Interesting Resolution test RED ONE

[Rodney West]

Adam Wilt just got his hands on the new DSC labs 4k test chart, and gave the RED ONE a quick test. I don't really think anyone really thought the RED was lying to us about the resolution of the camera, but seeing someone like Mr. Wilt confirm it is comforting.

 

The RED actually did better than I thought it would, I imagined the low pass filter plus the optics would of cut down the resolution a little more than it did.

 

http://provideocoalition.com/index.php/awi...um_resolved/P0/

 

More RED Res Testing: The Mystery(ium) Resolved

A new chart shows what this camera (and several lenses) can really do.

I obtained a new, 4K resolution test chart at NAB, and aimed RED ONEs running build 15 version 2.2.5 at it using four different lenses: a 50mm Super Speed, a 50mm Ultra Prime, a 18-50mm RED zoom, and a 24-290mm Optimo. Cutting to the chase: I?m pleased to report that I see detail extinction at about 3.2K, confirming the numbers RED and others have claimed.

[John-Erling Holmenes Fredriksen]

Really interesting, although it mostly just confirms what I've heard elsewhere. I guess he's got a good point about the reasoning behind the new 5K camera.

[Emanuel A Guedes]

Thanks for posting, Rodney.

 

Clearing up enough. Still funny though.

 

When the independent tester is someone from... Or used to handle with small video camcorders, at least.

 

However, it is proper to wonder if is there out here anyone who may contradict the independence of someone like this Mr. Wilt?

[Emanuel A Guedes]

Sorry, I meant: ...is there anyone out here...? :lol:

[Stephen Williams]

Thanks for posting, Rodney.

 

Clearing up enough. Still funny though.

 

When the independent tester is someone from... Or used to handle with small video camcorders, at least.

 

However, it is proper to wonder if is there out here anyone who may contradict the independence of someone like this Mr. Wilt?

 

Hi Emanuel,

 

Why would somebody with a reputation for accurate testing put his credibility at risk?

 

Stephen

[Patrizio De Sica]

I obtained a new, 4K resolution test chart at NAB, and aimed RED ONEs running build 15 version 2.2.5 at it using four different lenses: a 50mm Super Speed, a 50mm Ultra Prime, a 18-50mm RED zoom, and a 24-290mm Optimo. Cutting to the chase: I?m pleased to report that I see detail extinction at about 3.2K, confirming the numbers RED and others have claimed.

That said, what for do we need an Epic? Baaah :P

[Stephen Williams]

That said, what for do we need an Epic? Baaah :P

 

Hi,

 

Resolution is only one part of the story, DR & ergonomics probably more important.

 

Stephen

[Keith Walters]

Adam Wilt just got his hands on the new DSC labs 4k test chart, and gave the RED ONE a quick test. I don't really think anyone really thought the RED was lying to us about the resolution of the camera, but seeing someone like Mr. Wilt confirm it is comforting.

I am a little bit worried about this statement from Mr Wilt:

"it?s normal to talk about ?TV lines per picture height?, not lines per picture width"

 

Since when? TV cameras are normally restricted to fixed numbers of lines per picture height; 576 for PAL and 480 for NTSC, and 720 and 1080 for most HDTV. Until relatively recently there was no real way of changing the vertical resolution of a TV camera, even if you wanted to!

 

Horizontal resolution is/was an entirely different story; you could change that easily by simply using low-pass filters, and the theoretical maximum resolution possible for most TV systems was something that was aimed for but rarely achieved. Basically, vertical resolution was only limited by the optics, horizontal resolution was limited by a number of factors.

 

With film scanning on the other hand, normally the number of horizontal pixels per frame width is fixed, but the number of pixels per scan height varies, depending on whether the format is 2.35:1, 1.85:1, 16:9 and so on.

 

In the past, some professional video cameras were notorious for giving really impressive resolution results when pointed at test charts with repetitive patterns, but they were the only things they were any good at imaging! (Similarly, a lot of studio PAL/NTSC to RGB converters would do a bang-up job with a colour bar pattern and other static test signals, but were nowhere near as effective with live TV pictures!)

 

You don't really need precision printed resolution charts. You can do the job just as well by printing out a series A4-page-sized blocks of sinewave images, chalking out a suitably-scaled 16:9 rectangle on a wall, Blu-Tak-ing the pages up inside it, and adjusting the framing so that the rectangle just fills the sensor area. For example, a 4,000-line "burst" can be simulated by printing a 400-line pattern on one or more pieces of paper that are one-tenth the width of the "16" part.

 

The real test of any system is what happens when you tilt the pattern of lines slightly, easy to do with Blu-Tak.

[Brian Drysdale]

That said, what for do we need an Epic? Baaah :P

 

One reason given on REDUser was to allow shooting high speed 4k for VFX work. The RED ONE isn't perfect, so I'd imagine the Epic addresses some of the issues that can't be meet at that price point. As Stephen mentioned, there's more to a camera than just the resolution figures, as much as the marketing people love to shout them out. Unfortunately, there's also a limit to how much the lenses can resolve.

