Lines Per Millimeter Of Resolution Of The Red One Camera
Published : 7th September 2009
I tested the Red camera with the industry Century Precision Chart system. I used a 85mm Zeiss Super speed mark II lens at a distance of 14'2" at an f-stop of 2.8/4. Which put me at the 50x chart of table A. The results were not what the emotional Red people would lead you to believe. I took one 4k frame and enlarged it to study the chart using a macbook pro and preview.
The 85mm Zeiss Super-speed Mark 2 t1.3 tested at greater than 120lines per millimeter on a lens projector which far exceeds 35mm film or the RedOne. If someone has the Zeiss Lens Evaluator if would be great to know what the 85mm lens tests out at. But for my test the important thing is this lens double the resolution of both formats.
The chart for the center of the image at 4K was a lines per millimeter at 42.
The left side was the same 42.
The right side was not resolving the vertical but did resolve the horizontal which puts the right side of this build 17 Red one at less than 42.
I did the same test with 35mm 5217 and the same Zeiss 85mm lens and got better than 62 lines per millimeter in the center. I used a 50x Microscope to enlarged the single frame. The data conclusion is that the Red One camera with its 4k file has 1/3 less resolution than 35mm film. Contrary to the emotional red community the RedOne 4k camera is not equal to 35mm film.
Please replicate my test. I have the still files which are 48 MB each which are too large to attach. But we have uploaded them and Geoff is putting them someplace on the web.
Monday I will test the 5201 which should produce even high resolution than the 5217. It would be great if other people around the world who have access to the latest cameras could test these formats to get the data.
New York/Hollywood Roster
Michael... I'll let you tell Steven Soderbergh, Peter Jackson, Alex Proyas, Doug Liman (and a hundred others) that they are "being emotional". Maybe you had better check your methodology before calling them out like this?
Your results don't match the real world. Why do you think?
>>It would be great if other people around the world who have access to the latest cameras could test >>these formats to get the data.
I know the BSC was planning a major test sometime this yr perhaps someone on the list knows anything about it.
1st Asst Cameraperson - 1st Asst Kameramann - 1er Asst Opérateur
bvk-SSFV European based
Cinematography Mailing List Moderator - CML
There have been several independent studies (posted results) that have shown actual resolution of our 4K sensor to be 3.2K actual resolution. One major studio did many tests of 4K film scans and found the BEST they could get was less than 3.2K with ISO 50 film. So make sure your test is accurate (what was your methodology again?) before you post your "facts".
WA, NV & CA
Jim Jannard wrote:
>>There have been several independent studies (posted results) that have shown actual resolution of our >>4K sensor to be 3.2K actual resolution.
Another one was done in LA. a few months ago. The results should be out soon. Sometimes you have to wonder if the results are being suppressed?
Mako, Makofocus, S. Pasadena, CA
Jim Jannard writes:
>>On this board, c.com and RedUser.? Old news.
I have to say, Jim's figures pretty much agree with my empirical experience with both RED and 35mm film images, at least as far as Academy width goes. Full silent aperture is better. But Kodak set the original Cineon full res full aperture at 4096, or 3656 for Academy width, with the intention of it being an oversample.
Now, if you're talking about line pairs...
Personally, I think digital needs to blow right on past 35mm, and in a very cold, clinical, and unemotional way I'm very much looking forward to trying out RED's new modular cameras. Preferably either the FF35 or the 645, because I like a squarer aspect ratio.
"pondering the fate of those traded-in RED ONE's"
Santa Monica, CA
>> Sometimes you have to wonder if the results are being suppressed?
Having done a few of these and being in the middle of preparing some more web pages of tests I can say that nobody ever anticipates just how long something like this takes and the amount of work involved.
I'll be finishing the tests/comparisons that I have in the next week or so.
If anyone has any tests that they'd like to share then please email me details of the procedure and put the files on the FTP site :-
Site: ftp.cinematography.net or 188.8.131.52
Note the exclamation mark.
If you use a Mac you may need to use the numeric IP address but that's why it's there .
Geoff Boyle FBKS
mobile: +44 (0)7831 562877 www.cinematography.net
Particularly the recent post by Guy-Louis Mier, who has also accurately posted on this topic in the past.
Red.com, Ottawa, Canada
Michael Carmine wrote:
>> The chart for the center of the image at 4K was a lines per millimeter at 42.42 lines per mm is,
You're off by a vast factor. Your comment mixes line pairs per mm and lines per mm throughout. Even if you meant 42lp/mm it's still way out. You don't have to believe me when I measure around 3.2k lines, which is 144 l/mm or 72 lp/mm, but if you look back through the CML archives, you'll see Guy-Louis Mier measured very similar numbers.
