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 :)

Slowmo footage very noisy all of a sudden :/

Will Keir said:
Karl,

You seem to have a bit of wisdom but this get's frustrating. If the native sensor is 320 ISO, why do the people at RED, top down, including the engineers continually state it's a 800 ISO native camera? Why does Jim tell us Epic is a 800 ISO camera?

I've underexposed the camera at 800 ISO, maybe in practice you are right, you get great results at 320, but this is because your over lighting your shots and Epic's can handle more light better than less. I have shots that look totally bleach white yet didn't clip.

If you are getting no noise in dark situations at 2k, 300 FPS you are certainly doing something right. I have had similar problems to Dominik at 800 ISO. 2k 300 FPS in daylight, and the dark part of the rocks danced with flies.

Perhaps the 800 ISO rated native sensor is 5k 24fps and with slow motion work closer to 320 ISO at 2k 300 fps?
Click to expand...

Noise is at the deep (read dark/ left side of you histogram) ... for normal conditions, RED tells us 800 iso is fine leaving us with some noise in lowlights but not to much. But Shooting 2k at such high iso at high speed is another thing it's then zoomed in rather heavy noise... So when shooting highspeed is better to push your histogram to the other end / right / bright side. Then you use less of the actual dynamic range of the sensor but you stay away from the noisy end...
 
Well said Björn.

Björn Benckert said:
Noise is at the deep (read dark/ left side of you histogram) ... for normal conditions, RED tells us 800 iso is fine leaving us with some noise in lowlights but not to much. But Shooting 2k at such high iso at high speed is another thing it's then zoomed in rather heavy noise... So when shooting highspeed is better to push your histogram to the other end / right / bright side. Then you use less of the actual dynamic range of the sensor but you stay away from the noisy end...
Click to expand...
 
Dom, been reading along your thread as i have also had a few instances with heavy noise in pure daylight...and like you running at 300fps with around 1000 shutter. What i have found works is the black cal like you have done and also bringing the iso down to 320 or similar.(rendering 2K out at SD helps too as 1080 from 2K doesn't really work well. 3K is the minimum i recommend for finishing at 1080).

To comment on a few very weird things being posted and not directly to you Dom. 800 is the recommended 'native' iso of the MX chip by RED that in turn gives you the best DR of your scene above and below and in total. If it was a video camera shooting a baked in image like prores or XD's or similar you would find RED would place 0db at what 800ISO is. Saying it has a 'native' iso of 320etc is like saying the Alexa, F3 or any other camera has a different 'native' iso as to what is stated and doesn't really chime true because it's not a simple statement of fact in black and white. Moving your ISO to 320 is something i often do, in fact whenever it's night or a scene that will be pushing the mid tones and bottom end from underexposure i ALWAYS bring the ISO down and expose appropriately. It's simply just feeding the camera as much light as possible...and exposing with your eyes and meters at 320 will do this. BUT, it leaves almost zero room in the highlights so it must be done carefully. Same is true in reverse and you will have more highlight range but risk noise. I treat the ISO with RED for what it is, just metadata. Think of it as a floating ISO between around 320 and even 1600 all the time keeping an eye on your RAW meters for highlight and noise problems. But as a general working range i plant it in 800 as it's where i get the most performance from the sensor above and below. It really is about testing this all yourself. This really is the best method and something if you get your head around will get you perfect results every time as the biggest mistake people make with this camera is treating it like a video camera...it's not and this is the first sign of trouble down the road.
And this mention of using ND's to control highlights etc is just simply incorrect. ND's just reduce the amount of light that can reach the sensor and just enable you to use a different stop on the lens to reach it's sweetspot or desired DOF. I think it was just a misunderstanding of what ND's actually do. Why expose, then stick on an ND for your highlights and then pull the image back up in post? Expose properly to start. Noise and highlights are 100% all about exposure and where it's at within the scene and how it is on the sensor at it's RAW level and the range between the two with how you do this.
Personally i don't want to use Twixtor or any other post tool to recreate something the camera can and should do. I think it's bad attitude to take into a shoot unless technically necessary...instead of fix it in post...i prefer fix it in camera:)
2K from RED's have always been pretty ordinary but definitely exposing at about 320 and keeping a close eye on your clipping is key for high speed...or get a phantom:))
 
Will Keir said:
Karl,

You seem to have a bit of wisdom but this get's frustrating. If the native sensor is 320 ISO, why do the people at RED, top down, including the engineers continually state it's a 800 ISO native camera? Why does Jim tell us Epic is a 800 ISO camera?

