one was highlighting that issue. Everyone was too enamored with the fact they could shoot video through an interchangeable len.

Now, I am NOT comparing the C300 to the D90, what I am comparing is the experience. I learned a valuable and expensive lesson about believing the performance of a camera from videos I've seen online. I'm also not saying that the C300 might not outperform beyond what you believe as well, because when I saw much of the sample videos from the AF100, I was not impressed. In fact, I've even seen the test videos at a Panasonic booth on a Panasonic monitor with THEIR AF100 and was not impressed. Yet, when I had my AF100, with a few changes to the default settings, I had a beautiful image that performed excellent and that the people I've worked for absolutely loved. People have different requirements, different tastes.

But nevertheless, the sensor isn't the issue with the HL handling, it's the clamping in the color science and it CAN be fixed in post.
 
I think there is some misinformation here about clipping

[EDIT: Rick Burnett posted while I was typing, and he and I are maybe saying about the same thing]

First let's define "clipping." Clipping has to do with the pixels on your TV set, not the camera. Clipping means that two things have happened: (1) that some of the pixels on your TV set are at full white (the brightest they can be, the highest number that the video signal counts to, 255 in 8-bit color) and (2) that there is not much of a gradual transition to these white pixels from the pixels around them. It doesn't look right, or natural. Film, of course, transferred to video will have some white pixels, but the transition to them is usually gradual, not abrupt.

It might occur to you, then, that with a little grading you can "ruin" something that came off of film, making it clip. Actually, did you know that film clips, too, just like video, except in the opposite direction? Film clips in the shadows. Film photographers talk about "protecting the shadows" just as video takers talk about "protecting the highlights." It will all become clear if you think for a moment about a physical piece of film. When a frame of film is developed, the emulsion dissolves off of the frame, depending on how bright or dark that part is supposed to be. Pop Quiz: To represent a totally black object, the film frame would be (A) opaque (B) clear. Answer: B, clear. When part of the film is completely clear, with no emulsion left, light shines completely through, unabated. "But that's completely white," you say. Right, but this is a negative, remember, so when it is printed to another piece of film for viewing it will be completely black. As clipping can also be defined as when there is no texture in that part of the scene (it is completely white or black) then you can see that when you have no specks of emulsion left on the film there can be no texture. So film and video both clip but at opposite ends of the spectrum --- which makes sense when you consider that film is negative. On a side note, slide film, which is positive, responds similarly to video and clips in the highlights.

Now that we've got all that out of the way, we can move on to the notion of some video sensors clipping less than others. This doesn't happen. All video sensors respond to light linearly. That is, as the amount of light rises, the voltage will change the same amount. No sensor has pixels where the voltage starts to taper off. And so, they all hit the voltage that represents "white" with the same gradient --- natively. Notice that I said, the voltage that represents "white." Many of you know that photosites are like buckets, and they have a maximum capacity. To drive home the likeness to water, it's even called "full well capacity." My point is, is that the voltage that represents full white is not necessarily the same as the voltage that corresponds to full well capacity. It might be well below that, depending on the ISO you choose. Really, it ultimately ends up as a choice in the grading. Grading is simply mapping values from the camera pixels to values in the viewer's screen, and you can map them however you want. You could even map it oppositely, if you wanted a negative image, for example.

At this point, if I've explained it all well, and you haven't fallen asleep yet, it should be obvious what my next point is. The fact is that with video it's up to the image processing to have any sort of gentle roll-off into the highlights. No chip will give it to you. It will always give you a straight linear transition into white, which is usually too harsh and is called clipping. The application of a gentler roll-off can happen at any point after the signal is read off the chip, in post or within the camera itself by its own pre-applied looks (if not recording RAW).

There is a catch, however. If you're going to fix it in post, what you're basically gonna do is raise the brightness of the mids without raising the highlights. This change in relationship between the mids and highlights, to where they are no longer linear but tapered, is basically the same as saying you've applied an S-curve. It's a curve in the shape of an S (because, as I had failed to mention, you're also making a tapered transition to the darks from the mids). The catch is, if you're going to raise the mids in post, that means they must be darker when you're first recording them than what you eventually want them to be. That means that unless your on-set monitor is doing some kind of pre-translation for you, it's going to look like you're underexposing everything. The highlights will look right, but everyone's faces, for example, will be in the dark almost (probably two or so stops underexposed, because you want to raise them by a couple of stops in post).

