Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

RED 8K Sensor Test 2017

Great stuff Phil!

Such a valuable resource and contribution. Much appreciated.

Cheers,
David
 
Unless you have done similar tests, I don't think anyone truly appreciates the amount of work that goes into this type of testing. Especially when you have three OLPFs to test * all of the ISO levels. It is a lot of work.

Thank you for doing it Phil!

What is the light source producing all of the dots on the metal?
 
  • Thread starter
  • Moderator
  • #23
David Battistella said:
Great stuff Phil!

Such a valuable resource and contribution. Much appreciated.
Click to expand...

Thank you David, and thanks to everybody while I'm at it!


Gavin Greenwalt said:
During the noise chips
Click to expand...

Gremlins. But it's not a faulty clip or iris thing. I never use filters for tests like this. You actually can't use filters for a test like this.


Karim D. Ghantous said:
Very cool. And we all appreciate that you spent time and effort doing these tests.

I wonder - and perhaps it has been settled in another thread - but does RED use 18% grey or 12.5% as the mid point? What are your light meters calibrated to?

I notice that both sensors have 16+ stops of DR. That is obvious. I am curious though as to why DxO would rate the sensors at just under 15 stops? Perhaps it's because, while the clipping point is a hard fact, the amount of tolerable noise in the shadows is subjective.
Click to expand...

I go off of visible and measurable data and I also go to weird lengths to see where things really break. The clipping point is seemingly a hard fact, but it's actually weirder than that on all cameras. In this test it's a full RGB exposure on the Xyla with no channels clipping. It's actually not a daylight source, but it's fun to test both and see what channel each OLPF clips first. We're talking about the minutia found at the top level of the exposure before it blows out, but it's interesting there is that variance. Looking at the linear data is also more interesting as well, but that will be discussed later for sure.

You can't get much more precise than the three methods I use to measure DR, but it is something do willy nilly and you can't just use "any lens". First you need to find something that essentially doesn't vignette across your sensor and that in itself is not always easy. For fun this time around I use a few lenses because I just have that ability. For the content you are seeing I'm using the same fast very large format prime that I did the 6K tests with about 2 years ago. It's the best reference lens I have for these formats.

I use 18%, the charts use 18%, my meter is calibrated to 18%, I'm reading 18% on the RED cameras.


Stacey Spears said:
Unless you have done similar tests, I don't think anyone truly appreciates the amount of work that goes into this type of testing. Especially when you have three OLPFs to test * all of the ISO levels. It is a lot of work.

Thank you for doing it Phil!

What is the light source producing all of the dots on the metal?
Click to expand...

Thank you Stacey. It is indeed a lot of work. Plus the compositing of the different portions of the shuttered stops to avoid light contamination. That's where it got fun. This time around I made a program to do some assembly for me, which sped things up, but I also tested it about 3X more than last time.

The light source is LED based and multiple bulbs, just thought it was a good way of "showing gaps" of the clipping size for a side by side comparison.


The additional chaos with these tests are the Xyla itself, which is a light source and illuminates the lens and the camera. So that's all in a blacked out room, and I mean pitch black and shielding the LCD on the camera to block it's emission. The individual patch test luckily is more controlled than that, but eventually everything is reflecting or emitting light so you are working within physical limits.
 
  • Thread starter
  • Moderator
  • #25
Kemalettin Sert said:
Phil can you share youtube video as a downloadable?
Click to expand...

I'll upload a proper 8K version somewhere. The main problem is that's it's over 100GB even with 8K ProRes LT, so web sharing is hard.

The not-YouTube-recompressed version actually looks much, much better and that's where the fun is for content creators :)

I get back to LA later this week and can do that then.
 
Last edited:
Thankyou Dr Phil for what must been a phenomenal amount of work.

+) The chips on this test look more solid that the 2015 Weapon dynamic range test (from stop 10 and greater)

+) Do you think that IPP2 benefits Dragon more than Helium (noise in dark colours areas on the Dragon look less digital and cleaner than 2015 test)

+) Have you developed an OLPF preference on VV with IPP2?

Loved your Space Walk (another modest talent) btw :-)

AJ
 
First off, dragon looks incredible.

The xyla images are made from 16 frame averages, which I think gives us quadruple the signal / noise ratio of the single frames we get when recording realtime video. From my experience dragon fades into a very noisy shadow albeit with some signal there, while Helium clips to black before it gets so deep into the noise. While increasing snr dragon might have more dynamic range, yet I wonder if anyone would use those last two or three noisy stops or try to crush them to black a la helium?

Also helium obviously still has issues with green and magenta casts that dragon used to have, but now seems to have gotten over.
 
Phil, your dedication to this community is truly amazing! Thanks so much for taking what I'm certain is a massive amount of your time to get this done. The information is priceless!
 
  • Thread starter
  • Moderator
  • #32
Antony Newman said:
Thankyou Dr Phil for what must been a phenomenal amount of work.

+) The chips on this test look more solid that the 2015 Weapon dynamic range test (from stop 10 and greater)

+) Do you think that IPP2 benefits Dragon more than Helium (noise in dark colours areas on the Dragon look less digital and cleaner than 2015 test)

+) Have you developed an OLPF preference on VV with IPP2?

Loved your Space Walk (another modest talent) btw :-)

AJ
Click to expand...

Cheers Antony!

- Yep, and in motion this is out of camera no averaging or anything.

