A look at anamorphics on the mx and the possibilities

[Karim_Nassar]

This is a work in progress, input is appreciated.

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Thread topics:

1. Anamorphic misconceptions
2. Recommended editing workflow
3. Demonstration of lenses squeeze factors
4. Shot on the mx with anamorphic lenses
5. Articles and resources on the subject

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

1. Anamorphic misconceptions

A. When unsqueezing anamorphic footage you get more resolution

This is misleading as you do get more resolution when desqueezing the footage in terms of a bigger file, but not on actual captured detail resolution.

On film you would actually get more resolution when using an anamorphic lens, because more of the film area would get used:

But on digital cameras, depending on the size of the sensor, there is actually less of the sensor that is used to record the image.
Resulting in a resolution loss compared to the sensor full capabilities.

For instance on the mx:

The image is squeezed horizontally by the anamorphic lens but only a portion of the sensor is used to record it.

B. Doesn't the anamorphic lens squeeze more detail horizontally that is then revealed when unsqueezing the footage?

It does not.
Photoshop pixel grid view gives us a handy tool to see what is actually happening when you unsqueeze the footage:

The same pixel information is stretched horizontally, there is no new hidden detail gain.
There is a resolution gain horizontally but only in file size.

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

2. Recommended editing workflow:

A. Horizontal desqueeze (bigger file than shot) vs Vertical squeeze (smaller file than shot)

Desqueezing the footage horizontally/changing the pixel aspect ratio is the recommended workflow.

Even if there is no gain in detail resolution by doing that, and you are stretching the footage, there is no resolution loss that occurs compared to reducing the height of the footage.

B. When to desqueeze?

That should ideally be done at the finishing stage from your grading and finishing tools.

If you have to convert from RAW earlier in the post chain, in order of recommendation:

- 16bit DPX if all your post tools support it
- EXR
- 10bit DPX as the third choice (the old standard)

C. Which debayer or resolution settings?

When using the red rocket:

- full resolution full debayer quality

When using a software decode:

- a neutral filter is going to be the ideal choice most of the time

You don't want a sharpening filter like Lanczos at this stage in the game on what amounts to an image enlargement.
Same goes for a softening filter like Gaussian, you're already softening the image simply by de-squeezing.

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

3. Demonstration of lenses squeeze factors

A. 2x squeeze

Spherical 85mm red one mx in 4k ana mode :

85mm + Bolex 16f 2x squeeze lens on red one mx :

B. 1.33 squeeze

red one mx 28mm
red one mx 28mm + slr magic 1.33x

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

 

[Karim_Nassar]

5. Shot on the mx with anamorphic lenses

Haywire / hawk anamorphics

http://www.youtube.com/watch?v=KFV0Uvzpz0o

Lockout / lomos

http://www.youtube.com/watch?v=zLTOzu8N1Pg

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

6. Articles and resources on the subject

Articles:

The DI Difference of Haywire:http://www.lightiron.com/blog/di-difference-haywire

Ressources:

Open dcp:https://code.google.com/p/opendcp/

Dcp information:http://www.dcpinfo.com/

Vintage anamorphic lenses:http://www.eoshd.com/comments/forum/5-anamorphic/

 

[Freya]

I love anamorphics and I would like to discuss with you their implementation on the mx.

Any contribution to the following info is very much welcome.
(if I made a mistake please do correct me)

I would like to start with a misconception/unclear subject that I have seen many times, on any camera:

"You can desqueeze the footage and you get more resolution"

This is misleading as you do get more resolution when desqueezing the footage in terms of a bigger file, but not on actual captured detail resolution.

On film you would actual get more resolution because more of the film would get used:

But on a digital sensor, there is actually less of the sensor that is used to record the image, resulting in a resolution loss compared to its capabilities:

The image is squeezed horizontally by the anamorphic lens but only a portion of the sensor is used to record it.

It's not true of all digital cameras, it depends on the sensor. The Alexa studio for example, is more geared towards anamorphic shooting. It's about the native aspect ratio of the sensor, just like it's about the native aspect ratio of the film. (things vary with different film formats too)

The best sized anamorphic lens for the red would actually be a 1.33x anamorphic adapter which would take you from 16:9 to 2.40:1.