I've underexposed the camera at 800 ISO, maybe in practice you are right, you get great results at 320, but this is because your over lighting your shots and Epic's can handle more light better than less. I have shots that look totally bleach white yet didn't clip.

If you are getting no noise in dark situations at 2k, 300 FPS you are certainly doing something right. I have had similar problems to Dominik at 800 ISO. 2k 300 FPS in daylight, and the dark part of the rocks danced with flies.

Perhaps the 800 ISO rated native sensor is 5k 24fps and with slow motion work closer to 320 ISO at 2k 300 fps?
Click to expand...


I know this is confusing. Some may call it counter-intuitive.

Here's the deal. A single CCD or CMOS sensor has a FINITE sensitivity to light. Period. We usually refer to this specific sensitivity as a sensor's ISO or ASA rating. Some cameras get around this by sampling each frame twice for added dynamic range. The ALEXA does this by basically sampling each frame at low (320) and high (800) ISO and then combining each into a single high dynamic range frame. The EPIC does the same thing in HDRx mode. But even when a sensor reads out twice it is still only working "at it's best" at a FINITE point.

The EPIC is 320 ISO. Period. End of discussion. How do we know this? Easy, use the "RAW Check" function. You only shoot RAW with the EPIC. And only "RAW Check" gives you an (semi) accurate reading of RAW exposure. Set the EPIC to 800 ISO and hit RAW Check. Notice the exposure change by at least a full stop? Now set the EPIC to 320 and RAW Check. Notice how exposure does not change?

EPIC + RED MX = 320 ISO
 
I'll give that a try on the weekend Brigham, blackshading beforehand and staying with 3K instead of 2K, if i really need the extra slowmotion I ll twixtor the timteline.

will post some results when they're done :)
 
I think there's some confusion between the terms "native" sensitivity and "recommended" sensitivity. Like Adrian says, native ISO/ASA/EI is a fixed characteristic especially for a RAW camera that doesn't do any signal processing to bake in an image prior to recording. However, you may get "better" more aesthetically pleasing results by underexposing a small amount to protect highlights. That's all you're really doing by setting the camera at 800 ISO.

This is where DPs used to shooting on film have an advantage, as this is essentially what they've used to doing, re-rating film stocks to create and not using the box speed.
 
Thanks, I wasn't sure about that one, trying to figure it out myself. You got the idea of what I was saying, but found fault in the logic? So the deep blacks will just get nosier? They won't remain the same deep black and ND's take the highlights down?

Brigham Edgar said:
And this mention of using ND's to control highlights etc is just simply incorrect. ND's just reduce the amount of light that can reach the sensor and just enable you to use a different stop on the lens to reach it's sweetspot or desired DOF. I think it was just a misunderstanding of what ND's actually do.
Click to expand...
 
Yeah Dominik, looking forward to those tests. Hopefully you can do a few tests. 1) Seeking quality results 2) Trying to reintroduced the problems that caused the noise in the first place.

Thus, solving the enigma. :)

Dominik Muench said:
I'll give that a try on the weekend Brigham, blackshading beforehand and staying with 3K instead of 2K, if i really need the extra slowmotion I ll twixtor the timteline.

will post some results when they're done :)
Click to expand...
 
In regards to noise in slow-mo, from our early experience...it does seem that when you stress the sensor out of it's natural comfort zone (ie 24fps..up to maybe 96ps)...you are more likely to get noise. Anything out of this range, whether it's timelapse at slower rates then 24fps, or slowmo at 300fps (AND at 2K)...increases the odds of strain...and noise...and must therefore be treated more delicately. IE, black shade beforehand, and (very likely) underexpose to protect your highlights and pump more range into your blacks. But again, these are just early mental notes we've been making, and are not science...Just observations we've been debating over the course of our projects.