Red.com, Ottawa, Canada
Emmanuel SUYS wrote:
>>I know the BSC was planning a major test sometime this yr perhaps someone on the list knows >>anything about it.
BSC was screening some behind the scenes at their New Equipment Show last month. Representatives said end of April would be the earliest to expect to see results.
From March British Cinematographer:
Sue Gibson BSC President:
...there will be a series of presentations around the country, in London, Cardiff, Bristol, Edinburgh, Manchester, Glasgow, and the Screen Academies where we will show these tests along with case histories, explanation of workflows and panel discussions.
>> Having done a few of these and being in the middle of preparing some more web pages of tests I can >>say that nobody ever anticipates just how long something like this takes and the amount of work >>involved.
Having done much smaller tests I'd have to agree. They can be a HUGE time suck.
I believe that when Adam Wilt tested the RED he got 2.8k to 3.2k, with the quality of lens being the deciding factor.
This is where I find out if I recalled incorrectly, because Adam will most certainly pop up and say so if I'm wrong.
3.2K, close enough as makes no difference. Assuming the active area is 24.4mm wide, then we’d be seeing 1600/24.4 = 65.6 lp/mm (that's 1600 line PAIRS, film-style; not 3200 LINES, TV-style or telecine-style!). That's measuring to the point at which the CTF drops to zero.
Practically speaking I didn't see any of the factor-of-two variation that can occur when a square-wave test signal at the Nyquist limit is 90 degrees out of phase with the photosite pattern; many cameras *will* show a noticeable high-frequency detail variance with slow pans when looking at such a chart, but not Red One. Between the excellent (IMHO) OLPF in the Red One, and the deBayering jiggery-pokery in REDCine, the CTF extinction occurs before the raw-photosite-based Nyquist limit. Diagonals can show some aliasing but in practice it's not a factor, and horizontals and verticals are about as clean as you can get.
Adam Wilt / filmmaker, MeetsTheEye / writer, provideocoalition.com
On PVC, the first test posted by Adam Wilt put Red's resolution at about 2.4 to 2.5K (and that was a "conservative estimate", according to Wilt). That doesn't seem out-of-line with a theoretical estimate of what a Bayer pattern sensor that's 4096H total could resolve on a B&W resolution chart.
As digital still photographers will tell you, if you debayer it right, you can really sweeten it to make the best of a resolution chart. The interest was to see how far it could resolve up-the-trumpet to sizzle closer at 3.2K. The only thing agreed upon is that the Red's not 4K, as the marketeers would have it.
Besides the few posts on this list, and PVC, I haven't seen a lot of other places that have presented independent, empirical results. As for the test done here in LA, I was glad that Red was added last-minute, as at first I'd heard they would not be part of it. I think we'll see some surprising results there - fine detail, low light, motion artifacts, all of it very fairly tested - so I hope results will be available.
I'm less interested in the megapixel war as the Stills Camera market undertook (yes I know that's not all that defines resolution), but want to see better ergonomics, more dynamic range, less noise, and less compression artifacts (one has to admit, for as highly compressed as Red's 4K is, when you consider the Mb/sec rate, its remarkable what the wavelet shrinking of the raw has done). Are there negative aspects to all that? Sure, there are always tradeoffs - no free lunch.
Perhaps one can summarize the 2 paradigms as this: it used to be camera makers would build a camera to "how high can we aim?". Here's one that turned it on its head and said in many areas, "how low can we go?" (price, data rate, what we can call "4K", how much we leave to post).
By the time the dust settles on Red we'll be on the next round of cameras from all manufacturers. The only thing will be all these old R1's being used "cheap" in the industry the way the f900's are still used today. And the rendering speeds in post increasing to deal with it. Let’s see in 4-5 years what's the current distraction and what prevails. C'mon Red users, recycle those cameras with a trade-in.
I'll be more than happy to see Varicam 1080, f23's and f35's being the new "cheap camera" too - and shooting with an old Viper a vintage undertaking. Pull out a windup Bolex and get stares of disbelief.
LA based DP
Much as with Geoff's rather direct comments regarding the 'what's next?' the read, and Mike's comments here, I have to say I find a lot of the discussions on the technical abilities of the various cameras rather depressing.
I said before, and I doubt I was the first to say it; this is 'show business', not show perfection. It's all too easy to look for issues with any given camera, and see problems where there may not actually be any, as with my recent comments on the Red shot episode of Red Riding, which it's obvious now had nothing to do with the camera. Resolution is often not a key factor in how well a given image appears to be, although it can be...Lack of dynamic range on capture can cause serious problems, but with digital cameras being mainly wysiwyg can be less of a problem as exposure issues can be seen and dealt with while shooting.
