RED vs. 35mm vs. HVX over/under key/fill tests

[Charles Haine]

http://www.mediafire.com/?skdgshggnlu

 

So, I started a new discussion since the other wandered off into a debate about the future of 3D TV. The above link is to an SD file encoded in H.264; I have a 1080p ProRes file, but it's way too large to post.

 

This is a new file; I re-telecined the 35mm, and I re-exported the RED footage from REDcine using REDlog, and I also corrected as much as possible for the over and under exposure in REDcine, in order to mimic as closely as possible the workflow from the 35 (where the over and under are compensated for with Da Vinci). Da Vinci is a more powerful color corrector than REDcine, so the test isn't completely fair, but REDcine is the program that most will use for processing their RED footage (except for those with Scratch, of course), so I felt like it was fair enough for this test.

 

We shot 5219 for the over/under tests, 5229 for the key/fill tests, and Fuji Reala for the exterior shots. The RED was shot 4K 2x1. The HVX was used with a Letus adapter and Lomo spherical lens's.

 

Yes, it would be better to have a full 4K side by side; however, this was a test done for my advanced cinematography class at LACC, and many of those students will only be finishing their films to SD or HD anyway, but they are still considering whether to capture RED or 35mm, and this test is helpful for them.

 

I do not personally consider this test defniitive in any way; it's just another test in a long series of tests that will be done with any new camera to get familiar with it.

 

I share it simply because some people might be curious.

 

RED camera use was donated by Shawn Booth, an owner/operator from Hawaii with http://www.r3dguerilla.com/. We used Shawn's RED len's, a 300mm and an 18-50 zoom.

 

35mm camera's (Kinor with Lomo spherical and anamorphic glass) were donated by Bruce Taylor from http://www.indi35.com/.

 

Transfer we donated by Paul Korver at cinelicious. Transfer was done on an SD Ursa Diamond with Diamondclear then upresed to AppleProRes (HQ) 1080p. You can read more info on their upressing system at http://http://www.cinelicious.tv/.

 

35mm footage was shot on 5219 (the over/under tests), 5229 (the key-fill ratio tests), and Fuji Real 8593 for the daylight shots. Processed at Deluxe.

 

Maura Milan brought in her HVX with the Letus set up for OCT-19 mount. http://mauramilan.net/

[Cesar Rubio]

Charles:

 

You did not answered me on the other thread. But that's ok, the HVX footage (CCD's), is better saturated than the Red camera (CMOS).

 

I know that film is still better saturated and has better dynamic range than both CCD's and CMOS ...but shooting digital has other advantages that film lacks.

 

Anyways thanks for the tests, they say more than all the discussions of Red vs...(anything).

 

Cesar Rubio.

[Walter Graff]

"the HVX footage (CCD's), is better saturated than the Red camera (CMOS)."

 

Actually it all depends on how the cameras as set. One can not simply say one was more saturated and think it means anything when it comes to video cameras. It doesn't. One has to know if during the tests, the cameras were properly calibrated or just turned on and shot. If they were turned on and simply shot with, than any test means little as to abilitites for a particular shoot.

[Gunleik Groven]

Not commenting on the test itself, here - but more on the use of RedCine.

 

RedCine is (IMHO) not to be used for correction at all, but is more like the combined scanning/developing stage of a film process.

 

If you exported log files and did not CC them afterwards, then your results will have to be way off. It's like showing uncorrected filmscans.

 

Also, it seems you may not really have looked into how to make the images pretty noiseless on the RED at low exposures, but that (partly) relates to the above problem. I am not talking about "noise reduction", just a question as to how you used RedCine...

 

As to the HVX, it looks like the black ped was set way low to lose noise with a pretty high saturation value, but you cannot really judge the cam from those settings...

 

I'm curious as to how your 35mmm workflow was?

 

Just my 2c

 

 

Gunleik

Edited May 11, 2008 by Gunleik Groven

[Gunleik Groven]

(Couldn't edit the above, sorry...)

