Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Effective output resolution for RED Scarlet 3K Fixed

Erwin Vanderhoydonks

Erwin Vanderhoydonks

Well-known member
Joined
Jun 8, 2009
Messages
266
Reaction score
8
Points
18
Age
56
Location
Belgium
Website
e-vdh.be
Can somebody tell us the 'effective' output resulotion for the RED Scarlet 3K Fixed ?
The sensor resolution is 3072x1620. So I think that the RAW files resolution will also be 3072x1620. But after the DeBayer process, what will be the real effective resolution? What will be the usable resolution in Final Cut Pro after converting to ProRess 4:x:x ? People are talking about 2.5K but I find this a little low.
If you take an oldphoto camera (Canon 10D), it has a sensorresolution of 3152x2068 and the effective resolution is 3072x2048. 80 pixels less. Or is this not correct to think the same way about the RED sensor ?

Can someone give us a good answer ?

Thanks for your feedback...
 
At least as far as we know from previously announced specs, 3072 x 1620 is the effective recorded pixel format for the 1.89:1 digital cinema wide screen format. There may be added border buffers or look around area, but this is the actual recorded raw pixel format and you will be able to output a 3072 x1620 format frame as a 16 bit tiff at the highest quality level.
That is a different question than the effective optical resolution, measured not in pixels, but lines/mm or line pairs/mm. All Bayer pattern sensors optically resolve at best between 70% and 80% of their pixel resolution. But that is not the same thing as the pixel format of the image file.
 
Erwin Vanderhoydonks said:
Can somebody tell us the 'effective' output resulotion for the RED Scarlet 3K Fixed ?
The sensor resolution is 3072x1620. So I think that the RAW files resolution will also be 3072x1620. But after the DeBayer process, what will be the real effective resolution? What will be the usable resolution in Final Cut Pro after converting to ProRess 4:x:x ? People are talking about 2.5K but I find this a little low.
If you take an oldphoto camera (Canon 10D), it has a sensorresolution of 3152x2068 and the effective resolution is 3072x2048. 80 pixels less. Or is this not correct to think the same way about the RED sensor ?

Can someone give us a good answer ?

Thanks for your feedback...
Click to expand...

David Raspberry's response was spot on with the note that Bayer sensors resolve 70-80% of their total pixel resolution.

Now, as a RAW format, RED will output files that are 100% the pixel resolution of the sensor.

3K RAW Bayer will resolve down to between 2.09 and 2.4 K resolution. I expect that RED will deliver effective resolutions near the top of that range, but that remains to be tested.

So, the expectation is that RED 3K cameras will deliver plenty of resolution for both 1080p and 2K theatrical release.

Since most films today are being released at 2K, I think RED's 3K should be more than enough for the foreseeable future. Indeed a great many cameras used for high profile projects today "only" acquire 2K, like Arri's Alexa or less like the 1080p Sony HDCAMs used for Avatar.
 
Last edited:
Thanks for the feedback...
 
Ralph B. said:
http://www.reduser.net/forum/showthread.php?t=41121

This isn't definitive but it seems that the 70-80% might be older news at RED. They are really pushing in so many areas.

R.
Click to expand...

When I look at those graphs I see where RED improved things and its exciting...

but I also see a big drop off as tested resolution goes up.

Look at how there is a broad band at lower resolutions, and how that band narrows radically as resolution goes up. At some point the waveforms variation exceeds its average value.

One interpretation is that RED has found some good ways of getting usable image data out of the noise. Another interpretation is that the selection of subjects that can benefit from that level of resolution detail on this system gets more and more constricted, i.e. you need fairly high contrast detail to keep the camera resolving at the higher end.

In other words, I think that for ordinary picture material, i.e. not a test chart, 70-80% is still a good working number. Remember how a test chart is made: they are fundamentally high resolution black lines on white background.

By the time your camera system has stopped resolving that sort of detail, it has long since stopped resolving detail like individual strands of hair of nearly the same color or the texture of "solid" fabrics. I wouldn't call that usable resolution under most circumstances.


I will also note that there are other considerations. For example what resolution can your glass deliver? Most lenses can deliver in excess of 3K resolution... but 4k is really pushing it. You need fairly new glass and it has to be well maintained if you want to see 4K or higher resolution. (Time for another look at those S5i and Master Primes)

To be completely fair, eliminating or making account of such considerations could very well push these tests above what RED is claiming.

I don't want to understate what RED continually accomplishes, RED is a very exciting company.

They aren't magicians.
 
Most premium grade lenses can deliver 3k+ resolution on a 35mm sensor. Doing that on a 2/3" size sensor is a bit more difficult. Scarlet will probably be more limited by lens choice than the sensor/debayer process in most cases.
In any case, if the optics are up to it 3k format should easily resolve around 1100-1200 lines vertical resolution before downscaling, comfortably beyond the limits of 2k cinema projectors, and 1/3 or more better than any 1080p HD camera, most of which fall in the 600-750 line range.
 
David Rasberry said:
Most premium grade lenses can deliver 3k+ resolution on a 35mm sensor. Doing that on a 2/3" size sensor is a bit more difficult. Scarlet will probably be more limited by lens choice than the sensor/debayer process in most cases.
In any case, if the optics are up to it 3k format should easily resolve around 1100-1200 lines vertical resolution before downscaling, comfortably beyond the limits of 2k cinema projectors, and 1/3 or more better than any 1080p HD camera, most of which fall in the 600-750 line range.
Click to expand...

I forgot that we were talking about 2/3" cameras!

