S/N Ratio

[Kemalettin Sert]

whats this S/N RATIO means?MX sensor 66db and Dragon 80db says in Technical Specs..

 

[Martin Stevens]

S/N means the Signal to Noise Ratio.

 

[PatrickFaith]

Noise is a tricky thing, you know it when you see it but it's harder to to define then you'd think, how i think about these numbers - the the dragon is about 15 db drop in noise over the epic. So if you have a shot that has a bunch of noise (points/grain in the the greys), and if you had a dragon sitting next to a epic, the dragon would have 30 times less grain points then an epic (i.e. for every 1 grain on the dragon you would have 30 grains on the epic). I'm perty sure it works out to a 30x grain reduction (but then again, what is grain, and it is situationally depending on what your shooting).

 

[Antony Newman]

@K,

For a more in depth analysis of the Signal to Noise of the MX sensor - Alan Roberts analysis from Nov 2012 can give a good insight:

http://tech.ebu.ch/docs/tech/Tech3335_s04.pdf

If the Dragon sensor averages 14dB better S/N than the MX - my guess is that you won't get the 'blues' when you you shoot a Dragon. ;-)

AJ

Edit: If the S/N of the Blue Channel in the MX sensor trailed the other channels by 5dB ... and the new Dragon S/N is now +14dB - you Will now get more saturated Blues even in darker images.

 

[Blair S. Paulsen]

80db S/N... yummy!

80db S/N... yummy!

There are many clever ways to process the voltages coming off the sensor to enhance picture quality. That said, better signal to noise ratio is, to me, the foundation of better digital images. Why?

1) Low frequency details maintain clean edges making it easier for the eye to perceive them due to learned pattern recognition abilities
2) Bayer pattern reconstruction algorithms can interpolate more accurately due to less spurious variations
3) It can dramatically improve distribution codec performance (varies based on codec topology)
4) Areas with lower exposure look more natural as they do not carry a non-linear noise characteristic

Side benefit: Jim has noted that the Dragon's lower noise also allows the use of higher compression ratios in camera with results comparable to lower compression ratios on MX.

In short, virtually every measurement of image quality one might apply benefits from better S/N ratio - including the quality of color rendition. While it may be theoretically possible to create a Bayer pattern CFA reconstruction algorithm that is immune to noise, in reality noise reduces color precision.

I would further note, that while expressing dynamic range or latitude in stops is a useful shorthand, it is essentially a derivative of S/N ratio that can, in some cases, be misleading.

Cheers - #19

 

[Timur Civan]

lets just say the cleanest camera i've ever used ever was the Sony F3 it was 64dB S/N ratio @ 800ISO... The MX behaves like this at around 320ISO when boosted to 800ISO its percieved S/N drops a bit as you are boosting the "gain".

The thing to rember is S/N is exponential. 3dB cleaner is a lot cleaner. 80dB s/n is INSANE. If its 80dB @ what is likely 800ISO as a "base", and 2000ISO for a balanced middle grey, it means you will likely get a cleaner image @2000 iso than the current MX gets at 800. So dropping down to ISO800 on dragon, will be absolutely completely utterly spotless.

 

[BRANDON JAMESON]

lets just say the cleanest camera i've ever used ever was the Sony F3 it was 64dB S/N ratio @ 800ISO... The MX behaves like this at around 320ISO when boosted to 800ISO its percieved S/N drops a bit as you are boosting the "gain".

The thing to rember is S/N is exponential. 3dB cleaner is a lot cleaner. 80dB s/n is INSANE. If its 80dB @ what is likely 800ISO as a "base", and 2000ISO for a balanced middle grey, it means you will likely get a cleaner image @2000 iso than the current MX gets at 800. So dropping down to ISO800 on dragon, will be absolutely completely utterly spotless.

Our F65 is vritually noise-free.

 

[Matt Ryan]

Our F65 is vritually noise-free.

same with the F5/F55

 

[Timur Civan]

Yea. Same tech base as the F3.

It will be nice oo have the epic up to date.

 

[Blair S. Paulsen]

Agree that the F55 is virtually noise free, but I would add that the images can look a bit "air brushed" as if the noise reduction is also "smoothing" things a bit. Don't get me wrong, it makes excellent images with lots of DR and good color precision, but the noise killing does a minor amount of collateral damage to subtle textures. A worthy trade off in most cases, but not all.

Cheers - #19

 

[Matt Ryan]

Agree that the F55 is virtually noise free, but I would add that the images can look a bit "air brushed" as if the noise reduction is also "smoothing" things a bit. Don't get me wrong, it makes excellent images with lots of DR and good color precision, but the noise killing does a minor amount of collateral damage to subtle textures. A worthy trade off in most cases, but not all.

Cheers - #19

But at least results are uniformed from shooter to shooter vs. what we have now where everyone processes their own RAW images and some just flat out suck. That's one of the reasons RED gets a bad name is because a lot of people process the footage like crap. With Sony that starting place is already nice. I'm hoping Dragon alleviates that issue a bit with the new color science and possible image processing updates in RCX. It would be awesome if RED had a pre set "processing" applied to the RAW files in the SDK which have a tiny bit of NR as well as some unsharp mask to get those great images out of the box. Of course being RAW you could always unselect this "processing" if you'd like to apply your own and go from straight RAW. But I honestly think RED should make that "processing" similar to Adobe Camera RAW does to DSLR images, then there should be less hating coming from others and better direct images for those who have a hard time now.