 

In the professional world, the Epic has a relatively average price for what you'd assume is a camera intended to be a top end camera.

[DJ Joofa]

Horizontal resolution is/was an entirely different story; you could change that easily by simply using low-pass filters, and the theoretical maximum resolution possible for most TV systems was something that was aimed for but rarely achieved. Basically, vertical resolution was only limited by the optics, horizontal resolution was limited by a number of factors.

 

I don't think the above quote is entirely correct. Unlike a typical (Fourier) spectrum of an image that results in both horizontal and and vertical frequencies (2D spectrum), the way the TV scanning work for a traditional TV system (say NTSC), there is only *one* frequency. The same single frequency represents both horizontal and vertical detail. So if you put a low pass filter on that single-dimensional frequency spectrum you are going to get rid of some vertical (typically oblique) detail in addition to horizontal detail.

Edited May 11, 2008 by DJ Joofa

[Patrizio De Sica]

One reason given on REDUser was to allow shooting high speed 4k for VFX work. The RED ONE isn't perfect, so I'd imagine the Epic addresses some of the issues that can't be meet at that price point. As Stephen mentioned, there's more to a camera than just the resolution figures, as much as the marketing people love to shout them out. Unfortunately, there's also a limit to how much the lenses can resolve.

 

In the professional world, the Epic has a relatively average price for what you'd assume is a camera intended to be a top end camera.

Yeah? And where's their mission statement? The REDONE mission statement... In the garbage?

[Brian Drysdale]

Yeah? And where's their mission statement? The REDONE mission statement... In the garbage?

 

I thought that RED was a camera manufacturer not a political party. They appear to be creating a range of cameras for different market sectors, just like any other manufacturer. To be successful they have to meet the demands of the maximum number of customers in the markets they're addressing and make a suitable level of profit.

 

From my understanding, you will be able to buy upgrades for the REDONE (including I believe the sensor) and you can still shoot films suitable for quality 35mm distribution with the camera.

[John Sprung]

I am a little bit worried about this statement from Mr Wilt:

"it?s normal to talk about ?TV lines per picture height?, not lines per picture width"

 

Since when? TV cameras are normally restricted to fixed numbers of lines per picture height; 576 for PAL and 480 for NTSC, and 720 and 1080 for most HDTV.

This bit of confusion goes way way back to the earliest days of TV history. There are horizontal scan lines, which is how old time analog TV samples the picture. There are black and white line pairs on test charts. The term "TV lines" actually meant the black and white test chart line pairs. It's not such a great choice of words.

 

Film got there first, and the film guys were the first to measure resolution. They used line pairs per millimeter, which seemed convenient, because you could pick up the physical film, and measure it under a microscope. Film grain is random, so it didn't matter at all which direction the lines went.

 

The use of line pairs per picture height came about because in the early days of TV, they didn't have a standard physical size to go by, like with film. So, they used picture height, no matter which direction they had the lines going. Picture height was always less than the width, so you could set up to match the height of a chart, then turn it, and not lose any out the sides.

 

 

 

 

-- J.S.

[Phil Rhodes]

Errrr 3.2K off a 4K sensor counting light/dark pairs is more than Nyquist even if you assume lens, OLPF and bayer effects are zero, which they are emphatically not.

 

What does he really mean here?

 

P

[Patrizio De Sica]

I thought that RED was a camera manufacturer not a political party. They appear to be creating a range of cameras for different market sectors, just like any other manufacturer. To be successful they have to meet the demands of the maximum number of customers in the markets they're addressing and make a suitable level of profit.

 

From my understanding, you will be able to buy upgrades for the REDONE (including I believe the sensor) and you can still shoot films suitable for quality 35mm distribution with the camera.

You are right. Now they?re acting like any other brand. Arri, Sony, Panasonic, etc. They will be protecting their top end camera. They will be protecting their middle range from Scarlet. Is this what they had in mind when they signed with their early supporters?

[Michel Hafner]

Errrr 3.2K off a 4K sensor counting light/dark pairs is more than Nyquist even if you assume lens, OLPF and bayer effects are zero, which they are emphatically not.

Nyquist = 2K line pairs. He gets 1.6K line pairs.

[Brian Drysdale]

You are right. Now they?re acting like any other brand. Arri, Sony, Panasonic, etc. They will be protecting their top end camera. They will be protecting their middle range from Scarlet. Is this what they had in mind when they signed with their early supporters?

 

Some of the early supporters thought you could run around using the RED One like a HVX 200, regardless of people telling them at the time that it would be much larger and heavier.

 

My impression was that RED promised that the RED One wouldn't be obsolescent over a short period (I assume like the consumer/prosumer cameras, the professional cameras seem to have longer lives - the F900 is still going strong). The Epic doesn't do that to the One, it has different features and the RED One is still selling at a low price and is due to have better firmware soon.

