ER on RED...

[David Rasberry]

I wonder if the Vantage rehoused Leica glass would cover FF35.

 

[Don Woods]

I thought the look was consistent with the pervious episodes. I really love how the Red renders skin ones looked great. Look forward to the last 6 episodes.

 

[M Most]

I think that's a slight oversimplification. In ample light, one can get the advantages of the larger sensor without any change in DOF by stopping down.

Considering that much high end television production is currently lit to about a 2.8 on a recording medium that's at least one stop faster than a Red (i.e., 500 speed film, for the most part), I consider it unlikely that cameramen would choose to shoot a Red at, say, 1.7 or less, which for what you're suggesting would be necessary if the effective shooting speed of the new sensor designs is not significantly faster than the current model. And even if it's technically possible, the choice of a 2.8 stop - or, in some cases, a 2.8/4 split - is based as much on what's generally considered the sweet spot of current lenses than the need to use fewer lighting units. Of course, if the new designs yield a sensitivity that's closer to 500ASA than the current 250-350 (depending on who's rating it), that would be a different story and a potentially good reason to adopt the newer technology.

 

[Blair S. Paulsen]

One stop from total domination

One stop from total domination

IMHO the switch to digital capture for most episodic TV is simply a matter of time. The RedOne will be chosen by many of them. For RED to garner the lion's share of that market greater sensitivity and/or better S/N ratio (kinda the same thing in practice) is the key.

Even when shot on sound stages with plenty of power, instruments, ventilation, etc the cost of bigger lighting packages matters - particularly when a lot of pages have to be shot per day and when camera moves make it harder to hide lights - not to mention location shooting.

Because of the S35 sized sensor (which allows the use of Optimos and other high end optics), oversampling, RAW color space and manageable files sizes made possible by RedCode the RedOne is a popular choice for DPs. With another stop of useable sensitivity it would own the episodic market. Just my 2 cents.

Cheers-

 

[Roberto Lequeux]

I think that FF35 is really going to kick a lot of ass, more DR means better everything. And you do not need to say goodbye to your PL glass. This has been talked about all over, and even here several people already mentioned it, including Jim. You can shoot S35 on the FF35 "just" to get the additional DR while keeping your S35 workflow + glass... Bigger photosites mean more sensitivity or at least less noise which also translates to more actual DR, and even a bit more sensitivity sort of speak as far as I comprehend, plus more light period... I mean... it is going to rock and there are no drawbacks at all. Period. I wish people could understand that.

 

[M Most]

IMHO the switch to digital capture for most episodic TV is simply a matter of time. The RedOne will be chosen by many of them. For RED to garner the lion's share of that market greater sensitivity and/or better S/N ratio (kinda the same thing in practice) is the key....... With another stop of useable sensitivity it would own the episodic market. Just my 2 cents.

Blair and I agree on most things, but not on this.

Those looking to technical specs as primary reasons for industry acceptance, particularly in high end episodic work, are looking in the wrong place. The F900 is arguably the most successful 24p camera ever used in television. It's used to shoot almost every multicamera sitcom on the air and has been for a number of years, and that represents a LOT of shows. But the technical specs are nowhere near the film stocks that it essentially replaced. What got it acceptance was availability, adaptability, strong support from major rental houses, a simple, accepted post path, and reliability. Red, to this point, has fallen short primarily in 2 of these areas: major rental house support, and a simple post path. Increasing dynamic range is very nice and very forward thinking, but it's not in itself going to lead to much wider acceptance. Personally, I think what's really missing is a device to do real time or faster than real time file conversions (complete with metadata, of course), but at this point, only Red can really build such a device because the compression, encryption, and file format is proprietary. That would really address the post bottleneck that everyone is doing their own science projects to solve today. Nobody is going to "own" the episodic market because there are numerous choices, each of which is more appropriate for some things than others. There will be a healthy mix of equipment in general use, and yes, Red will undoubtedly be one of those. But it won't be the only one.

 

[Jeff Kilgroe]

Sorry for not giving the warning, guys. Long story.

I'm just bummed because my DVR somehow forgot to record the show!

I didn't know to tell my DVR to record the show. I'll take a look this week.

 

[David Birdy]

Good Points Mike.

