Welcome to our community

Be a part of something great, join today!

  • 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 :)

Let's talk sensor...

Status
Not open for further replies.

Brook Willard

REDuser Sponsor
Joined
Dec 28, 2006
Messages
5,392
Reaction score
3
Points
38
Age
41
Location
Burbank, CA
Website
www.brookwillard.com
Alright, let's talk about sensors here. Not Monstro... because we know nothing about it and it won't be around for at least a year after Epic arrives. Just Epic and the "X" sensor it'll ship with.

The place to start is with the RED ONE. Here's what we know:

The RED ONE full active pixel array is 4,900 x 2,580. The extra pixels on the sensor were originally specced to be used for black calibration, etc.

The maximum useful sensor area of the RED ONE is 4,520 x 2,540, sized 24.4mm x 13.7mm.

The maximum CURRENT recording resolution is 4K: 4,096 x 2,304. This was always specced as the primary shooting resolution of the camera: 4K. It is what we all think of when we think of 4K these days and is what I will base most of my calculations off of. The 4K recording area of the sensor is 22.11mm x 12.44mm.

To compare, three-perforation Super 35mm film gates are 24.89mm x 14.0mm. Granted, there are multiple gates available... but this will be the figure I use for the sake of this thread.

There is, of course, four-perforation Super 35mm film - sized at 24.89mm x 18.66mm.

When all relevant formats are put into a diagram, this is what it looks like:

5_1215155417.jpg


There are a few things to take away from this:

First, the RED ONE 4K recording area is significantly smaller than a normal three-perforation Super 35mm film gate. While there is generally some cropping applied to a film negative before it reaches final distribution, that figure changes and is not measurable for the sake of this discussion.

So, for the sake of this discussion, the RED ONE 4K we've all become so familiar with is only 78.93% the size of a normal three-perforation film gate. This will surprise some, but there is indeed a subtle field of view conversion factor with the RED ONE when shooting 4K. A 50mm lens on a RED ONE in 4K will have the apparent FOV of a 63mm lens on a standard three-perforation gate.

This will surprise some, but it is the nature of the beast. It has not bothered many over nearly a year of field production. That's not to say I like it, just that I accept it.

Another thing to note is the relation between the 4.5K look-around area of the RED ONE versus the 4K recording area. The look-around area is 21.65% larger than the recorded area. Something to remember.

The amount of "look-around" on a film camera varies on the viewfinder optics, ground glass and position of the moon in relation to the sun. Regardless, just know that that area is indeed outside of the size of the film gate.

So this is how it all stands today. There are good things and bad things about all of this... but it is what we have with the RED ONE. The great thing about the Epic is that it's an opportunity to improve upon the RED ONE with all of the lessons we [and RED] have learned along the way.

So, Epic.

The only sensor specification we have for Epic at this point is that its 5K recording resolution is planned at 5,120 x 2,880. Since RED will not put an optical viewfinder with spinning mirror shutter on the camera, one must assume that there will be a look-around area to this camera as well.

Since Epic is a thoroughly professional camera specifically designed for film production [not ENG/EFP, not to replace a video camera, not to be a great upgrade from an HVX, etc], I personally feel that every opportunity to make it the best cinema-style camera possible should be taken. Since we're only talking about the sensor in this thread, that's where we'll start.

While the RED ONE 4K FOV has not bothered many in 16:9 shooting, that is not to say that it isn't perfect. For the Epic, the actual 5K recording area absolutely should be the full size of a Super 35mm film gate - no more and no less.

That would mean that the 5K recording area in 16:9 mode should be 24.89 x 14mm.

Based on the RED ONE's look-around area's size in relation to the recorded area, Epic's look-around area should be 21.65% larger than the recorded area. That means that the sensor's look around area should be 27.47 x 15.45mm. By the same math, its resolution should be 5.6K: 5,650 x 3,178 pixels.

Having a sensor that large will result in vignetting in the look-around area... just as the "look-around" area in an optical viewfinder on a film camera will vignette on certain lenses. Regardless, the sensor must be that large in order to be the same size as Super 35mm film while still retaining the look-around area that people love so much.

When all of these dimensions get put down onto the same diagram, this is what we get:

5_1215155468.jpg


Epic is yellow. I was too lazy to add labels.

With Epic added, it's still very clear that RED is lacking something that film still has: a four-perforation frame. While it may seem old fashioned, many commercials still shoot 4:3. More importantly, many projects shoot anamorphic.