If you are concerned with resolution, you also need to take the resolution that the lens can deliver into account too. Anamorphic lenses can often be softer than flat lenses traditionally. (Partly I suspect because they tend to be waaaay older lenses). Theres a whole bunch of new anamorphics coming on the market so it will be interesting to see if they are different.

Freya

 

[Karim_Nassar]

It's not true of all digital cameras, it depends on the sensor. The Alexa studio for example, is more geared towards anamorphic shooting. It's about the native aspect ratio of the sensor, just like it's about the native aspect ratio of the film. (things vary with different film formats too)

The best sized anamorphic lens for the red would actually be a 1.33x anamorphic adapter which would take you from 16:9 to 2.40:1.

If you are concerned with resolution, you also need to take the resolution that the lens can deliver into account too. Anamorphic lenses can often be softer than flat lenses traditionally. (Partly I suspect because they tend to be waaaay older lenses). Theres a whole bunch of new anamorphics coming on the market so it will be interesting to see if they are different.

Freya

thank you for your contribution much appreciated. I will correct the alexa part and update my post with the 1.33 squeeze.

 

[Karim_Nassar]

The best sized anamorphic lens for the red would actually be a 1.33x anamorphic adapter which would take you from 16:9 to 2.40:1.

What are we aiming for regarding aspect ratio for the theaters?

When I do some research online I find these: 1.85:1 / 2.39:1 / 2.40:1
Yet the mx ws ratio is: 2.33:1 and the ana mode 2.44:1

Any of these formats will fit on a theater screen? It does not matter?

thanks

 

[Freya]

The standard aspect ratios for film prints were most recently 1.85:1 and 2.40:1 (2.39:1 is basically the same thing but they standardised on 2.40:1, in practice you probably can't tell the difference)

The 1.85:1 aspect ratio was a crop on the 4perf film print. 2.40:1 was projected with an anamorphic lens and used the full 4perf film area.

Okay now, most theatres have now gone over to 3D digital projection. Almost always it's a 2k projector (and when it's a 4k projector it's usually a 2k DCP or something else in the chain is 2k). So modern "cinema" is now 2k for the most part (just marginally better resolution than a 1080p blu-ray) In fact often the 3d setup is left on for 2d films which marginalises the image further.

To add to that the digital projectors are AFAIK natively 16:9, or something similar. Thus when you have a wide screen apsect ratio on a digital projection the output is heavily letterboxed. I don't know what the exact resolution numbers are (I don't really want to know as it's going to be something horribly low)

Hopefully if UHD TV takes off then 4k will find its way into the theatres but for now that's the rather nasty state of play.

Freya

 

[Freya]

...On the upside, it means that it doesn't matter what aspect ratio you shoot at, you just letterbox (or pillarbox) it as appropriate.

Freya

 

[Björn Benckert]

I love anamorphics and I would like to discuss with you their implementation on the mx.

I would like to gather information and experience on the subject in this thread.
Any contribution is very much welcome.
(if I made a mistake please do correct me)

I would like to start with a misconception/unclear subject that I have seen many times, on any camera:

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

"You can desqueeze the footage and you get more resolution"

This is misleading as you do get more resolution when desqueezing the footage in terms of a bigger file, but not on actual captured detail resolution.

On film you would actually get more resolution when using an anamorphic lens, because more of the film would get used:

But on digital cameras depending on the size of the sensor, there is actually less of the sensor that is used to record the image.
Resulting in a resolution loss compared to the sensor full capabilities.
For instance on the mx:

The image is squeezed horizontally by the anamorphic lens but only a portion of the sensor is used to record it.

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

"Doesn't the anamorphic lens squeeze more detail horizontally that is then revealed when unsqueezing the footage?"

No.
Photoshop pixel grid view gives us a handy tool to see what is actually happening when you unsqueeze the footage:

The same pixel information is stretched horizontally, there is no new hidden detail gain.
There is a resolution gain horizontally but only in file size.

