Digital Bolex weeks away

[Barry_Green]

The D16 shoots native 2048x1152 2k pixel resolution in 16:9 ratio in a super 16 sized format. It will produce a pixel for pixel 2k format DCI compliant cinema format image for standard wide screen and 2.39 anamorphic formats without upscaling and without the digital artifacts that plague most low cost HD video cameras including the BMC's. Will it optically fully resolve 2k on a test chart? No. Neither will a 2k cinema projector. Neither will a 35mm film release print as projected in most theaters. No pixel based image format can optically resolve its full pixel resolution either as a camera or as a display period. It is not physically possible. It is a pointless criticism.

It is most certainly NOT a "pointless criticism"! Either you don't understand this, or you are deliberately trying to obfuscate the situation. So let me lay it out exceptionally clearly, and hopefully you will understand what is being said.

If you use a 2048x1152 sensor to make an image, you will lose approximately 30% of its potential resolution due to the demosaic process necessary. If you use three 2048x1152 sensors to make an image, you will lose approximately 0% of its potential resolution due to the demosaic process, and will end up with a much sharper image. Alternatively, if you use a single 4096x2304 sensor to make an image and downscale it, you will end up with an image that is much sharper than the debayerd 2048x1152 sensor.

If you then display those images on a 2048x1152 2k pixel resolution display or projector, the one from the 2048x1152 single Bayer sensor will be substantially softer on the display. Undoubtedly, without a doubt, without a question, guaranteed, unquestionably, period and end of discussion.

The D16 is specifically designed as a replacement for an S16mm FILM camera.

Understood.

Its artifact free resolution is comparable to a 2k finish from scanned 16mm film.

Says who? Where are the test results? Where is the chart that shows that it is artifact free? That's what we've been asking for, and as far as I know it has never been provided.

Apparent sharpness and true resolution are two very different things. A couple of years ago a RED shooter did a comparison of HD video from his Epic that was debayered at full quality and down scaled to 1080p vs the same video debayered at half quality and downscaled. The impression for most was that the half quality sample was significantly sharper, but it was in fact a cruder less well resolved image.

True resolution is revealed by a resolution chart. We still anxiously await seeing what the real results are.

 

[Barry_Green]

One of Bloom's comments in his review of the D16 was that it showed no aliasing or moire artifacts, none, zilch.

And did he test for it using the standard method of using a properly-shot resolution chart?

You're really going to have to forgive me for not taking anyone's word on this. People raved about how gorgeous and "artifact free" the BMC camera was, until of course John Brawley shot a resolution chart and showed that it had horrific aliasing on par with a first-generation DSLR. And after that, people started seeing the artifacts everywhere.

So -- consider this a broken record, but -- the amount of aliasing and moire in the video will be revealed when the proper tool is used to demonstrate exactly what the camera actually does. Until then, it is speculation.

You guys are still thinking in terms of the limits of conventional compressed video codec cameras.

I can't answer for harddrive, but I can assure you that I absolutely am not. I am exceptionally well aware of and versed in how image processing works, throughout the entire imaging chain.

Not what is ultimately possible with an uncompressed raw digital cinema camera using the highest quality state of the art processing taking seconds per frame to encode uncompressed high quality 2k DPX sequences that are still in the 2MB+ per frame range for DI work.

You may not realize, but I was one of the very first buyers of a Red One, back in 2008. Serial #364. I'm quite well aware of the raw workflow, and the differences between raw and uncompressed.

 

[harddrive]

I can't answer for harddrive, but I can assure you that I absolutely am not. I am exceptionally well aware of and versed in how image processing works, throughout the entire imaging chain.

Indeed, and I'm in full agreement with everything else you say in the last two posts.

The whole point about RAW is that it gives *FLEXIBILITY* into the post workflow - it does not necessarily give much (if any) advantage in image quality terms, not if you get it right in camera.

And that flexibility comes at the expense of datarate and HAVING to do a lot of post work. That may be a price worth paying - it's horses for courses. But don't think it's a magic bullet that will miraculously make images from a mediocre sensor look fantastic.

 

[jdv]

Fascinating discussion from where I'm sitting... So if I'm understanding what Hardrive & Barry are saying, the DB16 will likely resolve much closer to "true" 1080 HD than true 2K?

Other than Razz, does anyone else plan on getting a D16? Again, if that crank handle charged the battery, it'd be more attractive to me, but in any case I do think the 16mm size has it's place (a heck of lot easier to get focus for things like wild life shooting or family events).

 

[harddrive]

So if I'm understanding what Hardrive & Barry are saying, the DB16 will likely resolve much closer to "true" 1080 HD than true 2K?

