Cinematography Mailing List - CML
    advanced

RED vs Film

My colleague and I recently completed a large thesis defence comparing the RED on Build 20 to Kodak Vision3 500T. We got some surprising results here and there (including clear examples of the RED rolling shutter issue) and thought to pass the whole thing along. So, while knowingly opening a can of worms for heated discussion, take a peek at the link:

http://www.benjaminverhulst.com/dop/RED_vs_Film.html

We shot the RED alongside full frame 35mm on a BL-4 and Super16 on an SR3 with matched Zeiss Super Speeds. The RED was rated at 320 ISO and 3200º, while the 5219 and 7219 was exposed for 320 ISO. We had a working print as well as a 4k scan down-res'd to 2k DPX sequence for the 35, while the S16 was telecine'd to HDCam Tape to compare different filming scenarios. The university's working on getting all 480 gb of raw files online at some point - if you're interested e-mail me off list and I can let you know if it ever gets up.

We'd really like to thank Rexford Metz, ASC for helping us out and Walt Rose at Fotokem for donating processing and transfer services. We'd also like to thank the Florida State Univ Film School.

Ben & Alex
Cinematographers
Los Angeles

Benjamin Verhulst
213.538.2006


Why don't the video clip image sizes match ... if you're trying to compare quality?

Why are the resolution tiffs so tiny?

Mako, Makofoto, S. Pasadena, CA


Argh! I know you meant well, and are working under budget restrictions, and are depending on the kindness of donated services. But if you are doing a comparison between apples and oranges, you need to clearly point out the specifics of how each image was captured and processed.

When you say "matched Zeiss Super Speeds" -- please elaborate. Was the S16 test shot with the same lens as the 35 test and RED test, or was it shot with an equivalent field of view lens, of a different focal length? Were the lenses all of the same generation? Internal flare and general lens contrast differences could account for many things. Even internal reflection in the camera could account for contrast differences.

The biggest red herring is the S16 material's post path. It was telecined to HDCAM (SR or crappy old HDCAM?) on what kind of telecine? Obviously a colourist was involved, which is why there is more dynamic range (even if it was a best-light transfer) than a vanilla film scan, which by nature isn't graded when captured.

Why on earth were the 4K scans down-resed to 2K when you are comparing them with the output of a 4K camera? As JPB pointed out here recently, you are shooting film in the foot by doing that.

I could go on -- it's crazy to say the RED "doesn't capture pure blacks" because there is no way of knowing the signal path of all the images -- how did they get to that waveform monitor? Were LUTs involved? Did you use filtration to "rate the RED at ... 3200K?" What filter?

As for the three shots of the woman at the window, it would have made more sense to grade the shots to each look their best -- there should be a lot more detail in the 35mm highlights unless the scan is bad.

It would be easier to compare the res chart TIFFs if you had removed the chroma, and corrected the exposure -- why the variations?

I'll stop now. I don't mean to be harsh, but this sort of test is terribly flawed. Good idea, but the concept and execution are lacking.

Jeff "keep trying" Kreines


Didn't you smell a rat when the 35mm footage came out looking worse than the 16mm?

Didn't you think that maybe, just maybe you had made a mistake somewhere along the line?

Why compare film 500T with RED 320D? Why not use Film 250D?

HDCam??

I'm going to stop.

Cheers

Geoff Boyle FBKS
Cinematographer
EU Based
Skype geoff.boyle
mobile: +44 (0)7831 562877
www.gboyle.co.uk
www.cinematography.net


Here's what I'm seeing :

The 16mm was transferred to HDCAM--which is an 8-bit format--and then back to an uncompressed 10-bit Quicktime. The two bits of information between 8-bit and 10-bit don't actually come back. They are lost forever. The bit depth of the DPX files is not stated. The RED's footage is hobbled out of the gate as it could have been graded directly from the original 12-bit R3D files instead of being converted to 10-bit Quicktime before grading. (I'm not sure what the point of the 35mm work print is since it can't be compared accurately to any of the other footage.)

