PYXIS 6K

BMCC6K (PYXIS sensor) + SIRUI 35mm Saturn 1.6x Anamorphic Full Frame (lens test) (Matteo Bertoli, Lisbon, Portugal)

"Everything was shot in 6K 3:2 (1.6x desqueeze in camera). Codec used was BRAW 12:1. Mix of 36 and 24p. I also used Hoya diopters filters +1 and +2 to get some close up shots (the minimum focus of 0.9m makes it hard to get close to the subject), together with my NiSi IRND 3 and 6 stops and a 1/4 Tiffen Black Pro Mist. Edited and graded in DaVinci Resolve with BUTTERY LUTs."

Hands on with the Sirui 35mm and the BMCC6K (PYXIS sensor) | Anamorphic on a budget! (Matteo Bertoli, Lisbon, Portugal)

Beautiful! I love that city and the footage does it justice.

BMCC6K (PYXIS sensor) + SIRUI 35mm Saturn 1.6x Anamorphic Full Frame (lens test) (Matteo Bertoli, Lisbon, Portugal)

Since the RS on the BMCC 6K FF (and presumably on Pyxis) is quite slow (25 ms in Open Gate), I can see this as a problem when using anamorphics, because nonspherical lenses tend to amplify distortion on the periphery of the image and hence make the wobble effects much more noticeable. I just can't tolerate that cheap effect on any motion picture.

As an example, see the wobble on 0:50, 1:18, and 2:21 on this video.

Do you guys think this could improve with the use of a gimbal or it's better to stay away from anamorphics and very wide lenses on this cameras if you want to avoid wobble effects like those?

Blackmagic Cinema Camera 6K rolling shutter speeds:
6048 x 4032 - 25.08ms
6048 x 3200 - 19.90ms
6048 x 2520 - 15.67ms
4832 x 4032 - 25.08ms
4096 x 3072 - 19.10ms
4096 x 2160 - 15.13ms
2112 x 1184 - 8.29ms
1920 x 1080 - 7.56ms

I wonder how the rolling shutter correction in Resolve's gyro stabilisation works with anamorphic lenses?

As an example, see the wobble on 0:50, 1:18, and 2:21 on this video.

Yes, those are some of the most pronounced RS examples I have ever seen, and it's not something that I get too hung up about either. The effect to me in any case, of anamorphic, 'open gate' and FF are so subtle, especially on YT , that here in this instance I think it would be wise to use the 2:4:1 crop, with this particular sensor more so, especially given it's FF, higher than 4K rez anyway and has appreciably better RS figures, in that mode.

And then:
Are you sure that BRAW is only 4:2:0? I get much better keying and colour selection with it than any 4:2:0 sources.

I'm sure, but consider that fact that in a bayer CFA there are also only one pair of color samples for every 4 photo sites as well.

Good Luck

That's correct, of course there's always interpolation going on.
But the fact that you have only one colour sample in blue and red per 4 photocells does not directly translate into storage as 4:2:0, which is YUV.

What is the rationale for so many white sample points when arguably two green samples work similarly well for determining luminance?

Sensitivity. If you ever had the pleasure to shoot B&W with a sensor without Bayer filters, you would have got up to 1.5 stops more. It's surprising how much light those filters swallow.

I understand that this is a theoretical advantage, but we haven't seen it in practice. There are bayer pattern sensors with better sensitivity. And not just the Sony and Canon sensors with crazy low light capabilities, but even many "normal" cameras. When I tested the original 12k, it was spectacularly average in this regard.

I understand that this is a theoretical advantage, but we haven't seen it in practice. There are bayer pattern sensors with better sensitivity. And not just the Sony and Canon sensors with crazy low light capabilities, but even many "normal" cameras. When I tested the original 12k, it was spectacularly average in this regard.

Given that BMD's 12K sensor has almost three times the pixel/mm density of the BMPCC6K's sensor, the fact that it maintains a "spectacularly average" level of light sensitivity is a pretty solid accomplishment.

White pixels are ALL colour when mosaic’d with RGB photosites. They aren’t monochromatic.

JB

Of course. As pixels they are colour, as photo sites or ‘sensels’ they are not.
The typical confusion if we are not precise in our language.

