Jump to content


Photo

The Mile High Club


  • Please log in to reply
22 replies to this topic

#1 Keith Walters

Keith Walters
  • Sustaining Members
  • 2219 posts
  • Other
  • Sydney Australia

Posted 08 September 2006 - 11:20 PM

From the information that is now appearing on similar forums to this one and other places on the Net, RED are claiming that their "Mysterium" sensor has a 14-15 f-stop dynamic range.

Can somebody tell me exactly what that means?

If a "stop" means a doubling of the incoming photon intensity in the normal photographic sense,
are they saying that the difference in brightness between the most dimly illuminated and the most brightly illuminated pixels of a captured scene can be as much as 2^15 (two raised to the 15th power) ie 32,768 times?

In other words, (assuming one could actually MAKE a lens this good) for every one photon striking the darkest pixel, there could be around 32 thousand photons striking the brightest pixel! Or do they mean something else?

Let's just put that into perspective for a moment.

Everything you watch on TV (either broadcast or from DVD etc) at some point is going to be processed and/or delivered with/as an 8-bit luminance signal. That means that the maximum possible number of brightness steps between the darkest and brighest pixels on the screen is 256. (A lot of Plasma and LCD TV manufacturers claim some remarkable dynamic range figures eg 10,000:1 for their products, but that just means you can vary the size of those 255 "steps", over a considerable range, really an attempt to give you the same range of overall brightness and contrast that you used to get with a regular CRT-based TV). It doesn't mean that that there are 10,000 possible brightness increments in the on-screen image!

If you think of the smallest possible brightness step as 1 centimetre, which is about as thick as the end of your little finger, then an 2.4 Metre (8-foot) household ceiling (which is the minimum allowed under most building codes) is pretty close to 256 little-finger widths.

So if the tip of your little finger represents one bit, the minimum brightness increment, then the floor to ceiling height approximately represents an 8 bit luminance signal, which is all you are ever going to see broadcast to your TV set.

Yet nobody is going to claim that 8 bits isn't enough. You see some stunning images on prime-time TV, which have been rather convincingly squeezed through that 8-bit bottleneck! So the trick is obviously to capture and compress the extremes of brightness that occur in the real world and massage them through that tiny transmission "hole". In a TV studio they normally go to a great deal of trouble to ensure that the lighting never exceeds an 8-stop range but you can do that because you're normally completely cut off from the extremes of the outside world.

And how big are these extremes?

Well using our building analogy, if you regard 8 bits as being the same as 8 stops, then if we have a camera with a 10-bit dynamic range, with the same iris setting as an 8-bit camera it should be able to handle a dark to bright difference of four times 256 = 1024. 1024 cm is about 32 feet - (which is about the height of a three-story block of units, the biggest you can normally build without providing an elevator). Without touching the iris or otherwise altering the exposure, you should then be able to take the camera back into the 8-stop-lit studio an capture images there, without them becoming excessively noisy.

Of course, for transmission the images can't remain the way they were captured, the 10-bit images will need to be adjusted down to the normal 8 bits, but the important think that is often overlooked is that the camera has to capture the dynamic range before it can process it!

What about 15 stops? Well that's 32 feet x 2^15 (32) = 1,024 feet! Which is a just little bit shorter than the empire state building!

If you consider the very maximum exposure that film emulsion can produce anything recognisable from as being 20 stops (sunlight reflecting of wet leaves or similar) well that would be 2 ^ 20 times, a little bit over a million. A million centimetres is 10,000 kilometres - about six miles - or the height of Mt Everest!

But getting back to the Empire State Building claimed by RED, this sure raises some interesting questions. Unless they use some sort of analog pre-processing a 15-stop dynamic range mandates the use of a 15-bit analog-to-digital converter. But according to their figures, if it's a 5K sensor it must have around 14 million pixels. At 24 fps that equates to 336 million 15-bit samples per second.

While you can get 8-bit ADCs that work at that speed, I've never heard of any 15-bit units that go anywhere near that fast. (A 15-bit ADC is 128 times the size and complexity of an 8-bit unit). The Arri D-20 sensor is divided up into 32 separate channels, each with its own (I think) 12-bit ADC.

So I'd love to know where RED are getting their chips made, and also what sort of signal processing technology they have than can process over 5 Gigabits per second? And why Jim Jannard only seems interested in applying all these technological breakthroughs to a cheap movie camera.