The footage was matched using MacBeth charts, which doesn't make sense to me. Shouldn't a DSC chart have been used instead? MacBeth charts are good for examining subtle shifts in common colours but are really more useful in print than in HD. DSC charts are specifically designed for matching cameras and footage. Most of the stills shown don't match each other in colour or contrast.

I get why 2k was used: that's the max Final Cut will handle. But the differing workflows aren't conducive to comparison at all:

(1) 2k DPX, HDCAM and Camera RGB colour spaces and bit depths are completely different. You're comparing 4:4:4, 27:9:9 and RED footage with one of several possible colour spaces baked into it.

(2) HDCAM cheats horizontal resolution as it's really 1440x1080, so resolution tests should take that into account.

The 35mm images consistently look marginal, and my suspicion is that's because they went through an additional step--from graded 2k DPX to uncompressed Quicktime, resulting in changes in colour and contrast between the codecs. (It's no wonder the RED looks the best on page 15: its footage has gone through the fewest destructive transformations.)

Page 16 shows just how flawed these tests are: the 16mm and RED stills match each other better than the 35mm still, which is crazy contrasty and has some sort of awful overcranked-video-knee look in the highlights. It looks like super white was clipped during a transformation between codecs or on the Final Cut Pro timeline.

Saying that blacks look muddier on the RED isn't really valid. Blacks are something that have to be set. WYSIWYG cameras set the pedestal internally before recording the image; the RED does not, allowing you the freedom to set black later in post. Digitized film blacks are set the same way.

On page 18 it is stated that 35mm's blacks are rich and strong while RED's are milky and grainy. This is because 35mm has more latitude than the RED, and the RED's blacks will look thin when trying to match them to 35mm because the RED doesn't have quite the same range. It's easy to make the RED's blacks look strong and rich but you won't see as much detail in them as you would 35mm. Saying that the RED's blacks are milky implies that they are baked in when they are not. If they are milky then they were not set correctly in post.

On page 11 it is noted that the 500T was overexposed by 2/3 stop "as is common in the industry." I'm not sure that's the case. Is it common to overexpose a little? Yes, although it's less common than it used to be. Is it common that everyone overexposes 2/3 stop? No. It's also common to rate 16mm stocks slower than 35mm stocks in order to crush the grain a bit.

It is stated that the RED sensor has a native colour temp of 5000k but can be set to 3200k, which is incorrect: the sensor is always 5000k, but the white balance flag can be set to emulate white under different circumstances.

On page 14 it might have been helpful to show the RED's response curve, which--like all digital formats--is a straight line. This chart implies that all "camera systems" have a toe and shoulder, and electronic cameras don't unless one is manufactured using an additional gamma curve (like Sony's Hypergammas).

Page 17 is just bizarre: none of the grades match each other in the slightest. It's mentioned that the 35mm film might show more detail if another brighter workprint was struck, which implies that we are looking at scanned workprint. We should be looking at the digital results of the Colour grade.

I appreciate the effort that went into this test, and it's a good subject for a student thesis. But there was a lot more going on here than was addressed and the differing workflows negated any value these tests might have had. Next time I'd recommend coming here and running your workflows by the group first. There are a lot more gotchas these days than there used to be when you'd just shoot film and print it.

And if someone wants to do a real test like this sometime, I would love to be involved. There's got to be a good way to compare all these formats and show their strengths properly.

-----------------------

Art Adams | Director of Photography
4 1 5 . 7 6 0 . 5 1 6 7

showreel -> www.artadams.net
trade writing -> art.provideocoalition.com

ICG, SOC, NWU


Just looked at some of the video clips. It would help if they were all framed the same way, and with roughly the same action, otherwise they are impossible to compare accurately. One of the darker shots had a lot of hands waving through it, which was very distracting.

The film footage, both 16mm and 35mm, looks really bizarre. The highlights are completed whacked. Since Fotokem's name is on this test it might be worth them doing a little forensic analysis as I know that they would never let footage that looked like this leave their facility. Clearly something happened farther down the pipeline.

Paging Mr. Colman, Mr. Ed Colman to the white courtesy phone...