I just made the connection of how I know you....I made a trip to Miami
with Al Francisco,(DP) and Jeff Hilligas,(Creative Director) to view samples of the Movie "Dale" we were working on for NASCAR in 2006.

It's great to see you here!

If your Post house needs any Red rental cameras, let me know!

Dave

 

[Daniel Browning]

Thanks for the reply, Mike. I always enjoy your posts here and elsewhere.

Considering that much high end television production is currently lit to about a 2.8 on a recording medium that's at least one stop faster than a Red (i.e., 500 speed film, for the most part), I consider it unlikely that cameramen would choose to shoot a Red at, say, 1.7 or less, which for what you're suggesting would be necessary if the effective shooting speed of the new sensor designs is not significantly faster than the current model.

Interesting.

And even if it's technically possible, the choice of a 2.8 stop - or, in some cases, a 2.8/4 split - is based as much on what's generally considered the sweet spot of current lenses than the need to use fewer lighting units.

Agreed.

Of course, if the new designs yield a sensitivity that's closer to 500ASA than the current 250-350 (depending on who's rating it), that would be a different story and a potentially good reason to adopt the newer technology.

I think that is true right now, immediately. It doesn't require newer technology. It can be proven with raw sensors we have had on the market (still cameras) for years.

The same design with the same technology, in a larger sensor size, will yield lower noise (i.e. higher ASA rating) when used at the same f-stop+AOV.

Newer technology would only build on that advantage. I'll explain what I mean with a RED ONE 2K/4K example below.

The reason why is pretty clear if you stop thinking about f-stops or single pixels, but step back and look at the whole picture, which is:

The total amount of light.

I know it's unusual to think of things that way because we're so heavily trained to work with things like f-stops, which is not the total amount of light, but just "light intensity per area". You need the area before you can know the total amount of light. Another difficulty is the common obsession with the performance of a single pixel in isolation (or in small clumps of "100% crops"). That's the standard unit of measurement, of course, but by itself it does not describe what will be seen on the screen.

The best way to look at it is like this:

Noise scales with aperture in low light.

Aperture, and not sensor size, will determine the amount of noise when light is not ample, AOTBE (all other things being equal).

By "aperture", I do not mean f-number (relative aperture), which is often confused for aperture incorrectly. Aperture is the result of focal length divided by f-number.

Since aperture describes DOF (except for bellows factor), then this is also true:

Noise scales with DOF in low light.

In other words, any two shots that have the same DOF will have the same noise, AOTBE. (In this post, I'm only referring to the DOF projected by the lens, not the second-order DOF factors such as resolution, processing, and display.) Here's an example with the RED ONE:

• RED ONE in 4K mode, 640 ISO, 50mm f/4.0: aperture is 12.5mm
• RED ONE in 2K windowed, 320 ISO, 25mm f/2.0: aperture is 12.5mm

If both were projected at the same size, then both would have:

• The same field of view.
• The same depth of field.
• The same total amount of light.
• The same aperture (12.5mm).
• The same amount of noise.
• The same "brightness".
• Different resolution.
• Different available frame rates.
• Different compression (maybe).
• Different read noise (maybe).

The reason they have so much the same, despite the fact that one is a much smaller sensor, is the fact that the total amount of light is the same. The amount of light is the same because the aperture is the same. The aperture is the same because the f-number is wider (f/2) on the smaller sensor (2K windowed). This causes it to have a higher intensity of light over a small area. Whereas f/4 on the larger sensor (4K) results in lower intensity of light over a larger area.

In the same way, FF35 in low light can have the same DOF for the same amount of noise (light):

• Scarlet FF35, 800 ISO, 80mm f/4, : aperture 20mm.
• Scarlet S35, 500 ISO, 50mm f/2.5: aperture 20mm.

Again, those two setups would have the same AOV, DOF, light, noise, aperture, etc.

If you shot Scarlet FF35 at the same f/stop+AOV you were using for Scarlet S35, then it would have more light and less noise. But it would have thinner DOF.

Of course, the obvious conclusion is this: a larger format with the same technology only increases light (and reduces noise) if you do one of two things:

• Use thinner DOF.
• Add more light.

If you're always in low light, then "add more light" is not a useful option, so the larger FF35 scarlet sensor will only equal the S35 scarlet and not surpass it. Shoot exteriors, though, is different, as one can often use less ND to get more light, so FF35 would surpass it.