My expectation would be somewhat worse. The generally accepted figures for Bayer are that final resolution will be of the order of somewhat less than 80% of the photosite count. So for a 1920x1080 sensor, think max output resolution of somewhere like 1500x850. Use the full frame (2048x1156), and it becomes more like 1600x900 - but then that has to be downscaled to 1920x1080 which can cause problems in itself.

If it's aimed at Digital Cinema, it's most likely to be the wider screen 2048x1080, so after debayering I'd expect around 1600x850. (In other words, 850lpph.)

BUT, a few caveats. With a 1920x1080 sensor the aliases start to become in band (ie can be caused by wanted detail, not detail finer than the system) and the only way to minimise that is with an aggressive OLPF - which Razz16 has stressed several times this camera uses. If that's true, the above figures are optimistic. The price of nearly complete alias suppression will be a further softening, and 800lpph is more likely, if that. We'll only know for sure when it's results are seen on a zone plate.

Other big caveat is that those figures only apply to luminance - not chrominance. A Bayer 1920x1080 sensor has red and blue arrays of 960x540 each - and that will limit it's chrominance resolution to about 1/2 of 1920x1080 in each axis. In other words, it's inherently a sort of 4:2:0 because of the sensor - regardless of processing, recording format etc. If the information is not there in the first place, you can't process it back.

If this sounds too theoretical, then look at Alan Roberts test of the XF100/105 - section 2.3.1 and Figure 1 - http://downloads.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP034-ADD65_Canon_XF100-105.pdf to see a practical test, and what the use of the 1920x1080 Bayer for 1080 output actually means. Basically, horizontal and vertical lines finer than about 700-800 lpph are giving coloured aliasing, corresponding diagonals are giving luminance aliasing. Use an aggressive OLPF to stop such aliasing (as Razz16 implies the DB does) and useful resolution must stop around the 700-800 lpph mark.

Compare with Alans zone plates of the EX1 and you can easily see what a good 1080 camera should look like.

....... but in any case I do think the 16mm size has it's place (a heck of lot easier to get focus for things like wild life shooting or family events).

But..... it depends on sensitivity. If we assume a sensor 4x the area of the DB, then for comparable DOF it would need an f stop 2 stops smaller. So f8 to match the DB at f4, yes? In other words, depth of field is dependent on sensor size and aperture. (All else equal.)

From all I've read, then low light is not one of the DB's strengths. I hear ISO 200 realistically, for a decent level of noise. Consequently, it's quite possible for a bigger sensored camera to likely be so much more sensitive that it CAN be used 2 stops smaller for similar noise levels, so achieve similar deep depth of field.

Whilst having the flexibility to be used much wider open with ND, and the possibility of shallow depth of field.

 

[combatentropy]

I can't tell whether y'all are flexing your math muscles for sport or if you actually think such small distinctions should sway someone's purchase. Next are you going to argue how many angels can stand on the head of a pin?

The Digital Bolex has the same res as the Ikonoskop and the Blackmagic Pocket.

 

[plnelson]

My expectation would be somewhat worse. The generally accepted figures for Bayer are that final resolution will be of the order of somewhat less than 80% of the photosite count. So for a 1920x1080 sensor, think max output resolution of somewhere like 1500x850. Use the full frame (2048x1156), and it becomes more like 1600x900 - but then that has to be downscaled to 1920x1080 which can cause problems in itself.

If it's aimed at Digital Cinema, it's most likely to be the wider screen 2048x1080, so after debayering I'd expect around 1600x850. (In other words, 850lpph.)

While we're on the topic of resolution, where can I see actual tests of the lens resolution of commonly-used lenses for video work? There are lots of discussions around the web of the spatial resolution of sensors. But coming from a still photo background where sensors are typically 24x36mm (i.e., the size of a 35mm film frame) I'm impressed by how much smaller video camera sensors are. Even "Super 35", used in many high-end cameras has a much smaller frame than that.

So as we enter the realm of 4K, cramming lots of pixels onto a small sensor, how do we know that the spatial resolution of the sensor doesn't exceed the capabilities of the lens? Most lenses give their best performance at the center, and usually when they're stopped down one or two stops from wide-open, so good lens tests are usually performed at a variety of f-stops and both center and edge. As a relative newbie to video where can I see some lens tests for common video lenses?

 

[harddrive]

I can't tell whether y'all are flexing your math muscles for sport or if you actually think such small distinctions should sway someone's purchase. Next are you going to argue how many angels can stand on the head of a pin?

Have you looked at the links I posted to Alans tests, showing actual results from the EX1 and the XF100 (former a "true" 1080 camera, latter typical of 1920x1080 Bayer)?