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

What editing workflow should be used then?

Desqueeze horizontally to get a huge file, grade, and then scale down according to the final output?
Or squeeze vertically giving a smaller more manageable file?

-----> colorist input needed <-----

This is dictated by your final use of the footage.
It makes more sense to desqueeze the footage horizontally as you do not lose vertical resolution. But you are stretching the footage horizontally.
By squeezing vertically you are loosing vertically resolution but you are not stretching the footage.

There also might be a quality gain by grading a bigger file when doing sharpening and what not and then scaling down.
But if a colorist could give some input on this would be nice.

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Desqueezing horizontally (bigger file than shot) vs squeezing vertically (smaller file than shot) in terms of resolution and detail

Is there a loss/gain in detail/sharpness/quality between the two techniques?

I wish I could test these things but sadly my vintage anamorphic lens is not sharp:
This is the image I get with a spherical lens in 4k ana mode (right click, open image in a new tab for full size):

And this is the anamorphic version (right click, open image in a new tab for full size):
Sadly my lens looses light and sharpness and casts a greenish tint, so I cannot accurately measure the possible gain/loss in quality by using any of the two techniques.

We need sharp anamorphic footage to judge the techniques properly.
But even if my anamorphic footage is not, out of curiosity I did this:

On the left I desqueezed it horizontally. 2816/2304 -> 5632/2304
On the right I squeezed it vertically and then upscaled the image to match the desqueezed image size. 2816/2304 -> 2816/1152 -> 5632/2304

Can somebody with a sharp anamorphic lens please do the same comparison?

No need to do the test.
For a fact "stretching out the image" to full height double width will give you a sharper image at the end resolution. However if you go for a lower work resolution or final resolution then downsampling the hight by a factor of two is a great thing to do as it helps the debayering (or actually cuts debayer issues in half). So thats what I do I shoot 5kana and scale to 3200 somthing to 1350 that is a great resolution to work form. I do not do much 4k work and it's far higher than 2k ... and the image quality is superb as the image got the downsampling.

 

[Karim_Nassar]

The standard aspect ratios for film prints were most recently 1.85:1 and 2.40:1 (2.39:1 is basically the same thing but they standardised on 2.40:1, in practice you probably can't tell the difference)

The 1.85:1 aspect ratio was a crop on the 4perf film print. 2.40:1 was projected with an anamorphic lens and used the full 4perf film area.

Okay now, most theatres have now gone over to 3D digital projection. Almost always it's a 2k projector (and when it's a 4k projector it's usually a 2k DCP or something else in the chain is 2k). So modern "cinema" is now 2k for the most part (just marginally better resolution than a 1080p blu-ray) In fact often the 3d setup is left on for 2d films which marginalises the image further.

To add to that the digital projectors are AFAIK natively 16:9, or something similar. Thus when you have a wide screen apsect ratio on a digital projection the output is heavily letterboxed. I don't know what the exact resolution numbers are (I don't really want to know as it's going to be something horribly low)

Hopefully if UHD TV takes off then 4k will find its way into the theatres but for now that's the rather nasty state of play.

Freya

thank you very interesting!

No need to do the test.
For a fact "stretching out the image" to full height double width will give you a sharper image at the end resolution. However if you go for a lower work resolution or final resolution then downsampling the hight by a factor of two is a great thing to do as it helps the debayering (or actually cuts debayer issues in half). So thats what I do I shoot 5kana and scale to 3200 somthing to 1350 that is a great resolution to work form. I do not do much 4k work and it's far higher than 2k ... and the image quality is superb as the image got the downsampling.

Interesting that squeezing the height helps with debayering, so that only works when exporting the raw file out of red cine x?
If done later, the debayering would have been already calculated right?

(can you please not quote the full post with the pics and just a phrase?)

 

[Andy White]

How are you posting the Ana files?

Are you exporting them from RCX or using native r3ds?

Rather than change the physical size of the file, you can change the pixel aspect ratio.

