Jump to content

The cheap RED aesthetic.


Recommended Posts

  • Premium Member

80 line pairs per mm * two lines per line pair * 24.9mm per image width is just under 4000 lines per image width. So just under "true 4k."

Like most of the people here and all of the people on Reduser, you're confusing an actual electronic sinewave that's already been generated, with an optical image of a pattern of pattern of lines that needs to be turned into an electronic sinewave.

 

The only time 4000 pixels could accurately sample 4000 lines is when the dark and light lines precisely line up with the columns of pixels. With any other configuration you are constantly going to get the situation where dark and light areas of the image straddle the same pixel.

 

A pixel that is dark on the left and bright on the right (or top and bottom) is going to store the same value as a pixel that is bright on the left and dark on the right.

 

You can't actually store a pixel which is dark on one side and light on the other; it can only ever store an average value. The nearest you can get to that is to split the pixel up into four smaller pixels, which is another way of saying that you need four times as many. (Twice as many both horizontally and vertically).

 

Bottom line: If you took a film camera with film and lens that could resolve 4000 lines, and slowly and carefully zoomed or tracked in on a 4000 line resolution chart, keeping accurate focus all the while, you would a get film strip of a continuously decreasing count of vertical lines, all the way down to where you couldn't focus any more. Every one would carry an exact copy of the currently focussed image of the chart, plus a bit of film grain, and nothing more.

 

If you repeated the experiment with a focally identical 4K video camera without an optical low pass filter, until the line count got down to 2,000 lines, you would get an unholy mess of "screen door" type Moire` artifacts.

 

This has been discussed many times before, and the same peuedo-technical non-sequiturs keep getting regurgitated.

 

Here, one again for your considering pleasure, is an example of the still unacknowledged Elephant in the Room:

 

http://www.cinematography.com/index.php?showtopic=38421&view=findpost&p=284406

 

The reality is, when RED say "about 3.2K" what they actually mean is: "Equivalent to a non-Bayer camera with 3200 horizontal pixel", which is basically equivalent to a film camera imaging 1,600 lines.

 

If you shoot a 4,000 line chart with a properly focussed 35mm movie film camera, the 4000 lines will be there on the negative, and they can be clearly seen under a microscope. Whether people are unable to (or choose not to) scan to obtain that actual resolution, is not the point, the point is, the resolution is in there for possible future use.

Link to comment
Share on other sites

  • Replies 103
  • Created
  • Last Reply

Top Posters In This Topic

Like most of the people here and all of the people on Reduser, you're confusing an actual electronic sinewave that's already been generated, with an optical image of a pattern of pattern of lines that needs to be turned into an electronic sinewave.

 

You're falling back on the same pseudo-technical talk, though. Yes, sampling theory says you need n*2 pixels to accurately sample n lines (or n/2 line pairs), but in reality no one is sampling things that perfectly and camera manufacturers, even the most conservative ones, allow for a little moire, because the lens will knock the resolution down to the point where aliasing is not a significant issue more often than not. The system needs to sample at n*2 pixels to get n "mathematically perfect" pixels, but the sensor itself can sample more aggressively and the lens will knock it down to something acceptable.

 

In stills mode, the 7D pushes 3100 lines per image width with aliasing (2500 without) according to published tests using a sharp lens. In theory it should only resolve about 5184 * .7 bayer efficiency /2 (nyquist sampling) or 1800 lines. But in practice it somehow does better, largely because a pixel does not have to be mathematically perfect to be worth anything visually. We don't sample perfectly according to theory in the real world. If we had to, a 2k film scan would resolve 500 line pairs or have tons of visible aliasing. But it resolves more than that with very little aliasing.

 

And then microcontrast enters the picture; the integral of the area under film's mtf curve, the most widely accepted measure of perceived sharpness, is likely no better (and probably worse) than digital's even if the resolution doesn't extend out quite as far. My own tests on still film bear this out. My sharpest 25 megapixel scans of velvia have detail nearly as fine as a 25 megapixel digital camera's in high contrast areas (and no aliasing). But image quality is subjectively comparable to maybe 5 megapixels, at the very best, because the stuff is so grainy and detail gets fuzzy fast.

 

That said, "true 3.2k" out of the red does sound totally impossible. That's substantially better than the 7d in stills mode despite a more aggressive OLPF and a significantly lower resolution chip. But who knows? Sampling theory is just that: theory.

Edited by M Joel Wauhkonen
Link to comment
Share on other sites

  • Premium Member

Can anyone point to a large sensor digital camera that does suffer from these very same issues? Does any digital cinema camera maker measure resolution how Keith is describing?

