My Scarlet Dragon arrived - noisy, but otherwise great!

[luigivaltulini]

To me the exchange here is all about employing strategies.

It all depends where you sit on the personal taste spectrum of noise. Red set out to replace film, not to make a video camera to broadcast the World Cup (sony is great for that). Red makes a cinema tool, (but if you want that Broadcast look just set it to 59.94 timebase.

So. If you want that grain that we all love from film you can start by rating and exposing from about 800-2000iso. It might be a creative look or a budget thing or whatever.

If you want sharp, clean noise free stuff, you rate lower250-640 and get yourself a lighting package to support that.

what is important to understand is that this is an extremely flexible system with a lot of creative room. But the camera does not make miracles. No camera really does.

Battistella

I fully agree with David.
It 'just a matter of choice ....
So every time you turn on with Alexa ProRes not good?! too much noise at 800iso ..!!!!!
I think we're exaggerating with these posts on noise ...
This image is not a comparison with Dragon.
But it is to say that everyone shoots to get a result, not on the basis of noise.
my thoughts ..

some frames Alexa 800 iso prores.. not use denoise

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]

 

[Matt Ryan]

Matt,

I doubt raft there are two grades of sensors. I mean, that would be a complete disaster as multi camera shoots could produce enormous problems.

Plus us that suggests there are two levels of sensor or two levels of customers, the ones that get noisy dragons and the ones that get clean ones. That implies some kind if conspiracy which I doubt is the case. I imagine RED's testing and diagnostics would want to ensure a consistency across all sensors and cameras.

I guess to to test your theory, you'd have to try to get a studio day booked and invite as many Scarlett and dragon owners in your area to bring their cameras in to do a controlled test. The parameters would be easy

1. Same firmware
2. Same temp
3. Same lens
4. Same scene

But I suspect it might be a bit complicated to set up.

Battistella

I'm not saying that it's a conspiracy. Sensors from different batches, running different firmware and blackshades, etc... That has room for variances in those scenarios (not major but still a tolerance level has a range). Not that some get the good ones and some not so good. What I'm saying is I can't put my finger on why one was cleaner than the other even though both have the new olpf. I was trying to flesh that out in text but it may have sounded like a conspiracy. It's not what I meant.

What I don't understand is why we can't have multiple user sensor calibrations so when I'm shooting in low light I can go back to the old olpf calibration and during daylight, new olpf calibration. All while maintaining the benefits of the new olpf and color. I think that would be a better solution and people can make sure they also have it calibrated for their shooting scenario. It's basically the same idea as having multiple black shade calibrations stored...

P.S. I amended my original post to try and remove some conspiracy theory confusion that I wasn't trying to make but could be taken as such.

 

[Phil Holland]

I've been watching, reading, and looking at the images in this thread for a moment. Thought I'd jump in and get my toes wet.

This problem exists folks! Red are working on it we assume. If it didn't exist why would Red be fixing it. They re fixing it for the big hollywood boys and the one man bands.

Remember Red is tackling the red "sparklies" at the low end. Underexposure is still going to lead to a noisy result. Be aware of what is actually being "fixed". From Jarred:

Yes it is pretty great but I dont want people to have unreal expectations... the color in the noise is substantially better... no more red sparklies and red tone. but there still is noise. And there is still the extra stop on the top over the old OLPF. We didn't do the easy thing and just add Noise Reduction. Graeme spent alot of time making the noise better.. making the color better in the noise and more neutral as you wander deeper and deeper into the gutter.

Thats one of the reasons we chose highlight protection over lower noise... Noise you can deal with if you are under-exposed. Some people like a bit of noise some people don't. Blown Highlights though... there is nothing subjective about that. Once its clipped, its clipped. No getting it back unless you hire some VFX guru to rebuild what is gone.

There certainly will be noise in underexposed areas. Myself and other have mentioned the real "key" here. Dragon is operating a lot like film. Higher the ISO rating to more visible texture you will get. If you are hoping for a device that shoots "the same image" at every ISO that device doesn't exist. Not yet at least. I just took a close look at the Sony A7s and had to explain to a dude why there's noise at the crazier ISO ratings. *facepalm

i think the main "issue" is (or may i say the main question) : IS there a difference in the cameras or how can it be that some footage is obviously "better" than from other cameras / user ...