 

[Mike P.]

There was a thread late last night called 'always learning' or something in recon. Jim basically said that the MX at 800ISO is cleaner than the D800 (current top-dog) at 800ISO **without Nikon's brilliant internal processing applied**. Natural conclusion was that RED/Jim was looking into some kind of RED-centric noise reduction/image processing for RCX... or at least, that is the direction the thread was going before it was deleted.

The thing is, if you had RED's/Graeme's own RED-specific noise reduction to fiddle with in RCX, you could potentially run MX at, say, 2000ISO and still have a useable image. Meaning you could protect your highlights even further on MX. It's all pie-in-the-sky stuff... I mean, I was still hoping Scarlet-X would get 4.5k WS.

 

[Blair S. Paulsen]

But at least results are uniformed from shooter to shooter vs. what we have now where everyone processes their own RAW images and some just flat out suck. That's one of the reasons RED gets a bad name is because a lot of people process the footage like crap. (snip)

Don't take this the wrong way Matt, but if the "advantage" is that my shots are more likely to look like everybody else's (more uniform), I'm thinking that talent and a unique creative eye are getting shortchanged. Not sure that's the better ecosystem for me. Beyond that, is cinematography to be judged only by fidelity to scene?

FWIW, one of the joys of working in RED's ecosystem, is all the treatments one can apply to the acquisition data set in "developing" the RAW image. Whether in Scratch, DaVinci or the ever evolving RedCineX Pro; that trip through the "virtual darkroom" is a key part of the image making process to me.

IAC, if you care to, you can create presets in RCXP and load them into camera - great for multi-cam shows.

BTW Matt you are spot on that poor processing by folks that aren't up to speed has hurt the RED brand. It could be argued that the user base is both RED's greatest strength and its greatest weakness. I'm not looking to offend anyone, but for decades high end gear was virtually always handled by experienced pros who would have a clear and specific understanding of how to get the desired results.

Ultimately, I'm just ecstatic to have a camera system with a robust signal path that supports a huge variety of image manipulation techniques, many of them non-destructive.

Cheers - #19

 

[Matt Ryan]

Don't take this the wrong way Matt, but if the "advantage" is that my shots are more likely to look like everybody else's (more uniform), I'm thinking that talent and a unique creative eye are getting shortchanged. Not sure that's the better ecosystem for me. Beyond that, is cinematography to be judged only by fidelity to scene?

FWIW, one of the joys of working in RED's ecosystem, is all the treatments one can apply to the acquisition data set in "developing" the RAW image. Whether in Scratch, DaVinci or the ever evolving RedCineX Pro; that trip through the "virtual darkroom" is a key part of the image making process to me.

IAC, if you care to, you can create presets in RCXP and load them into camera - great for multi-cam shows.

BTW Matt you are spot on that poor processing by folks that aren't up to speed has hurt the RED brand. It could be argued that the user base is both RED's greatest strength and its greatest weakness. I'm not looking to offend anyone, but for decades high end gear was virtually always handled by experienced pros who would have a clear and specific understanding of how to get the desired results.

Ultimately, I'm just ecstatic to have a camera system with a robust signal path that supports a huge variety of image manipulation techniques, many of them non-destructive.

Cheers - #19

Noise reduction and post sharpening "processing" won't make everyone's footage look the same. Does every still images look the same, because they are all starting with Adobe RAW processing or some other form which applies NR and sharpening for the specific camera? It's giving everyone a clean "base" start. I'm just saying it would be nice for those who complain and don't know how to handle RAW properly for a really easy "base". Obviously lighting, exposure, composition, movement, and final grade etc is what makes any footage stand apart from one another. I just feel like it's a cleaner and more solid starting point than what we have now. I'm not saying a color LUT is applied, just the noise reduction and sharpening. Color is up to the user, so that means your footage will look different from the next. I don't think that means everyone will have the same looking footage, but that's just me. Plus if it's an option you can just turn it off and have the same old full control.

 

[Robert Ruffo New]

But at least results are uniformed from shooter to shooter vs. what we have now where everyone processes their own RAW images and some just flat out suck. That's one of the reasons RED gets a bad name is because a lot of people process the footage like crap. With Sony that starting place is already nice. I'm hoping Dragon alleviates that issue a bit with the new color science and possible image processing updates in RCX. It would be awesome if RED had a pre set "processing" applied to the RAW files in the SDK which have a tiny bit of NR as well as some unsharp mask to get those great images out of the box. Of course being RAW you could always unselect this "processing" if you'd like to apply your own and go from straight RAW. But I honestly think RED should make that "processing" similar to Adobe Camera RAW does to DSLR images, then there should be less hating coming from others and better direct images for those who have a hard time now.

That's a dumbed down version. If other people are too dumb to learn how to grade Red files that's their problem. I don't want auto-anything.

 

[Matt Ryan]

That's a dumbed down version. If other people are too dumb to learn how to grade Red files that's their problem. I don't want auto-anything.

Then you can uncheck the box in Redcine. Don't understand your issue. If you don't like it, don't use it....