So the catch is, is that if you expose so that it looks right on the set, it will look overexposed if you try to apply curves in post.

Which brings us to the final wrap up, why do some cameras look more film-like? Well, as we have said, because their images were graded correctly afterward. But as you may know you can't make all cameras look the same, no matter how good a grader you are. The reason is, is that sensors do differ in one regard: dynamic range. Some have 9 stops, some over 13. The more dynamic range you start off with, the more you can stretch the curves in post to make it look more like film. Sensors with less dynamic range will start to show more noise when you bring up the mids. You also need enough bits to avoid banding. I'm not sure how much, but if the camera's own image processing hasn't helped you out a little bit by applying some curves before recording to 8-bit, then you'll have too much stretching left to do. If you're dealing with RAW, I think you'll want at least 10 bits. This is a function of the analog-to-digital converter, not the image sensor itself. You could slap an 8-bit ADC on a sensor with 12 stops of dynamic range, and you could put a 16-bit ADC on a sensor with only 9 stops (It's a waste, either way). I would add, if you still haven't given up on me and are still reading, that if the ADC converts logarithmically instead of linearly, then you need fewer bits. In fact, 10-bit log is like 19-bit linear.
 
Last edited:
That was a fantastic explanation. Thank you very much Rick and combatentropy!

Although this can easily turn into one of those old discussions of art vs tech and which should give up to wich. And as Rick said, I too, think there's something about the C300 processing highlights algorithms that clearly work in favor of more pleasant images, regarding of how they got there to do this. Same as with the Alexa, clearly.

It just bugs me of why a Red camera is so easy to give you this ugly highlight without a more pleasant rolloff or why not come up with some sort of processing matrix to make this more easy on the eye. Because if a much cheaper camera can do this, why not them? Surely, image quality/beauty must be primordial to them, no?

Anyway, combatentropy, thanks so much for your wisdom on this, I only wish one day I can understand this technology with such clarity!
 
Nuno Dias said:
It just bugs me of why a Red camera is so easy to give you this ugly highlight without a more pleasant rolloff or why not come up with some sort of processing matrix to make this more easy on the eye.
Click to expand...
Because a Red DOES NO PROCESSING.

That's the whole point of a Red One.

It records RAW.

It doesn't process the signal in any way. It leaves all processing to post.

If you don't like the way Red's footage looks, it's because of the way the individual colorist processed it.
 
Barry_Green said:
Because a Red DOES NO PROCESSING.

That's the whole point of a Red One.

It records RAW.

It doesn't process the signal in any way. It leaves all processing to post.

If you don't like the way Red's footage looks, it's because of the way the individual colorist processed it.
Click to expand...

amen.
 
One thing about film is that it is inherently non-linear.

Development does not, in my experiance dissolve any of the emulsion but converts silver halide into metalic silver which forms the grain structure.

I'll say no more, but reffer intersted readers to "The Negative" by the Blessed Ansel Adams

Dave
 
Barry_Green said:
Because a Red DOES NO PROCESSING.

That's the whole point of a Red One.

It records RAW.

It doesn't process the signal in any way. It leaves all processing to post.

If you don't like the way Red's footage looks, it's because of the way the individual colorist processed it.
Click to expand...

I just want to clarify one thing, even RAW cameras do quite a bit of processing. What they don't do is bake-in a demosaic and white balance to what they record, but they do do this for the monitoring signal.

Even the RAW file is not recorded untouched right off the sensor. For example, when you black balance, you create base level sensor data that is used in-camera to denoise the image up-stream of recording.
 
Last edited:
Barry_Green said:
Because a Red DOES NO PROCESSING.
Click to expand...

That's clearly in your head isn't it? Its called Red-Raw for a reason: Its not raw!

