Interview With Peter Jackson's DP

[LighthouseMEdia]

Okay Guys we finally got the podcast with Richard Bluck posted at FreshDV.com (http://www.freshdv.com) for all of you who want to check it out.

[Mardi_Gras]

Very upfront and quite matter-of-factly. I like that he echoed the impression of most of the people on this forum, being, the RedOne offers the indie filmmaker the creative freedom in concentrating on telling the story as against struggling with a substandard tool. I see it as a win/win situation for us and the industry as a whole.

[Álex Montoya]

What does he say about latitude? English isn't my native language....

[mjeppsen]

He says around 9 stops. I understand that there have been improvements made since then and it now rates around 12 stops. I'm not being a Red apologist, simply mentioning what I've been told (by someone who would know).

[Finner]

I could not download this interview but if Richard Bluck is saying 9 stops of lattitude I would see it as a pretty bad thing. David Stump said 8-9 stops and if Richard Bluck came to the same conclusion it would be pretty low performance. Red said just after David Stumps tests that it had been fixed and the camera had more then 11 stops but if Richard says he found it to have 9 that would be fairly low performance.

I just wish RED would let a few independent DOP's do a full set of tech tests on the camera so this could be all cleared up.

[Shawn Nelson]

9 stops? Shoot, that's not good. They shot a bit more than 3 weeks ago with pre-production units, but well after they "fixed it" to be 11 stops. Are we back to only 9? Good latitude is probably one of the things I am most looking forward to in Red, I hope the spec of 11 stops doesn't slip

[Craig Bowman]

From what I know about the PJ shoot, all the camera had enabled was start and stop. Over the next number of months as more of final features get locked down and the downloadable firmware updates bring all of the features online in the camera, then I'll feel more comfortable checking the final numbers on Red's dynamic range.

[Finner]

Sure Craig but the low dynamic range problem was said to be taken care of long before the PJ shoot so the cameras should have been in the 11 range like it has been stated they are by the RED team. The thing that concerns me is that every non RED employee that uses the camera so far is saying that it has a dynamic range of 9. This is quite a big discrepancy.

[david farland]

Well said Finner.
I saw that mentioned yesterday and was going to post a please explain.

I can feel a 'shoot out' coming on!

DF

[Finner]

Let me make sure that I am not misunderstood here. I am not trying to start a fight or talk badly about about the RED team. On the contrary I was highly impressed by every RED representitive at the booth and the footage looked incredibly stunning. All I would really like is before I pay for my RED is a camera shootout against other cameras. I assume this will happen when the camera is ready to start shipping. So I am not accusing RED of anything I just am confused that RED says 11 stops and pros using the camera are saying 8-9?

[Graeme Nattress]

You don't measure stops by "eyeballing" an image. You measure it with a test chart, and the best chart for that is the Stouffer 13 stop wedge.

Then you have to define what you mean by stops, and that's a can of worms.

I define stops by saying that it's the difference between peak white (ie no more detail in the highlights) and dimmest shadows (where the noise swamps the detail in the shadows).

As you get darker and darker in the image, the data becomes non-linear as you get more quantisation noise.

That gives us two different measure of dynamic range - linear dynamic range and non-linear as we go into the noise floor.

We measured Frankie and saw > 10 stops of linear dynamic range, and another stop of non-linear dynamic range in shadows giving about 11.3 in total. In that respect I don't see much different in the stuff we've shot with Boris and Natasha.

Graeme

[david farland]

..............

The cameras were malfunctioning during the PJ test.

Stump did his tests on dramatically earlier firmware on Frankie.

11.3 is the number you should expect.

They did their piece with the malfunctioning cameras because that is what was available at the time. They aren't afraid of showing the "bad" with the "good". They aren't hiding anything.

I'll leave out specifics of what actually happened on the PJ test for the people who were actually involved.



Graeme,

You bettter explain that to Brook cos he's saying the cameras were malfunctioning and Richard Bluck says there has been an increase in range to what he was using.

Cheers,

[Graeme Nattress]

Well, even with the sensor noise that we had in the PJ shoot, I can clearly see detail in the first step of the wedge (so it isn't quite clipping) and I can see the subtle difference between step 33 and step 34. Each step is 1/3 of a stop, and I think in the newer prototypes I'd be able to make out that last step more clearly and perhaps the next one too. You've got to do some very extreme levels in photoshop on the 16 bit tiff to see this 11+ stop dynamic range, but that's the nature of the beast.

Remember this though - Boris and Natasha are early working prototypes, and although they made utterly superb images in the hands of PJ and Richard, they're still early prototypes, with foibles and problems. Nobody who saw the footage pointed out any issues we ourselves hadn't seen and know that the engineers back at the lab are either working on fixing or have fixed. Never-the-less, superb images.... Let's not forget that.

Graeme

[Jarred Land]

Graeme,

You bettter explain that to Brook cos he's saying the cameras were malfunctioning and Richard Bluck says there has been an increase in range to what he was using.

Cheers,

Brook wasnt there so any comments by him of the camera malfunctioning are hearsay, and the extent of his exposure to the footage has been in the theater like everyone else... he has never seen the raw files from NZ or any additional footage from Boris or Natasha so take Graeme's explanation as the authority.