Now obviously one can shoot a 4:3 extract out of a RED ONE 4K frame for the sake of anamorphic. There is a conversion, of course, but many people are accepting of it. Ever see that put down on paper?

5_1215155487.jpg


For those playing at home, shooting a 4:3 extract on a RED ONE in 4K is 44.43% the size of a four-perforation film gate. That's a 225% conversion factor. Ouch. Think 16mm film.

Now when RED enables the inevitable 4.5K anamorphic mode, the size will be 53.88%. Relative to the Arri D-21, the RED ONE 4:3 extract is 59% the size. So this will get better... but it'll never be good.

Accordingly, since the Epic is a truly professional cinema camera, it needs to have a real 4:3 four-perforation sensor. No more conversion factors, no more nonsense, just a real film-sized sensor.

Besides, one cannot help but to point out the irony of a camera named "Epic" that can't shoot anamorphic properly.

To match a four-perforation film gate, the Epic's 5K recording area should be 24.89 x 18.66mm. By this math, the resolution would be 5,120 x 3,840.

Next, the sensor still needs to be large enough to have a proper look-around area. By the 21.65% conversion used on the RED ONE, the Epic's sensor should be 27.47 x 20.61mm with a full resolution of 5,650 x 4,238. That's 23.9 megapixels, kids.

Put into our favorite diagram, here are the final formats:

5_1215155532.jpg


Nice. Now that's how it should be.

That said, it's only a 23.14% pixel density increase. If the Epic's 5K sensor was exactly the same size as the RED ONE's 4.5K sensor, there would be a 28.43% pixel density increase. So not only should this size and resolution increase make the sensor more comparable to film, it would also result in a lower pixel density which means higher sensitivity and lower noise. We're not talking about shoehorning a zillion pixels into a small space... just making the space larger.

Those are my thoughts on the sensor. Obviously RED has their lips zipped about any specifications beyond the marketing materials we received at NAB, so it's hard to know where things stand on the camera's development.

Regardless, this is what I want out of this sensor. I may have buggered up some math here, but I think that the logic is sound. I just want a properly film-sized recording area with look-around. I'd prefer the sensor to be four-perforation instead of three-perforation as well.

Thoughts? Anybody from RED care to comment?
 
  • Thread starter
  • Moderator
  • #3
No, let's not. We're talking about the sensor in a camera that's right around the corner, not a sensor that has zero specifications and won't drop until 2010.

To those that are curious, I wrote this thread on an internet-free laptop today before Monstro was announced. It's unfortunate that I wasn't able to post this thread before the announcement.

Either way, I'd like to keep it on topic.
 
Yeah that yellow outline looks pretty Monstro to me.

Brook, I think what you have explained so well is the only logical approach on which to base a no-nonsense digital cinema camera like Epic. It covers all the bases and provides the maximum flexibility in shooting formats that could be made available.

I'd still like to see the spinning mirror shutter and optical EVF to take care of the look around issue, but I am slowly realising that this mechanical way of thinking does not fit into what Red is about, and what they are doing, which is pushing technological boundaries.

Have you done the calculations to determine whether a 5120 x 3840 pixel Bayer sensor will give us a true 4K output when debayered? This is what I have been wondering but don't really understand the math.
 
  • Thread starter
  • Moderator
  • #5
That sort of question is a good one for Graeme. "True" 4K or not... it matters not to me. Frankly, I'd imagine that most of Epic's post options will remain 4K for some time. If he writes a great debayering algorithm that puts out a [near] pixel-perfect 4K file... that's awesome.

I just hope the sensor ends up the right size.
 
the problem is that people look at the 4k red one #, and say it's a 4K, or 4.5K footage, or when they add the rest of the pixels would be 4.5K 16:9 footage.
it's not it would be more like a little less than 3K footage that is not as good as a 3K with no crop factor.
it would be much better to use a 3K 4:3 full frame sensor. if people were more smart!!!
 
All I care to, and can say, is thank you Brook. This is informative, perhaps one of those true great posts.

I will follow where this goes. I wish I could contribute more to this thread.
 
Nit-picky point, Brooke.

Lower pixel density only results in lower pixel density per pixel, with the same sensor technology.

Eg. comparing the 1ds III to the original 1ds, the Mk. III is less noisy, when compared at the same rez.

Increasing pixel density and downrezzing to the baseline rez can reduce noise.

If EPIC's 5k sensor was the same size as RED ONE's, and used the same sensor technology, we would expect a lower noise image at 2k or 4k.
 