Do you really think the differences are just "small distinctions"!?! Seriously?

The differences very roughly equate to those between 720 and 1080 (it would be 1 megapixel and 2 megapixel for stills). I don't call that a "small distinction", and even less when there's so much talk now of 4K!

As for whether or not it should sway a purchase, then I'll leave it to any prospective purchaser to make their own decision. But I do think they should know the facts (which is why Barry and I are so interested to see the D16 on a chart, ideally a zone plate). That purchaser may well then decide other factors outweigh resolution issues, and go ahead. That's fine - it's their money, and yes, resolution isn't everything. But at least they are buying with proper knowledge of what they're getting.

 

[harddrive]

So as we enter the realm of 4K, cramming lots of pixels onto a small sensor, how do we know that the spatial resolution of the sensor doesn't exceed the capabilities of the lens?

But even with 4K, that equates to about 8 megapixel in the stills world, small by stills standards. And 720 is roughly equivalent to 1 megapixel, 1080 to 2 megapixel - pretty low resolution by stills standards. Which is why (in general) lenses for video don't need the same resolution they do for stills. Where it can become an issue is with long zooms, and with video they need to keep focus whilst changing focal length - not necessary with still photography.

And whilst video sensors have typically been far smaller than those for stills, a change is being forced by 4K, such cameras typically having larger sensors than in the past. That's got as much to do with diffraction limiting issues restricting usable aperture range as much else though.

 

[Prendi]

Jeez, they are still planning to make this thing? And Razzmatazz is still plugging it. Seriously, the world has moved on...

 

[pulpfiction007]

Jeez, they are still planning to make this thing? And Razzmatazz is still plugging it. Seriously, the world has moved on...

They've actually already made it, they are shipping as we speak.

 

[Razz16mm]

My expectation would be somewhat worse. The generally accepted figures for Bayer are that final resolution will be of the order of somewhat less than 80% of the photosite count. So for a 1920x1080 sensor, think max output resolution of somewhere like 1500x850. Use the full frame (2048x1156), and it becomes more like 1600x900 - but then that has to be downscaled to 1920x1080 which can cause problems in itself.

If it's aimed at Digital Cinema, it's most likely to be the wider screen 2048x1080, so after debayering I'd expect around 1600x850. (In other words, 850lpph.)

BUT, a few caveats. With a 1920x1080 sensor the aliases start to become in band (ie can be caused by wanted detail, not detail finer than the system) and the only way to minimise that is with an aggressive OLPF - which Razz16 has stressed several times this camera uses. If that's true, the above figures are optimistic. The price of nearly complete alias suppression will be a further softening, and 800lpph is more likely, if that. We'll only know for sure when it's results are seen on a zone plate.

Other big caveat is that those figures only apply to luminance - not chrominance. A Bayer 1920x1080 sensor has red and blue arrays of 960x540 each - and that will limit it's chrominance resolution to about 1/2 of 1920x1080 in each axis. In other words, it's inherently a sort of 4:2:0 because of the sensor - regardless of processing, recording format etc. If the information is not there in the first place, you can't process it back.

If this sounds too theoretical, then look at Alan Roberts test of the XF100/105 - section 2.3.1 and Figure 1 - http://downloads.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP034-ADD65_Canon_XF100-105.pdf to see a practical test, and what the use of the 1920x1080 Bayer for 1080 output actually means. Basically, horizontal and vertical lines finer than about 700-800 lpph are giving coloured aliasing, corresponding diagonals are giving luminance aliasing. Use an aggressive OLPF to stop such aliasing (as Razz16 implies the DB does) and useful resolution must stop around the 700-800 lpph mark.

Compare with Alans zone plates of the EX1 and you can easily see what a good 1080 camera should look like.

But..... it depends on sensitivity. If we assume a sensor 4x the area of the DB, then for comparable DOF it would need an f stop 2 stops smaller. So f8 to match the DB at f4, yes? In other words, depth of field is dependent on sensor size and aperture. (All else equal.)

From all I've read, then low light is not one of the DB's strengths. I hear ISO 200 realistically, for a decent level of noise. Consequently, it's quite possible for a bigger sensored camera to likely be so much more sensitive that it CAN be used 2 stops smaller for similar noise levels, so achieve similar deep depth of field.

Whilst having the flexibility to be used much wider open with ND, and the possibility of shallow depth of field.