-----------------------

Art Adams | Director of Photography
4 1 5 . 7 6 0 . 5 1 6 7

showreel -> www.artadams.net
trade writing -> art.provideocoalition.com

ICG, SOC, NWU


So the ASC, BSC, Roger Deakens, Darius Khonji and every other great DP is wrong and you are right? That is your conclusion?
Don't give up your day job.

Jeez ... We do this for a living.

Dan Bronks
DP
UK
www.danielbronks.com


>>and the RED's blacks will look thin when trying to match them to 35mm because the RED doesn't >>have quite the same range.

And also RED has a 12 bit AD which from my experience has an effect on the tonality or separation of tones in the shadows, log or no log. It will be very interesting to see Epic with 14 bit AD with 12 or 14 bit recording especially if noise levels have been reduced.

The "mapping" of the AD is rarely discussed and it would be useful to learn more of the trade offs and black art utilised by different manufacturers lest I have the wrong idea

Mike Brennan
DP
Melbourne London


After almost 5 year in the HD and digital cinema discussion I'm really surprised to see the same topic as the beginning ... comparison between film and digital! I really don't care about which is better.. they are different . period.

What about the language changes that Digital Cinema even HD bring to this business..

Rodrigo Llano
Cinematographer


>> After almost 5 year in the HD and digital cinema discussion

Star Wars Attack of the Clones commenced shooting June 2000 so I make it 8-9 years ago that the discussion began cooking ...But when it comes to the history of the introduction of electronic motion picture cameras, I'm not one to make a broad distinction between Progressive HD cameras and newer Digital Cinema Cameras.

Hopefully as time goes on there will be a greater significant difference in the two than there is now
Mike Brennan
DP
Melbourne/London


>>After almost 5 year in the HD and digital cinema discussion I'm really surprised to see the same >>topic as the beginning

We're not starting that topic over again. A couple of film students posted their tests, and that was very generous of them. Of course it was very generous of us to point out where the tests went horribly wrong, but that's part of the learning process. As Dave Stump says, "It didn't used to be rocket science. Now... it's rocket science."

Let's not punish people for trying to contribute.

-----------------------

Art Adams | Director of Photography
4 1 5 . 7 6 0 . 5 1 6 7

showreel -> www.artadams.net
trade writing -> art.provideocoalition.com

ICG, SOC, NWU

 

>Benjamin, thank you for s


style="margin-top: 0;">haring your work with the CML. It looks like you took a great deal of effort to accomplish this, and congratulations on sharing this.

However: There are many issues with the paper, and since you have posted publicly, perhaps this will help you in the future.

1. There is no peer review: Since this is an academic paper it is your responsibility to have it reviewed by your peers, before submitting it. The feedback from your peers, or even better, those more advanced, helps to avoid publishing a paper with many mistakes. Crediting those who reviewed the paper also goes to helping your credibility. I didn't see any peer review credited, and I am a little surprised your academic adviser did not suggest you seek peer review.

2. You mention publications in your bibliography, but do not seem to cite them as either supporting or contradicting sources for some of your observations. For example you have made a statement about the resolving power of 16mm, but what do others conclude is the limit of the resolving power of 16mm? I could respond that I have a copy of a Redlake Hycam 16mm rotating prism highs speed camera operating manual. According to the copy the original manual contained a clip of film shot with the Hycam showed a resolution of 68 lines per millimetre. Granted this is both an owners manual and a sales brochure, but this claim was made in 1965. One would believe that the resolving power of film has improved in the past 44 years. (Also note that the rotating prism optical system limits the resolution attainable compared to a standard intermittent motion film camera.) If you had stated what Kodak, or others have defined the resolution of 16mm film as being, then I would have a basis for understanding how well your test was accomplished. As it is, we have no way to evaluate your test results. Additionally MTF seems to be measured in Cycles per mm (at least on the info supplied by Kodak :

>
http://motion.kodak.com/US/en/motion/Products/Production/VISION3_Color_Negative_Films/

style="margin-top: 0;">VISION3_500T/tech5219.htm

An explanation on how to convert that to line pairs per mm value would have been helpful.

3. There is a fair bit of terminology, and descriptions, encountered early in your paper that is not quite correct, leading me to wonder about the validity of the rest of your paper, and your conclusions. In your descriptions of the technology you show a very general (at best) understanding of the film process, lacking in expertise and familiarity.