 

No matter how you measure resolution, the fact of the matter is that the Red One has more resolution than any other digital camera out there.

 

If that is not correct, then show us an example.

Link to comment
Share on other sites

As I already said, the resolution-debate has very little to do with the "RED-look"-issue. But maybe it's important to discuss due to the different approaches in the industry.

 

RED clearly markets it's camera as a 4k/4.5k/5k-camera - nothing less. They shoot test-charts in labs and give "k"-results to determine actual output-resolution (which already simplifies the issue) and justify the approach of using a sensor with rather small photosites and compression to handle the amount of pixels.

They claim 3.2k for 4k, 3.7k for 4.5k and 4k actual resolution with 5k acquisition - that's VERY optimistic, just like the dpreview-results for lpph with DSLRs often resolving beyond nyquist... There is something on the charts - but that's not real resolution.

It's difficullt finding non-RED-sponsored results on the net but the images itself already give an idea how a correctly aliasing-filtered (0% contrast at 4k!) and bayer-interpolated, non-sharpened image looks like and from judging the RAW-files on the net the newer REDs are not different - no surprise since the technical/optical background remains the same. The ALEXA is not different, there is no such thing as "ARRI-magic" or "RED-magic" - but the marketing is different - the ALEXA is marketed as an excellent HD/2k-camera and judging from the images, it's true. But what is RED? There is no 3k-workflow, you can go either HD/2k or 4k according to DCI. The sample I've posted showing the actual quality-gain from going the (still not cheap) 4k-route and what ends up on the screen is not very uplifting, IMHO.

Here are a few samples showing the resolving power on a test chart (these are high-contrast, b/w, partly aligned patterns - ideal for digital):

http://provideocoalition.com/index.php/awilt/story/more_red_res_testing_the_mysteryium_resolved/P1/

 

2400 lines horizontal resolution are a good result and the ALEXA with 2880 photosites in the horizontal axis will have trouble reaching this result. But what matters (resolution/sharpness)-wise is the MTF of aliasing-free 1080p/2k-output. Don't get me wrong, watching a 4k-projection from RED looks sharp, especially when coming a from a horrible 35mm-print. But is the "4k-argument" really valid in case of RED? I highly doubt it.

 

The size-argument (especially in 3D-rigs) with RED EPIC is valid, so is the high max. frame rate - the resolution-argument propably not.

Link to comment
Share on other sites

  • Premium Member

I'm not saying that resolution is the be all and end all of an image. But there are standardized ways of measuring resolution, and that's with test charts. Camera makers, to DP's to the BBC all use these.

 

If they are overly optimistic, fine, but then they are overly optimistic for all digital cameras, not just the Red.

 

Some people write paragraph upon paragraph of why the Red really isn't 3.2K. Fine, perhaps they're right. But then show us a digital camera that does better. If the Red isn't really 4K or 3.2K, then what is the F35 or the Alexa?

Link to comment
Share on other sites

I agree that we like get lost into too technical discussions which have little relevance in reality anymore. They're only important to differenciate technical developments and marketing.

 

I'm not saying that resolution is the be all and end all of an image. But there are standardized ways of measuring resolution, and that's with test charts. Camera makers, to DP's to the BBC all use these.

 

If they are overly optimistic, fine, but then they are overly optimistic for all digital cameras, not just the Red.

 

Some people write paragraph upon paragraph of why the Red really isn't 3.2K. Fine, perhaps they're right. But then show us a digital camera that does better. If the Red isn't really 4K or 3.2K, then what is the F35 or the Alexa?

 

Read more: http://www.cinematography.com/index.php?showtopic=46954&st=80#ixzz1F4Ns7M4k

 

The standardized, professional way is MTF - resolution over contrast - tangential, sagittal, wavelength-dependent... All these net-samples I've posted are not professional enough, either - they can only give a very basic idea.

The high-res sensor of the RED is more of an technical approach than a real performance criteria. As I said, on a test chart, the RED EPIC can outperform the ALEXA, I'm pretty sure about that - regarding MTF in the final HD/2k-ouput propably not. Another story are the other IQ-parameters which more or less suffer the small photosites-approach of RED. When your producers pay for a 4k DI, they also pay for 35mm and unless you only shoot >>200ASA a (propably slightly degrained) 35mm-acquisition will offer superior performance (higher extinction MTF, dynamic range, color transitions) without the "cheap RED-aesthetics" this thread was about in the first place.

 

RED has it's advantages, the "K-argument" is hardly one of them.

