F5: FS Noise et al.

M

morgan_moore

Major Contributor
I am talking FS7 here.

Various bloggites, members etc are pushing for NR in EI mode.

Of course if it is switchable off I don't care if it is added as long as the image is kept 'pure'

But seriously I think if you are getting noise from this camera today you are..

-exposing incorrectly
-shooting in a coal hole

I would encourage you all to do your own tests..

1) set a 'normal' shot - maybe a person holding a grey card lit with a single kino or CTB corrected readhead
2) light it to expose as you think for is correct for 5.6 (aka vary the distance of the instrument from the subect)
3) shoot the scene at 11,8,5.6, 4, 2.8
4) take all shots and equalise them in your grading suite aka if there is a grey card in the shot vary the 'development' until each card sits at the same level.
6) evaluate noise

I bet the 2.8 or 4 shot is super clean.

Firstly - no one wants to hire a camera op with a 'noisy needs a load of post work' camera. If you have bought the camera sustaining the noise myth shoots yourself in the foot.

So the F4 shot is clean? What does this mean? well a load of stuff about LUT implementation (Id love luts on Custom) and various other things that may impact your post workflow like using an 'ettr or T-1' LUT

Should you use CBrowse/EI then all that is handled.

IMO there are better things to pressure Sony for.

If a DP uses phrases like 'I have never been a paint by numbers shooter," etc then the camera is probably some missing some LUT management (in pre and post) to force them to shoot clean - currently the camera has to be painted by numbers - do I like this? - no - but scaring everyone with tale of noise wont change that - campainging for good LUT management and EI choice sustain in non sony software will.

I consider the 'thick negs' from these two tests to be free of noise, with no noise reduction in camera or post.
My equalised images. only noisy on the 'thin' exposures.
https://dl.dropboxusercontent.com/u/33167160/Mugshot-jpg12.jpg
https://dl.dropboxusercontent.com/u/33167160/ei.jpg

 
Last edited:
Hi Sam,

I've done some further testing today, SLOG3, EI 800 with a well exposed image. I found the images too noisy for my personal liking. I liked what your were doing with Custom Mode and using Slog as a gamma.

Unfortunately most have my day has been taken up with the 2nd camera Sony sent me out to compare with my own for the magenta bug. I did firmware updates on both cameras. It was also too windy here in Birmingham to do much more testing.
 
Until I see a bracket test various Tstops equalised in post I will not accept your phrase 'well exposed'

Can you see that my three equalised images are similar in tonality and DR but wildly varying in noise?

S
 
To me noise on the FS7 is a non-issue. I'm more concerned getting a good look out of the camera and shoot high frame rate without to many artefacts. Noise on the FS7 is less than the F3 so I'm good.
 
morgan_moore said:
Until I see a bracket test various Tstops equalised in post I will not accept your phrase 'well exposed'

Can you see that my three equalised images are similar in tonality and DR but wildly varying in noise?

S
Click to expand...


Slave Driver ;)

More controlled tests in the next few day.
 
I agree noise is largely a non-issue, but having shot in a few coal holes lately I'd like the option to switch it on.

It's funny, when the C300 was released the blogashpere was full of people saying how natural it looked - what total rubbish that was...
 
Sam, the noise is not a myth. It will depend how you expose YES, but it will also depend on the overall dynamic range of the scene as well as how much of the scene has dark areas, and from those dark areas how much gradation is present.

Your described test does not test the noise of the camera accurately.

If fact what you've described does the opposite, it illustrates how to circumvent it. Basically you've overexposed by 1-2 stops then restore the levels in post which yields a +6db to +12db improvement in signal-to-noise. So you take the default 57db signal-to-noise ratio of the camera and turn it into 63db to 69db. It's why many DP's rate cameras lower to begin with, so that noise is effectively reduced and especially with scenes where they would want to preserve detail in the shadows.

A better test is to go shoot a 18 step grey scale chart and look at the first 4 darkest tones. You will see the noise for what it is at the camera's native signal-to-noise ratio throughout the entire dynamic range of the camera.