Given that BMD's 12K sensor has almost three times the pixel/mm density of the BMPCC6K's sensor, the fact that it maintains a "spectacularly average" level of light sensitivity is a pretty solid accomplishment.

To what gain? (Not camera gain, consumer gain.) It certainly doesn't resolve 3x picture detail. There are 8K bayer pattern sensors that match or exceed it in real world resolving power and light gathering. Even those who promote the sensor say it's not about the resolution. They fall back to undefined claims of "better color" and then refuse and/or fail to demonstrate it.

I'd love for there to be new tech and techniques that provide demonstrable real-world benefits over the old bayer pattern, but at least so far, I haven't seen them.

I'd love for there to be new tech and techniques that provide demonstrable real-world benefits over the old bayer pattern, but at least so far, I haven't seen them.

I've seen it from Foveon sensors but those have also terrible drawbacks and will probably never work for video.

Yeah why the hell would they spend tens of millions of dollars and years of R&D on their own unique custom sensor?

Makes no sense at all.

JB

And then:
Are you sure that BRAW is only 4:2:0? I get much better keying and colour selection with it than any 4:2:0 sources.

I'm sure, but consider that fact that in a bayer CFA there are also only one pair of color samples for every 4 photo sites as well.

Good Luck

Color subsampling and a Bayer pattern cannot be compared one-to-one but pure image information-wise the resolution of a Bayer pattern lies somewhere between 4:2:2 and 4:4:4, closer to 4:2:2 than 4:4:4.

my goodness. All this tech talk!

The Pyxis is one of those cameras that, if I didn't have an UMPG2 or any experience with internal NDs, I'd probably get. It's a really nice sensor and the Full Frame would allow me to see all the glorious character of my FF lenses. But man, is it hard to give up the quickness of internal NDs and the lower-light capabilities of a P6kPro.

(I'm REALLY hoping to test out a Ursa Cine when they are available to rent!)

Yeah why the hell would they spend tens of millions of dollars and years of R&D on their own unique custom sensor?

Makes no sense at all.

JB

That's an "Appeal to Authority" fallacy. It doesn't actually answer the question. We know that a company full of really smart people have bet on this tech. Maybe they did so for quality control, maybe its for cost savings, maybe its for any number of reasons we're not privy to.

I'm not asking that question, though, because as a consumer, it doesn't really matter. The question I'm asking is: What's the real world, demonstrable benefit to shooters over other shipping choices? There are cameras using bayer pattern sensors that are equal or better in any objective criteria I've seen. If there is an advantage to shooters other than price, what is it? I genuinely want to see it, because while I'm not a big fan of the Ursa form factor, if it actually improved PQ in a meaningful way, I'd have to at least consider it.

So, I do very much look forward to testing the new Cine 12K because if it can deliver close to the promised dynamic range, and it yields commensurate improvements to latitude, then for $15K that is a camera to be reckoned with.

There is no advantage whatsoever.

We have been here before. Last time this happened this guy was really just trying to drive his monetized blog / you tube channel and didn’t take kindly to when his methodology was questioned.

viewtopic.php?f=2&t=135032&start=100#p732550

JB

We have been here before. Last time this happened this guy was really just trying to drive his monetized blog / you tube channel and didn’t take kindly to when his methodology was questioned.

And that's an "ad hominem" attack.

If a camera/sensor has a picture quality advantage, it should be demonstrable. It's incumbent on those making such claims to prove it. Why is asking for a demonstration of "ultimate in image quality" an unreasonable request?

FYI, this thread is about Pyxis.

How `bout that Pyxis!

Sorry everyone.

These discussions can be rather pointlessly into weeds that in the end don't matter. As an autistic analysis/synthesis person myself, while theories are great, practical is *always* real world.

And sadly we don't have much other than John's use of the Pyxis to judge from.

But from that, and his comments having actually used the camera, and looking over the specs, the size at NAB ... that might very well ve a very good rig for *me*.

But I won't have a more solid view until after it's shipped for awhile, and we get real world uses posted.

Hope that's soon ...

Sent from my SM-S908U using Tapatalk

These discussions can be rather pointlessly into weeds that in the end don't matter. As an autistic analysis/synthesis person myself, while theories are great, practical is *always* real world.