It's all quite fascinating, and I look forward to seeing the results of all this in due course.
  • 0

#2 Anders Holmstrand

Anders Holmstrand
  • Basic Members
  • PipPip
  • 10 posts

Posted 08 September 2006 - 11:58 PM

Brilliant, Keith - well written and very entertaining! Even, somehow, convincing...if not also depressing....

I still have high hopes for RED and think Jannard is a real peach for even trying.
  • 0

#3 Jim Jannard

Jim Jannard
  • Basic Members
  • PipPipPipPip
  • 361 posts
  • Cinematographer

Posted 09 September 2006 - 01:00 AM

Keith... or is that "bullandgoose"?

We claim greater than 66db. And that we can get better with some sensor tweaking. That's it. Others have said 14-15 stops. If one of our guys repeated it, sorry. Too early to say a stop range.

Nice to see you post your real name this time. And I acknowledge receipt of your personal email.

Jim
  • 0

#4 Keith Walters

Keith Walters
  • Sustaining Members
  • 2219 posts
  • Other
  • Sydney Australia

Posted 09 September 2006 - 01:34 AM

Keith... or is that "bullandgoose"?

We claim greater than 66db. And that we can get better with some sensor tweaking. That's it. Others have said 14-15 stops. If one of our guys repeated it, sorry. Too early to say a stop range.

Nice to see you post your real name this time. And I acknowledge receipt of your personal email.

Jim

"or is that "bullandgoose"?"

I'm sorry, I've got no idea what that means.

"Nice to see you post your real name this time. "

Obviously you have me confused with somebody else. I've only ever made a small number of posts here and they've always been under this name. I used to have an older account under a made-up name, but I've long since forgotten what the signin name was and I hadn't posted anything for years, just been a "lurker". When they started the crackdown on user names earlier this month it was easier to start afresh. Until recently you couldn't open an account with a "freebie" e-mail address either , but that appears to have changed.

"And I acknowledge receipt of your personal email."
I don't remember sending one to you. (Unfortunately my mailbox didn't save copies of what I have sent)



Well anyway, 66dB means a sensor output voltage range of about 1,400 to 1 or about 11 bits. Somewhat more realistic, and not all that different to what you get from current model HDTV cameras.

I must confess I'm a trifle concerned about some of the extravagant claims I've been hearing on other forums. Every time I hear reports of "noiseless" images I have to ask does that mean "Images without any noise" or "Images where the noise has been removed".

When are you going to post some stills either on your site or somewhere we can access them?
  • 0

#5 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11941 posts
  • Other

Posted 09 September 2006 - 07:03 AM

Hi,

I think the point here is not how they've managed to process a 15-stop range, but how they've managed to image it in the first place. A 15-stop sensor, especially a 4k by 2k Bayer that size, would be truly revolutionary. What's actually going on here is that they don't really have a 15-stop sensor, they have a 66dB sensor, which is actually possible.

Jim, I'll be in Amsterdam on Monday and I'll look at what you have then. But whatever you've built technically, you are a master of information management and coralling the groundswell of opinion!

Phil
  • 0

#6 Jason Rodriguez

Jason Rodriguez
  • Basic Members
  • PipPipPip
  • 175 posts
  • Cinematographer

Posted 09 September 2006 - 10:47 AM

What's actually going on here is that they don't really have a 15-stop sensor, they have a 66dB sensor, which is actually possible.


I think you need to make a distinction between dynamic range and actual S/N ratio.

Is it 66dB of S/N? . . . if so, that's VERY high for an HD camera . . . most of the sensors on the market today advertising 66dB of dynamic range only have actual S/N ratios in the low 50's which makes them somewhat noisy in comparsion to a sensor with an actual S/N ratio of 66dB.

Edited by Jason Rodriguez, 09 September 2006 - 10:47 AM.

  • 0

#7 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11941 posts
  • Other

Posted 09 September 2006 - 11:03 AM

Hi,

Yes, quite. The problem with comparing numbers here is that it really depends what you consider within dynamic range, and what you consider to be so buried in the noisefloor that it's unusable. I saw up to 10 stops dynamic range from basic cameras such as the XLH1 and HD100 in tests, but the bottom two were hopelessly noisy. For this reason, a cooled sensor (with attendant reduction in noise) may appear to have a greater dynamic range; what it actually has is more usable shadow range.

Astronomers have the luxury of averaging successive exposures to reduce apparent noise. We don't...

