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 :)

Why aren’t more camera sensors optimized for anamorohic?

Jon Pais

Well-known member
Joined
Jun 12, 2021
Messages
287
Reaction score
134
Points
43
The popularity of anamorphic lenses seems to be skyrocketing now that affordable sets from the likes of Atlas, Sirui, Venus Optics and Great Joy are becoming plentiful and receiving excellent reviews. Unfortunately for anamorphic shooters, it would appear that most camera sensors aren't optimized for the lenses, with Panasonic, the Kinefinity Mavo LF and ARRI cameras being notable exceptions. Blackmagic's Pocket 6K anamorphic mode crops heavily into the sensor, resulting in an area equal to micro four thirds, while the RED Komodo 6K s35 sensor has a pretty huge 30% crop in 4:3 anamorphic mode. So it was with great interest that I read this article from 2020 about Netflix's puppet fantasy The Dark Crystal, shot on modified REDs specially designed with a 1:1 ratio square extraction mode (4320 x 4320) from which to de-squeeze the anamorphic image shot on Cooke glass. Lots of shows and movies are being shot in anamorphic these days, including with RED cinema cameras, so I’m wondering why more sensors aren’t optimized for the lenses. I also think that more filmmakers would be interested in a Komodo with a sensor optimized for anamorphic than in a monochrome version of the same camera.

https://ymcinema.com/2020/03/12/net...ed-red-camera-with-cooke-anamorphic-i-lenses/
 
The popularity of anamorphic lenses seems to be skyrocketing now that affordable sets from the likes of Atlas, Sirui, Venus Optics and Great Joy are becoming plentiful and receiving excellent reviews. Unfortunately for anamorphic shooters, it would appear that most camera sensors aren't optimized for the lenses, with Panasonic, the Kinefinity Mavo LF and ARRI cameras being notable exceptions. Blackmagic's Pocket 6K anamorphic mode crops heavily into the sensor, resulting in an area equal to micro four thirds, while the RED Komodo 6K s35 sensor has a pretty huge 30% crop in 4:3 anamorphic mode. So it was with great interest that I read this article from 2020 about Netflix's puppet fantasy The Dark Crystal, shot on modified REDs specially designed with a 1:1 ratio square extraction mode (4320 x 4320) from which to de-squeeze the anamorphic image shot on Cooke glass. Lots of shows and movies are being shot in anamorphic these days, including with RED cinema cameras, so I’m wondering why more sensors aren’t optimized for the lenses. I also think that more filmmakers would be interested in a Komodo with a sensor optimized for anamorphic than in a monochrome version of the same camera.

https://ymcinema.com/2020/03/12/net...ed-red-camera-with-cooke-anamorphic-i-lenses/

I don't get it, raptor has an 8k 1:1 mode 4320x4320 built in, but that's still a massive crop, 1.90x which is alot more than raptor 6:5 mode, but yh, raptor at the very least seems to be otimized for anamorphic, i can't speak to komodo.
 
RED's VV format is very, very anamorphic friendly. But there's certainly been some optimized format related things in DSMC3 that some either like or won't like.

The good. Above S35 height sensor and support of every anamorphic squeeze ratio you'd like want. And their VV and S35 formats desqueeze into 4K spec requirements by various entities. So A+ there.

The weird. Based on feedback here and the FB groups, some still want to window down further for higher speed frame rates. DSMC2 had this ability. DSMC3 locks you into "best practices" in modern times which also creates quality results. I'm cool with the locked in mindset, but some won't be.

Now that full frame anamorphic glass is being released, some have expressed a desire for FF35 3:2 sensors. Will be interesting to see if RED ever goes that route or an even larger format sensor that would just mitigate that all together, which is what I would prefer. Generally I prefer the wider sensor as spherical filming is still pretty much what most of anything still is by a heavy margin.

2X and 1.8X work fantastic on VV. 1.5X fantastic through pretty good. 1.6X is in the same category there. I have yet to be inspired by any of the newer 1.33X solutions and 1.25X is a rarity. Mainly for me anamorphic is about 2X through 1.5X for a while. Which isn't exactly shocking as it's been my main choice for 20+ years.

I could see RED making something like a modern take on the Gemini sensor again, which was/is rather anamorphic friendly. And many love that sensor.
 
The next question is probably whether we can get an anamorphic sensor with non-square photosites... :)
 
Please explain...