So for "bloggers" or "professionals" to say that exposure is wrong can be true in some cases indeed - but go ask them to shoot a high dynamic range scene and not get any noise present in the darker areas of the frame. If you crush it in post which is one solution, then you lose the detail.

In the end it's not so simple to summarize with blanket statements.
 
Last edited:
starcentral said:
Sam, the noise is not a myth. It will depend how you expose YES, but it will also depend on the overall dynamic range of the scene as well as how much of the scene has dark areas, and from those dark areas how much gradation is present.

Your described test does not test the noise of the camera accurately.

If fact what you've described does the opposite, it illustrates how to circumvent it. Basically you've overexposed by 1-2 stops then restore the levels in post which yields a +6db to +12db improvement in signal-to-noise. So you take the default 57db signal-to-noise ratio of the camera and turn it into 63db to 69db. It's why many DP's rate cameras lower to begin with, so that noise is effectively reduced and especially with scenes where they would want to preserve detail in the shadows.

A better test is to go shoot a 18 step grey scale chart and look at the first 4 darkest tones. You will see the noise for what it is at the camera's native signal-to-noise ratio throughout the entire dynamic range of the camera.

So for "bloggers" or "professionals" to say that exposure is wrong can be true in some cases indeed - but go ask them to shoot a high dynamic range scene and not get any noise present in the darker areas of the frame. It you crush it in post which is one solution, then you lose the detail.

In the end it's not so simple to summarize with blanket statements.
Click to expand...

Den - everything you say is true, sort of.

The thing is Slog has kind of opened up the sensor for users to do as they will - in a 709 camera - the manufacturer just clamps out the noise floor in the profile. This 'free' approach to letting the user do as they will with the sensor puts more onus on the user to make intelligent choices - choices that the manufacturer has been making for us for years.

'Marketing hype' of over speeding spec sheets (Grugs thread) may push users towards some bad choices. Or users may just make bad choices with no help from marketeers. If sony called the camera 10stop, 1000ISO who would buy it - no one - but is a 10 stop 1000ISO camera good? Absolutely. A 10/1000 camera that if you shoot in 14stops at 4000ISO you get a usable but noisy file - even better!

What is a noisy camera? Well we all have our own bar -here is mine..

Interview with a kino and two dedo kicks - completly clean
Day EXT - completely clean
Day Window lit INT with no additional lights - mild noise
Domestic night INT no additional lights - visible but acceptable noise

That is my 'bar' and the FS7 crosses it.

My feeling is that users are not getting their cameras to cross the bar outlined above - purely through user error.

Another way of looking at it - the competition.

Red Epic MX (tested by me) - my tests of this rated it as 320 Iso and basically unusable without full cine lighting - completely different camera unusable for corporate/doco work - a fine cinema camera
BMC4k (I shot a feature) - truthfully represented by the manufacturer as 200-400. Again a nice cinema camera unusable for doc/corp but fine for cinema
Arri Alexa - Ive not tested
C500 - Ive not tested.
Dragon - not tested
My observations of third parties footage would suggest that the Alexa is equal and the Canon might have a stop.

What got me was the 'Professional DP' who basically said he wants to expose by 'feel' and then noted/moaned 'noise' on the camera. This DP if he must 'expose by feel' needs a DIT to sort out his LUTs so that the camera works like that.. which it can with the right monitoring set up. The thing with a Red is you put it on 320 and expose by feel you get a clean file, no special LUT from your DIT. The FS7-800LUT can be too light to expose by 'feel'

To use the FS7 without a LUT the image need to appear bright. Concerningly bright. Director throw his toys out of the pram bright. If you use HG800LUt and EI 3200 you will fall off your horse.

-If you rigidly use the WFM you will be fine.
-If you build your self a 1000/10 LUT and expose 'by feel' you will be fine.

You may have seen my 4000ISO image here.. http://www.dvxuser.com/V6/showthrea...ge-etc&p=1986504876&viewfull=1#post1986504876

Any camera you can head into the woods (I needed a torch to find the controls on the camera) with a small Led light and shoot a pretty corporate.. is not noisy.

Once again I feel the noise noise can scare producers and is poor business strategy for owner ops who want to do business with their investment.
 