Phil
  • 0

#8 Jim Jannard

Jim Jannard
  • Basic Members
  • PipPipPipPip
  • 361 posts
  • Cinematographer

Posted 09 September 2006 - 11:15 AM

We claim, and apparently to those that have seen the footage, deliver an extremely clean 66db S/N ratio. There has been NO noise reduction on the footage we are showing. None. A simple debayer, color profile and color timing by the Assimilate people. David Stump can verify the process. The footage now speaks for itself. I will sign off this board due to time constraints. I just don't have the time to monitor and properly respond to all forums. My best to you all. Hope you have a chance to see the footage as we find new venues to show it.

Jim
  • 0

#9 Hal Smith

Hal Smith
  • Sustaining Members
  • 2280 posts
  • Cinematographer
  • OKC area

Posted 09 September 2006 - 02:07 PM

From the information that is now appearing on similar forums to this one and other places on the Net, RED are claiming that their "Mysterium" sensor has a 14-15 f-stop dynamic range.

So I'd love to know where RED are getting their chips made, and also what sort of signal processing technology they have than can process over 5 Gigabits per second? And why Jim Jannard only seems interested in applying all these technological breakthroughs to a cheap movie camera.

Agilent (old Hewlett-Packard Test and Measurement Division) has gear that can handle 5GB/S standing on its head.

Jim has the sort of resources that if he needed an engineer who could design 5GB/S ADC's, he'd drive over to Palo Alto, park in Agilent's lot, wave his checkbook in the air, and drive home with the best hyper high-speed ADC engineer in the world in his passenger seat.

66dB = 10 ^ 66/20 = 1995.26 ~ 11 stops latitude, very close to 5218 and impressive for a solid state sensor.

Keith, I'm curious, what's your technical/scientific background? I'm an old retread Physics teacher (MS in Teaching of Physics) with a broadcast engineering consulting business for the last 18 years.
  • 0

#10 Keith Walters

Keith Walters
  • Sustaining Members
  • 2219 posts
  • Other
  • Sydney Australia

Posted 09 September 2006 - 06:55 PM

Agilent (old Hewlett-Packard Test and Measurement Division) has gear that can handle 5GB/S standing on its head.

What do you mean by "handling" though? 5GB/S is just the rate the raw data comes in, actual real-time video processing at that data rate would be a prodigious task, particularly for something that has to be portable and able to run off batteries.

Getting the software bugs out of something that runs that fast would also be a major challenge.
It's an interesting point that at present there doesn't appear to be a single manufacturer of HD digital set top boxes who has managed to get his software working reliably! Consider that they are receivers only, working with a much lower data rate, and the size of that market...

Keith, I'm curious, what's your technical/scientific background? I'm an old retread Physics teacher (MS in Teaching of Physics) with a broadcast engineering consulting business for the last 18 years.

As far as video camera technology goes, I have been involved in several projects over the past 20 years that required sometimes major redesign of commercial CCD cameras. Initially these were automated surveillance applications.

A side effect of all this was that I had to put in a lot of in-depth study on exactly how CCD (and CMOS) sensors operate, what their limitations are and so on, as we struggled to get the last ounce of performance out of them. Later when I worked for Panavision Australia I put this to good use designing video tap cameras, among other things. For a while I was also responsible for the maintenance of their Betacam fleet, although they got out of that in the early 90s.

Like other technically literate people, when somebody from a small firm with no track record in the field announces that they've suddenly made a major and long-desired technological breakthrough that has thus far defeated the combined intellectual muscle of the likes of Sony, Kodak, Philips and all the other "usual suspects", my first reaction is to ask: "That's nice. But HOW did you do it?"

I'm not saying he can't possibly do it, all I (and quite a few others here) are really asking is HOW is he doing it.

Jim has the sort of resources that if he needed an engineer who could design 5GB/S ADC's, he'd drive over to Palo Alto, park in Agilent's lot, wave his checkbook in the air, and drive home with the best hyper high-speed ADC engineer in the world in his passenger seat.

I have no doubt Mr Jannard has the resources to do lots of things, he just seems remarkably reluctant to tell us what they are, in anything but the vaguest terms at any rate. In any event, he'd cerainly need to do more than to "drive home with the best hyper high-speed ADC engineer in the world," he'd also need a place for him to put his designs into actual hardware. It's not just a matter of the actual design, in general he would also need a way to fabricate higher-performance chips than are currently available. If his engineer failed to deliver the goods, he could just terminate his employment, but you can't get a refund on a state-of-the-art wafer fab facility quite so easily. In any event, I doubt that even HE could afford to build one of those.

