13+ stops....

[Nick Ambrose]

That's okay Stuart.

The trouble with saying a camera is 11.3 or 10+ stops measured 'our' way is they are simply numbers.
People have no frame of reference that they can gauge that number....11.3 stops
....well maybe a stouffer chart on a lightbox viewed through a photoshop grade.
It's a marketing number that give me no comparative measure to another camera.
This isn't fair to the public....well who said marketing was fair!!
But I greatly respect you and Graeme as men of science.
And if either of you tell me a test (or range of tests) to compare the Red's ability to measure latitude response can't repeated on other cameras fairly, (meaning it's more right than wrong)..well I'll believe you.
D

I would say it's more than fair --- they post their results, and methodology so if you follow it exactly, you should get the same results. Who could say fairer than that ?

You could say "well, for my purposes, it's only good to x and y" but the fact it, you should be able to reproduce RED's numbers (if you do it right

 

[Jannard]

Consider them reference numbers. You know what you get from a RED ONE. Mysterium-X is a stop better. Monstro is two stops better than Mysterium-X.

Jim

 

[Gunleik Groven]

which should imply higher bit depth, or did I get that wrong?

 

[JanneJansson]

Is Monstro and further development of Mysterium-X, or is it a completely new puppy?

 

[Gunleik Groven]

would seem like it...

The real fun happens when Mysterium hits the after-market...

Imagine iPhones with Mysterium and RAW in three years time...

That's when (fill in prediction here...)

 

[Sanjin Jukic]

Consider them reference numbers. You know what you get from a RED ONE. Mysterium-X is a stop better. Monstro is two stops better than Mysterium-X.

Jim

Great Jim.

That means I would first upgrade my RED1 to Mysterium-X to get one stop more,

then later I'll upgrade my RED1 to Monstro and get even two steps more.

And what a wonderful world will be in that case just because I'm

using various lens mounts (S35 and FF35) on my RED1 as you know well.

 

[Daniel Browning]

Ymmv

Ymmv

so the RED ONE is supposed to have 11 stops (10+ on the new chart) but after people tested it, it only really has about 8 stops....

Say a cup of Turkish coffee has 11.3 ounces in it. Most people don't drink the last 1-3 ounces because it's too strong, depending on their personal taste and circumstances. For direct-to-DVD, you can drink all 11.3 and it tastes great. For theatrical 4K projection, some only drink 8 ounces. For a slow-motion greenscreen in uncorrected tungsten light, some people only drink 6.

with all the hype i thought we'd see 16+ stops.

It would have been polite to suppress your yawn. Even a 5-15% improvement, like DSLR manufacturers manage every 18 months, is enough to get most people sit up and pay attention. Actually, many times the improvement comes at a cost to low ISO dynamic range (e.g. 30D->40D, 400D->450D).

Here RED is predicting a 400% improvement.

maybe they'll wow us with awesome ISO ratings.. like super clean ISO 10,000

Keep in mind that very fabric of light itself has photon shot noise, and RED ONE is already very close to the maximum theoretical limit for a Super35 sensor. So going with larger sensors will allow them to improve the photon shot noise without breaking the laws of physics.

Read noise, on the other hand, is an area that RED can improve at any sensor size.

The point of the camera is not about getting 'absolute' figures of the RedOne in laboratory conditions.

Laboratory conditions are the only ones that are objective. Everything else involves some measure of personal opinion. One person may think a little noise reduction is fine, and they'll leave some low frequency detail in the shadows, another might think that the loss of color in the shadows ruins the image. There are so many factors that you have to make the measurement as objective as possible, as RED has done.

I'd opine that the most widely accepted results comparing different products would come from an independent third party, rather than any specific manufacturer.

Agreed, ideally there would be dozens of owners who have a $35 test chart and publish their own conclusions and test data.

I opt dpreview do some tests...

Please don't. Dpreviews tests are technically inconsistently done and notoriously flawed.

Ah, the voice of "reality".

DPReview publishes camera reviews that are superior in most ways than all the other review sites, and I love the site for that. But there are some areas where they make very basic and fundamental flaws in measurement, then add flawed interpretation on top of that. Dynamic range is one such area that DPR is greatly flawed. For example, they found almost 13 stops of dynamic range in a 12-bit camera. DPR mixes in all sorts of subjective factors, like using biased raw converters instead of measuring the raw data. On top of everything else, they don't publish their testing data (raw files).

When my 5Dm2 gets here, I'll be doing my own tests and publishing them.

So maybe the problem is the assumption that some objective, absolute number or set of numbers is a good way to compare different imaging systems.

I think it's a good start. One can say "RED ONE is measured to have 1:1 SNR at 11.3 stops, but I only use 10.3 stops because I need 1:2 SNR" or "I measured 11.3 engineering dynamic range on my RED ONE, but I only use 7 stops because of vertical pattern noise caused by radio frequency interference from shooting next inside an electrical sub-station."

Scientific imaging applications, like some astrophotography, don't really need to know how the data looks to a human. So in that case, RED's 11.3 stops number would be wholly sufficient to explain the capability of the camera. But here we're generally more interested in how something looks to a human.

So in the end, a number can only tell one part of the story. There is no number that will describe how the human eye sees pattern noise (for example) or how that noise looks after resampling to DVD output resolution.

I've shot ISO 800,000 with my Canon 20D and got 18 stops (15 usable) of dynamic range at YouTube output sizes. It's not pretty, but you can easily see the test subject. Of course, for 12x18 prints, I stick to ISO 200 and 8 stops or so.