[Poi Boy]

You can't be impressed with the images one minute and then when someone puts an eyeball figure on dr say "oh my god how terrible". If you like the images, you like the images and the DP definately liked the images.
Aloha
-A

[Brook Willard]

Brook wasnt there so any comments by him of the camera malfunctioning are hearsay, and the extent of his exposure to the footage has been in the theater like everyone else... he has never seen the raw files from NZ or any additional footage from Boris or Natasha so take Graeme's explanation as the authority.

Bingo. I always made it clear that what I was saying was merely what I was told. I wasn't a part of the test. You can actually see me saying exactly that in the post that David quoted to "bust" me. :)

[Shawn Nelson]

Okay, I am confused...is Red 11 stops or 9?

[Brook Willard]

11.3 stops. I'll leave commentary on what happened in NZ to the guys with red names.

[Poi Boy]

Shawn,
Didn't you rave about the footage at NAB ? Yes; Then let it go dude. I'm taking Graeme's number, mine goes to 11 and the images are awesome.
-A

[Brook Willard]

Now seems like an appropriate time to post this chart. Here are the results Jim posted from the test Graeme described.

http://homepage.mac.com/brookwillard/redfaq/dynamicrange.jpg

[Graeme Nattress]

That's exactly what you should expect from the shipping cameras also. Perhaps better. Notice about 10 stops of very linear dynamic range, and just over a stop more of non-linear dynamic range in the shadows.

Also, depending on lighting conditions, you may also be able to grab an extra stop or so in the highlights with the "Highlight" control in REDCINE. On an over-exposed Macbeth chart where I had to top two white squares blown out, I managed to pull it back so I could see both of them clearly differentiated again. The "Highlight" control uses under-exposed data in non-clipping channels to guess missing data in the clipped channel, so it's not real dynamic range, but it can look very real, so I don't include it in my figures. If it works, it's a bonus. If it doesn't work for your shot, you've got good numbers anyway.

Graeme

[Brook Willard]

That feature absolutely blew my mind when you showed it to me, Graeme. Big kudos once again...

[david farland]

Graeme,

I'm not going to blow sunshine up your backside.....you're too high on my respect chart!

.....jeez and all in the one sentence!!

I think you're being asked to defend you figures in a real live test, a continuation from the CML one.

Never mind the substance of what actually was, Bruck's comments by others will be typical as soon as you sell the first 100. I very much agree with your comments about eyeballing vs test.

Get in early I would say.

Cheers,

DF

PS: Brook, do you really think I was trying to bust anyone?

[Alexander Nikishin]

Further testing on the dynamic range of Red will be done, don't worry people.

Until then, the most concrete proof we have is the stouffer wedge, not a he said she said claim.

[Brook Willard]

PS: Brook, do you really think I was trying to bust anyone?

Nope, that's why I put the smile in there. :)

[Evin Grant]

Having played with Redcine and Graeme's exposure & highlight recovery tool for the four days of NAB I'm quite cofident that extracting the full 12 stops of the 12 Bit depth will be a non issue on all but the most aggregious exposure errors. But that's just my optomistic opinion, I'll let Graeme give to you by the numbers.

[Craig Bowman]

Sure Craig but the low dynamic range problem was said to be taken care of long before the PJ shoot so the cameras should have been in the 11 range like it has been stated they are by the RED team. The thing that concerns me is that every non RED employee that uses the camera so far is saying that it has a dynamic range of 9. This is quite a big discrepancy.

Yes it is isn't it! Look forward to seeing for myself.

[Graeme Nattress]

Thanks Evin!

The highlight control will give varied results. On some footage it will not help at all, on others you may claim a stop or so. In either case, it's a trick to allow you to fully access the data that's there in RAW, and doesn't reflect the DR of the sensor. I'd prefer we just stick to the basic definition of DR and just know in the back of our heads with have these tricks up our sleave.

Graeme

[Evangelos Achillopoulos]

Hello to all red followers

The Stouffer T4110 transmission step wedge is the best way to measure the latitude of a camera.

But the exposure alone without noise to be taken account is half job. Please read the following link which is one of the best software that analyzes camera performance in a scientific way (and its very cost effective…).

http://www.imatest.com/docs/tour_q13.html

We have calibrated our Varicam with it and the results were stunning. As for the latitude measured with the Danes-Picta TS28D transmission step wedge which is 13,6 stops and we measure (for the Varicam) 10,2 stops usable latitude above noise.

Graeme keep trying, you are there on the right track, the revolution IS started.

Regards,

[Graeme Nattress]

Thanks - we use the Stouffer in house, and it shows just over 11 stops before the noise swamps the signal.

Graeme

[Evangelos Achillopoulos]

Are you measure it with imatest?

[Graeme Nattress]

Nope, I wrote my own little program that extracts the values from the chart and tells me when I get too much deviation from linear in the shadows.

Graeme

[Brook Willard]

How would a different [slower] debayering algorithm affect the detail you're pulling from the shadows?

[Graeme Nattress]

Absolutely. That's one of the benefits of raw.

Graeme

[baro]

The "Highlight" control uses under-exposed data in non-clipping channels to guess missing data in the clipped channel

How does that work? How can you guess the clipping channel from the non-clipping ones? Do you use the color information in the area surounding the clipping area?

[Graeme Nattress]

To discuss further, I'd have to get into algorithmic details, and I'm not prepared to do that on a forum. It's all secret sauce stuff.

Graeme

[baro]

understandable