If you're already pushing your puller to the limit with your existing DOF, and you can't add more light, then FF35 can only equal your existing noise+DOF. If you accept thinner DOF or add more light, then you can get less noise for the same DOF, but that's not going to happen in every situation.

I think this is all pretty clear if you forget about the nitty gritty (f/stop, pixel size, etc.) and focus on the big picture: total amount of light.

 

[M Most]

If you're always in low light, then "add more light" is not a useful option, so the larger FF35 scarlet sensor will only equal the S35 scarlet and not surpass it. Shoot exteriors, though, is different, as one can often use less ND to get more light, so FF35 would surpass it.

First, let me return the compliment. I learn an awful lot from your posts, in particular this one, and they force me to look at the underlying logic behind what I've often found experientially to be true. So thanks for that.

On this first point: while it's certainly true that it is relatively easy to get more light when shooting a day exterior, doing so also brings other issues into play, because shooting to a deeper stop also means more DOF (not always desirable, but usually not a problem on a day exterior) and more practically, it sometimes means more fill is needed in order to keep the contrast ratio in check. Sometimes you can do this with grip equipment alone, but in many situations, it's time to break out the big lights to make this happen. That means time, power, heat, and numerous other nasty things that are often not welcome when you're trying to move quickly. So like many other things, it becomes a tradeoff.

If you're already pushing your puller to the limit with your existing DOF, and you can't add more light, then FF35 can only equal your existing noise+DOF. If you accept thinner DOF or add more light, then you can get less noise for the same DOF, but that's not going to happen in every situation.

And therein lies my major concern when going to larger formats. Not only does focus become more critical, it is much more noticeable when it's less than perfect.

 

[David Mullen ASC]

Let's speak practically here -- most DP's are used to lighting interiors to an f/2.8 at 500 ASA, and have had to bump up their light levels for the RED at 320 ASA, even more if they put a blue filter on the camera.

So while you could shoot at f/4.0-5.6 split on a FF35 sensor and get the same DOF of f/2.8 on an S35 sensor, if you match field of view and distance... that's a stop and a half light more to come up with, or a stop and a half higher ASA rating.

So for a DP used to lighting to an f/2.8 at 500 ASA, they'd have to rate the FF35 EPIC between 1250 to 1600 ASA in order to stop down to f/4.0-5.6 and maintain the same DOF -- if they didn't want to change their light levels.

We have no idea at this point what the sensitivity is of the EPIC nor how it looks at higher ASA ratings, so we can't really know for sure whether this depth of field issue will be a problem or not. It may be no real problem at all, or it may be a big problem if the camera turns out to not be much faster than the RED ONE. Or maybe people will only switch to 6K FF35 mode when they shoot their wide shots, when focus-pulling is less challenging.

On the other hand, shooting in anamorphic in 35mm has the same depth of field challenges as FF35 and some people either like the shallower focus look aesthetically (though as Mike Most pointed out, it can be problematic for fast TV shooting schedules) or they just get used to using more light and stopping down to compensate, feeling that the extra work is worth the extra quality.

I think the bigger issue will be what sort of optics come into widespread use for FF35 filmmaking, or whether more people will just use S35 cine lenses and crop the sensor most of the time.

 

[Blair S. Paulsen]

TV is going to be Epic

TV is going to be Epic

Like Mike I have learned a lot from Dan Browning's posts, props for posting your knowledge on this too often derided forum.

I accept Mike's different viewpoint on the prospects for RedOne adaptation in the episodic world. I would point out that sitcoms often have a different aesthetic than dramas like ER and I can understand why the F900 caught on. Sony didn't get where they are without knowing a few things about marketing their wares.

For the RedOne, or more likely Epic, to achieve the world domination I so boldly predicted three things are needed beyond a 500ASA rating:

1) Real time (or even anything better than 3X real time) solution for RedCode file output that processes the full 4K HD (2,160 by 3,840) frame from an S35 sensor (or S35 crop of a FF sensor) with a full quality debayer of the 4K file (not an extraction) and then downrez to 1080 DPXs (or even half float Open EXR :nerd: ) at a reasonable cost (under $30K) with full metadata retention. Think Scratch-Cine, the next generation.