I wonder what lenses are out there right now for 2/3" that resolve 3K? Does anyone actually know? I'd start by guessing at Zeiss DigiPrimes, but I just don't know the 2/3" lens market that well.
 
Premium super 16mm and 16mm lenses will do well. Especially those of fairly recent vintage with modern coatings.
 
Still vs. Moving

Still vs. Moving

In a Bayer sensor camera quite a bit of sharpen is used, so you can make the MTF curve just about any shape you want to for the spec sheet.

But you don't want to over sharpen a movie camera since it will in part un-do the OLPF blur to avoid aliasing.

What looks "just right" is a still camera is too much sharpen for a movie camera.

One problem is that people are TOO close to their monitor, you should judge the sharpen adjustment by being 3x the diagonal of the monitor image away and set it to the minimum setting that results in a MOVING image where you see no aliasing on high contrast edges that are in focus.

Another issue in movie cameras is that you don't want the lens to resolve just one pixel, since the red and blue pixels are spaced wider and have a lower fill factor than the green pixels in a Bayer sensor. Thats one reason you need the OLPF, to span the space between the red to red and blue to blue pixels, otherwise you get chroma moire.

So the lens would only give you about 0.7 to 0.5 the pixel resolution above 60% MTF, and with some over sharpen you might read 70% to 80%, but the more you over-sharpen the more aliasing you will see at prime stop on high contrast edges, so you get aliasing "runners", more visable after down size from 3K to 1920 or SD if you do not blur the images first before you down size them.

At any rate, you should not adjust the image sharpness settings just looking at a still frame, rather you should look at the images resized to their final end use size since if you over sharpen at 3K then down size to HD you will get more noise and aliasing than if you under-sharpen at 3K then downsize to HD for end use.
 
Alexander Ibrahim said:
I forgot that we were talking about 2/3" cameras!

I wonder what lenses are out there right now for 2/3" that resolve 3K? Does anyone actually know? I'd start by guessing at Zeiss DigiPrimes, but I just don't know the 2/3" lens market that well.
Click to expand...

Don't forget about the RED mini-primes, either. They are literally made to be used with these cameras.
 
I doubt the S16 lenses will resolve to 3K. But old 16mm glass could always make a great deal.

The Fixed and miniPrimes seem to be exceptional quality - should do just fine for 3K.
 
Good 16mm lenses are typically sharper than most 35mm lenses.
Cinematechnic sold an older Angenieux 12-120 some time ago that measured 200 line pairs per mm on a lens projector, most of these old lenses would fall more in the 120 lp/mm range. More than enough for Scarlet. Jarred mentioned in a post that the mini-primes were measuring in the 120 lp/mm range. This was considered a "soft" lens for 16mm, and wide open they were a bit soft. The later T2.1 HEC version is sharp with good contrast wide open. The Cooke 9-50mm is also a great older lens with excellent quality.
Digi-primes are specified at 56 lp/mm at 90% mtf, a very high contrast ratio at that resolution. Typical lens specs use 50% mtf as the reference.
You can find many images shot with the G1/GH1 and old 16mm primes if you do a search. Some of them look excellent in every respect. One site that has a good range of examples is www. old lens.com. Here is a sample of a 60's vintage Kodak Ektar. This lens is sharp enough to clearly show the aliasing limits of the G1's 14 mp frame.

P1030491c.jpg
 
Alexander Ibrahim said:
I wonder what lenses are out there right now for 2/3" that resolve 3K?
Click to expand...

I've got a suspicion that the RED mini primes and the 2/3" fixed will get you there. REDs whole philosophy is built on high resolution and their lenses so far fall right into that performance parameter. I would entertain any evidence that RED would go in another direction, but I haven't seen it. They have shown a want and a capacity to deliver fantastic glass at a fantastic price and I expect them to only go further down that path.
 
Ralph B. said:
I've got a suspicion that the RED mini primes and the 2/3" fixed will get you there. REDs whole philosophy is built on high resolution and their lenses so far fall right into that performance parameter. I would entertain any evidence that RED would go in another direction, but I haven't seen it. They have shown a want and a capacity to deliver fantastic glass at a fantastic price and I expect them to only go further down that path.
Click to expand...

I expect the mini-primes will be an excellent value compared to what else is out there new or used. M4/3 format shooters have pushed manyof the better used 16mm primes to higher prices than these will sell for. The fixed brain will cost less than many high quality 16mm or 2/3"HD zooms do used. Think of it as a fully robotic remote control zoom lens with a free camera attached.
Only if you need very long or broad range zooms or very fast lenses, or just like playing with old glass like me, do you really need to consider the cinema model over the fixed.
 
Lots of issues with resolution that I've never thought about. Good thread.
So here is a practical resolution puzzle: I use an FX-1(1080i HDV) to shoot live theater. I edit in standard def and use the larger hdv frame to do 'close-ups' by resizing them in the sd frame. The editing software automatically resizes the hdv to 44% to fit the sd frame and I can go up to about 80% before the video gets soft.

Sooo, if I shot with the Scarlet fixed at '3k' and edited it in a 720p frame, how much could I 'blow up' a 3k frame and still have clean video? Or, to put it another way, would a 720p frame use approx. 50% of a 3k frame? Would it be something like 1440/3000 (1440/2,500) or should I be using line pairs as the measurement?

And given that some resolution is lost in HDV due to encoding/compressing, and of course the FX1-1 lens won't match the fixed scarlet's-I'm pretty sure-...

So factoring this stuff in, any guesses as to how much I'l be able to 'blow-up the frame?
 
You must log in or register to reply here.
Back
Top