I have no doubt he is going to produce some sort of camera, although how he is going to achieve this at the price quoted remains the major mystery.

66dB = 10 ^ 66/20 = 1995.26 ~ 11 stops latitude, very close to 5218 and impressive for a solid state sensor.

It all depends on what he actually means by "66dB".
(I made a mistake in my calculation by the way.
60dB is about a million times power gain, another 6dB is another four times giving 4 million total, the square root of that is about 2,000 as you say, just short of 2,048 which represents 2^11 or 11 stops. For some reason I decided 6dB represents a doubling of power, not voltage!)

Unfortunately "latitude" is something of a weasel word. With film, the exposure latitude is somewhat arbitrary, because even when you get past the quoted figure, there is still information being recorded on the emulsion, which could be extracted by advanced film scanning technologies such as the "double exposure" system of the Arriscan. The main point here is that there really is no point where the emulsion "stops working", it's more a matter of what level of distortion is considered acceptable. Any "mile high" pinpoints of the image tend to get "rounded-off" rather than "sliced-off" which is visually much more acceptable.

With an electronic sensor, once past the quoted exposure latitude, there is usually nothing recorded but a white blob, and no amount of post-processing now or in the future will ever recover anything from those areas.
  • 0

#11 Hal Smith

Hal Smith
  • Sustaining Members
  • 2280 posts
  • Cinematographer
  • OKC area

Posted 10 September 2006 - 12:19 AM

What do you mean by "handling" though? 5GB/S is just the rate the raw data comes in, actual real-time video processing at that data rate would be a prodigious task, particularly for something that has to be portable and able to run off batteries.


Agilent has logic analyzers that can look at 4GB/S data streams and give complete analysis and readouts of what's coming down the line. For instance: http://www.home.agil...00844.0/pc.html . They've got Ethernet test gear that can analyze 10GB/S networks in real time. http://advanced.comm...5989-1556EN.pdf . It looks to me like Jim could be using off the shelf 10GB/S Ethernet hardware and chips internally to spit one heck of a lot of data around in his gear.

RED could be using parallel processing where they pull the data off the chip simultaneously from different zones, process it, and then interleave it which is a very straightforward task in the digital domain. That would buy them more time to real-time process their data.

Unfortunately "latitude" is something of a weasel word. With film, the exposure latitude is somewhat arbitrary, because even when you get past the quoted figure, there is still information being recorded on the emulsion, which could be extracted by advanced film scanning technologies such as the "double exposure" system of the Arriscan. The main point here is that there really is no point where the emulsion "stops working", it's more a matter of what level of distortion is considered acceptable. Any "mile high" pinpoints of the image tend to get "rounded-off" rather than "sliced-off" which is visually much more acceptable.

I'm not certain I'd agree totally, to me the latitude of a media is the reasonably linear or curvilinear (AKA gamma curve) portion of a sensitivity curve. There may be some recoverable data in film blown out highlights, etc. but I think that's the realm of applications where you're pushing the media to the very specialized extremes found in areas like intelligence analysis. Maybe saving the occasional irreplaceable shot in a documentary, etc

I don't think we're coming from radically different viewpoints - I'm just a little more willing to give Jim the benefit of the doubt as to whether or not his team is clever enough to be leveraging the technology that exists in other state-of-the-art digital applications like high speed telecommunications.
  • 0

#12 Keith Walters

Keith Walters
  • Sustaining Members
  • 2219 posts
  • Other
  • Sydney Australia

Posted 10 September 2006 - 01:48 AM

Agilent has logic analyzers that can look at 4GB/S data streams and give complete analysis and readouts of what's coming down the line.

Those devices really only comment on the health or otherwise of the signals being received, and to a large part can be built using the same components as the communication links they are designed to test.

It's another thing altogether to process real time high definition video, particularly at the resolution stated.

In any event, the purpose of my original post was to clarify what sounded like an outrageous claim, which as it happened, RED deny actually making. So, as has so often been the case here and on other forums, it was either somebody's overactive imagination running away with him, or more likely somebody who simply had no idea what he was talking about.
  • 0

#13 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11941 posts
  • Other

Posted 10 September 2006 - 09:48 AM

Hi,

Actually if you listen to that YouTube video, you'll hear someone mention a highlight being 15 stops hotter and holding detail.

