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 :)

Does Scarlet has 3 2/3" sensors? Or only one?

ericyoung said:
The most negative aspect of the electronic rolling shutter of RED's CMOS choice IMHO is not the jello effect which RED seems to control very well, but the issue with strobe and short duration lighting. How will Scarlet and other RED cameras deal with that?
Click to expand...

But film has the exact same issue with strobes or other short-duration light or images. Muzzle flashes on firearms a good example too. If a strobe or a muzzle flash goes off after a mechanical shutter has begun to wipe across the film plane, we see a partially-exposed flash or strobe.

The only real difference is that with film, the edge of where the exposed light ends in frame is a lot softer due to the edge of the mirror shutter being defocused. The edge of the shutter also wipes across the film plane at an angle instead of perfectly horizontal as it does on the RED CMOS.

Ultimately, I think the best real-world solution would be an adjustable rolling shutter that could be dialed from 0 (or the equivalent of a global shutter) up to some value that induces the desired amount of skew. To take it one step further, what if instead of a horizontal scan-line reset pattern, the user could define a reset pattern mask, to control how the electronic shutter behaves. We could achieve skew that is nearly identical to that of a mechanical shutter camera if the reset can be made to roll across the sensor plane on a 45 degree angle or such. ...I've suggested this before, but I don't think the technology is here yet.

That said, I'm glad RED has chosen the rolling shutter path. In most situations it helps give the camera's image a nice organic feel and more film-like motion that we just don't get from a CCD or global shutter imager. It will be nice when the reset times can be made faster though. Supposedly, Mysterium-X is about 40% faster. That was a number that was thrown out a few months back.
 
well said jeff

i am also glad scarlet will have rolling shutter, but only because i have confidence in RED to do as they say and eliminate the rolling shutter artifacts that we see in almost every other current CMOS cameras on the market. when rolling shutter is done right, it is the best
 
David Mullen ASC said:
Concerned is the wrong word -- obviously the people who made "Slumdog Millionaire" were OK with the picture quality from a single 2/3" 2K sensor... but I find it hard to believe that a single 2/3" sensor in a $3000 camera would outperform a 3-CCD 2/3" camera that costs $150,000. But until the camera is released and tested, we can't know for sure. I could be wrong.

But for now, let's say be hopeful that since that single Scarlet sensor is 3K, it should be able to generate a decent 1080P image, comparable to a 1.9K 3-sensor camera, at least in terms of resolution. But in terms of noise and sensitivity, the F23 is around 500 ASA and very low-noise even in 3200K balance -- if the Scarlet performs equally well at 500 ASA in 3200K, noise-wise, that would be a great achievement. We'll see.
Click to expand...

David,

Wouldn't raising the Nyquist limit from oversampling also be an advantage. From the resolution charts I've seen (mostly sub-$20k cameras) the MTF falls off significantly after 300-400lp/ph. However, the higher resolution cameras seem to hold that number higher, often near or past 540lp/ph.

Since the Redcode recording can also capture the detail, does the sensor's ability to not be impacted by Nyquist low pass optical and digital filters through 1080p and 2k add to the advantage?

David
 
If you hold a good MTF above 540lp/ph, you're into the land of aliasing, and that's not good. Ideally you want as high an MTF below Nyquist, with as low as possible above Nyquist. Optically, that's not going to happen. Digitally, you can super-sample and use a digital downsample filter. However, digital filters can ring, be sharp or soft, allow aliasing or not. There's not one perfect digital filter.

However, digital filters to downsample in post give you immense flexibility and choice that you don't have otherwise.

Graeme
 
Yes, however doesn't the optical low pass filter and the digital filter both attenuate the signal above Nyquist? The problem, as I understood it, was that when you Nyquist limit IS 540lp/ph, those filters will also attenuate a good bit of the signal below 540lp/ph. By having a Nyquist well above 540lp/ph, you can't you have a high MTF at 540, and the attenuation occurs in a spatial frequency range you don't care about? Does that let you make the optical low pass filter stronger? Isn't that one of the RED (and for that matter Canon 5DM2) advantages?
 
Yes, optical filters are "slow" and to get sufficient attenuation at high frequencies, you need to bring the filter in low so it can work effectively. That is indeed why sampling at very high resolutions, pushing the optical filtering needs higher, is a good thing. However...

It relies on good downsampling. RED One post production tools allow for good downsampling, and as you record full resolution raw, you can choose your downsampling in post.