Last edited:
Now one thing I have not tested is the FS7 verses the 'large pixel' 1080 cameras (FS100, FMW F3 I own both) or any 3chip 2/3 1080 cams.. there is some chance that that are genuinely cleaner at high ISOs because their 'light catching buckets' are 4X the size on a like for like sensor.

S
 
We're definitely not comparing apples to apples here.

You can not use the technique of rating the camera lower to improve signal-to-noise, and then say "see the camera is not noisy".

If you chose to overexpose by 1-2 stops and readjust in post for the purpose of improving the noise levels - it comes at the expense of losing some of your highlight latitude.

Really your argument is misleading and you should not be saying "expose properly" because that assumes using 32IRE for grey - what you should be saying is to work around any noise "overexpose and fix in post" and this is EXACTLY how and why the CineEI mode exists, because CineEI mode is the tool that helps you achieve that overexposure with ease.

Looking at your mug shot, the middle image where you exposed "properly" is not bad and quite acceptable. The ETTR image is the best one and cleanest because you overexposed and adjusted in post. You lost highlight latitude doing this which isn't always a problem, but its still very true.
 
Apples bananas whatever.

In one thread - you are saying - just match your meter and the camera, here you are saying that to do that is somehow cheating!

I am indeed saying 'this is a great 10stop/1000ISO camera' - sure the manual says it is a 14 stop 2000 ISO camera, but I didn't read the manual.

Im also saying I think it matches up to anything else you can buy - pretty much - because all the manuals lie (apart from BMC which happens to say 200ISO, 10 stop on the box!)

Simply I think my FS7 is a nice camera and your 55 surely nicer - lets enjoy that - which is possible with a bit of trial and error

Most important let us knock our clients socks off with pretty pictures.

S
 
One day someone on this board will say - yep I tried that and it works - thanks for your efforts - up to that point I might just not post.. just it is a nice social scene and a good method to avoid editing!
 
..or they might say something actually clever like - ' i tried it and due to knee compression at the top of the image I think you are wrong - have a look at my tests.."
 
You are quite confused at the light meter post. If you want to use a meter to find your middle grey value for a known IRE value on your waveform monitor a light meter can help you do that. Most people trash meter use and in reality you don't need a meter to tell you anything - use your waveform monitor or false color option and expose away.

But what you are talking about here is exposing "differently" than the camera's ISO. ie. over or underexposing and pushing/pulling in post.

In your initial post you blatantly say and accuse members that "you are exposing incorrectly" if you are getting noise. I have a problem with blanket statements like that which are in fact false and leaving certain facts out of the equation and will lead others into thinking that if they do expose properly and get noise they are doing something wrong. So I decided to challenge you on it.

If you really want someone to thank you for your post then I would conclude in your post that "exposed normally you will get a bit of noise but is Sony so its super clean and acceptable - but if you want a really clean image and super blacks than simply rate your camera lower so you overexpose everything by one stop but NOTE that it comes at a loss of your overall dynamic range and you will lose a stop off your highlights."

This isn't about FS7 vs F55 or anything, this talk about exposure is the same with ANY camera. Let's just be clear about exactly what is happening in these "tests".
 
Fair play - you are right about losing highlights in exchange - I probably make too many assumptions on what I consider to be obvious.

So by exposing a bit right..

1) you lose highlight headroom
2)you get cleaner shadows

Im just saying 'hey using it like this is pretty sweet - maybe not on a chart or whatever but for the shots Ive actually done of real scenes'

Frankly In daily semi intelligent use with normal lighting, reflectors and all the gubbins most of us have I think we can get a sweet picture. as sweet as a C300 or whatever.

The tone Ive taken from some posts is that the camera is flawed/broken/needs new firmware, need NR in post, needs NR in EI mode whatever.. I dont feel that to be the case.

The cameras rock - and coming from me that is saying something - I hate cameras!

Of course Id like better lut management, RGB waveforms, onboard cappacino, 2000FPS 8k, but the camera in my hands right now is not broken. It is the best solo op camera on the planet.

S
 
Last edited:
You must log in or register to reply here.
Back
Top