 

You're sorta right in exporting LOG if you're a doing a CC afterwards, but if this is your final CC and you're not doing a print from it, you'd be much closer to a "realword" result, if you exported the files as REC709.

 

G

[Cesar Rubio]

"the HVX footage (CCD's), is better saturated than the Red camera (CMOS)."

 

Actually it all depends on how the cameras as set. One can not simply say one was more saturated and think it means anything when it comes to video cameras. It doesn't. One has to know if during the tests, the cameras were properly calibrated or just turned on and shot. If they were turned on and simply shot with, than any test means little as to abilitites for a particular shoot.

 

Walter:

 

I've worked with other CMOS sensor cameras and all of them deliver unsaturated images out of the camera.

 

CCD based cameras offer better color rendition than CMOS out of the bat. If you Color Correct (CC) CCD images, they look even better.

 

The Red camera produces good images too, but they have to be heavily CC in post to deliver usable images. Not with CCD's, you could use the images with out any CC if needed or wanted.

 

The same for film, it has to be heavily CC to get usable images....that is why I prefer CCD's over both film and CMOS.

 

CR.

[David Mullen ASC]

CCD cameras that are designed to produce LOG images, like the F23 or Viper, can create just as flat and low saturated images as a CMOS camera that need to be corrected later for a saturated Rec 709 broadcast image.

 

It's not really a CCD vs. CMOS issue, it's a Rec 709 gamma encoded recording versus RAW and/or LOG recording issue. The Dalsa is a 4K CCD Bayer camera rather than a 4K CMOS Bayer camera like the RED and the RGB LOG images created from the Dalsa RAW files need to be corrected just as much as the RED images do.

 

A prosumer camera like the HVX200 does not create LOG images so you're seeing the saturation of images recorded to look correct for monitor viewing from the start, not images that need some sort of post color-correction to look correct.

[Cesar Rubio]

CCD cameras that are designed to produce LOG images, like the F23 or Viper, can create just as flat and low saturated images as a CMOS camera that need to be corrected later for a saturated Rec 709 broadcast image.

 

It's not really a CCD vs. CMOS issue, it's a Rec 709 gamma encoded recording versus RAW and/or LOG recording issue. The Dalsa is a 4K CCD Bayer camera rather than a 4K CMOS Bayer camera like the RED and the RGB LOG images created from the Dalsa RAW files need to be corrected just as much as the RED images do.

 

A prosumer camera like the HVX200 does not create LOG images so you're seeing the saturation of images recorded to look correct for monitor viewing from the start, not images that need some sort of post color-correction to look correct.

 

David:

 

I can record in Log, Gamma or linear with StreamPix using Cineform Raw with my Pike F-210c camera, and I have not seen what you are saying.

 

CR.

[Patrick Neary]

The same for film, it has to be heavily CC to get usable images....that is why I prefer CCD's over both film and CMOS.

 

CR.

 

Not.

[Cesar Rubio]

Not.

 

Not? What?

 

Be specific please.

 

CR.

[Patrick Neary]

Just not.

 

You can shoot 16 or 35 with minimal care and one-light a beautiful print, no cc.

[Cesar Rubio]

Just not.

 

You can shoot 16 or 35 with minimal care and one-light a beautiful print, no cc.

 

That's why I hear. I really don't have any experience with film. So you might be right.

 

CR.

[David Mullen ASC]

All I am saying is that there are CCD cameras that produce flat milky images that need color-correction in post to look correct for monitor viewing. A LOG recording is supposed to look that way, like a Cineon LOG scan of a film negative does, it's not meant for direct TV viewing, it's meant for color-correcting and film-out work. The point of a LOG scan is to optimize the number of bits for the dynamic range of film, to preserve as much luminence information as possible. It's not meant for direct viewing in any form of "correct" contrast or saturation, just as a film negative is not designed for direct projection, but for the gamma of the print material.