As Lucas says frame wide wavelet compression causes high frequencies to be attenuated, rather than causing blocking as with block based compression (DCT, etc). But you really don't want to concatenate compression as that will give rise to serious problems, especially when the compression level is high. So why get into a shower-room 'mine's bigger, longer, thicker that yours' comparison competition when they all can do the same job, but differently?
Many years back I worked with Peter Swinson at Cintel, when I was technical director there, and he has a real love for film and quoting its absolute potential capture resolution, but then did some great experiments showing how 4K images could be compressed to below SD storage sizes, and still have far more image detail than an SD captured original.
Nyquist does make a serious difference. But Peter also like to look to the future and suggest that you need to capture images for tomorrow, when better display technologies exist, and can show more detail and resolution. I think no 'business' would ever finance such an approach.
It's happened in the past with TV programs shot on 35mm because there was no serious alternative. But it's very unlikely to happen again, and while film negative can capture high resolutions, the process required to undergo distribution printing causes a lot of high frequency attenuation, with most normal prints, rather than those used for premier screenings, showing a lot less than 2K resolution (approx 1.2K), as has been shown repeatedly, and yet, the random nature of film grain, and the human mind's ability to remember detail from one 24th of a second image to the next, means these measured numbers don't really tell the story.
And as most images are moving, this same effect holds true, in a different way, with digital images - captured and projected.
All I'm saying is it's good to know the foibles of any camera system to help get the most from it, but please don't try to hold up a set of figures as being the reason to shot with a given camera or not - they just don't mean that much to the final image.
The Art & Science of Digital Imaging
Mark Doering-Powell wrote:
>>I'll be more than happy to see Varicam 1080, f23's and f35's being the new "cheap camera" too - and >>shooting with an old Viper a vintage undertaking. Pull out a windup Bolex and get stares of disbelief.
Well said. What's around the corner is very encouraging- and still there was another PLX2000 "film" festival this year.
The Nuys, CA
>> On PVC, the first test posted by Adam Wilt put Red's resolution at about 2.4 to 2.5K (and that was a >>"conservative estimate", according to Wilt).
The article clearly says "I'm pleased to report that I see detail extinction at about 3.2K, confirming the numbers RED and others have claimed. " Why did you choose to ignore the second half of that test?
Jannard isn't selling subprime loans, nor did he concoct a Ponzi scheme. He has made a pretty successful attempt to make a cutting edge motion picture video camera at a very competitive price. I would think you would want someone like this in the game. Even if you choose to not like his product, that product is a driving force and will heat up the market. You will benefit.
>> I'm less interested in the megapixel war...
What....? Weren't you just being critical of RED marketing a 4k camera? So you're interested in a megapixel war, but...not interested?
>> ... it used to be camera makers would build a camera to "how high can we aim?". Here's one that >>turned it on its head and said in many areas, "how low can we go?" (price, data rate, what we can call >>"4K", how much we leave to post).
I'm not sure if you meant that to be as hostile as it sounds. Is there something wrong with making a product that punches higher than its weight class? Isn't that a basic rule of the free market?
I am using the Century Precision Optics Chart and all numbers relate to that system. Arriflex has a new no industry system which will be put forward to SMPTE technical committee this year. My numbers are right on and I do not work for RedOne.
There are other people that are getting the same numbers throughout the world. Pixel lines do not equal resolution lines.
>> but then did some great experiments showing how 4K images could be compressed to below SD >>storage sizes, and still have far more image detail than an SD captured original
I remember those Peter Swinson demos at IBC and elsewhere very clearly. I think those memories are one of the reasons, combined with the evidence of my own eyes!, that I'm less obsessive about avoiding compression than I once was. It's one thing to compress the bejasus out of a 1.4K picture and something totally different to do the same thing to a 4K image.
As the saying goes "the proof of the pudding is in the eating and the RED tastes pretty good”
Geoff Boyle FBKS
mobile: +44 (0)7831 562877 www.cinematography.net
How sharp, is info you sought a test you've done before, you thought.
They will say you are wrong. Your answer should be "so long" Your own test is the reality you've got.
I do tests to help decide on the complete package that I might use for a project. Resolution is part of the testing. I use the same tests every time. It is the basis on which I decide things like sensitivity, latitude, colours, and resolution. My tests work for me. I have seen tests done by others and often walk away thinking I learned little, an image is produced but do I have the info I need to actually fully utilize the equipment on a project. Presumably, others viewing my tests would think the same thing.
Many people have tested equipment and had results that are different than the "specifications" stated by the manufacturer. Surprise surprise. Imaginative specifications from those selling stuff, and selectively quoted by those who have investments to protect. Should your results be less, they will tell you that your results are wrong, your methods are wrong, and you are wrong. Consider the source, and the product's history, then only trust your own tests. The 2K results that many people have found are absolutely real for those who did those tests. And they will proceed knowing those results are correct. The 4K, or 3.2K or 2.8K, or whatever number results are equally true for the people who did those tests. I would hope they are not just repeating what they "heard" somewhere but actually did their own testing.