And you're right, I don't like Red's image, not because I grade it bad, just personal taste, can I have it here? And listen, you're a moderator, don't fuel a fire that was already extinguished! I posted images shot by Steven Soderbergh, not just mine! But that miserable clipping was all over both mine and his! But maybe Steven's colorist should get some schooling from you... apparently.

Further more, Red's specs are in the very least an alternative truth... I've read Alexa's white papers, can't on the Red... but if you like it, fine. Power to you. But don't be inconsiderate when someone doesn't.
 
This is the beauty of RAW for those that know how to use it, and the downfall for those that don't. It's gives you absolute freedom to do what you want, but at the same time, you have to know how to use the tools to make it shine.

Look at photoshop, how many people use photoshop and you can tell all they did was drop some built in filter by itself and marvel at how amazing it looks. Yeah, I played with emboss in like 1985 :) The true master of photoshop is where you see something, as a seasoned photoshop user myself, and it BLOWS YOU AWAY. This happens often to me. I've seen pictures taken with the cameras I own processes with the software I own and I have NEVER been able to do something that looks the same.

That's the art.

The same thing is true with the C300, the Epic, the Alexa, the F3. Some people will argue that the Epic will give you the most freedom with your image. I agree with that. At an expense, no doubt, even if in just space on a disk. There are people that shooting on an Epic or Scarlet-X will be a waste of money. They'll just use some preset, and/or throw it through magic bullet. Not that there is anything wrong with that, as the consumer of client will be the ultimate judge, but all that extra you are spending on RAW is probably going to waste.

I absolutely agree that the art is important, but art for just the sake of art is subjective. When I write scores for films, I put my heart into it. I do the best I can do, but I do it within a budget and toolset that makes sense to the clients I work with. Most people making money off their work HAVE to look at the business/technological aspect as well.

That's why the DSLR revolution, and I am going to call it that, hit hard for a lot of people. People were coming in with a camera 1/10 the cost of what they had producing images that while on many levels had technical inferiorities, where subjectively more pleasing to people. Suddenly it looked like "hollywood" and people loved it. They didn't care about some noise in the trees when they were looking at something else in the scene.

THAT is the reality check. The 'filmic' look people try to join with technology is not about technology. It's the aesthetic of the moving image. Just because a camera has a horrible rolling shutter does't mean you can't use it in situations like a talking head that never exhibit the problem. It's about knowing where those limits are in the technology and staying within them.

Let me give an altering opinion on the C300. Scroll down the page and look at the screen cap of the lady looking to the left:

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

That is a clip from the C300 in one of the films.

Look at the blocking in the face. Look at the softness in the image. Look at the banding in the purple/white blog to the left of here. Look at the grain in the noise in the background that is very similar to the AF100. Now, I have NO CLUE how this image came to be, and it certainly is LOWER quality than other samples I have seen, but don't you think it would have made more sense to grab a shot from the camera to show on the main page that actually looked good from a quality standpoint?

Compare this to this:

http://tonacitran.com/wp-content/uploads/2011/06/invinciblestill.jpg

Which red epic. Again, color grading should not be compared as it's an artistic choice. I find the Red Epic image resolution superior in those two shots. I also think the HL handling of the metal tabs on her head to look perfect.

I imagine all of us could go back and forth with any of the examples finding better and worse from ALL cameras talked about.
 
Nuno Dias said:
That's clearly in your head isn't it? Its called Red-Raw for a reason: Its not raw!

And you're right, I don't like Red's image, not because I grade it bad, just personal taste, can I have it here? And listen, you're a moderator, don't fuel a fire that was already extinguished! I posted images shot by Steven Soderbergh, not just mine! But that miserable clipping was all over both mine and his! But maybe Steven's colorist should get some schooling from you... apparently.

Further more, Red's specs are in the very least an alternative truth... I've read Alexa's white papers, can't on the Red... but if you like it, fine. Power to you. But don't be inconsiderate when someone doesn't.
Click to expand...