Link to comment
Share on other sites

  • Premium Member

These threads tend to become rather circular:

 

Red isn't really 4K >> Actually Red claims it's 3.2K >> Well Red isn't really 3.2K either >> But this test chart says it is >> That test chart is flawed for measuring resolution >> So what digital cinema camera has better resolution >> Resolution doesn't matter >> Then why make a big deal that Red isn't really 4K >> B/c film has more resolution >> How much resolution does film have >> At least 4K >> But isn't the Red 4K ... and so it goes.

 

BTW, I agree with you that resolution is one small aspect of IQ. And I think that all its competitors have the advantage in terms of color rendition.

Link to comment
Share on other sites

  • Premium Member

Did all these resolution discussions come up before Red came into the digital cinema camcorder market? Resolution is the Red agenta, and we are happily chewing on it:)

 

Sure, Dalsa started the whole "4K RAW = 4K?" argument a few years before Red. And the arrival of the 24P HD Sony F900 started the whole "is HD good enough?" arguments and the attack on 35mm started with reports about how 35mm print projection was barely 1K, etc. This is nothing new, it's just been ramped up. But megapixel fetishing isn't unique to the movie business, look at the still camera market.

Link to comment
Share on other sites

  • Premium Member

Can anyone point to a large sensor digital camera that does suffer from these very same issues? Does any digital cinema camera maker measure resolution how Keith is describing?

 

 

Of course they don't. If they did, they would have to admit that their cameras still come up short.

 

But you're missing the point.

The problem has always been where people continually assert that such-and-such a camera out-performs 35mmm film in such-and-such a fashion. The issue has always been one or more of the following:

 

A. The physics of the camera's sensor construction simply do not allow the performance claimed.

B. The "proof" that is offered up in the form of test charts and so on, does not really prove anything, or worse, tends to back up exactly what the critics are saying!

C. The Internet tends to get choked up with "Testimonials" that are either painfully tendentious (eg proclaiming the "democratization of the media" which translated means approximately "at last; any below-average Joe with zero talent and even less money can now make blockbusters.") or simply do not stand up to scrutiny of the end product.

 

 

No matter how you measure resolution, the fact of the matter is that the Red One has more resolution than any other digital camera out there.

 

If that is not correct, then show us an example.

Red MX, Epic?

 

This another annoying movement of the goalposts. The Epic is certainly getting more action on big-budget sets these days, (although I can't help noticing that it's almost entirely for 3-D productions where using electronic cameras has clear advantages that may outweigh any loss of image quality), but the point is, the Epic is NOT the Red One. The Epic is probably the camera the Red One was hyped up to be.

 

The Epic may well be the product that sees J. H. Jannard's camera finally company gain some street cred with people who really matter, but the RED One so far hasn't really revolutionized anything, apart from rabid Fanboy-ism.

 

Yeah, but I get it: The polite people are the ones most likely to know what they're talking about :rolleyes:

Link to comment
Share on other sites

At this point aren't we getting into technical semantics that don't matter that much? 4K or 3.2K, the fact is, the Red provides plenty of image information for most of us who use it, who are producing 2K or 1080p end products.

 

I don't agree that the Red has democratized filmmaking or whatever, but it even if it doesn't quite live up to its own marketing, it certainly has given those of us in the lower and middle tiers of filmmaking and videography a better camera than we've seen before at this level.

Link to comment
Share on other sites

  • Premium Member

Thanks, Keith, for writing this up yet again. I thought about doing it, but didn't have the time....

 

Perhaps this could be the start of a sampling fundamentals FAQ.

-- J.S.

Yeah, but maybe somebody else can do it; I'm pretty much over the subject.

We really need somebody who:

 

A. Is credibly competent to write it

B. Has free accesss to the various cameras and equipment involved.

C. Has no vested interest (overt or covert) in the outcome

 

When I started proving how sh!thouse the actual dynamic range of some high-end cameras actually is a couple of years back, (using scientifically verifiable measurement methods, not just looking at a monitor and pulling results out of my arse) the "welcome" mat suddenly disappeared from my sources of equipment.

 

You Know: "Impartial testing is supposed to confirm out marketing department's claims, not refute them! I thought you understood that Mr --------".

Link to comment
Share on other sites

  • Premium Member

At this point aren't we getting into technical semantics that don't matter that much? 4K or 3.2K, the fact is, the Red provides plenty of image information for most of us who use it, who are producing 2K or 1080p end products.

 

I don't agree that the Red has democratized filmmaking or whatever, but it even if it doesn't quite live up to its own marketing, it certainly has given those of us in the lower and middle tiers of filmmaking and videography a better camera than we've seen before at this level.