 

If your camera has a LOG mode and the image doesn't look milky, low-contrast, and desaturated without a LUT or color-correction applied, then it's not LOG, it's gamma encoded video.

 

The HVX200 has a number of gamma modes, some flatter and lower in contrast than others, but not a true LOG emulation mode.

[Cesar Rubio]

...If your camera has a LOG mode and the image doesn't look milky, low-contrast, and desaturated without a LUT or color-correction applied, then it's not LOG, it's gamma encoded video.

 

David:

 

Probably you are right in this one. Norpix and Cineform are working correcting the Log recording because now it looks the same as the Gamma one.

 

When its done I will post a new image recorded in Log.

 

Thanks,

Cesar Rubio.

[David Mullen ASC]

That said, it's a grey area because LOG mode in a digital camera is just a simulation of a LOG scan of film by manipulating the gamma, it's just a specific type of gamma set-up.

 

Cameras like the F23 have gamma modes that come close to a flat LOG look, but still fall within certain broadcast specs, like HyperGamma. And it has a "S-LOG" ("Sony LOG" I assume) mode that is similar to the Genesis cameras "Pana-LOG" mode and the Viper camera's LOG mode, though they all are slightly different than a film Cineon LOG scan.

[Frank Barrera]

That's why I hear. I really don't have any experience with film. So you might be right.

 

CR.

Oh he's very right. There is no "might" about it. I (as well as thousands of others) have shot many beautiful one light to video transfers of 16 and 35 film images that needed little to no correction what so ever. Don't get me wrong, I think HD (RED,F23,EX-1,HVX200 etc) is awesome and look forward to more and more advances but let's be serious here. Film negative is the standard that all electronic formats seek.

[Sam Wells]

If the discussion is focused on RAW capture the "off the bat" and "needs correction" are not really the right terms as "off the bat" in terms of saturation, gamma, color balance presumes a processed image...

 

A good CMOS DSLR can produce very good saturation in jpeg, jpeg-tagged RAW or RAW straight out of a good RAW converter.

 

-Sam

[Charles Haine]

Not commenting on the test itself, here - but more on the use of RedCine.

 

RedCine is (IMHO) not to be used for correction at all, but is more like the combined scanning/developing stage of a film process.

 

If you exported log files and did not CC them afterwards, then your results will have to be way off. It's like showing uncorrected filmscans.

 

Also, it seems you may not really have looked into how to make the images pretty noiseless on the RED at low exposures, but that (partly) relates to the above problem. I am not talking about "noise reduction", just a question as to how you used RedCine...

 

As to the HVX, it looks like the black ped was set way low to lose noise with a pretty high saturation value, but you cannot really judge the cam from those settings...

 

I'm curious as to how your 35mmm workflow was?

 

Just my 2c

 

 

Gunleik

 

Please see the other discussion (RED vs. 35mm shoot-out) in this forum. It describes the workflow for all processes.

 

Additionally, it has links to the footage output REC. 709 and color corrected. I did the above linked output in LOG instead of 709 at the request of some RED owners. The belief was that it would show more latitude (the REC 709 output fell apart about a stop and a half over and under), and it did, indeed, show more latitude to output in log. However, as you can tell from the footage it also looked pretty noisy.

 

Fascinating conversation about CCH vs. CMOS, thanks for the info.

[DJ Joofa]

[A] Cineon LOG scan of a film negative does, it's not meant for direct TV viewing, it's meant for color-correcting and film-out work. The point of a LOG scan is to optimize the number of bits for the dynamic range of film, to preserve as much luminence information as possible.

 

 

The scan is never log, it is always linear, it is the negative data that is already log. There is a difference between the two. Therefore, the linear scanning is not optimizing the number of bits. It is the response of the film to the incoming light that is doing that. The relationship of film data to incoming light is logarithmic in nature and a typical DI scan just *linearily* scans that already non-linear log range.

