Anamorphics on RED?

David Sweetman · September 16, 2006

By way of speculation, would it be possible to use anamorphics on RED? By what I understand the gate (sensor) would be different proportions...and just to clarify, I've heard it is to be super-35-sized, that means 4:3 (or close, point is it's not widescreen) right? - anyway I just want to know if it's theoretically possible is all.

Greg Lowry · September 16, 2006

By way of speculation, would it be possible to use anamorphics on RED? By what I understand the gate (sensor) would be different proportions...and just to clarify, I've heard it is to be super-35-sized, that means 4:3 (or close, point is it's not widescreen) right? - anyway I just want to know if it's theoretically possible is all.

The RED sensor is 24.4mm x 13.7mm, which is 1.78:1 (16x9). To use anamorphics with standard 2:1 squeeze, you'd have to crop or window the imager left and right down to the standard scope frame aspect ratio with the resulting loss of pixels. This would result in quite a small section of the sensor being used. As a result, the focal lengths of the anamorphics would effectively be increased. All in all, not a very happy situation. RED is promoting a top-bottom cropped frame approach, and given the aspect ratio of the sensor that makes more sense. At one point I did the math of the left-right sensor crop w/ anamorphics versus the top-bottom crop with spherical lenses. More RED sensor real estate is used for the latter which settled the debate for me. Unless you like the anamorphic look better.

Edited September 16, 2006 by Greg Lowry

Greg Lowry · September 16, 2006

Addendum to my previous post.

It would be more accurate to say that the field of view of the anamorphics would be decreased when using a left-right cropped or windowed portion of the RED sensor.

I'd also add that RED's approach may actually be more closely aligned with DCI's approach to 2.4:1 widescreen because I don't think DCI's spec calls for anamorphic projection lenses so anamorphically photographed content will ultimately be projected spherically (after being unsqueezed for the DI or master of course). I stand to be corrected on that.

Edited September 16, 2006 by Greg Lowry

John Allardice · September 16, 2006

The math works out this way

you'd have to vertically crop the width to 3048 to get a 1.2 pre-anamorphic aspect ratio

so 3048*2540 = 7,741,920 pixels

shoot spherical & crop the height to 1884

4520 x 1884 = 8,515,680 pixels

its a 9.99% increase in picture area. At that difference, I think it comes down to a matter of asthetic preference, as to whether you want your anamorphic artifacts or not.

As a sidebar, I still think it'd be a great idea to be able to tell the Red to pre-crop your picture to specific parameters BEFORE ENCODING, thereby saving your 27MB/s just for the info you intend to use.

Jim Jannard · September 16, 2006

its a 9.99% increase in picture area. At that difference, I think it comes down to a matter of asthetic preference, as to whether you want your anamorphic artifacts or not.
As a sidebar, I still think it'd be a great idea to be able to tell the Red to pre-crop your picture to specific parameters BEFORE ENCODING, thereby saving your 27MB/s just for the info you intend to use.

We agree with your math and conclusion. We are considering offering the option to crop before recording, although if we do it is likely to be an upgrade after release (free, of course). But the question begs to be asked if saving a little disk space (in REDCODE RAW) is worth giving up post choices?

Jim

David Sweetman · September 16, 2006

Thanks for the responses, that clears up a lot. So I don't expect the RED will see much anamorphics use...10% seems a sizable resolution hit to take unless you really wanted the bokeh and flares, plus it would make the lenses even longer? so even harder to get wide than on 35mm anamorphic.

Would it negate the low-light benefit of digital, with the light requirement of anamorphic lenses?

John Allardice · September 17, 2006

We agree with your math and conclusion. We are considering offering the option to crop before recording, although if we do it is likely to be an upgrade after release (free, of course). But the question begs to be asked if saving a little disk space (in REDCODE RAW) is worth giving up post choices?

Jim

It wasn't actually a question of expanded disk space, but more a question of lowering the compression ratio, with the cropped picture being only 67%(anamorphic) or 74%(spherical) of the original size, it would theoretically lower your 27MB/s from 11.7:1 compression, to 7.9:1 compression (or 8.7:1 compression) .....that's gotta be worth something :)

Jim Jannard · September 17, 2006

It would be visually lossless either way. I'd vote for the elbow room.

Jim

Greg Lowry · September 17, 2006

It would be visually lossless either way. I'd vote for the elbow room.

Jim

I'd vote that way too.

Greg Lowry · September 17, 2006

Would it negate the low-light benefit of digital, with the light requirement of anamorphic lenses?

The "low-light benefit of digital" is, of course, ultimately a matter of the ISO rating of the sensor, but anamorphic lenses are often slower and have less DOF than their spherical counterparts, so that's definitely a factor if you're considering low-light work.

David Mullen ASC · September 17, 2006

Panavision has been developing a set of 1.34X anamorphic lenses for the 16x9 Genesis (and it would work on 3-perf 35mm) to achieve 2.40 without cropping. I heard a rumor that Joe Dunton was considering building a similar set, but with front anamorphic elements rather than rear anamorphics like Panavision is going to do.

Trouble with rear anamorphic elements is that it's harder to get it as sharp (although I guess since they only need to achieve a 1.34X squeeze, it's easier to make it sharper) but also that you don't get the classic anamorphic distortions and horizontal flares, which for some people is the appeal of anamorphic lenses.

Greg Lowry · September 17, 2006

FYI ...

The DCI spec for 4K 2.39:1 is 4096x1714; 2K 2.39:1 is 2048x858.

Maybe it would be best to adopt these formats for 2.39 (2.40) spherical capture too?

Edited September 17, 2006 by Greg Lowry

Greg Lowry · September 17, 2006

I pose the question in my previous post realizing that the DCI format applied to capture does not maximize the available spatial resolution of the RED sensor for 2.40:1.

4096 x 1714 = 7,020,544 v. 4520 x 1884 = 8,515,680 pixels. A big difference. But perhaps there is value in a standardized workflow.

My usual preference would be to go with max spatial res, so it's just a question for discussion.

Keith Mottram · September 19, 2006

Okay, I many be confused here, but surely by using anamorphics you will gain 'recorded' information, in the same way you would when shooting to film. Now obviously you sacrifice some of the chip area for this, but if it is only around 9% surely you would end up with a sharper projected image as well as enjoying any aesthetic benefits. Whilst I get a slight feeling of deja vu here this is a somewhat different case than the old anamorphics on S16mm or am i missing something?

keith