4. Your methodology is very suspect for example:
a. Why have a 35mm work print struck from the 35 neg, if you are not going to do the same for the other two formats/mediums?

b. Why is the 16mm going through a different process as the 35mm? Once this happens, how does this validate or invalidate your results?

c. Why is the 35mm being downrezzed to 2K?

d. Why aren't you matching the field of view of the lenses used on each format?

e. You don't supply enough info about the lenses: Comparable Zeiss lenses is not sufficient. Was the same lens used on each camera? Serial numbers? What about the Telecine? There are many different Telecine systems, and they are not all equal, why are there no specifics about the Telecine system used (or the operator?)

f. Why are you using Final Cut and quicktime? Are there no other options? Are these the best workflow, or only what you had available to you?

5. Your conclusion is weak:
a. There are many kinds of "digital", and many many times Digital is shot and displayed without "colour correction", additionally you do not seem to draw a difference between reversal film, which was designed to be directly projected, and negative film which has always had the process of "correction" designed into the workflow.

b. Although in your conclusion you recognize that there are other tests to make, you have failed to recognize any mistakes or any realizations about your methodology that may have influenced the results or conclusions.

c. How come Digital will get better in the future, but film will not? Kodak currently offers (at least) Vision 2 and Vision 3 films, each a different "generation" of film technology, analogous to different generations of sensors or even firmware builds.

I could also point to many other flaws and assumptions (on Page 12 for example) where you state that "Thanks to the digital intermediate colouring process almost all big budget films now go through, a cinematographer often doesn’t have to live with mistakes made on the set." No supporting info, and didn't you state in your introduction that the RED now allowed the "average prosumer" could now enter the independent arena. What has the big budget film have to do with the indie world, and do you know for a fact that digital intermediates are being used to fix mistakes made on the set, where other means wouldn't have been used in the past?

I don't wish to browbeat you, because it was a brave thing to post such a paper, and you are to be commended for taking on the topic. A good review of the methodology and assumptions before undertaking this test, as well as a good review prior to publication, could have gone a long way to making it a more valuable paper, with stronger more valid conclusions as well as better raw data gathered. You put in a lot of work testing, it’s a shame if falls so short.

Best wishes.

Steven Gladstone
New York Based Cinematographer
Gladstone films
http://www.gladstonefilms.com
917-886-5858


Thanks for taking it easy on us ...

I thought I had made it clear this was an honours thesis, I don't want to claim to be something we're not.

The tests were designed to mimic what actually happens in the real world, not to compare the absolute best scenarios for each format. Hence the 35mm was on a 4k scan downres'd to 2k (10-bit dpx) (when was the last feature you worked on that scanned the entire thing at 6k downres'd to 4?) and the S16 was telecine'd (for comparison to what many low-budget films and most film schools work with - one-light to charts).

And yes, there were definitely some constraints with donated services, etc. That being said, I think we're doing something you don't see that often - comparing formats in ways they're actually used).

The image sizes are different because the 35mm was full-frame academy and the RED is the equivalent of Super35 (yes, this affects the resolution to some extent). The exact same lens was used on the 35 and RED at the same stop and focus with the exception of strobing tests when the cameras were rolling at the same time. The Super16 was using a lens from the same set with an equivalent field of view (by 'matched' I mean all lenses were re-glassed at the same time).

For the resolution tiffs - that's literally the size of the screen grab after it's blown up to be able to see the limits of the resolution. As far as colour correcting the res footage before comparing, we found it made no real difference and we wanted to provide unprocessed images whenever resolution was concerned.

We never shot with any glass in front of the lens and the RED was 'rated to 3200' in the metadata. We used 500T because that's what most people would shoot in this situation. Again, we aimed to compare real life scenarios, not 'best for the camera' scenarios.

The workflow was designed to be as pure as possible - the R3Ds were exported from RED Cine using the RGB LUT standard as a quicktime 10-bit uncompressed 4:4:4 log file. The resulting files were minimally adjusted in Colour to match charts, but there were no LUTs applied after being exported from R3D.