In terms of recent Ana (5K Epic) movies, the Total Recall remake springs to mind - and Haywire was Ana on the Red One (Light Iron did that - you might find this interesting - http://www.lightiron.com/blog/di-difference-haywire

 

[Karim_Nassar]

I have updated the thread with a new more complete diagram for pixel ratios and new squeeze diagram.
New editing workflow recommendation were added thanks to Jeff Kilgroe

 

[Karim_Nassar]

In terms of recent Ana (5K Epic) movies, the Total Recall remake springs to mind - and Haywire was Ana on the Red One (Light Iron did that - you might find this interesting - http://www.lightiron.com/blog/di-difference-haywire

thank your for this!
I will process all that information and add it to the thread later

 

[Karim_Nassar]

In terms of recent Ana (5K Epic) movies, the Total Recall remake springs to mind - and Haywire was Ana on the Red One (Light Iron did that - you might find this interesting - http://www.lightiron.com/blog/di-difference-haywire

I went through the article interesting read thank you. Added to the thread.

Before I complete the thread with the info I have 2 questions:

1. Why did they did this?

​We have to desqueeze the X to 1.5x = 42244.
We have to desqueeze the Y .75 = 1728

This leaves us with the desqueezed resolution of 4224x1728

Which gives the same 2.44:1 aspect ratio as when using a 2:1 ar except you lose vertical resolution.

2. What is this "center extraction"?
​
This Because the picture is 2.44, we are a little too big for cinema, so we have to crop about 1% off every frame and do a center extraction of 1% in order to make it fit the academy 2.40:1

thanks!

 

[charles lim yi yong]

in terms of resolution real estate.
Frankly i find the way the red epic handles the X2 anamorphic lens, really wasteful.
If anything the sensor is more suited for 1.33 anamorphic , but there is no option for it.

 

[Karim_Nassar]

in terms of resolution real estate.
Frankly i find the way the red epic handles the X2 anamorphic lens, really wasteful.
If anything the sensor is more suited for 1.33 anamorphic , but there is no option for it.

There is indeed no option for it, but on the mx when shooting at the 16:9 mode you get 4096/2304 pixels of the sensor utilized which is the maximum possible on the mx.
No waste there, and when using the 1.33x aspect ratio even when halfing the height you get a 4096/1732 pixels or a 5447/2304 image when using pixel aspect ratio correction.
It is big and only minimal cropping is required. (illustrated in topic 5)

On the epic you should get even more.

And you have monitors that provide custom aspect ratio squeezes so if you have one you are good to go.

I have the la7200 coming monday via rental, we will see how good of a footage I will be able to pull out of the mx.
I expect it to be pretty good :wink5:

 

[Peter Majtan]

Changing the pixel aspect ratio doesn't do anything to the actual footage aside of changing how it is displayed.
This was useful when shooting analog 35mm film during DI, as you would then film-out back to 35mm using the entire area of the film frame. Such a film is then projected using anamorphic glass which has the same effect as using the pixel aspect ratio.

This is however completely irrelevant for digital projection as all DCI-spec projectors have native resolution of 4096x2160 (or half x/y for 2K DCI).
Considering it is now impossible to buy DI film stock (effectively the film is dead for all given purposes other then archiving) - I do not recommend the pixel aspect ratio route. Why?

It is true that un-squeezing the ANA footage to double of the horizontal resolution does not magically add any "real" resolution, however there are many processes in the postproduction that do just that. The most obvious one is sharpening. Most people are aware the RED does not add any sharpening to the RAW footage (in contrast to ALL other non-RAW cameras, such as Alexa in ProRes mode). RED however highly recommends to add sharpening in post and it leaves it up to the user to chose both the amount and the algorithm/technology used for this. But regardless of choices one makes - this is where having more pixels makes all the difference. Another common process that benefits from having more pixels is color correction, where is reduces the "posterization" of gentle color gradients - such as in a sky, or smooth surfaces like car bodies...