I have participated in the installation of 35mm film projection systems in small to medium size theaters, up to about 40 ft wide screens that included projecting resolution chart films. None ever exceeded 850 lphh. So if the D16 at 2k makes 800 lpph which is realistic with cine quality glass, does so without the aliasing common to most low end cameras because it has a higher fill factor sensor, a proper critically matched OLPF and is not limited to poor quality debayer algorithms like real time single chip video cameras then I and many others will be quite happy. I expect the D16 to resolve about like the Red Epic does shot at 2k S16 crop. Maybe a little better because it is optimized for 2k and the Epic does lose a bit from compression though it is still raw.


Suzy and I went to the Metropolitan opera's Live at the Met broadcast of Prince Igor last week in a 4k theater. Probably shot on Sony HDcam 2/3" 3-chip 1080p cameras and finished on tape from the look of it. The D16 looks significantly better on a large screen than that show did.
As to sensitivity, current gen DSLR's produce their best quality widest DR raw images at ISO speeds between 200 and 400 by DXOmark's tests. Their test cutoff point for ISO vs quality is at 18dB S/N ratio or 3 stops effective DR. No camera they have tested exceeds about 3200 ISO with the Nikons being the current champs. Thinking in extremes might give the wrong impression about how the D16 performs in low light, though that is not its primary strength.

Sample NY times square shoot by Brian Morgan:

 

[starcentral]

Razz with all the razzing I have to say you are the most calm and collected dude on the planet.

Kudos.

 

[mik303]

Razz, it really doesn't matter how good your arguments are, there are haters out there that are just going to say bad things about the camera. People are regularly trashing RED, who has done nothing but kick ass for years in the film industry. I understood the criticism when the DB was vaporware, but now that it's actually out in the wild, my perceptions have changed. Personally, I was a huge fan of Ikonoskop, but they were a bit pricey. I'm watching the digital bolex closely and am very interested. The images and workflow look great for shooting films. The price range is realistic as well. Unfortunately, I can't just drop 4+ grand to pre-order the camera and accessories and afford to wait for 8+ weeks (hopefully) to get the cam. But I'm watching and when the production catches up with the demand, I may make the purchase. The future looks bright for DB, so keep up the good work.

 

[Prendi]

They've actually already made it, they are shipping as we speak.

Ah OK, congrats to them. I never see anything from this camera apart from a few threads on here that usually involve Razz shouting at someone about pixels.

Kudos to them finally making it happen.

 

[Samuel H]

Razz with all the razzing I have to say you are the most calm and collective dude on the planet.

Kudos.

+1
I'm definitely not in love with the dbolex concept, but I will say this: you could certainly consider an alternative career as a buddhist monk

 

[starcentral]

+1 +1

 

[Razz16mm]

+1
I'm definitely not in love with the dbolex concept, but I will say this: you could certainly consider an alternative career as a buddhist monk

Taoist monk in my case.

 

[combatentropy]

Low light doesn't look like a problem, from this footage.

Philip Bloom said this is not a low-light camera. But from what I've seen so far, my bet is that it is not any worse at ISO 800 or ISO 1600 than 16mm film is at ISO 800 or ISO 1600. In other words, grainy, yes, but if it was okay for you in your 16mm days (for those of you who had 16mm days) then it will be okay for you now.

What it won't be is super-mutant low-light like the Canon C100.

Some have the hair stand up on the back of their head if they see any grain at all in video footage. These same people complain that video still doesn't look quite like film and use expensive computer software to artificially add grain to their video footage.

If I didn't mind rolling shutter, I'd say get the Canon C100. I think it's the best mix of ergonomics and picture quality at a price that is not stratospheric. However (begin rant) rolling shutter is a huge deal to me and I can't believe a single manufacturer lets a camera out the door with it. "But global shutter reduces light sensitivity, dynamic range" is like making a car to get 50 MPG by leaving off all the windows and body panels. It ain't done! Only a pixel-peeping purist would pine for 4K. How can you be worried about 4K vs 2K, and not be a purist about motion, about lines in your picture being off by several pixels with even slow pans, with legs and arms and lips and anything else that is moving, being distorted --- forever! For a practical production person who has to make money today, your choices are limited, and you might have to settle on rolling shutter, and for most things people won't notice. But for purists who want 4K, how can you not also be a purist about global shutter! Global shutter should come before 4K! (end rant)

Among the global-shutter cameras available today (a short list) the Digital Bolex looks like my best choice. Is there any better bang for the buck?

 

[16mman]

Some have the hair stand up on the back of their head if they see any grain at all in video footage. These same people complain that video still doesn't look quite like film and use expensive computer software to artificially add grain to their video footage.

I love a bit of digital grain now and then. It can add character and texture to footage. Mark my words, one day we'll be nostalgic for it : ) It's funny how the industry is in this loop: fight defects in new technology, embrace and try to replicate them, repeat.