…just as food for thought. Would it make sense to desqueeze directly on the sensor with photosites that are taller than wide? Probably not so much because the aspect ratio would be hard-coded to one specific anamorphic lens and the lenses’ effects would be mostly compensated also.
Or, could the anamorphic effects be achieved with a sensor with non-square pixels and regular spherical glass, instead having more expensive anamorphic optical designs in each lens? And even make good use of CMOS sensor smear for the characteristic wide flares?
 
Now that full frame anamorphic glass is being released, some have expressed a desire for FF35 3:2 sensors. Will be interesting to see if RED ever goes that route or an even larger format sensor that would just mitigate that all together, which is what I would prefer. Generally I prefer the wider sensor as spherical filming is still pretty much what most of anything still is by a heavy margin.
- Phil Holland


Marc Shipman-Mueller seems to think that 3:2 is a good choice for both s35 and LF:


The ALEXA 35 sensor has a 3:2 aspect ratio, just like the sensors of almost all our other digital cameras. Yes, it is true, the ALEXA Classic, XT, SXT, Mini, LF, Mini LF and AMIRA all have a 3:2 aspect ratio sensor; the only exception is the ALEXA 65 with its 2.1:1 aspect ratio sensor.

Why 3:2? Because it is a very good compromise between a very wide sensor, which you would want for shooting with spherical lenses for a 2.39:1 deliverable, and a very tall sensor, which you would want for shooting with 2x anamorphic lenses for a 2.39:1 deliverable. Those are the two most extreme shooting formats for sensor width and height.

So what is 4:3? 4:3 is a left over term from the film days, defined originally in 1917 by the SMPE (before they had the “T” in their name) as the aspect ratio of the film negative's Silent Aperture. Later, this would also be called Super 35. In 1932, the AMPAS defined the smaller Academy Aperture, which also has a 4:3 aspect ratio, known afterwards as Academy Regular 35. When the first ALEXA became available with a sensor taller than 16:9, we marketed it as a “Cinematic 4:3” sensor, even though it really had a 3:2 aspect ratio, in order to differentiate it from the 16:9 competition. But, really, there is nothing 4:3 about it, nor is there any modern deliverable that requires 4:3.

So what sensor area do you need for 2x anamorphic lenses, if not 4:3? For anamorphic 2x lenses you need a 6:5 aspect ratio area on the sensor. The target deliverable aspect ratio for 2x anamorphic lenses is 2.39:1 (don't get me started with that 2.35:1 stuff; it has been 2.39:1 since 1970!). Since a 2x anamorphic lens will squeeze the 2.39:1 image by half, you end up needing a 1.195:1 area on the sensor. For simplicity's sake, we rounded that up to 1.2:1, which is the same as 6:5, to stay with full integers.

I can’t imagine wanting to deliver wider than 2.39:1…
 
Last edited:
Marc is generally correct with the mindset in my mind, but they also are in a different situation with camera bodies. They don't offer and entry level body and really only are making two sensors for the general public, one of which is their current flagship. And in my perspective if you are looking at ARRI bodies or even Sony V2, you are mainly looking at V-Raptor.

The big ones for me and everybody are a sensor that covers these spherical ratios:
- 1.9:1
- 2:1
- 1.85:1 (most just extract from something closeby now)
- 16:9 (1.78:1)
- 2.39:1
- several other taller 2.....:1 variants

And for anamorphic varients:
- 6:5 (2X, needs to be at least S35 in height)
- 4:3 (2X, needs to be at least S35 in height, 1.8X)
- 1:1 (2X, needs to be at least S35 in height)
- 3:2 (1.8X, 1.3X)
- 16:9 (1.5X)
- 1.9:1 (1.3X and 1.25X)

All of that is within V-Raptor and relative to the ideal format sizes you'd want.

For full frame, in 2023 it will desired for 1.5X and 2X beyond S35 in height a bit more. The resolution disparity aside, the LF and Venice sensor do give you 3.94mm and 2.33mm taller, but also a sensor that is 4.26mm and 5mm narrower. Leaning into the concept that well beyond 90% of all production is spherical, that wider sensor ratio is likely more ideal.

But V-Raptor today is taller that the Alexa 35 in height and significantly wider.