I stand on fairly firm ground saying I've probably seen the most Dragons of anybody in this thread. I've also experienced a good level of different people shooting with it. I'm going to say I'm 100% certain that there's no "camera variation". These cameras leave Red's HQ with a hefty amount of QC and you must remember the sensors are calibrated and QC'd to specification both at the time of "baking" and after they have been QC'd in setup. If they don't pass, they get sent off to that great trash bin in the sky. They don't end up in production cameras.

That leads to your second point. Which is actually the right thing here. In this thread there have been good exposures posted and one shot that was approximately 3 stops under. Maybe 2.5 stops. Depending. Anyways.

Really everybody here needs to study what they are seeing and test themselves. I've written a fair amount on how to expose properly and get good predictable results. Read those notes: http://www.reduser.net/forum/showthread.php?92999-Red-Dragon-Info-and-Data-Sheets


I'm not writing that stuff because I enjoy making technical graphics. I'm hoping it can help lead to people capturing better images and have a better understanding on how to utilize the tool in their hands. It's a powerful tool once you get the hang of it. Just like any camera really.


Okay, now an entirely on topic side note. One thing I will say is I'm not a big fan of chroma noise, so in general I tend to use the denoise feature in Redcine-X Pro on all ISOs. That's more of a me thing and how I like my images to look. If you want a more "filmy" look don't do this. If you don't like colorful noise, squelch it out with that slider.

 

[David Battistella]

Fair enough Matt,

I get what you are saying. Sometimes we type and post quickly and things can be taken out of context.

I'll modify my posts as well.

Battistella

I'm not saying that it's a conspiracy. Sensors from different batches, running different firmware and blackshades, etc... That has room for variances in those scenarios (not major but still a tolerance level has a range). Not that some get the good ones and some not so good. What I'm saying is I can't put my finger on why one was cleaner than the other even though both have the new olpf. I was trying to flesh that out in text but it may have sounded like a conspiracy. It's not what I meant.

What I don't understand is why we can't have multiple user sensor calibrations so when I'm shooting in low light I can go back to the old olpf calibration and during daylight, new olpf calibration. All while maintaining the benefits of the new olpf and color. I think that would be a better solution and people can make sure they also have it calibrated for their shooting scenario. It's basically the same idea as having multiple black shade calibrations stored...

P.S. I amended my original post to try and remove some conspiracy theory confusion that I wasn't trying to make but could be taken as such.

 

[Ryan Hamblin]

The noise to me is not the concern as much as the subtly of difference between mx and Dragon. don't get me wrong I love RED... hell I bleed red... literally ( bad joke sorry ). For me and the shooting situations I am typically in ( controlled and ample lighting at hand ) that it has been very difficult to find situations in which the 10000.00 upgrade is a tuff pill to swallow when I can add another couple lights to the package in situations when I know that its gonna be an issue. It tends to look like that is less detail to pull back in the blacks which to me has always been the more of an issue for me. Which is why I almost always now shoot with an Ultracon filter unless I will be flaring the lens.

 

[David Battistella]

Looked at the R3D's

The original shot is (as phil stated) 2 to 2.5 stops under. How does this exacerbate the problem.

you need to push an underexposed image just to get it about "right", then maybe you have to bump it another stop to really make it work. So you just "turbo gained" an image by 3.5 stops to get it where you "want it", but now it feels noisy? Yes. Because you exposed a 4000ISO image at 800ISO.

The bookshelf.

That noise at 800iso shows up in a REDLOG film ungraded image. The way to deal with that one is easy.

Process this shot at 250iso with curves. Drop the noise, put back the blacks and you are golden.
You want to flatten out the middle grey? Process at 800 and you'll have some more colour noise in the blacks.

processed @250 ISO

Processed @800ISO

 

[David Battistella]

When you see and feel how much room you now have in the highlights you will never look back. That is just a fact. The thing just doesn't clip. You don't have to worry your ass off so much about clipping. The thing you mention is a bandaid but the Dragon highlight extras are the real deal.

I would never have said this until I actually tried it. It holds the highlights so much better and the colour is really good.

Battistella

The noise to me is not the concern as much as the subtly of difference between mx and Dragon. don't get me wrong I love RED... hell I bleed red... literally ( bad joke sorry ). For me and the shooting situations I am typically in ( controlled and ample lighting at hand ) that it has been very difficult to find situations in which the 10000.00 upgrade is a tuff pill to swallow when I can add another couple lights to the package in situations when I know that its gonna be an issue. It tends to look like that is less detail to pull back in the blacks which to me has always been the more of an issue for me. Which is why I almost always now shoot with an Ultracon filter unless I will be flaring the lens.