For whatever it's worth...
 
Dominik,
I don't want to hijack your thread but here is some high frame rate sample test I ran this summer in day light from 2K to 1K.
I didn't perform a black shading before, just ETTR at 800 iso.
On the 1K samples I use 5.1 compression wich mean their is even more headroom to reduce the noise level.
http://vimeo.com/45830834

Best
B.
 
Will Keir said:
Karl,

You seem to have a bit of wisdom but this get's frustrating. If the native sensor is 320 ISO, why do the people at RED, top down, including the engineers continually state it's a 800 ISO native camera? Why does Jim tell us Epic is a 800 ISO camera?

I've underexposed the camera at 800 ISO, maybe in practice you are right, you get great results at 320, but this is because your over lighting your shots and Epic's can handle more light better than less. I have shots that look totally bleach white yet didn't clip.

If you are getting no noise in dark situations at 2k, 300 FPS you are certainly doing something right. I have had similar problems to Dominik at 800 ISO. 2k 300 FPS in daylight, and the dark part of the rocks danced with flies.

Perhaps the 800 ISO rated native sensor is 5k 24fps and with slow motion work closer to 320 ISO at 2k 300 fps?
Click to expand...

Lol. There's multiple amount of repeat information around now, and I can see why it's confusing. I'm not a cinematographer like some, just an indie shooter who performs his own tests and makes his own conclusions so I'm just one voice among many. You can rate whatever you like - all Im saying is IF you are getting compression artifacts in the shadows at ISO800 then you need more signal, It makes sense to rate it at 320 and lose some highlight detail but protect the shadows from compression/noise. It's all about balancing it, and that can change shot to shot. In daylight you might need the protection that ISO800 provides, but on a nighttime shoot you could rate at 320 and you still may not get near to clipping.

There are times I will rate Epic at 800 (mainly outdoors/daylight) but I will also rate at 320 at night to limit the noise (especially if shooting tungsten) and as per the thread I linked to... Ive had awful results shooting high fps at 800. 320 isnt a guaranteed solution, but it can help. Anything where compression is an issue (like high fps) it's good to ETTR as much as you can. Remember ISO simply offers you a viewing curve to adjust exposure so you may in fact end up exposing the same regradless of which ISO you choose. Also to confuse matters more your viewing curve (redlogfilm, redgamma, or a user selected curve) will also affect your monitoring path and so your decisions on how to expose.

As Graeme said in his other post - they recommend 800 as a starting point. However as he states, Raw view is almost identical to ISO320, so clearly there is a curve/gain adjustment to get it to 800. That's normal, Im sure all sensors are boosted by the manufacturer to hit the optimal level for signal/noise, and then they declare that as the ideal ISO. All Im saying it that the recorded histogram/signal is the one you see at 320. You can test this yourself. You can set the camera to ISO800, expose and then for the hell of it change your ISO to 2000, or 250 and hit record - it wont change the recorded signal. That's fixed at around 320. Switch to RAw view on either playback or record and see what you're getting. The whole point of RAW is ISO isn't baked in.

Personally I think it should be - we'd get much better results at ISO2000 if we recorded that specific histogram instead of one a few stops under - if that makes sense.
 
Brice Ansel said:
Dominik,
I don't want to hijack your thread but here is some high frame rate sample test I ran this summer in day light from 2K to 1K.
I didn't perform a black shading before, just ETTR at 800 iso.
On the 1K samples I use 5.1 compression wich mean their is even more headroom to reduce the noise level.
http://vimeo.com/45830834

Best
B.
Click to expand...


That was really cool footage.. probably the best highspeed stuff I seen from the epic. So where did you set your iso, just curius?
 