Dude, what is your problem? There is image processing going on, but Barry's overall point is true. Red's RAW format gives a lot more control in post, which means when you see crappy images you can't necessarily blame the camera. Actually, the Epic takes such amazing images, that you can't really ever blame the camera.

But if you've taken the time to read Arri's whitepapers, then you should also be familiar with the fact that the Red One was released essentially as a work in progress. So depending on sensor version, RedCine build and settings, post decions and of course DP, you can get any variety of looks... from awesome to horrible.

The Red recording RAW gives you enough rope to hang yourself. The Alexa recording ProRes tries its best to keep you off the gallows.
 
Last edited:
BobbyMurcerFan said:
Dude, what is your problem? There is image processing going on, but Barry's overall point is true. Red's RAW format gives a lot more control in post, which means when you see crappy images you can't necessarily blame the camera. Actually, the Epic takes such amazing images, that you can't ever blame the camera.

But if you've taken the time to read Arri's whitepapers, then you should also be familiar with the fact that the Red One was released essentially as a work in progress. So depending on sensor version, RedCine build and settings, post decions and of course DP, you can get any variety of looks... from awesome to horrible.

The Red recording RAW gives you enough rope to hang yourself. The Alexa recording ProRes tries its best to keep you off the gallows.
Click to expand...

+1,000,000
 
This is all starting to get a bit absurd. Here are tons of great cameras to choose from. Look at the guns we get to select from these days! But NONE of them will make ANYBODY a better film maker based of specs and white papers. I see crap work out of every camera and I see great work out of every camera. And guess what? Only we, the shooters give a flying crap about clipped highlights. Just try and explain clipping to your wife or girlfriend. You will have as good a chance getting the point across to your dog. Why, because they only care about the story when watching something, as do the rest of general viewers.
Why don't we start looking at these cameras as a blessing? Just how long ago was it that the HVX200 or the DVX100 was the great cinema tool? Look how far todays cameras have come. Come on, I've seen things shot on a Canon T2i that are more pleasing to the eye than some things shot on RED Epic. It's who is behind the camera that makes the real difference. Newer cameras will come out before we even get use to todays cameras. Just grab one and shoot. Hell, I just picked up a Canon XF105 for run and gun and I'm blown away by it. And!!!!!! Stop fricken pixel peeping 8 inches away from your computer monitor! Who watches anything that close. When you evaluate a compressed vimeo or youtube test, BACK OFF YOUR COMPUTER AND WATCH IT LIKE YOU WOULD IN THE BAR!
 
This guy exposed as I was talking about (watch the video): http://reduser.net/forum/showthread.php?59024-TOIA-RED-MX-Latitude-test-2-HI-CONTRAST

You can see he protected the highlights and then raised the mids in post. It could look more film-like if he also raised the contrast of the mids and maybe desaturated a bit. But at least there's not the tell-tale highlight clipping.

Here's another example of how you can make something look a lot like film if you grade it right: http://vimeo.com/17424690
 
BobbyMurcerFan said:
Actually, the Epic takes such amazing images, that you can't really ever blame the camera.
Click to expand...

Do you realize this how abstract this stamtent is? Amazing images, on who's eyes? Do you mean, technically amazing images? Do you mean that as soon as you grade them and they fall to pieces its not the cameras fault anymore, but the Dp's or the colorist? So, in your line of thinking, one has to use the files from the Epic, as they are out of the camera and straight onto cinema screens!

The Epic can only ever dream of looking as gorgeous and perfectly structured as the Alexa does or the C300. Give me a 5D any day over any Red, that's how much I trust Red...

But to each is own, that's the beauty of this all industry. Because if Red was the only manufacturer in the world, this would be a sad, sad industry. I would probably be much happier baking cakes then.

By the way, I don't give Red another 5 years. Probably much sooner than that.
 
After letting the dust settle... and getting over my "spec lust"... i'm quite looking forward to the C300 coming out and hopefully getting my hands on one. The form factor looks great, to me, and the image is sublime (in the right hands, of course). I get the feeling that you will have to really nail exposure when rolling, to get it looking as good, but most of us should be used to that anyway.
 
Status
Not open for further replies.
Back
Top