And Geoff - that was exactly our thought on the 35 footage at first, but the working print looked fine, great even (again, one-light to charts). Unfortunately I don't have a way to show you that, but you do get a bit more exposure latitude when the dpx is graded.

Our next step would be to grade the images from the 'raw files' and compare them, but we're having trouble thinking of a way to do so fairly. So much of what the 'best look' for an image is personal preference.

Just got to Art - You're right, I can't argue with a whole lot, just clarify.

We shot Macbeth Charts, 18% Grey, Grey Scale and RGB/CMY Charts and made sure they all matched. DSC is a good idea. You're correct on the HDCam, but it's still pretty obvious the resolution is limited by the grain, not the image process.

The 35mm was exported to 10-bit uncompressed quicktime from dpx, BUT it had almost exactly the same colour and contrast before. The highlights were clipped from the get go and we were confused by that as well. Ideally, we'd really like to get another scan, but didn't want to push the issue as it was donated.

On the blacks, our point was that even though there may be more information in the blacks in the raw r3d than in properly transferred film, you lose most of it when you set the blacks to their proper level. Essentially that film has greater exposure latitude.

The film was rated at 320 to keep the comparisons even, if 320T was a common film standard, we'd have used it. Didn't mean to suggest it's what everyone does, just that it is common.

We should've been more specific on the white balance - yes, it is meta-data.

As for page 17 - in an early version of the paper we were comparing the projected workprint to the digital files because the scan was so strange and this made it through somehow. The scan is from the neg.

The YouTube video was designed to see it in motion, not as much for actual colour comparison. The formats were shot separately in order to use the same lens.

Dan - our conclusion is meant to say that film is a stronger format right now (I'll shoot on it any day), but a 4k R3D file has more resolution than a 2k dpx sequence and digital is getting better very quickly.

--

Thanks for the critique, it's what we were hoping for.
In keeping with the spirit of comparing these formats in ways they are actually used by most of the film world and not just in ideal situations, how would you design the post workflow?

If anyone's interested in a more in-depth comparison starting over (or with the raw files we already have) I'd love to be involved as well.

Best,
Ben

Benjamin Verhulst
Cinematography
Los Angeles
213.538.2006


>> Let's not punish people for trying to contribute.

For obvious reasons this is the spirit of CML. Thank you Art.

When you flame someone, particularly a student who is trying to contribute, it just shows what you are made of.

I can remember back when we were on CompuServe. And there just a few hundred of us. And the thought of getting flamed never crossed my mind. So I learned to ask questions, get help, and take advantage of the incredible resource we have thanks to Geoff.

And on that thought, Thanks As Always Geoff.

Jim "Where do I contribute again?" Dollarhide
CML member since the Beginning . . .

Jim Dollarhide
Director/Cinematographer
I live in Mississippi and I love it.
601-853-4252

style="margin-top: 0;">http://www.dollarhide.net


Benjamin Verhulst wrote:

>> And Geoff - that was exactly our thought on the 35 footage at first, but the working print looked >>fine, great even (again, one-light to charts).

Which has absolutely nothing to do with whether or not the scan was good, or whether or not your post path took a good scan and made it look awful. It means your negative was ok, that's all.

I won't deal with the other stuff. Art has done a good job.

Jeff Kreines


I'd like to thank everyone for their responses to this thread, all better than my grumpy old man one!

Art & Steven, a special thank you for your replies.

Yes this probably should have gone to students list but it's got a great response here.

Oh and Benjamin, if you're at the RED day then come and say hello.

Cheers

Geoff Boyle FBKS
Cinematographer
EU Based
Skype geoff.boyle
mobile: +44 (0)7831 562877
www.gboyle.co.uk
www.cinematography.net


Sorry, wrong. The RED 4K image is 22mm across, same as Academy not Super-35. That would be more like 24mm across.

Mitch Gross
Applications Specialist
Abel Cine Tech


>> c. Why is the 35mm being downrezzed to 2K?

Because that's what his editing system will handle. I don't think this part is that big a deal unless the test is entirely about resolution.

I wonder how hard it would have been to put everything through a 2k DPX pipeline? He certainly could have done that with the RED footage. He'd still have to figure out what went wrong in that final conversion process.