It is really important to know what you wish to find out and develop a test method that gives you that info. For long term comparisons, your own testing needs to be accurate and repeatable. Once you have that, then you will learn. The most important result of any testing is the information learned being translated onto the screen to tell a story containing images that are creatively appropriate. If things look as you wished they should look, then the info you got from your tests is correct for you since you utilized the info to create the images you wanted. If not, find out why, which you cannot do if you merely believe someone else's tests.
An advertised 4K camera that your tests show has 2K resolution? Fine, for you it is a 2K camera, nothing more. A result many have seen. Remember though, if you ignore other's tests, expect that they should ignore yours. Which would be preferable to some salesmen telling any cinematographer they are wrong and their tests are wrong because the results don't match the "generally agreed" result or the advertising result. Blindly trusting anyone with a financial interest is a very questionable practice.
Resolution tests are easy and when done with consistent methodology are a very accurate comparison tool that can be used over a long term.
Trust your own testing, not some else's
Director of Photography
>>The 2K results that many people have found are absolutely real for those who did those tests.
One benefit of resolution testing, as you point, is that they're easy and generally very accurate. They're also not likely to achieve a false positive result where the measurement gives a higher reading than is either possible or accurate. However, there are many things that can limit the number you achieve while testing resolution, so I always like to think of any test as giving a minimal value, rather than a maximal value. There's nearly always something legitimate you can do to get a higher number out of a resolution test (better lens, sharper T stop, more accurate focus etc.)
Indeed, it's always best to perform your own testing, but that does somewhat put the onus on yourself for your testing to be accurate and repeatable, and also not to make mistakes with your math on any calculation you perform. If you get a test result that is way off what other competent individuals are getting, you need to re-look at your methodology and math and ensure you've not made a mistake.
Of course, every time we buy a product, we're trusting someone who has a financial interest in that product. But trust cuts both ways and there's no benefit for any manufacturer to fabricate test results, especially on something as easy to measure and verify like measured resolution.
Red.com, Ottawa, Canada
Andrew Sobkovich wrote:
>>Resolution tests are easy and when done with consistent methodology are a very accurate comparison >>tool that can be used over a long term. Trust your own testing, not some else's."I'd add that resolution >>tests are just a slight part of it.
I was surprised to see most of the camera tests with digital "motion" cameras to be almost still shots, especially considering that most of the problems I've seen related to those cameras and especially the RED were motion related: weird pumping focus in some parts of the picture, rolling shutter, stuttering. I think a lot more should be done on that side of things since compression and the introduction of electronic shutter gives a lot of new artifacts to understand and master.
>>There's nearly always something legitimate you can do to get a higher number out of a resolution test >>(better lens, sharper T stop, more accurate focus etc.)
Gosh, the testing sequence I use is the same one I have used for well over 3 decades so I'm pretty confident in it. I changed the resolution portion about 15 years ago (the chart I used previously went missing) but it has remained unchanged since. I might not have done everything to get the highest possible numbers but instead was looking for results that would translate to the screen during day to day shooting. That is one of the differences in what we are interested in. You are looking for the highest numbers while I am judging the capability of tools to put what I want on the screen. Resolution is a usually a small part of the overall consideration of equipment for a project. For me, the RED test results of a 2K image is merely one of many considerations.
My results are not way off what others are getting as similar results of 2K have been posted on CML many times before. Since the release of the RED, similar results have been seen. While the "ninjas" get their knickers in a twist about any results outside of their beliefs, I know you are above that. I do appreciate that the results are not results that the manufacturer likes nor the same as they have advertised. I understand that. Graeme, I am not questioning your results, just as you are not questioning my results. It's OK for you to have different results.
Trust your own tests. I do. So should everyone else.
Director of Photography
Of course there's more to sharpness than limiting resolution.
Check out "The Pinhole Camera Revisited, or, The Revenge of the Simple-Minded Engineer" :
Apparently you can optimize a pinhole for maximum resolution, OR you can optimize it for a sharper-looking image!
If you want to get into it more, with the same sort of glass lenses most of us use (Art and his Lensbabies aside, grin), look into Zeiss' "How to Read MTF Curves", with the handy url :
There are two PDFs, there; download the first, then follow the link to "Part 2".
Adam "happily cutting 4K Red One material in a 1920x1080 timeline and
shooting 4K green screen inserts for XDCAM HD programs, and darned if
it doesn't look mighty nice whatever the numbers say" Wilt /
Filmmaker, MeetsTheEye / writer, provideocoalition.com
Read the conclusion, so to speak!
This link refers to
tests were shot by Graeme Natress ...