So if IQ (in particular the sharpness / resolution) is of main concern, I recommend the following workflow with ANA footage:

- de-squeze the ANA footage by doubling the horizontal resolution
- apply first-light color correction and sharpening (these two steps can be combined in tools such as the RCXP)
- apply any image correction if necessary (stabilization, reframing, etc...)
- process any VFX
- master color correction
- crop the footage to your final aspect ratio*
- scale the footage down to DCI horizontal spec (4096 or 2048)
- fit the final image inside the DCI frame via letterbox (black bars on top/bottom of the frame)

This will result in max IQ - even if your final delivery is only 2K, or even HD...
The downside is the large amount of data and processing power (time) required for such a workflow.

Having said all that - in case of the Epic (and particular the Dragon sensor) - the ONLY reason for such a workflow would be if you insist on using ANA lenses for either the specific "look" (i.e. oval bokeh/flaring, coating, etc.) or the FOV these lenses provide (which in the past was unmatched by spherical glass, but this is no longer the case now). This is because using "normal" spherical glass and cropping the FF of the Epic sensor to WS will result in better IQ, not to mention that spherical lenses are sharper (and faster) then ANA equivalents...

* With digital projection there is no longer ANY restriction on the image aspect ratio, as in the end you need to fit the image into DCI spec frame (4096x2160 or 2048x1080) and it is up to you how much of the black bars do you add to your footage. The only nuisance is for the projectionists who have preset aspect ratios programmed for zooming and mechanical curtains...

 

[Karim_Nassar]

* With digital projection there is no longer ANY restriction on the image aspect ratio, as in the end you need to fit the image into DCI spec frame (4096x2160 or 2048x1080) and it is up to you how much of the black bars do you add to your footage. The only nuisance is for the projectionists who have preset aspect ratios programmed for zooming and mechanical curtains...

thank you very much, very interesting read, updating the thread.

I did a little research on what you said and came up with this:

Remember, theaters have fixed size screens and some have moveable masking to accommodate scope or flat. That means, even though you could have a DCI compliant resolution, it may not fit on the screen.

source:
http://dcinemaforum.com/forum/index.php?topic=104.0

So to be safe I assume we should fit the academic 2.39 standard?

 

[Karim_Nassar]

updated thread with new diagram and much more comprehensive topic number 5

 

[Jeff Kilgroe]

in terms of resolution real estate.
Frankly i find the way the red epic handles the X2 anamorphic lens, really wasteful.
If anything the sensor is more suited for 1.33 anamorphic , but there is no option for it.

What do you mean there's no option for it? When you shoot with 1.33X anamorphics, you want a 16:9 aspect frame. After the 1.33 de-squeeze (or stretch, as it should be called), you get a 2.36:1 image. 4K 16:9 on the RED One actually handles 1.33X anamorphics very well.

There is indeed no option for it, but on the mx when shooting at the 16:9 mode you get 4096/2304 pixels of the sensor utilized which is the maximum possible on the mx.
No waste there, and when using the 1.33x aspect ratio even when halfing the height you get a 4096/1732 pixels or a 5447/2304 image when using pixel aspect ratio correction.

Exactly. Hence the reason there is no specific option for it. It would be an option by name only or another label on what already exists.

On the epic you should get even more.

We do. Vertical resolution on EPIC is 2700 pixels high or another 396 lines of resolution. For 1.33X anamorphics, we would shoot 4800x2700 to get a 16:9 frame on EPIC. Although, most of the time, people will just shoot the 5KFF area of 5120x2700 to give a little extra side to side shift in post and guaranteed 2.40:1 framing. After the de-squeeze, we get 6810 x 2700 images to work with if we don't want to throw away any detail. They make for nice super-sampled 4K at the end of the post chain, even though they're a bear to work with along the way.

1.33X anamorphics are somewhat of an enigma anyway. They only give a portion of the optical effects inherent to anamorphics as we think of them and other than the optical characteristics, there is no reason to shoot anamorphically on the current selection of digital sensors. Or film stocks, for that matter.

 

[Christopher Mills]

so.. remind me

so.. remind me

which legacy lenses are 1.33x?
thanks, guys