>>D. Why aren't you matching the field of view of the lenses used on each format?

This is a biggie.

>>f. Why are you using Final Cut and quicktime? Are there no other options? Are these the best >>workflow, or only what you had available to you?

I suspect they are. HE'S A STUDENT!

-----------------------

Art Adams | Director of Photography
4 1 5 . 7 6 0 . 5 1 6 7

showreel -> www.artadams.net
trade writing -> art.provideocoalition.com

ICG, SOC, NWU


Mitch Gross wrote:

>> Sorry, wrong. The RED 4K image is 22mm across, same as Academy not
>> Super-35. That would be more like 24mm across

Maybe he shot the new 4.5K mode to get Super-35 (24mm) across?

Also, some scanners scan Academy at "4K" at actually 3656 x 2664 pixels, so even if he shot 4K (4096 / 22mm) and 35mm Academy (also 22mm), the file sizes could be slightly different, though the field of view would have been the same. However, if the 35mm Academy had a narrower view on the same lens, he must have shot in 4.5K mode.

David Mullen, ASC
Los Angeles


David Mullen writes:

>> some scanners scan Academy at "4K" at actually 3656 x 2664 pixels

Just FYI, that's based on 35mm full silent aperture being 4096 pixels across, minus the soundtrack area. It's known as "Cineon Full Res". 1828x1332 being of course "Cineon Half Res".

Tim Sassoon
SFD
Santa Monica, CA


Benjamin, please keep contributing. You're heading in the right direction. There's much to learn, though, and CML is a damned good place to learn a bunch of it.

-----------------------

Art Adams | Director of Photography
4 1 5 . 7 6 0 . 5 1 6 7


Art Adams wrote:

>> I suspect they are. HE'S A STUDENT!

Thanks for the answers Art, but what I was getting at were points and issues that he did not really cover as to why the choices were made. These need to be made. It is perfectly valid to use Finalcut Pro, but you have to explain why. "We chose Final Cut and colour as our workflow because it is in widespread use in the indie world, and it is what we had reasonable access to", makes it an informed choice, and explains the workflow, highlights either a weakness or a strength of the process.

Allows the reader to understand the thought process and results better.

Steven Gladstone
New York Based Cinematographer
Gladstone films
http://www.gladstonefilms.com
917-886-5858


>> Allows the reader to understand the thought process and results better.

Got it. That makes sense. Final Cut Pro is not the best way to do tests like this because, frankly, it's a bit buggy; but when it's what you have available then that's what you use, and it's probably important to tell people that.

-----------------------

Art Adams | Director of Photography
4 1 5 . 7 6 0 . 5 1 6 7


Thanks for the criticism and not just shutting it down completely, you have no idea how appreciated it is.

We knew of some of the flaws in the methodology going in, there are many being pointed out which we should've thought of and a couple which never would've crossed my mind.

I'd like to offer a couple quick points and then do a quick overview of our image process and note the flaws that have been brought up to make sure I'm on the same page. Obviously we were not nearly as clear on our processes as we needed to be.

Also - I threw a single dpx file online (I'll put a whole set on later) to see if anyone can find more detail in the 35mm highlights than we could.

http://www.benjaminverhulst.com/dop/RED_vs_Film_files/rvf_test_flt_txtd_111009.0002012.dpx

A couple quick points:

style="margin-top: 0;">• We didn't use Final Cut for the project with the exception of editing the clips you see on the site, the files were only processed with Apple Colour, Apple Shake and RedCine.

• The RED footage was shot on Build 20 and 4k 2:1 (the sensor is advertised as 24.4mm by 13.7mm Super35 on the RED site - I'm not sure if the entire sensor width is used at 2:1).

• We used the same lens on each camera for the same shot to minimize disparities, hence the different fields of view (a bad choice, perhaps).