 

A good CMOS DSLR can produce very good saturation in jpeg, jpeg-tagged RAW or RAW straight out of a good RAW converter.

 

-Sam

 

The maximum attainable saturation (closest approach to the corresponding spectrum color) for a material having a specified dominant wavelength and visual efficiency (luminance factor) will be attained if the material has a spectrophotometric curve which is everywhere either zero or unity, and which has, at most, two transitions between these values within the region of visible radiation.

 

Since, most materials will not have this type of reflectance response (i.e., the characteristic should display only one continuous reflectance band or only one continuous absorption band in the range of visible wavelengths) the recorded saturation will always be (relatively) poorer.

 

Difference in chromaticity are less apparent for dark colors than for light colors. Therefore, the perfection of color reproduction is less important in areas of smaller luminance values.

[John Sprung]

CCD based cameras offer better color rendition than CMOS out of the bat.

CCD and CMOS both use the same photoelectric effect in silicon, so there should be no difference in how they handle colors. Any difference would have to happen before the photons get to the silicon. In the case of single chip cameras, there could be differences in the RGB filter dyes on the surface of the chip -- the Bayer pattern or stripe layer. In three chip cameras, it would be the dichroic layers that could vary. In general, the dichroics of a three chip should have less crosstalk than the dyes of a single chip.

 

 

 

-- J.S.

[Charles Haine]

Just a quick note to clarify some more about how the footage was processed.

 

In the telecine, we weren't going for applying a "look" per se, but were balancing our highlights and our shadows neutrally and trying to correct as well as possible for over and under exposure by watching our scopes for clips in either direction.

 

Of course, this was on a CRT and with hardware scopes, so the appearance of that footage (trancoded to H.264 and shown on an LCD) is altered a bit.

 

In REDcine I didn't have scopes, so I was using the historgram to bring the shadows and highlights as close to clipping as possible without loosing detail.

 

I think all of this footage could look better, personally; richers shadows, a little color contrast between shadows and highlights, some NR and some fleshtone diffusion would certainly help. But for the sake of the test, I wanted to leave it alone, since we were most interested in seeing where we started to loose information in the overs and unders.

 

One of the students asked if we were less likely to underexpose on the RED, since we'll have an accurate monitor and viewfinder (the viewfinder is gorgeous on the RED, i thought), but in reality it's rare that you underexpose by accident; generally it's only when there just isn't enough light. So, it's still interesting to see how the RED responds. Part of the reason we shot some of that street footage at the end was to see how the RED holds up on a common situation; a production that wants to shoot in the streets at night, guerilla style. And the RED held up wel.

 

In general, I was very impressed with the resolution and dynamic range of the RED; probably not quite as wide in latitude as most 35mm stocks, but then again, you aren't always using all that latitude either. Sure, it's nice to have it, but it would be nice to shoot 65mm for the resolutoin as well; even 35mm is a compromise of sorts between image quality and economics. The RED, with it's image quality sacrifes (and that isn't meant to be an insult, like I said, 35mm also has image quality sacrifies; even spherical 65mm does when compared to Imax, say), delivers tremendous images for the price.

 

If a ratio, of price to image quality was created and that is what was used to judge a format, the RED would win, hands down, among the formats that I have yet shot in terms of an owner/operator model. With rental rates where they are right now, on some jobs (high contrast setting like the desert, 35mm film-out desired, short shoot), 35mm might still cost less, especially with places like indi35 doing such great package deals for students.

 

I also had a tremendous time working with the camera; the assistant having a monitor with a good sense of focus, having an accurate sense of exposure on the monitor, all things that were a pleasure. Not necessary if you're used to shooting film, but conveniences that were appreciated.

 

I'm not sure if it's necessarily right for my students (right now, you need a very powerful computer to deal with the footage that doesn't really fit in every students budget), but it was really exciting to see it hands on.

 

thanks to everyone who commented, it's been interesting to hear other peoples thoughts on the footage.