And sadly we don't have much other than John's use of the Pyxis to judge from.

But from that, and his comments having actually used the camera, and looking over the specs, the size at NAB ... that might very well ve a very good rig for *me*.

But I won't have a more solid view until after it's shipped for awhile, and we get real world uses posted.

Hope that's soon ...

Sent from my SM-S908U using Tapatalk

I could be wrong, but I haven't heard anyone claim that you should expect a PQ difference from the Cinema Camera 6k. It's my understanding it is the same image processing pipeline from sensor to recorded signal. At best, one might hope for slightly less noise due to better thermals, but again, I haven't heard anyone, much less BMD, claim this. (Someone please correct me if I'm wrong.)

The Pyxis is really all about the form factor.

And sadly we don't have much other than John's use of the Pyxis to judge from.

I haven’t used it. I haven’t seen it.

JB

And then:
Are you sure that BRAW is only 4:2:0? I get much better keying and colour selection with it than any 4:2:0 sources.

I'm sure, but consider that fact that in a bayer CFA there are also only one pair of color samples for every 4 photo sites as well.

Good Luck

Color subsampling and a Bayer pattern cannot be compared one-to-one but pure image information-wise the resolution of a Bayer pattern lies somewhere between 4:2:2 and 4:4:4, closer to 4:2:2 than 4:4:4.

I agree, subsampling and demosaicing are completely different processes. Braw from the 12K is especially complex. Storing Braw as 4:2:0 is brills, 4:2:2 would simply be a waste of data. There are already 50% more color samples than color photo sites using 4:2:0.

Good Luck

Process.png (26.86 KiB) Viewed 8791 times

I agree, subsampling and demosaicing are completely different processes. Braw from the 12K is especially complex. Storing Braw as 4:2:0 is brills, 4:2:2 would simply be a waste of data. There are already 50% more color samples than color photo sites using 4:2:0.

It would certainly make sense to store RAW as 4:4:4 from a Bayer filter. While 4:4:4 would introduce some redundancy it will give the maximum maintainable original color information. For BM filters it is a bit more complex but would also store it 4:4:4

It won't be a waste, the compression algorithms will compress redundancy just fine.

Yeah why the hell would they spend tens of millions of dollars and years of R&D on their own unique custom sensor?

Makes no sense at all.

JB

That's an "Appeal to Authority" fallacy. It doesn't actually answer the question. We know that a company full of really smart people have bet on this tech. Maybe they did so for quality control, maybe its for cost savings, maybe its for any number of reasons we're not privy to.

It's not an appeal to authority fallacy. The human vision system is more deft. Acutance is the appellation for edge contrast which is apart from resolution, the overused simplification for people looking to ascribe one number or one term to all things qualitative. DR is another.

So you don't see it, the sensor tech. You claim to be looking for it, but not in the things that would reveal it honestly. You are on the record for publicly making known your agenda to disprove it. It's very clear, you are not open minded with altruistic intentions, so that's why it's being stated here.

The contrast with people who do pursue truth and science to its limits without preconceived biases, recognize the bar being set, without non-productive agendas, an example of such I found here; a helpful and detailed dive into the human vision system that finishes with an ironic conclusion about how many "K" is needed for the human vision system, acuity seen from inches to yards, and what "the right number" for target "K" should be. It's different than the "spectacularly failed" conclusion above.

https://patents.google.com/patent/US20180367775A1/en

JB

My bad John!

Sent from my SM-S908U using Tapatalk

I agree, subsampling and demosaicing are completely different processes. Braw from the 12K is especially complex. Storing Braw as 4:2:0 is brills, 4:2:2 would simply be a waste of data. There are already 50% more color samples than color photo sites using 4:2:0.