- I think IPP2 (it's various functions) and RWG/Log3g10 benefits all RED cameras. I've been testing footage aggressively since July 2016. For most the biggest immediate difference will be super saturated, monochromatic light sources, out of gamut colors will be more "wrangled". But this whole IPP2 thing provides more of a 3D quality and more of a "tonal thickness" overall from M, Mysterium-X, Dragon, and now Helium. Graeme's put a lot of work into this.

- My preference hasn't changed much over the years. I'm fond of the STH because I can manage what that means from a lens and lighting side of things. I more or less would recommend the Standard OLPF for general use if you are looking for better light sensitivity while maintaining better highlight sensitivity and overall color over the LLO.

On a side note, and I should preface this with "I'm weird", Dragon and Helium are two different digital film stocks and the OLPFs also are sub-film stocks for me as they mean different things when it comes to my Rated ISO selection and general artifact protection. The STH, while lower in optimal ISO performance, provides the safest image. The Standard get's you much of what's good about the LLO with a smidgen of what's good about the STH and is a great choice.


Anton Shavlik said:
First off, dragon looks incredible.

The xyla images are made from 16 frame averages, which I think gives us quadruple the signal / noise ratio of the single frames we get when recording realtime video. From my experience dragon fades into a very noisy shadow albeit with some signal there, while Helium clips to black before it gets so deep into the noise. While increasing snr dragon might have more dynamic range, yet I wonder if anyone would use those last two or three noisy stops or try to crush them to black a la helium?

Also helium obviously still has issues with green and magenta casts that dragon used to have, but now seems to have gotten over.
Click to expand...

The purpose of the average and the median filter as well as the extremely overt push to ISO 51200 (which these cameras can't do internally, nor should you do it) is mainly there for visibility's sake. When I post a test like this it goes out to all sorts of folks on calibrated and uncalibrated displays. Helium has same very nice rich, deep, and smooth shadow detail. Those "noisy stops" should never make it to say a skin tone exposure. They should always be in the bottom 0=30% for shooters and colorists alike. Not to mention you can light and in some cases expose and avoid all that entirely. The green-ish tint in Heilum's shadows is real, but no less real than say old Fuji film stocks. Heck, Kodak is sometimes uniformly green and corrected out. So it's not so scary, but something to be aware of. I haven't experience magenta casts however, certainly not like Mysterium-X.

(Side note and personal note, I'm all about Dragon and Helium, and not-so-much about Mysterium-X. It's a kickass sensor by it's own right, but Dragon and Helium are better in almost every way)

steve green said:
Phil, your dedication to this community is truly amazing! Thanks so much for taking what I'm certain is a massive amount of your time to get this done. The information is priceless!
Click to expand...

I like it here. And REDuser is an important community for people who make a living in the art, craft, and science behind cinematography. I hope this helps with images people are trying to create. I make my living using these cameras and I'm certain everybody else does too. Knowing exactly what these tools can and can't do is important. The good news is for a while now, these have been industry leading tools and if utilized properly it shows.
 
Hi Phil,

Thanks for taking the time.

A few Q'S:

1. Is your rec709 transform based on ipp2 or something different?

2. Is it fair to posit here that we are saying VV dragon has more DR than Helium ?

Or...

Is the DR just distributed differently?

Does size matter in the way we are quantifying ?

Example: 6K slice of helium vs 6K slice of Dragon VV - most would say DR not affected, but since we may be talking oversampling Magic at times...???

I'm not personally worried - we are in a time of digital cinema sensor gold with all these great options available to us. I'm just pleased as punch all around with my imaging tools these days.

Hope to join the "double 8K" club like you soon with the 8K VV sensor I've had a deposit in for since NAB 2015, and the 8K Helium inside my Epic W (which I'm absolutely loving after getting used to it and diving headfirst into a RedWideGamutRGB/Log3G10 +. Transform LUTs pool.
 
  • Thread starter
  • Moderator
  • #34
Dan Kanes said:
Hi Phil,

Thanks for taking the time.

A few Q'S:

1. Is your rec709 transform based on ipp2 or something different?

2. Is it fair to posit here that we are saying VV dragon has more DR than Helium ?

Or...

Is the DR just distributed differently?

Hope to join the "double 8K" , club like you soon with the 8K VV sensor I've had a deposit in for since NAB 2015, and the 8K Helium inside my Epic W (which I'm absolutely loving after getting used to it and diving headfirst into a RedWideGamutRGB/Log3G10 +. Transform LUTs pool.
Click to expand...

1. It is based on "one of the final" Rec.709 Transform LUTs, which to my knowledge you don't have as of yet. But, hey, coming soon to a camera near you :)

And if Graeme gives me permission I'll put it online.

2. Fair is an interesting word. These are two of the best sensors in the industry as a whole at the moment, and maybe even after certain other entities come to the party. I already know both of these are better than one thing on the horizon in terms of DR and color, so a big "whew" from an investment standpoint. Dragon can have more DR, but if you rate Helium appropriately with what I've written about, i.e. ISO 800 vs. 1600, you will find in practical use they are alarmingly similar. The best is which "digital stock" and format size you prefer in reality, and yes it kills me to say that. Take a close look at those Skin Tone and Color examples at the end of the test video. There's so much there and honestly you should watch them 100 times to fully see it.
 
Thanks for the test, Phil!

Those VV skin tones look amazing... Clean and consistent. Helium definitely need a bit of tuning though, I can see subtle green color gradation in the models skin.
 
Thanks for doing this, Phil!

Seems from this test that LLO and STH produce consistent colors when switching sensors, but STD colors look noticeably different when switching sensors.
 
You must log in or register to reply here.
Back
Top