Brice, thanks for posting that short film. It really looks great, and (almost more important) very clean. It´s nice to see what´s possible, and what to strive for.
Compared to my own 2K slo-mo, this has a lot less "dancing flies" (or whatever it´s called). I think the subject here lends itself better to the EPIC 2K slowmo than what I usually shoot. Throwing the background completly out of focus helps eliminate the "dancing flies".
I often shoot action sport with wider lenses and I struggle especially when it´s trees in the background. I´m also curious about your settings.
 
I've never seen a bee poop/pee before. Awesome.

If you are currious about what glass was used.

"The camera is a Red Epic with Leica glasses.
Especially, 180mm Apo Telyt, 60mm macro Elmarit, 50mm summicron, and 19mm Elmarit."


Brice Ansel said:
Dominik,
I don't want to hijack your thread but here is some high frame rate sample test I ran this summer in day light from 2K to 1K.
I didn't perform a black shading before, just ETTR at 800 iso.
On the 1K samples I use 5.1 compression wich mean their is even more headroom to reduce the noise level.
http://vimeo.com/45830834

Best
B.
Click to expand...
 
Last edited:
Dominik,
I actually noticed than 2K HD was more tricky than 2K WS due to the available compression rate.


To answer about the settings,
all was shot at 180 shutter at 800 iso using the best compression ratio I could get (except for the 1K format).
My procedure was the following.
First I did expose (at 800 iso my midgrey) with the Epic internal tools (zebra and old color)
Then used the magnify function pointed on the darker area of the frame and switch in raw view mode.
By using the magnify mode the compression rate goes a bit more shitty you can easly spot on the noise dancing.
Then I adjust the noise level by opening more the iris until the noise level get more pleasing.
Back to 800 iso view the frame apear of course a bit more flat and sorta over exposed (without cliping)
Et voilà.

Fact is, in Dominik samples, the frames are much more challenging for the sensor than the insects shots (not that much details and no back ground).
Dominik's clip=Wider view with lots of details (water in movement +high light reflection in the back ground+black rocks at the front and the talent in the middle of that).
Maybe the noise issue Dominik is noticing is more an artifacts problem due to the high complex DR.
Especially under the Ausies sun, where you can get sun burn in 20 mn. ;-)
You can't go wrong by shooting more flat, maybe in Dominik case using a lowcon filter combine with a graduated filter (for the dark area) woul have help a bit.

Just my two cents.

By the way thank you guy's your words.
Best
B.
 
Nice footage Brice,

You hit the nail on the head with your methodology. This info is kinda spread over 2 threads now but DOF is definitely a factor for the compression engine. I've shots lots of 2K with shallow DOF and the compression has held up fine. Shoot wide angle with a complex background, high DR and you'll probably be in trouble.


Will, note that even though Brice did not rate the camera at 320 he did basically the same thing by overexposing more than he would have normally at 800 (protecting the noise floor from compression blockies.).
 
Karl H said:
Nice footage Brice,
Will, note that even though Brice did not rate the camera at 320 he did basically the same thing by overexposing more than he would have normally at 800 (protecting the noise floor from compression blockies.).
Click to expand...
Karl, it's not exactly the same as rating at 320, even if a bit over expose, the fact the camera is rate at 800, the high light still have more head room than at 320.
Like Mr Natress explain many times on this board, 800 iso is a good starting point to begin with.
At the end their is no definite rules, every single shot has it's problematic.
It's always a compromise between protecting your noise floor and your high lights, and find the best compromise for what you want to acheive.
But I'm totally with you about dof, wide angle and high DR.

Ok, I'm off....
Dominik let us know if you solve your noise issue under the Ausie's sun.
Best
B.
 
I don't know if this theory makes any sense at all, but i was thinking about this for a time, please correct me if I'm wrong: Mx sensor is rated native 800 ISO in Epic, when using the whole S35mm area.. in order to achieve more FPS we sacrifice resolution, reducing at the same time the sensor size to do so.. Soooo, isn't a "smaller" sensor less sensitive and prone to grain at high ISO with low lights??.. don't really know the maths for this, but i'm really sure we loose between 1 and 2 stops of DR when going for the Slow Mo stuff..
 
You must log in or register to reply here.
Back
Top