Good Luck

Nailed it, BRAW 12K stored as 4:2:0 from 50% more R,B sites, patent confirms this.
That would equate to 6K 444, right? And what of 8K? 422? (Basing purely on pixel quantities)

With all the talk about the rolling shutter on the PYXIS and all. I got reminded that 10-Years Ago I shot Rosary from May 9-11. A lot of it was primarily handheld, and we used the Original Blackmagic Cinema Camera 2.5K EF on it. That had a rolling shutter as bad as the PYXIS, or slightly worse. The max frame rate was 30 FPS after all. And, when that camera was announced in 2012 it was $2,995 just like the PYXIS costs now.

https://www.blackmagicdesign.com/media/ ... 0120416-08

I'll share the music video below. Many of you have seen it if you have been on the Forum for a decade. But, I think it is proof that the PYXIS is capable of shooting exactly the same thing today. In many ways the PYXIS is a wonderful evolution from that first BMCC2.5K.

I'd be more than happy to use the PYXIS to shoot the above music video today knowing how it compares to the original BMCC2.5K.

an example of such I found here; a helpful and detailed dive into the human vision system that finishes with an ironic conclusion about how many "K" is needed for the human vision system, acuity seen from inches to yards, and what "the right number" for target "K" should be.

"should be" is not a "scientific argument", because we're not talking about the acuity of the human eye, we're talking about cinema. And what's the ideal resolution of that? This guy doesn't know, and neither does anyone else, because there is none, "cinema" is not a surveillance camera or microscope with a practical requirement. There is no threshold. And no ideal. The typical 35mm release print comes in at 600-700 lines (assuming the projector is in focus and the bulb isn't at the latter stages) and actual cinema has been known to exist at the sub-VHS level. There's at least one career I know of which would not have advanced if the filmmaker had not shot on pixelvision, which turned unexceptional material into something mysterious and elusive.

And we won't even go into perceptual anomalies, of the kind Steve Yedlin demonstrated, where the lower resolution image is mistaken for the higher one.

However folks want to answer the party in question, I don't think "science" is the refutation.

I agree, subsampling and demosaicing are completely different processes. Braw from the 12K is especially complex. Storing Braw as 4:2:0 is brills, 4:2:2 would simply be a waste of data. There are already 50% more color samples than color photo sites using 4:2:0.

Good Luck

Nailed it, BRAW 12K stored as 4:2:0 from 50% more R,B sites, patent confirms this.
That would equate to 6K 444, right? And what of 8K? 422? (Basing purely on pixel quantities)

I think it’s very very misleading to talk about 422 and 420 equivalents. Conflating those ratios of photosites with the connotations of encoded video doesn’t make sense.

It doesn’t mean 6k grades better than 8k from the same camera.

JB

Braw is 12 bit 4:2:0 end of story, whatever connotations folks bring to the table reflect their own prejudices and opinions, not facts.

Good Luck

Braw is 12 bit 4:2:0 end of story, whatever connotations folks bring to the table reflect their own prejudices and opinions, not facts.

Totaling of BRAW pixels is soft matter for a basis of prejudice and opinion, but I get John's point.

I think it’s very very misleading to talk about 422 and 420 equivalents. Conflating those ratios of photosites with the connotations of encoded video doesn’t make sense.

It doesn’t mean 6k grades better than 8k from the same camera.

JB

Indeed! Hey, JB and I agree on something!

While ratios can apply to anything, we all know that in the context of cameras they are almost exclusively used to refer to video encoding, so it's confusing at best, deceptive at worst, to use them to refer to sensors.

However folks want to answer the party in question, I don't think "science" is the refutation.

[/quote]

I should have said, in his opinion.

I wonder how the rolling shutter correction in Resolve's gyro stabilisation works with anamorphic lenses?

I tested this about two weeks ago with the BMCC 6k and the Greatjoy 1.8x anamorphic lenses (35, 50, 85). The rolling shutter removal works just like with any other lens. However, I found it works best when you divide the focal lengthof the lens by the desqueeze factor. So for example 85mm/1.8=47.222.
So I would enter 47mm in the lens profile on the BMCC when using the Greatjoy 85mm 1.8x.

Which made me think, with all the affordable cine zoom lenses being available and them not having any electronics to communicate the focal length, it makes them quite a lot harder to work with if one plans to use gyro stabilization.

I'd be more than happy to use the PYXIS to shoot the above music video today knowing how it compares to the original BMCC2.5K.

Tim, whilst you can definitely see a fair amount of RS in your video, it's not really objectionable or noticeable to the average viewer. The point here though about the Pyxis (or rather the same sensor on the BMCC 6K, in that case) was specifically it's use with anamorphics; which I think should be avoided with a sensor this slow.