Image Flow:

35mm

A1 - BL4 Shooting 5219 (V3 500T) Academy (framed for 1.85) exposed at 320 ISO, Zeiss T1.3 SuperSpeed Lenses

Flaws - Didn't do registration test on BL4, some take issue with overexposing 5219

A2 - Developed normal at Fotokem, Workprint made for reference, Scanned at 4k (we believe on an Arriscan), Delivered by Fotokem as 2k DPX File Sequence with NO colour correction

Flaws - Point of Workprint? (originally for resolution tests, found no way of examining resolution without destroying the print)

Why downsampled to 2k? (We actually expected 4k files back, but upon receiving them were told that a 4k scan always meant downsampling to 2k as an anti-aliasing feature)

Please see link above for one of the dpx files.

A3 - 2k (2048x1556) DPX Sequence (RGB, 16 bit) imported into Apple Colour. Colour set to linear, 'printing density 0-1023.'

Flaws - We found it strange the files were 16 bit Linear (Logarithmic looked horrible), DPX sequences are normally 10-bit Logs, yes? We almost never get to work with scans, so this was the steepest learning curve for us.

A4 - Sequence slightly colour corrected to match 18% grey and Macbeth charts.

Flaws - Could've used DSC Charts, etc; obvious room for human error

Why not colour correct shots? (The idea was if we matched the charts, we could objectively compare the formats. The biggest flaw to me in this thinking is not setting the blacks, but there are obviously more.)

A5 - Sequence exported from colour as 2k DPX (original file settings), then exported to 10-bit Quicktime Uncompressed using Apple Shake.

Flaws - Lose Bit-depth, changing colour spaces. (Our thought was getting all the formats into the same file type, bit depth and colour space would make for better comparisons. It seems this is where we went wrong the most.)

A6 - Tiffs for Paper were created by exporting from Quicktime.

Flaws - Bit-Depth, Colour Space, etc. of Tiffs

16mm

B1 - SR3 Shooting 7219 (V3 500T) Super16 (framed for 1.85) exposed at 320 ISO, Zeiss 35mm T1.3 SuperSpeed Lenses.

Flaws - No Registration test, issue with overexposing 7219, had to use different lens from set to match 35mm and RED field of view

B2 - Developed normal at Fotokem, Telecine'd one-light to charts (18% RGB/CMY) to HDCam Tape

This is how the FSU Film School gets its S16 developed and delivered (it's then converted to ProRes unless it's a VFX shot, but that's another conversation).

Flaws - Used HDCam, not HDCam SR; not actually 1920, but 1440x1080 b/c of HDCam; only 8-bit 3:1:1, only a one-light

B3 - Capture HDCam as 10-bit Quicktime Uncompressed

Flaws - Going to 10-bit just adds 2 empty bits of depth; same for going to 4:2:2 from 3:1:1

B4 - Imported into Apple Colour and slightly colour corrected to match 18% grey and Macbeth charts.

Flaws - Same as colour correcting 35 (A4)

B5 - Exported from Colour as 10-bit Quicktime Uncompressed, Tiffs for paper
created by exporting from Quicktime

Flaws - Bit-Depth, Colour Space, etc of Tiffs

RED

C1 - RED One Build 20 (Don't have the camera report with me for exact sub-build), 4k 2:1 (framed for 1.85), 3200°, RedCode36, Zeiss 35mm T1.3 Superspeed Lenses

Flaws - Different sensor size than Academy 35 (We shot this in October and Build 21 and 4.5k weren't out yet), different field of view from 35mm

C2 - Used REDCine to export as 10-bit Quicktime Uncompressed with Camera RGB
Setting, Left Meta-data unchanged

Flaws - Should've stuck with R3D for Colour Correction, why RGB? (We considered it the most established standard and least artificially enhancing)

C3 - Imported into Apple Colour and slightly colour corrected to match 18% grey and Macbeth charts.

Flaws - Didn't set blacks, same problems as A4

C4 - Exported from Colour as 10-bit Quicktime Uncompressed, Tiffs for paper created by exporting from Quicktime

Flaws - Bit-Depth, Colour Space, etc of Tiffs

This was an undergrad paper presented to a 3 person faculty committee. The paper underwent revision after the committee before being complete, but the testing could not be changed. The main conclusion by the committee was that the next step should be colour correcting to similar levels to compare exposure and colour.

We've graduated, so the official paper is all said and done I'm afraid, but we still have all the original files and would be interested in re-doing the post workflow.