Oddly I think I posted here a while ago, before the LF's release that if ARRI made the 16:9 variant of ALEV III for the LF that would have been the perfect move for them, though the max frame rate would have suffered. Mainly because a good chunk of Alexa 65 productions were using the 16:9 mode for 5K and I think many were hoping for that. Here's what that would have looked like against V-Raptor:
https://phfx.com/tools/formatCompar...lexa+65+Alev+III&formatB=1.78:1&focalLengths=

And just for reference here.

V-Raptor 8K VV and A35
https://phfx.com/tools/formatCompar...a+35&formatB=4.6K+3:2+Open+Gate&focalLengths=

V-Raptor 8K VV and LF
https://phfx.com/tools/formatCompar...+Mini+LF&formatB=4.5K+Open+Gate&focalLengths=

V-Raptor 8K VV and V2 8K
https://phfx.com/tools/formatCompar...Venice+II+8.6K&formatB=8.6K+3:2&focalLengths=

RED has 6K and 8K S35 cameras as of now for DSMC3, which do have anamorphic support, but again with the majority of production being spherical, work far more ideally for that. In earnest though really on V2 8K has an edge for the resolution and small format size gains for anamorphic, but in every other format VV has the edge. LF is negligible IMO due to the height and resolution disparity and I think VV has the edge in pretty much every other format as well.

Everybody is moderately working within the general image circle coverage spec and figuring out what to do with it. VV hits a nice sweet spot for me when it comes to spherical and anamorphic as well as just an ideal format for most finishing and shooting scenarios. And with the high frame rate now it's very useful for darn near anything. But I personally wouldn't mind an even bigger sensor both in height and in width, but I also wouldn't mind a higher resolution one either. That's more about what I'm looking for though. I think I was also one of the few looking for an 8K solution prior to VV even being a thing and everybody thought I was crazy then too.
 
Marc is generally correct with the mindset in my mind, but they also are in a different situation with camera bodies. They don't offer and entry level body and really only are making two sensors for the general public, one of which is their current flagship. And in my perspective if you are looking at ARRI bodies or even Sony V2, you are mainly looking at V-Raptor.

The big ones for me and everybody are a sensor that covers these spherical ratios:
- 1.9:1
- 2:1
- 1.85:1 (most just extract from something closeby now)
- 16:9 (1.78:1)
- 2.39:1
- several other taller 2.....:1 variants

And for anamorphic varients:
- 6:5 (2X, needs to be at least S35 in height)
- 4:3 (2X, needs to be at least S35 in height, 1.8X)
- 1:1 (2X, needs to be at least S35 in height)
- 3:2 (1.8X, 1.3X)
- 16:9 (1.5X)
- 1.9:1 (1.3X and 1.25X)

All of that is within V-Raptor and relative to the ideal format sizes you'd want.

For full frame, in 2023 it will desired for 1.5X and 2X beyond S35 in height a bit more. The resolution disparity aside, the LF and Venice sensor do give you 3.94mm and 2.33mm taller, but also a sensor that is 4.26mm and 5mm narrower. Leaning into the concept that well beyond 90% of all production is spherical, that wider sensor ratio is likely more ideal.

But V-Raptor today is taller that the Alexa 35 in height and significantly wider.

Oddly I think I posted here a while ago, before the LF's release that if ARRI made the 16:9 variant of ALEV III for the LF that would have been the perfect move for them, though the max frame rate would have suffered. Mainly because a good chunk of Alexa 65 productions were using the 16:9 mode for 5K and I think many were hoping for that. Here's what that would have looked like against V-Raptor:
https://phfx.com/tools/formatCompar...lexa+65+Alev+III&formatB=1.78:1&focalLengths=

And just for reference here.

V-Raptor 8K VV and A35
https://phfx.com/tools/formatCompar...a+35&formatB=4.6K+3:2+Open+Gate&focalLengths=

V-Raptor 8K VV and LF
https://phfx.com/tools/formatCompar...+Mini+LF&formatB=4.5K+Open+Gate&focalLengths=

V-Raptor 8K VV and V2 8K
https://phfx.com/tools/formatCompar...Venice+II+8.6K&formatB=8.6K+3:2&focalLengths=

RED has 6K and 8K S35 cameras as of now for DSMC3, which do have anamorphic support, but again with the majority of production being spherical, work far more ideally for that. In earnest though really on V2 8K has an edge for the resolution and small format size gains for anamorphic, but in every other format VV has the edge. LF is negligible IMO due to the height and resolution disparity and I think VV has the edge in pretty much every other format as well.