The Sony FX9, of a similar sensor size and RS, that is used extensively in broadcast reality handheld here, and I cut a huge amount of, is never a problem neither nor particularly noticeable. This is why I say I never get too hung up about it. But those particular worse examples in the video with the anamorphic lens, would almost certainly be a QC fail, were it for broadcast.

I think the Pyxis is a fine camera otherwise and I too would have no hesitation shooting on it; just not with an anamorphic lens.

I don't know, folks.... The piling on doesn't seem altogether fair. If film remains the gold standard, as it is for some of the best commercial filmmakers working today, consider that its color rendition could not be described as more "accurate" than digital; if anything, probably the reverse. Emulsions have obvious color characteristics, e.g., Kodak v. Fuji.

As noted earlier in the thread, there were strong claims made for the Foveon sensor, but that didn't go anywhere commercially.... If there are advantages to BMD sensor technology -- even if at a level beyond the color grading capabilities of most of its customers -- it's not all that unreasonable to ask for a demonstration.

My own view -- okay, nobody asked, but never mind -- is that human vision is relatively crude (compared, say, to hearing) and that even middling sensor technology, combined with post-production, has already surpassed human powers of discrimination in the realm of moving pictures. And of course art relies on limitation. There may be a case for a shooter or colorist to want 18 stops, but there's no such case for a viewer. A painting (or a movie) with 18 stops and a palette of a billion colors? Who would want it? And who needs it?

It's no secret that visual arts go the *other* direction: reduced color, reduced resolution, reduced DR:

The Pilgrimage to San Isidro.jpg (845.71 KiB) Viewed 7254 times

DP-14286-001-2.jpg (967.05 KiB) Viewed 7254 times

My own view -- okay, nobody asked, but never mind -- is that human vision is relatively crude (compared, say, to hearing) and that even middling sensor technology, combined with post-production, has already surpassed human powers of discrimination in the realm of moving pictures.

Vision is vastly more complex than hearing, but yes, short of dynamic range, many shipping cameras are essentially "good enough" to deliver images that meet or exceed perceptive abilities. (Humans can see aprox. 10 stops of DR at any one time, but because we assemble a scene in our mind out of many discrete "samples", effectively we perceive about 20 stops in any one scene.

And of course art relies on limitation. There may be a case for a shooter or colorist to want 18 stops, but there's no such case for a viewer. A painting (or a movie) with 18 stops and a palette of a billion colors? Who would want it? And who needs it?

It's no secret that visual arts go the *other* direction: reduced color, reduced resolution, reduced

Again, true, but ideally the artist decides how to "limit", otherwise it's not an artistic decision. When it comes to cameras, I'd prefer that they faithfully record as much information as possible and allow me to make the creative decisions rather than have a dispassionate hunk of metal and plastic impose a "look" on me.

And I guess that ultimately gets to the heart of why I keep banging on about vague manufacturer claims. There is no "magic" to how a camera works, not even "Black Magic". Since its IPP is imposing decisions, I want to understand what they are, and how far they deviate from an accurate representation.

PS. I've always found it interesting that most people readily agree that display devices need to perform within a narrow band of acceptable ranges - and would never dream of grading on an "inaccurate" monitor - yet those same people are fine with having no idea how accurate a camera records color in the first place.

Vision is vastly more complex than hearing

I'm not talking about the mechanism or the internal data processing. I mean ability to make discriminations. Compare the complexity of music to the visual arts and you'll see what I mean.

Again, true, but ideally the artist decides how to "limit", otherwise it's not an artistic decision.

And so much the better if it isn't.... Not relevant to the discussion, but external constraints exist at every turn, and without them, the work will be deadly boring or incomprehensible. I could cite examples, but this thread isn't the place.

The majisterial view of the would-be artist, particularly in a medium so deeply tied to commerce and the popular taste, and with so much yet to be proven (where's all this art again?), suggests the triumph of vanity over craft.

PS. I've always found it interesting that most people readily agree that display devices need to perform within a narrow band of acceptable ranges - and would never dream of grading on an "inaccurate" monitor - yet those same people are fine with having no idea how accurate a camera records color in the first place.