So my two big question -

1) Any thoughts on what's up with the highlights in the 35? Again, see the dpx link above.

2) Would it be worth anything if we re-did the post workflow?

Best,

Benjamin Verhulst

Cinematographer
Los Angeles


>>A3 - 2k (2048x1556) DPX Sequence (RGB, 16 bit) imported into Apple Colour. Flaws - We found it >>strange the files were 16 bit Linear (Logarithmic looked horrible....

The file you supplied is not 16 bits, regardless of how Shake is identifying it. A 16 bit file of this pixel size would be almost 20MB. Yours is about 12MB, a typical size for a 10 bit scan. If I load the image in Assimilate Scratch (I just happen to have that handy), it identifies this file as 10 bit log, which I would suspect is the case. I took the liberty of colour correcting it two ways - interpreting it as linear (as you claim it to be, but not as it appears to be), and also as 10 bit log (corrected while viewing through a film print preview LUT, with that look baked in). I see a bit of detail in the highlights, but I have to say that I do think there's something fishy about the scan, as it required some fairly extreme
correction to yield this image. But in either case, I think in the future, you might do well to have a professional colourist look at the images when you've got questions like you seem to have. And I'm sure Rex would recommend the same thing.

I put the two images in a Mobile Me gallery at this address:

http://gallery.me.com/mdmost/100015

The log image will appear a bit flatter due to the toe and knee effect of
the baked in film print LUT.

Mike Most
Technologist
Woodland Hills, CA.

"Postworld" blog at http://mikemost.com


The RED footage was shot on Build 20 and 4k 2:1 (the sensor is advertised as 24.4mm by 13.7mm Super35 on the RED site - I'm not sure if the entire sensor width is used at 2:1).

That is the physical size of the whole sensor, not the active image area when shooting 4K. The photosites are 5.4 microns, which means that the width at 4096 would be 22.1184 millimeters. This information is fairly easily available.

Mitch Gross
Applications Specialist
Abel Cine Tech


Fair point. Way to harsh and grumpy. Sorry.

Dan Bronks
DP
UK


Benjamin Verhulst writes :

>> 1) Any thoughts on what's up with the highlights in the 35? Again, see the dpx link above.

As Mike says, that DPX is 10bit, but it's not a log scan. The highlights seem to have been clipped and then maybe graded...

If you bring that frame into Shake and attach a Histogram node, you can see the hard edges on the right indicating clipping. You can also see that the clip point is different in each channel - the white point has probably been adjusted after whatever caused the clip.

For those following along at home it looks like this:

http://lewissaunders.com/cml/rvf_test_histogram.png

A log scan would look much flatter, and there should be much more highlight detail, and I'd expect a bit more room in the shadows too.

I'm guessing that your facility has effectively done a rec709 one-light for you, which makes perfect sense for a student project - dealing with log scans in FCP and Colour is not at all easy!

Unfortunately you'd have to have the film re-scanned to see what's really there...

Lewis Saunders
Compositor | 2D TD
London


Lewis Saunders wrote:

>>As Mike says, that DPX is 10bit, but it's not a log scan. The highlights seem to have been clipped >>and then maybe graded.....I'm guessing that your facility has effectively done a rec709 one-light >>for you, which makes perfect sense for a student project - dealing with log scans in FCP and >>Colour is not at all easy!

After examining the image again, I agree with you. This seems to be basically a video gamma encoded, somewhat timed image.

> >Unfortunately you'd have to have the film re-scanned to see what's really there...

Or recover the original 10 bit log scan that might have been done prior to creating this image. Either way, what they have is likely not representative of the actual photography, and certainly nothing that should be used for comparative purposes or for drawing any conclusions from.

Michael Most
Technologist
Woodland Hills, CA.


Thanks very much Mike and Lewis.

We definitely had serious reservations when we saw the clipped highlights in our waveforms, but when we talked to the lab they seemed to be sure it hadn't been colour corrected at all. We'll be in contact with the lab and see what we can do!

I'm hoping we'll be able to incorporate the criticism and come back with a project that shows a true comparison.

Benjamin Verhulst

>Cinematographer

Los Angeles




Copyright © CML. All rights reserved.