He's on drugs. I will challenge them on this tomorrow...

Phil
  • 0

#14 Hal Smith

Hal Smith
  • Sustaining Members
  • 2280 posts
  • Cinematographer
  • OKC area

Posted 10 September 2006 - 10:21 AM

Those devices really only comment on the health or otherwise of the signals being received, and to a large part can be built using the same components as the communication links they are designed to test.

Agilent has been using DSP for twenty years to analyze signals, they're so far ahead of the curve that their ten year old gear is often better than other company's brand new gear.

Admittedly, if there was an Agilent/HP Anonymous, I'd be court ordered to attend meetings. I've got a shop full of their gear and a 90% complete collection of all the versions of their first product, the 200A/B series of audio oscillators, including a pre-WWII 200B identical to the one's HP sold to Disney for tuning up the Fantasound sound systems for "Fantasia".
  • 0

#15 Mr. Shannon W. Rawls

Mr. Shannon W. Rawls
  • Basic Members
  • PipPipPip
  • 127 posts
  • Producer
  • Lemiert Park(Hollywood), CA.

Posted 10 September 2006 - 10:25 AM

It's like a witchhunt.

Why can't everybody just let the company do their thing. Kickback and enjoy the footage?
  • 0

#16 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11941 posts
  • Other

Posted 10 September 2006 - 03:13 PM

Hi,

> Why can't everybody just let the company do their thing. Kickback and enjoy the footage?

Because we recognise that we are, in some way, being deceived.

This is not good. We are not consumers. We need equipment that works in a known way. This cloak and dagger stuff is extremely unwelcome.

Phil
  • 0

#17 Thomas James

Thomas James
  • Basic Members
  • PipPipPipPip
  • 844 posts
  • Camera Operator

Posted 10 September 2006 - 03:34 PM

So when are you going to buy into Red Digital Cinema Shannon?
  • 0

#18 Mr. Shannon W. Rawls

Mr. Shannon W. Rawls
  • Basic Members
  • PipPipPip
  • 127 posts
  • Producer
  • Lemiert Park(Hollywood), CA.

Posted 10 September 2006 - 05:37 PM

As soon as it's ready. In the meantime I've bought into what's available today.
  • 0

#19 Tomas Koolhaas

Tomas Koolhaas
  • Basic Members
  • PipPipPipPip
  • 334 posts
  • Cinematographer
  • los angeles

Posted 11 September 2006 - 10:21 PM

As soon as it's ready. In the meantime I've bought into what's available today.


Hi Shannon,
Well isn't life strange apparently we find ourselves on the same side of this one haha....why did they lock your topic givng JJ kudos??, I wanted to add my support to that sentiment!!, all this hostility is a complete mystery to me too, especially when many of the same critics will probably be clamouring over each other to use the RED when it gets released (if it is as good as it sounds...either way it's not even built yet so everyone just needs to relax and withhold judgement!!).
Cheers.
  • 0

#20 Mr. Shannon W. Rawls

Mr. Shannon W. Rawls
  • Basic Members
  • PipPipPip
  • 127 posts
  • Producer
  • Lemiert Park(Hollywood), CA.

Posted 12 September 2006 - 12:04 PM

Tom,
They probably locked it because it had no nutritional value to this forum, and I agree. I just had to vent a bit. *smile*


Yep you're right, the same critics will be in line trying to buy one if all works out well. The thing is, they know the controversy will be long forgotten by then.

The internet is devilish sometimes! lol
  • 0


Tai Audio

Paralinx LLC

Abel Cine

Gamma Ray Digital Inc

Ritter Battery

Media Blackout - Custom Cables and AKS

The Slider

Opal

Technodolly

FJS International, LLC

Wooden Camera

Broadcast Solutions Inc

CineTape

Metropolis Post

Rig Wheels Passport

CineLab

Aerial Filmworks

rebotnix Technologies

Willys Widgets

Visual Products

Glidecam

Aerial Filmworks

Tai Audio

Abel Cine

CineLab

Wooden Camera

rebotnix Technologies

Media Blackout - Custom Cables and AKS

Gamma Ray Digital Inc

Rig Wheels Passport

Ritter Battery

Technodolly

Broadcast Solutions Inc

CineTape

Visual Products

Glidecam

Paralinx LLC

Opal

Metropolis Post

Willys Widgets

FJS International, LLC

The Slider