A display device is a copy machine. The camera is a palette. Don't over analyze that. It's just a metaphor.

I know that if my display device is calibrated to the same standard as the viewer's device, when I adjust my palette they will experience the art as I intended.

... There is no "magic" to how a camera works, not even "Black Magic". Since its IPP is imposing decisions, I want to understand what they are, and how far they deviate from an accurate representation.

I do enjoy these discussions on aesthetics versus reality.

I would think in large part the secret sauce is the magic and it’s called ‘secret’ for a reason. I’d wager if you were in charge of defining the image processing for a camera manufacturer, you would strive to do the same thing from you aesthetic and you wouldn’t feel compelled to share your IPP and your intellectual knowledge with us.

I hate to cut your argument off at the knees, but I’m grateful you stated your position clearly. I hope you rethink your assumptions. We don’t want to capture reality in a bottle, we want to work with an impression of reality that we admire or suits the purpose.

It’s not terribly difficult to be reasonably comfortable with the different looks of Canon versus Sony versus Panasonic and so on. They all can look different out of the camera. The last thing we want in a camera is a ‘perfectly accurate’ reproduction of reality. We want art and beauty (or horror). Look at those paintings above; that’s a complete artistic fantasy in terms of colour and light levels, but that’s a good thing because we are moved by the painting, we understand what the artist intended.

PS. I've always found it interesting that most people readily agree that display devices need to perform within a narrow band of acceptable ranges - and would never dream of grading on an "inaccurate" monitor - yet those same people are fine with having no idea how accurate a camera records color in the first place.

Yes, it’s interesting and reflects basically the shared requirements of two different worlds:

My understanding is that the broadcast world and commercial advertising pays a lot of attention to meeting colour standards consistently (even though I can see significant variations in clips on my television).

The cinematic world which could care less about most real colours but does an amazing job of grading a scene at least have a consistent look subject to the intentions of the director which can change from scene to scene, or not. A Wes Anderson painterly film would be an example. These artists are painters.

I’m sure their grading processes use faithful grading monitors as a foundation for splashing about in their chosen rainbow. Just as a painter may begin a painting by covering the entire canvas with a wash of their preferred ‘white’ paint. The faithful monitor is there to ensure that the graded images are viewed as the creator intended them.

. If there are advantages to BMD sensor technology -- even if at a level beyond the color grading capabilities of most of its customers -- it's not all that unreasonable to ask for a demonstration.

The problem is it’s a disingenuous ask. The last time it happened with this poster it was asked as if from a genuine notion of discovery. And to even draw BMD to appear to be endorsing his already decided upon agenda and views. Theres only a pretence of true “scientific” discovery and method.

Hes trolling. I’m just calling it out.

JB

If you think of the most striking and emotionally impactful images, none of them are accurate.

Cinema is not visually accurate.

Accuracy is a road to banality.

No digital camera is 100% accurate.

JB

If you think of the most striking and emotionally impactful images, none of them are accurate.

Cinema is not visually accurate.

Accuracy is a road to banality.

No digital camera is 100% accurate.

JB

I am trying to keeping out of this senseless debate; and trying to focus on the Pyxis, so as not to be accused of talking bollix about what I don't know again.

But this is John Brawley at his his Goddamned best. And ain't that the truth!

And The Truth is that no camera, either film or digital capture, actually "sees" color.

Both only record light falling on the capture mechanism, and then employ brilliantly designed processes to recreate, or perhaps more closely stated, mimic the "scene" colors.

So for digital capture, every camera series will have a slightly to greatly different color recreation than another camera. As both the hardware and the math used in the processing will vary.

And as all of us humans vary infinitely from each other, we're naturally going to prefer different recreations.

It's trying to match one recreation of scene color to another that gets difficult at times.

Personally, I do like most BM color recreations. And the Pyxis by specs seems to hit everything I need/want, including price point.

So I am waiting impatiently for results of actual use to get posted.

Sent from my SM-S908U using Tapatalk

And then this.

https://en.wikipedia.org/wiki/Metamerism_(color)

Whereby the camera is an extra element that also perceives colour differently.

JB