 

[David Battistella]

I fully agree with David.
It 'just a matter of choice ....
So every time you turn on with Alexa ProRes not good?! too much noise at 800iso ..!!!!!
I think we're exaggerating with these posts on noise ...
This image is not a comparison with Dragon.
But it is to say that everyone shoots to get a result, not on the basis of noise.
my thoughts ..

Those noisy ALEXA shots are so...."filmic" How many times did you hear that one? Thanks for posting these examples.

Battistella

 

[Álex Montoya]

I had no idea that the Alexa was so noisy @ISO800

 

[Ryan Hamblin]

Alexa is noisy at 800 for sure. I haven't shot a lot of Alexa, but I love how much it sees into the blacks. Would I ever choose it over RED, even my poor little MX scarlet? not a chance in Hell. I know Alexa has its place that place is just not with me. Its a great looking camera for sure though

 

[Jan Reiff]

Good Post David.

GIVE THE DRAGON FIRE !

 

[David Battistella]

This is why I talk about an exposure strategy.

There is capture and process. The bookshelf is properly exposed, but how you process will affect the texture. I'm convinced this whole conversation will be gone with the next release with the color noise gone.

We we have been looking at all of this under an extreme microscope and not from a "finished graded" perspective and that is also part of what makes people think it's noisy. Log images are the starting point for grading. Puts some looks or LUTS on many of these shots and the images snap right into a really nice place.

battistella

 

[Jan Reiff]

2 more stills ... not really in focus , sorry, raining and it was a fast thing

https://flic.kr/p/odph1h

https://flic.kr/p/nW66gM

both without any noise reduction, no ADD
and, every jpg compression increases the dirty look, ... this shots in the native file are much better and finer in grain

it is fact, dragon will never be a Sony 7S, seeing beyond darkness (without noise) . 7S : other concept.

so, if you want to see more (for example here in my frame more of the wood of the wall): you have to add light. very simple.

or you expose it right and buy THIS "noir" mood ...

in this case of the entry of our studio a slight NR would delete all noise for sure, ... the rest remains dark, because it is ...

i will do the same with the new firmware and i am sure i can push a bit more, not the fine grain is the problem, it´s the red noise that limits the possibilities

(with ADD btw this shots were close to clean)

i (and lot of others i know) did Alexa shootings, and the Alexa is "noisy", never get why there is the myth that it is SO clean ... not true.

some Sony´s can see more in the dark, or some Canons. Fact. But the overall image, ... i will prefer Dragon.

 

[Philip Lanyon]

I'm just disappointed that this is even a discussion. I feel like I HAVE to do this upgrade more than being really excited by it. Not just the noise and the olpf thing, but the upgrade booth at NAB 2013, the year and a half wait, the price drop on the epic just after I bought it. I've been a red supporter from the beginning. I certainly would have been one of the early purchasers of the Red One if I had any money back then, an HVX was all I could afford and even that was split three ways.

I am appreciative of what Red has done for me and this industry but lately its been a little hard to keep the faith.

Just some inspiring words from Red I think would help. Like old times. Rally the troops. Let us know (again) that the future looks bright for all of us who push to achieve a cut above and are willing to take risks to get there.

 

[Paul J Steinberg]

I stand on fairly firm ground saying I've probably seen the most Dragons of anybody in this thread. I've also experienced a good level of different people shooting with it. I'm going to say I'm 100% certain that there's no "camera variation". These cameras leave Red's HQ with a hefty amount of QC and you must remember the sensors are calibrated and QC'd to specification both at the time of "baking" and after they have been QC'd in setup. If they don't pass, they get sent off to that great trash bin in the sky. They don't end up in production cameras.

Well at least one made it past that airborne trash bin. When my Dragon arrived back the red noise was really bad. Red tech support was great. They confirmed the issue and RMA'd it right away, replacing the sensor again as the one that was in the camera didn't meet spec. The new(est) sensor I have is much better but not in a state that I'm comfortable using it. There are many shots that it would work but there are many that would show off the noise. I'm finding anything brown really problematic including black/brown skin and dark wood. I'm just uncomfortable bringing a camera on set only to be shooting a brown coffee cup on wood. Once the new firmware arrives, this should go away, just no news on that front for the past few weeks.