  • Thread starter
  • Moderator
  • #9
Fair enough, that technical stuff is not my area of expertise. I see bigger pixels capturing more photons, resulting in lower noise in low light. That said, my primary concern is sensor size.
 
that's suppose to be the final ratio 4.5K 16:9. but it would be better if the diagram was of the actual image size after the resizing.to have a real perspective.
 
You can't get 16:9 4.5K, it doesn't work like that. The moment You crop it to 16:9 You go back to 4K... And if You do end up choosing to shoot 2.40:1 in the first place - why would You go 16:9 later? Just a thought...

PS: Any news from Apple support?
 
i said the "final" not the current or soon to be.
it's suppose at the end, to be 4.5K 16:9 (maybe 60fps - that one might flop)
 
too many reasons, i cant talk about all of them.

i'll give u some, a full frame sensor doesn't have a crop factor.
that will get u the best image possible, and a faster camera, and can use a real shutter and can record raw and faster the work and the process of the camera, and eliminate any artifacts and mistakes in the algorithm when removing the crop factor, better faster work flow, u don't have to re-size the image.
u'll have more view in the view finder and the monitors, so u can put all the info in the dark area outside the image, and have a clear view of the content being filmed and view all the info and settings at the same time.
u can record in 4:3 for 2.4 and use the rest of the image to ur liking.
a camera capable of shooting 4:3 with an option of 2.4 would be really stable recording data in 2.4 mode.

someone else can take it from here? and adding some more reasons.
 
By "full frame" sensor, you must be referring to the 8-perf 35mm still camera format (VistaVision).

There is no crop factor if you are comparing 4x3 Super-35 to 16x9 Super-35 cine formats, except vertically.

Anyone used to shooting in 35mm for filmmaking doesn't really have to deal with any crop factors when using the RED at 4K, other than the fact that at 4096 pixels across, you have an Academy-width image, not a Super-35 width image.

I'm not sure what you are referring to in regards to having to "resize" the RED recorded image. It's 4096 pixels across in 4K recording mode, which is a standard file size for D.I. work. And it has to be converted from RAW Bayer to RGB anyway, so the notion of a true "size" is a bit nebulous anyway.

I think the oversized sensor is a useful idea -- I like the lookaround area on the monitor and viewfinder outside the 4096 recorded area.

As for 4x3... I'm all for it because I like shooting anamorphic movies, but I also recognize that anamorphic makes up less than a third of all 2.40 shooting, so it's a niche within a niche. If someone just made some decent 1.34X anamorphic lenses, I'd live with a 16x9 sensor, which satisfies the majority of shooters.
 
Just simple deductive reasoning applied to Jim's announcements tells us EPIC should have a larger sensor than RED One.

EPIC is said to be 5K with the ability to derive pristine 4K+ from the Bayer data. RED One is 4K~4.5K with the ability to deliver pristine 3K or a bit over.

Jim has said that the Mysterium X upgrade for the RED One will still leave the camera at 4K. I don't think RED will be producing the Mysterium X in multiple densities. So assuming the same photosite density across all Mysterium X sensors, we could assume that EPIC has a larger sensor. The NAB literature and announcements said "full frame" sensor for EPIC. It's a vague description, but I also took it then as a larger size than the RED One. However, I'm still thinking that EPIC will have a 16:9 proportioned sensor for some reason.

I'm not sure what to make of the Monstro announcement. It's cool, yes, but what does it mean, if anything? For now I have many other concerns and desires to talk about more relevant products.
 
the "usable" recording area is not the same as the playback.the 4K or 4.5K would is not usable until it's re-sized, maybe someone else can explain it better.
the recorded footage has the information for a smaller size footage, it's the way the sensor works, it has a 1.5 crop factor, it's not full frame(it's doesn't mean a size, it is how the sensor works). the crop factor sensor is much smaller than the recorded footage, all the extra pixels are for holding more information more or less, u can't use the 4K or 4.5K footage without resizing to a 66%, or close give or take, to get the image.

u can see pictures and videos, or maybe tutorials that would be more useful to see how a crop factor sensor works, and see some pictures takin by a dslr, in full scale, and see them again scaled down, and u'll notice the difference, same thing with the camera footage, if u look at the 4k image(or part of it) and see the same part re-sized, u would notice the difference. u cant miss it.

i meant a full frame sensor, didn't mention the size, a full frame sensor with half the size would make a better image.
full frame sensors are 4:3.
 
Status
Not open for further replies.
Back
Top