2) Industry acceptance. This includes rental house support along the lines of what Panavision offers - this business is about service as well as hardware. It also includes post houses playing to the strengths of a data centric pipeline that leverages RedCode RAW rather than trying to force feed their existing workflows.

3) Now imagine an SSD the size of an iPhone that holds 30 minutes of footage (with a milder compression ratio than RedCode 36) and you've beat the crap out of 1000' film mags or SRW-1 decks in terms of ergonomics.

No offense to Zeiss Digi-Primes but the F900 with its 2/3" chips is essentially a news camera on steroids even if you do tether it to an SRW1. The F23 or F35 is a budget buster for most episodics. A proper digital cinema camera that uses Optimos, SSDs and RAW color space will appeal to DPs and once the "science project" label fades...

If you accept that all acquisition will have to become data at some point during the post process then any film or tape based infrastructure is just marking time. Assuming that there is no "image magic" in a particular intermediary media, the cost savings of a data centric pipeline will eventually triumph in the episodic arena. Assuming the cost per minute is low enough tape might last a few more years since it provides a dead easy archive right from jump - if Sony wants to sell SRW-1s for $10K and 5800s for $15K with tapes at $1/minute then they can stay in the game - if not, the bean counters will run to data based workflows as soon as the kinks are ironed out.

Both film and professional tape are handled using carefully refined procedures well understood by the trained workforce. Once data reaches that same level its inherent advantages in speed and cost will win.

Cheers-

 

[Roberto B]

the F900 with its 2/3" chips is essentially a news camera on steroids

this is quite a statement.. that cam was a dream became true till red1.. 7 or 8 years of total dominance.. a legend.. how dare you?

 

[M Most]

Let's speak practically here -- ....
We have no idea at this point what the sensitivity is of the EPIC nor how it looks at higher ASA ratings, so we can't really know for sure whether this depth of field issue will be a problem or not.

David makes the most important point of all here. At this point in time, March 2, 2009, 9:45 AM EST, none of this stuff exists. Anything said here is prognostication, and that kind of approach can either be largely correct or completely wrong, depending on the eventual reality. Everything said today is based on either current or past experience, or a belief in technology that isn't finished yet. As David points out (and I alluded to), if the new sensors can be sensibly rated considerably higher than the current ones - certainly not an impossibility - a lot of these points are moot. Not all of them, as the laws of physics won't be broken by Red or anyone else, so a larger sensor will still require a larger lens for full coverage - but certainly the points regarding shooting stops and DOF will all be answered in time. But that time is not right now.

 

[Stephen Williams]

Moco on the RED has also become another area of considerable adoption

Hi Ted,

For shots that dont require multi passes with shutter sync that is true.

Stephen

 

[Steven Caesare]

An interesting note on how the show was graded: you'll see a lot of clipped highlights and crushed shadows. Those of you who watch the show will recognize the look, but to those viewing it with a strictly technical eye... that's all in the grade.

I'll post some technical details about how we shot the show [etc.] when I get permission.

I'm glad you said this.

I also wished there was a way to see this without it going through the par-for-the-course MPEG2 stuffed in as little bitrate as they can manage pipeline that even the HD sat channels force...

 

[Daniel Browning]

Let's speak practically here...

Thanks for the post, David. I'm glad you haven't (yet) tired of responding to my posts, I value your contributions.

We have no idea at this point what the sensitivity is of the EPIC nor how it looks at higher ASA ratings, so we can't really know for sure whether this depth of field issue will be a problem or not.

I agree that we can't know how much better EPIC will be. But my position isn't based on that. My point is that it only has to be the *same* technology as RED ONE in order to satisfy that condition. In other words, it will be capable of higher ASA simply due to the fact that it's bigger; any improvements in technology will be separate from and in addition to the size advantage.

Technology is one part of the equation, but the more important thing is light itself: how much total light is falling on the sensor. That should form the baseline of our expectations.

Everyone knows that diffraction scales with sensor size:

Smaller sensors show diffraction at wider f-numbers. Larger sensors can shoot much narrower f-numbers before diffraction affects the image. The reason is that larger sensors are enlarged less for display (lower reproduction magnification). It so happens that when diffraction is the same for different sensor sizes (different f-numbers), the physical aperture is also the same. It so happens that the total amount of light is also the same, AOTBE.