I'm only mentioning this because I sent my MX in for upgrade based on the glowing reviews on Reduser, no one had mentioned this problem that seems to have shown up with the new OLPF. If I had known about this problem I would have held off until it was fixed and save myself a ton in rentals. This does effect peoples livelihoods.

 

[Elan Gunos]

Why are the people who receive Dragons like to under expose every shot they post? And then complain about the noise. Is shooting without lights or light source a new trend I don't know about?

 

[Elan Gunos]

Can't talk about the future too much, everyone will ditch their Dragon orders and order the new camera they're developing instead.

I'm just disappointed that this is even a discussion. I feel like I HAVE to do this upgrade more than being really excited by it. Not just the noise and the olpf thing, but the upgrade booth at NAB 2013, the year and a half wait, the price drop on the epic just after I bought it. I've been a red supporter from the beginning. I certainly would have been one of the early purchasers of the Red One if I had any money back then, an HVX was all I could afford and even that was split three ways.

I am appreciative of what Red has done for me and this industry but lately its been a little hard to keep the faith.

Just some inspiring words from Red I think would help. Like old times. Rally the troops. Let us know (again) that the future looks bright for all of us who push to achieve a cut above and are willing to take risks to get there.

 

[Mark Britton]

Why are the people who receive Dragons like to under expose every shot they post? And then complain about the noise. Is shooting without lights or light source a new trend I don't know about?

Because if you expose via histogram on a Dragon not set on RAW display for histogram, you'll probably underexpose by 1-2 stops. If you have color bars kissing the right side of the histogram display, on, say ISO800 with DragonColor/RedGamma 4, you're at least one stop short of any of the stoplights coming on, and possibly as much as 2 stops before you'll see red in the overexposure goal post. Quick examples in the attachments below (ignore the apparent exposure of the image on the monitor, my smartphone was doing auto exposure. Note histogram, note f-stop on the attachments only.

• At f8.0 ISO800 DC/RG4, the histogram was just touching the right side of the display. Note how much the left goal post was lit up. Underexposed, I think.
• At f5.6 ISO800 DC/RG4, stoplights were still not being hit, no over exposure goal post. But now underexposure goalpost is emptied. Can I go farther?
• Yup, probably! At f3.8 ISO800 DC/RG4, the blue stoplight popped on (green came on earlier) and we're starting to see some red in the overexposure goalpost

When set on RAW, the right side of the histogram seems to better line up with where the stoplights and overexposure start coming on. But there is a rather large gap on the underexposure end of the display on RAW still. So, not perfect. Histogram could use some tweaking to better represent exposure, in my opinion.

 

[Elan Gunos]

Thanks for the explanation Mark. I guess it's time to bust out my old light meter.

Because if you expose via histogram on a Dragon not set on RAW display for histogram, you'll probably underexpose by 1-2 stops. If you have color bars kissing the right side of the histogram display, on, say ISO800 with DragonColor/RedGamma 4, you're at least one stop short of any of the stoplights coming on, and possibly as much as 2 stops before you'll see red in the overexposure goal post. Quick examples in the attachments below (ignore the apparent exposure of the image on the monitor, my smartphone was doing auto exposure. Note histogram, note f-stop on the attachments only.

• At f8.0 ISO800 DC/RG4, the histogram was just touching the right side of the display. Note how much the left goal post was lit up. Underexposed, I think.
• At f5.6 ISO800 DC/RG4, stoplights were still not being hit, no over exposure goal post. But now underexposure goalpost is emptied. Can I go farther?
• Yup, probably! At f3.8 ISO800 DC/RG4, the blue stoplight popped on (green came on earlier) and we're starting to see some red in the overexposure goalpost

When set on RAW, the right side of the histogram seems to better line up with where the stoplights and overexposure start coming on. But there is a rather large gap on the underexposure end of the display on RAW still. So, not perfect. Histogram could use some tweaking to better represent exposure, in my opinion.

 

[Robert Hart]

Way back when, in the late 80s, I shot some 16mm Fuji neg stock. I found it rendered the tropical ocean blues and sky nicely but the reds on the jetty piles were noisier than the other colours.

Maybe the Dragon pendulum has swung the other way. Has anyone tried biasing their lighting towards warm or in daylight or daylight mix, using a warm filter, say 81EF through to 85B, then grading the image back to balance to hopefully reduce some noise in the red channel as used to be similar best practice for the earlier sensors which were blue deficient and would noise in the blue channel. Just a thought.