The problem is, certain people are claimimg that certain cameras have more recoverable resolution than 35mm film, and therefore have equivalent or better archival ("future proofing") potential.

While it is certainly true that anything that improves on the maximum available resolution from the currently used 1920 x 1080 HD cameras is worthwhile, the actual available improvement appears to be somewhat exaggerated.

Because the response is so minimal, the only place the "extra" resolution is likely to be discernable, is on a resolution test chart.

Link to comment
Share on other sites

  • Premium Member

Keith,

 

I believe I uderstand your beef with resolution charts. It's that the lines get closer at a predictable rate and follow a simple pattern.

 

This makes it easier for the lines to be reconstructed by interpolation. It also makes it so that you will have a point after the Nyquist limit is hit where the lines will align perfectly with the sensor grid, so extra resolution appears.

 

A simple thought experiment would be to have a completely black frame except for one vertical white line that would project on the sensor as would the white valley of 3K resolution lines.

 

This white line will always hit at least two colums of pixles, it could even hit three. That means a 3K res line will be interpreted by the sensor as a 2K (4K/2) or a 1 1/3K (4K/3) line. Now processing can make things look better, but probably not perfect.

 

Now let's make the line so it projects on the sensor as would the white valley of 4K resolution lines. This white line will always hit one or two colums of pixles. That means a 4K res line will be interpreted by the sensor as a 4K (4K/1) or a 2K (4K/2) line. So in the case of when the line lines-up exactly w/ one row of pixles, you have 4K resolution, but that's really the result of a happy accident. The gradual nature of resolution charts makes these types of happy accidents more likely and adds to measured resolution.

 

I believe this is the issue that you're talking about.

 

So claiming resolution of 4K sensor above Nyquist, i.e. greater than 2K, involves some processing trickery and fortuitous alignments that may not carry over to real life shooting.

 

Is that correct?

Link to comment
Share on other sites

  • Premium Member

If they can produce an 8K Epic, THEN they will have something that equals or at least approaches the useable resolution of film.

Well whaddya know, Sony are making an 8K camera

 

Downscaled to 4K it's true, but 8K downscaled is going to look a lot better than straight 4K.

 

But of course everybody knows a 4K sensor out-resolves 35mm film, so why are they bothering to use four times as many pixels?

 

No doubt all will revealed at NAB.

First it was Regency with the first pocket transistor radio, then Jannard with the RED One. Sony just can't walk away....

Link to comment
Share on other sites

  • Premium Member

So claiming resolution of 4K sensor above Nyquist, i.e. greater than 2K, involves some processing trickery and fortuitous alignments that may not carry over to real life shooting.

 

Is that correct?

Exactly correct.

That's not to say that this won't necessarily be worthwhile, in fact a tiny of image enhancement can result in a large subjective improvement.

But it's still not equivalent to photochemical film, which achieves its resolution with zero enhancement or guesstimation.

 

There was a product, legendary in the world of compliance testing laboratories, from a well-known Korean electrical manufacturer.

The product was a refrigerator.

Most modern fridges are designed so that the hot gas from the compressor is routed through pipes run up the sides of the cabinet, to keep them slightly warm and so avoid condensation.

This manufacturer thought they could save money by simply fitting a couple of cheap heating elements instead, that could be switched in and out by the same controller that operated the defrost cycle.

However, they then found that this fell foul of the Minimum Energy Performance Standards (MEPS) regulations currently being introduced into many countries, relating to energy efficiency in electrical appliances.

There is a special test standard for refrigerator/freezer MEPS which prescribes test procedures, such as what items shall be placed in a test fridge, the number of times per day the door shall be opened by a mechanical actuator and so on.

The designers of the fridge then came up with an imaginative solution to the poor MEPS ratings:

The controller chip was programmed to recognize the characteristic pattern of door openings and closures of the test standard, and if it detected this sequence, it would switch off the cabinet heaters and invoke other economizing measures!

 

A similar potential problem exists with testing cameras and other video equipment.

Test patterns are useful because they provide a standard reference which is easily interpreted by relatively non-technical operators.

Unfortunately, because of their repetitive nature, it is possible to design software that disguises or eliminates imaging artifacts, but only on static patterns. The ideal test chart would have a slight erratic "wobble" introduced by a small unnbalanced motor or the like

Link to comment
Share on other sites

Jeez, the thread that wouldn't die.

 

Keith, that fridge example really made me laugh! :lol: <--me laughing!

 

It makes me wonder just how many other things are designed to get around the regs in just that manner. It must be a lot once you think of what is at stake.