Noise scales with f-number in the same way:

Smaller sensors show noise at wider f-numbers. Larger sensors can shoot much narrower f-numbers before noise affects the image. The reason is that larger sensors are enlarged less for display (lower reproduction magnification). It so happens that when noise is the same for different sensor sizes (different f-numbers), the physical aperture is also the same. It so happens that the total amount of light is also the same, AOTBE.

Of course, that's only true if the technology part of the equation is similar enough. Sometimes the technology (low read noise, high quantum efficiency, high full well capacity, etc.) can be so much greater in a small sensor that it is enough to overcome a difference in size. That doesn't happen very often.

For example, the Canon 5D1 (FF35) is 3.5 years old, and it had pretty poor technology even for the time when it was new. (The only sense that it had "high" technology was that Canon found a way to make it so cheap.) The 5D1 sensitivity is only 0.61 electrons per 12-bit ADU per square micron. The newest Canon 50D (similar to S35 size, not FF35), on the other hand, has much higher technology, including gapless microlenses and much lower read noise. It's sensitivity is almost double: 0.100 e-/12-bit-ADU/um^2. However, the old, original 5D1 has a tremendous size advantage: *more* than double the size. So despite the older technology, when printed at the same size, it still has less noise than the 50D.

The 5D2, on the other hand, has technology much more similar to the 50D, so it isn't just somewhat better (like the 5D1), but much better: double the performance.

If RED came out with the RED FOO camera, which had the same (old) technology as RED ONE, just in a FF35 size, I think it would stack up like this:

• RED ONE in 2K windowed, 250 ISO, 25mm f/2.8: aperture is 9mm
• RED ONE in 4K mode, 500 ISO, 50mm f/5.6: aperture is 9mm
• RED FOO (see above), 1250 ISO, 80mm f/9: aperture is 9mm.

There are at least three additional factors to consider:

• Everyone else puts higher technology in smaller sensors.
• Compression.
• Unequal comparisons.

Higher technology in smaller sensors.
In still photography, the largest sensors tend to have the lowest technology and worst performance per area. They typically have half the quantum efficiency per area of the smallest cameras and twice as much noise per area. They are two times worse per area, but have 25 times more area. So in the end they are still 12 times better. Eric Fossum, inventor of CMOS sensors, has claimed that Canon's large sensors, for example, are very simple compared to even their own digicams, and he hoped that they would do more in the future.

I think that the reason smaller sensors are on the leading edge of technology is because their massive volumes have the highest absolute return (billions) and allow extreme amortization of R&D. The total market for MFDB is 6000 per year, whereas mobile phone sensors are in the hundreds of millions per year.

Therefore it would be normal to expect that EPIC FF35, Scarlet FF35, etc. will not have quite the same level of technology as their S35 variants. If RED finds a way to put similar technology in all sizes, then they will deserve credit for bucking the trend.

Compression.
The second factor is compression. Higher ISO ratings will be more difficult on the compression engine, and the result may be visible. My guess is that the effect of this will be within 1/3 stop if it's visible at all.

Unequal Comparisons.
This factor is huge, everywhere. In still photography, most people use deeply flawed methodologies for comparing different cameras. These flaws are seen much less frequently in cinema because there is more emphasis on the final output. For example, you might hear that one camera A looks better when you measure something in the D.I. process (e.g. resolution of a still image), but what you actually see on the screen in movement is worse than camera B, so camera B is better. Some still photographers, such as the editors on the luminous landscape web site, DPreview, magazines, etc., will say the opposite: camera A looks like it has less noise in 100% crops on the screen, so it's better, even though the actual 20x30 prints from camera B have less noise. (Silly, I know, but very common.)

So it's important to be aware of the most common flaws in comparisons:

• Unequal spatial frequencies
• Unequal sensor sizes.
• Unequal processing.
• Unequal expectations.
• Unequal technology.

Unequal spatial frequencies.

The most common mistake of this type is to compare 100% crops from cameras of different resolutions. This is magnifying one to a greater degree than another. It would be like using a 2X loupe to examine one and an 8X loupe to examine another. Or examining one on a 50-foot silver screen and the other on a 2-inch mobile video player. You have to compare them at the same size to judge the technicals.