YMMV.

 

[Emanuel A.]

Here RED is predicting a 400% improvement.

Well posted.

Got the (S)35mm sensor sized, DR is the key.

 

[Deanan]

When my 5Dm2 gets here, I'll be doing my own tests and publishing them.

There you go, DBreview.com

 

[Emanuel A.]

Well posted.

Got the (S)35mm sensor sized, DR is the key.

I forgot to refer without mentioning the FF35 2:1 anamorphic coverage but that's another subject.

 

[Dominic Cochran]

Usually one or more of the manufacturers objects to the findings because of the product setup or analysis used.

Shocking.

 

[Dominic Cochran]

I should add... over a large enough sample size, it doesn't matter what analysis is used as long as it's the same for each product.

Do you see why?

 

[david farland]

This is just noise here...
Jim

I think it's more a soft point you guys aren't defending.
You guys got quite a bit of mileage out of 11.3 stops and when it comes to the crunch I should consider them just reference figures...not absolutes.
My hope is that you can defend it better than that....My idea of a defence are any comparative tests I asked Stuart/Graeme to choose.
Why shouldn't I know where my camera stands in comparison to other camera's range? Am I missing some truth here?
Respects, D

ps: can I've a job at Red?

 

[Jannard]

The 11.3 stops for the RED ONE are absolute measured numbers by Graeme and a couple of independent 3rd parties. No excuses. The results were published. But measuring DNR in the lab, while completely real, does not necessarily transfer to the real world of shooting. It is not our job to do comparison measurements. No one would believe us anyway... like you- now.

Jim

 

[Evin Grant]

One thing not taken into account here is that most other "Video" cameras (F23/35/D21) have DSPs that process the image and crush the blacks and may use higher internal bit depths and then record in Log formats that artificially stretch the apparent DR. It's next to impossible to objectively compare DR of two cameras that have so significantly different processing paths. Of course personal taste really does have an enormous impact too. Noise is much more tolerable when it's organic and granular so a particular camera may have a lower measurable S/N ratio but a perceptually grater DR.

 

[david farland]

...There is no number that will describe how the human eye sees pattern noise (for example) or how that noise looks after resampling to DVD output resolution....

Entirely the point of having a controlled comparative test.
The absolute figure doesn't mean much on it's own but in respect to other cameras.
It may tell you a relative potential element of the camera...like is it more forgiving than a F23 in a high contrast sunny scene?

D

 

[David Mullen ASC]

I think to some degree this difference between measurable dynamic range in the lab and real-world experiences of shooters is partially due to the difference of measuring a RAW or LOG recording versus the visible range in the final color-timed image where the contrast has been set to the taste of the cinematographer.

Lately I've been shooting on the Genesis and switching in the DIT tent between the LOG ("raw") signal and one with a basic LUT where the contrast and black levels are set to what I consider good-looking, rich, etc. -- and it's interesting how much dynamic range is lopped off when applying this basic LUT to create a good monitor image.

I saw the same thing when I was shooting on the RED -- there was more dynamic range in the RAW files, but once I color-corrected them to what I considered a "pleasing" contrast, some of the detail in the high and low ends were crushed or clipped. I think it's inevitable that you'll want some of the low-end signal information to be crushed down in order to get rich blacks and less noise in the shadows. Hence why a measured 11-stop range in a lab becomes a working 9-stop range on the movie set when looking at a normal HD monitor displaying a simple conversion to Rec 709. Truth is that it is probably better anyway to work and light within a narrower dynamic range (i.e. light a little flatter) so that your RAW recording has more information than you lit for, giving you some leeway in final color-correcting.

 

[david farland]

It is not our job to do comparison measurements.Jim

True...but don't you ever wonder what's your customers experience is regarding latitude/lighting etc in comparison to other cameras?

No one would believe us anyway... like you- now.Jim

not true..why wouldn't I believe your comparative tests as much as I believe your 11.3 results of an absolute test. Of course I do.

It's what you don't test that interests me too.

D

 

[Jannard]

I think to some degree this difference between measurable dynamic range in the lab and real-world experiences of shooters is partially due to the difference of measuring a RAW or LOG recording versus the visible range in the final color-timed image where the contrast has been set to the taste of the cinematographer.

Lately I've been shooting on the Genesis and switching in the DIT tent between the LOG ("raw") signal and one with a basic LUT where the contrast and black levels are set to what I consider good-looking, rich, etc. -- and it's interesting how much dynamic range is lopped off when applying this basic LUT to create a good monitor image.

I saw the same thing when I was shooting on the RED -- there was more dynamic range in the RAW files, but once I color-corrected them to what I considered a "pleasing" contrast, some of the detail in the high and low ends were crushed or clipped. I think it's inevitable that you'll want some of the low-end signal information to be crushed down in order to get rich blacks and less noise in the shadows. Hence why a measured 11-stop range in a lab becomes a working 9-stop range on the movie set when looking at a normal HD monitor displaying a simple conversion to Rec 709. Truth is that it is probably better anyway to work and light within a narrower dynamic range (i.e. light a little flatter) so that your RAW recording has more information than you lit for, giving you some leeway in final color-correcting.

Try a curve instead of lopping off the tops and bottoms...

Jim

 

[Patrick Jennings]

wow.... this thread should probably be closed...

i really didn't want to start a recurring/boring/over-done back and forth between idiots, half of the RED staff, a billionaire CEO, the moderators and David Mullen.