The 5D2 "downsamples" in camera. And it does a measurably poor job of it.

Graeme
 
That is one intense discussion you guys are having..I'm trying to comprehend it!
 
Ok...since I recently started learning the limitations of Photoshop...I'm curious.

All digitial still cameras support down scaling. Do they use a Lanczos restricted sinc, sinc, bicubic, bilinear? I would have thought they had enough time to use a form of sinc.

From various posts I thought the part of the Canon's problem was sampling every third row, a much lower frequency than its two optical low pass filters. Is it algorithmic in the camera?

Which scaling filter does RED use?
 
I think the in-built scaling in most cameras is very basic. It's hard to do good scaling in real time. When we looked at test images to see what the 5D2, for instance, was doing for it's downscaling to "1080p" it looked to be skipping 2 out of every 3 rows vertically, and horizontally, it was just soft, not reaching anywhere near 1920 lines.

In the RED One, the scaling is done with nice, decent sized filters, using all the rows and columns that make up the RAW image.

Graeme
 
The row skipping was probably do to an inability to process 18M photo sites 30 times per second. I believe most cameras with oversampled CMOS sensors are less than 8Mp.

Wouldn't that be different that a higher frequency noise introduced in bicubic down scaling? I would have imagined RED and Canon used Lanczos. However, that is just a WAG.

Informed information would be very much appreciated!
 
Lanczos is not the best filter because there "is no best filter". What RED uses is a good filter for the purpose, and is tailored to the known nature of data from our sensor.

Row skipping 2 out of 3 also means that effective skew is 1/3. It also allows a 10fps sensor to be read at 30fps. It's also a very cheap way to downsample. There are many reasons why you'd use it, but quality isn't one of them.

Graeme
 
It wouldn't be. Just curious how RED does it, since it is the first widely used oversampled video camera without those compromises (at least that I'm aware of). However, I can understand if it is proprietary.

Thanks,

David
 
Well, with RED shooting RAW, you do it in post - where there are a variety of filters available in the RED tools, or if you make 4k RGB, you can downsample them whatever way you want in whatever app you want. We do do a special 4k to 2k combined demosaic downsample, and that algorithm is proprietary though.

Graeme
 
Not being a Red owner, but anxiously awaiting, cash-in-hand, for the Scarlet's release, are there any publications regarding Red's workflow that any of you could recommend? I read these boards, and while I understand a portions of it, I start to get lost in some of the Red-specific technicalities.

When you purchase your camera does Red provide you with detailed instructions and applications of shooting RAW? It sounds like when you get a Red camera you're also introduced to an entirely new workflow. Does Red explain and extrapolate this new workflow?

In cunjunction with purchasing a Scarlet I'm also going to be investing in new post hardware. I want to make sure I'm buying all the right equipment to compliment Red's workflow.

Besides these forums, is there anywhere I can go to learn more (details) about Red's workflow, system requirements, etc?

Thanks!
 
I remembered this thread because I was re-watching Heat (great movie) and look what pops up:

vlcsnap52270.png


Rolling shutter partial exposure... :)
 
David, the flares sometimes flash the whole screen and sometimes partially.

The edge of them are always inclined, so looks like the rotary shutter (perhaps not, too sharp of an edge)... So that led me to believe it was partial exposure.

David, do you know with what Anamorphic lenses Heat was shot? It isn't listed on IMDb.

I love the anamorphic look in Heat. I could spot some barrel distortion on some interior wide-angle shots, but only on shots the camera was really close to the actor (probably a very short focal length).

Thanks.
 
Impagliazzo said:
David, the flares sometimes flash the whole screen and sometimes partially.

The edge of them are always inclined, so looks like the rotary shutter (perhaps not, too sharp of an edge)... So that led me to believe it was partial exposure.

David, do you know with what Anamorphic lenses Heat was shot? It isn't listed on IMDb.

I love the anamorphic look in Heat. I could spot some barrel distortion on some interior wide-angle shots, but only on shots the camera was really close to the actor (probably a very short focal length).

Thanks.
Click to expand...

Panavision’s Primo Lenses
Super Hight Speed T1.4 50 mm Lenses
E SERIES 100mm/180mm at T 2.8
400 mm at T 3.5 and [11:1 and 5:1] Zoom lenses at T 5.6

Apparently!
 
You must log in or register to reply here.
Back
Top