 

J.J. gives an explanation why the Scarlet is configured the way it is on REDUser. Basically, the price would be a lot higher for an interchangeable lens camera because that market is so much smaller than the one for a camera with a fixed lens.

[Keith Walters]

Nyquist = 2K line pairs. He gets 1.6K line pairs.

Excuse me. I think I may have blundered into the wrong universe again.

Quickly, is the ratio of the circumference of a circle to its diameter: 2.718281828459045235360287471352,

or is it that other number, you know, the one that's approximately:

3.141592653589793238462643383279502884197169399375105820974944592307816406286208

998628034825342117067982148086513282306647093844609550582231725359408128481117450

284102701938521105559644622948954930381964428810975665933446128475648233786783165

271201909145648566923460348610454326648213393607260249141273724587006606315588174

881520920962829254091715364367892590360011330530548820466521384146951941511609

"Tart" or "pudding" or something like that?

[Guest Glen Alexander]

Excuse me. I think I may have blundered into the wrong universe again.

Quickly, is the ratio of the circumference of a circle to its diameter: 2.718281828459045235360287471352,

or is it that other number, you know, the one that's approximately:

3.141592653589793238462643383279502884197169399375105820974944592307816406286208

99862803482534211706798214808651328230664709384460955058223172535940812848111745

0

28410270193852110555964462294895493038196442881097566593344612847564823378678316

5

27120190914564856692346034861045432664821339360726024914127372458700660631558817

4

881520920962829254091715364367892590360011330530548820466521384146951941511609

"Tart" or "pudding" or something like that?

 

yes but everyone always rounds it off to 22/7

[John Sprung]

Nyquist = 2K line pairs. He gets 1.6K line pairs.

The theoretical Nyquist limit is N/2 = 2K. But unlike a digital filter, the OLPF can't have negative coefficients, and must begin rolling off at N/4 = 1K to be completely out by N/2. There are no "brick wall" OLPF's.

 

In practice, there are two things happening here:

 

1. Instead of being all the way out by the Nyquist limit, they accept a small amount of aliasing, and make an OLPF that pushes past the theoretical limit just a little. This is a subjective judgment, which is why I think DP's should have their choice of OLPF's.

 

2. They don't get 100% modulation transfer at 1.6K, but they do get enough to still see the lines. It's a second subjective judgment call as to where in the top octave you think you run out of useful detail.

 

As we've already seen here, we can get into endless pointless arguments about this stuff.

 

So, I'd propose that if we want to compare chips and cameras numerically, let's do it based on the only numbers that can be determined simply and indisputably: Just count the photosites, and state what kind of color filter pattern or separation is used.

 

 

 

 

-- J.S.

[Keith Walters]

yes but everyone always rounds it off to 22/7

Twenty - two over 7....?

I always thought there was something fundamentally wrong with the universe.

 

Doesn't it ever bother you that so much of our science and technology is based on irrational numbers?

[Guest Glen Alexander]

Twenty - two over 7....?

I always thought there was something fundamentally wrong with the universe.

 

Doesn't it ever bother you that so much of our science and technology is based on irrational numbers?

 

ah no, people often try to impose their own logic on the real world to make sense of it.

 

the calculation of pi doesn't even include any local spatial or temporal anomalies..

 

if you calculate pi - acos(-1.0) you will see your machine's numerical precision limit .

[Keith Walters]

the calculation of pi doesn't even include any local spatial or temporal anomalies..

 

if you calculate pi - acos(-1.0) you will see your machine's numerical precision limit .

Eh? Are you saying your machines still have a numerical precision limit?

How crudely quaint.

[Guest Glen Alexander]

Eh? Are you saying your machines still have a numerical precision limit?

How crudely quaint.

 

 

ha ha. well if you choose to adhere to IEEE numerical definitions or choose to ignore them when you compile your code. :ph34r:

 

i still don't see how one gets an ASA of 10,000

Edited May 15, 2008 by Glen Alexander

[Keith Walters]

ha ha. well if you choose to adhere to IEEE numerical definitions or choose to ignore them when you compile your code. :ph34r:

"Compile"? Quantum computers do NOT compile. Plan 12 if I am not mistaken?

Sorry, where are my manners? Would you like some food? Some bananas perhaps? :rolleyes:

i still don't see how one gets an ASA of 10,000

 

Perfectly feasible. 10,000 ASA is about 5 stops better than the nominal 320 ASA (since 32 x 320 = about 10,000)

 

Assume that a setting of 320 ASA only just uses all 12 bits of the ADC.

If one then closes the iris down 5 stops one will only be using 7 bits of the ADC.

Shift those 7 bits 5 bits to the left and voila one has a full contrast picture again.

Place 5 stops of ND in the Matte Box, open up the iris five stops again and there you have it: 10,000ASA, with 7 bits dynamic range. How useable that is depends on what you are shooting.