Everybody is moderately working within the general image circle coverage spec and figuring out what to do with it. VV hits a nice sweet spot for me when it comes to spherical and anamorphic as well as just an ideal format for most finishing and shooting scenarios. And with the high frame rate now it's very useful for darn near anything. But I personally wouldn't mind an even bigger sensor both in height and in width, but I also wouldn't mind a higher resolution one either. That's more about what I'm looking for though. I think I was also one of the few looking for an 8K solution prior to VV even being a thing and everybody thought I was crazy then too.
Thanks, Phil!
 
there is a lot of movement in Anamorphic world and things are getting rally divers. I am not sure that new offering 1.33x anamorphic form Viltrox is step in good direction for general use. but for all the Raptor/Monstro owners this can be interesting option as new Viltrox is VV lens and it should cover Raptor entirely. Now i dont know if 1.33x should be considered real Anamorphic but for sure usig entore sensor could produce some interesting 3D pop. Just 6 months ago such idea was far out. we will see more new anamorphic avenues soon


https://www.youtube.com/watch?v=jopt-5x32dU
 
Couple of reasons......Because the average audience member prefers the look of spherical and the majority of home viewers hate 2:40 aspect ratio, ana drastically increases your VFX budget, adds difficulty and cost to your post workflow, requires an additional pan and scan delivery ($7k) which is the only version that will get distro unless your a heavy hitter film maker, the spinning, buzzie bokah is almost universally not liked on a side by side comparison on a big screen that was done in 2009 at Paramount. And just think about some of the best cinematography in the last 40 years and how much of it was ana?
 
The next question is probably whether we can get an anamorphic sensor with non-square photosites... :)

Nikon's D1X had photo sites with a 1:2 aspect ratio, if you recall. They were definitely not thinking of anamorphic when they did this. But if you can find a cheap D1X, maybe try it with some anamorphic lenses and see what you can get.
 
Couple of reasons......Because the average audience member prefers the look of spherical and the majority of home viewers hate 2:40 aspect ratio, ana drastically increases your VFX budget, adds difficulty and cost to your post workflow, requires an additional pan and scan delivery ($7k) which is the only version that will get distro unless your a heavy hitter film maker, the spinning, buzzie bokah is almost universally not liked on a side by side comparison on a big screen that was done in 2009 at Paramount. And just think about some of the best cinematography in the last 40 years and how much of it was ana?

Ana was always a hard sale. But man you make it sound like someone is getting married to the stripper :)
 
I'm fresh off a week of industry events as it's the season of such things and part of the reason I'm on filming hiatus until next month. Notably ending the week on the ASC Lens Day. Lots of talk of future endeavors on the sensor side among companies and optics manufacturers. A few manufacturers were in town during the week and not even at the event, some sit downs and lunches were had. Nice to provide feedback early on really.

I don't think we'll be getting "weird shaped pixels" again. Not rectangular at least. People are chatting about curved sensors, which is both interesting and unnerving as it pertains to optics. But I'll keep an open mind. I'm currently unsure if the juice is worth the squeeze when you consider all potential variables at play. But if it works and works well, we will certainly see it incoming.

All the new glass was shown in recent days. With some discussion about some fascinating new lenses coming in 2023. We are a bit spoiled for choice now and that will continue.

Remington highlights many of the considerations for anamorphic being avoided for some projects. In the 2000s there were a lot of considerations on the VFX side which were occasionally large mountains to climb to integrate seamlessly. I still enjoy anamorphic a lot, but I don't enjoy many of the anamorphic lenses on the market now. I'd say maybe 20% of what's on the market is even appealing to me and getting me to choose that over quality spherical lenses is another battle. But it would depend on the project for sure.
 
Karim, thanks for the D1X hint. I knew I had seen its Bayer sensor schema picture somewhere in the past but couldn’t find it. It would be a nice and inexpensive testbed for sure, but first, the resolution is too low to reveal any significant differences to regular sensors, and second it seems like all software that processes these images already desqueeze the picture before you can grab the raw non-square data.

The reason why I brought up non-square pixels is that when I first read the topic of this thread, “sensors optimized for anamorphic”, followed by the fact that all they did was cut off the sides, I thought: that’s a bit meager, isn’t there more you could do to optimize a sensor for ana? Well, after all there is probably not much else that really makes sense, but then I doubt that it’s even worth optimizing a sensor for it at all because you can get close to the same result with the regular sensors without sacrificing an entire camera just for ana use.
 
Last edited:
Back
Top