Unequal sensor sizes.
If you are trying to prove that "large pixels" have less noise than "small pixels", and you use two different sensor sizes to prove your point, you failed. Your point would only be true if two sensors of the same size proved your point. (Of course, a small sensor with large pixels will never beat a large sensor with small pixels, so the common myth that large pixels are better is easily proven false.)

Unequal processing.

The most common mistakes here are to rely on in-camera processing (e.g. video formats or JPEG) or trust that the raw converter will treat two different cameras the same way (none of the commercial ones do). For example, most converters use different amounts of noise reduction for different cameras, even when noise reduction is set to "off". To make matters worse, even if a raw converter is used that can be proven to be totally equal (e.g. dcraw), the method it uses might be better suited to one type of sensor (e.g. strong OLPF, less aliases) more than another (e.g. weak OLPF, more aliases). So you have to be aware of the possibilities for inequality.

Unequal expectations.

This may or may not be a flaw depending on your viewpoint. If you expect that a camera that costs 50% more and is 50% larger should have 50% better performance in the same low light, then you will be disappointed. On the other hand, if you only expect to it be at least the same as the cheaper/smaller camera, then that would be equal expectations, and you will likely be satisfied.

Unequal technology.

If you compare a 5-year-old camera to a 5-day-old camera, it will not be surprising to find one better than another. In one sense, it will never be possible to compare any two cameras with completely equal technology, because even unit-to-unit manufacturing tolerances of the same unit will cause there to be inequalities. It's common to find one Canon 20D with less noise than another Canon 20D, even if absolutely everything else is the same. Units vary. I don't think that means we should give up on testing altogether, just that we should be aware of this potential factor.

Everything said today is based on either current or past experience

I don't think that's a necessary disclaimer. My past experience tells me that RED's 4 liter engine (RED ONE) tends to have about twice as much horsepower as their 2 liter engine (RED ONE 2K windowed). This matches the theory of how horsepower tends to scale with engine size (noise scales with sensor size), and it also matches my experience with other engine manufacturers. So if RED says they're coming out with a 9 liter engine (FF35), I will expect it to have 2.25 times as much horsepower as their 4 liter engine. Or maybe slightly less (2X vs 2.25X) since these things don't usually scale linearly. Or maybe slightly more (2.5X) since Jim said the technology will be superior.

In short: based on theory (light) and technology (my experimentation with still cameras of various sensor sizes), I say with certainty that RED FF35 can be rated at least twice the ASA and still yield a similar noise level to the RED ONE when both are shown at the same size and resolution.

 

[Brook Willard]

I'm glad you said this.

I also wished there was a way to see this without it going through the par-for-the-course MPEG2 stuffed in as little bitrate as they can manage pipeline that even the HD sat channels force...

Will work on that as well. I'll make a new ER thread when the day comes, as this thread is on a completely different topic now.

 

[Justin O'Neill]

Wow Daniel, you did a really good job of explaining the characteristics of sensor sizes in this post.

I may have to email this to a couple colleagues who still can't quite grasp the concept.

 

[M Most]

[*] RED ONE in 2K windowed, 250 ISO, 25mm f/2.8: aperture is 9mm
[*] RED ONE in 4K mode, 500 ISO, 50mm f/5.6: aperture is 9mm
[*] RED FOO (see above), 1250 ISO, 80mm f/9: aperture is 9mm.

I'm certainly not arguing with the math or the theory. However.......

In your hypothetical case, which I agree illustrates an "all other things being equal" scenario, my problem is that in order to make all of this work, we would need lenses whose sweet spot is at a considerably higher stop than most cameramen feel it actually is. While I understand that the total amount of light (and the actual light required on the set), and thus the noise produced by the sensor (as well as the effective DOF at a similar focal distance) is common between 250ISO with a 24mm at f2.8 and a 1250ISO with an 80mm at f9, I can't get past the notion that I really don't want to be shooting at f/9, regardless of the other characteristics, because the lens performance is just not the same as it is at f2.8 or f4. Perhaps this is just an artifact of having lived in the 35mm world for too long, or perhaps it's more significant than that. Time will tell. It's certainly.......... interesting.