Link to comment
Share on other sites

  • Premium Member

Well whaddya know, Sony are making an 8K camera

 

It says 8768 x 2324.... With square photosites, that would be a 3.77:1 aspect ratio. The article talks about Bayer, but that sounds more like the traditional Sony vertical stripe idea. It's a muddled report, we'll have to wait for solid info. But if 2324 vertically is correct, it barely makes the Nyquist limit for 1080p.

 

 

 

 

-- J.S.

Link to comment
Share on other sites

  • Premium Member

It says 8768 x 2324.... With square photosites, that would be a 3.77:1 aspect ratio. The article talks about Bayer, but that sounds more like the traditional Sony vertical stripe idea. It's a muddled report, we'll have to wait for solid info. But if 2324 vertically is correct, it barely makes the Nyquist limit for 1080p.

 

-- J.S.

Does it actually say it has square pixels? It might be more like the Genesis sensor, which has square pixels made out of 6 rectangular sub-pixels.

They could also be providing out-of-frame viewing like an optical viewfinder.

 

 

"It's a muddled report, we'll have to wait for solid info."

 

What?! I don't understand; you heard what the man said:

 

"What does the image look like? The pictures from the prototype, on a large 4K monitor, were stunning. Rich blacks, incredible detail, superbly sharp. I’m guessing the native ISO sensitivity is around 800-1000 — just quick speculation in a dark test room and a tabletop set."

 

What more evidence do you need :P

 

Actually there have been threads here and elsewhere about this camera, dating back over 4 years, so it would seem Sony aren't in any great hurry to get it out the door. Guess they want to get it right first time :)

Link to comment
Share on other sites

Sure, Dalsa started the whole "4K RAW = 4K?" argument a few years before Red. And the arrival of the 24P HD Sony F900 started the whole "is HD good enough?" arguments and the attack on 35mm started with reports about how 35mm print projection was barely 1K, etc. This is nothing new, it's just been ramped up. But megapixel fetishing isn't unique to the movie business, look at the still camera market.

 

Yes, we did have resolution discussions even when we were only watching 600 lines PPH projections when we were lucky in cinemas. The first Sony HD features looked pretty bad but it wasn't resolution that bothered me. It was the color and overexposure. The audience didn't seem to care though. It looked as sharp as a 35mm projection and the audience didn't complain. That's the definition of good enough, isn't it?

 

In the still camera camera it's beyond the point of ridiculous, they have 1um pixels now and even their basic ISO contains a lot of noise reduction and produces a very low dynamic range. At least Red chose a large sensor for a large resolution. I hope we don't get to the point where the megapixel number will be as importand as in the digital still camera market.

 

I believe we have come at a point where almost all commercial camcorders provide acceptable sharpness (ignoring compression). The agenda has to shift to aesthetics at some point. Less people complain about the F35 compared to Red One when it comes to color response and there is discussion on the subject. That's a very good sign. When a technology establishes itself and the discussion shifts from specs to aesthetics, it's a sign of maturity.

Link to comment
Share on other sites

  • 8 months later...

I think the problem lies in trying to make digital look like Film. Thats like making apple juice to taste like wine. The last 2 Star Wars films were shot digitally with the sony's in the early 2000's and they looked great. Avatar looked great on the Sonys as well. From Peter jacksons blog the make up artist commented about having to make the skin redder for the Epic camera i found that strange.

Link to comment
Share on other sites

I think the problem lies in trying to make digital look like Film. Thats like making apple juice to taste like wine.

 

You mean like cider?

 

The last 2 Star Wars films were shot digitally with the sony's in the early 2000's and they looked great. Avatar looked great on the Sonys as well.

 

To be fair tho, avatar and those 2 star wars films were mostly made inside a computer.

 

From Peter jacksons blog the make up artist commented about having to make the skin redder for the Epic camera i found that strange.

 

The secret to Red cameras seems to be designing things around the right shades of brown I reckon. Set your film in autumn and film the leaves, lots of brown set decoration but be careful, don't overdo the brown or make your browns too much, a sort of cinamon colour seems to work great! I notice many cheap commercials seem to have picked up on this lately. Great for chocolate and coffee commercials in paticular. You do need to be careful about it tho and not get carried away and be careful what shade of brown you are using, but from what I see the red seems to be all about the browns strangely enough.

 

Basically the moral of that story is love your red and embrace the browns.

 

love

 

Freya

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Forum Sponsors

Metropolis Post

New Pro Video - New and Used Equipment

Gamma Ray Digital Inc

Broadcast Solutions Inc

Visual Products

Film Gears

CINELEASE

BOKEH RENTALS

CineLab

Cinematography Books and Gear



×
×
  • Create New...