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 :)

Dragon Vs Alexa Low Light Tests - Someone?

Martin Stevens said:
It looks like Dragon is better than Film, but not at low ISOs like 200 and 500 etc.
To match the high-light handling of film, ISO 3200 etc. might need to be used.
Click to expand...

The test above is showcasing Epic with Mysterium-X, not Dragon.
 
Mike P. said:
The Alexa images were taken from 12bit prores, not arri-raw (14bit, if memory serves). I wonder how much smaller the gap would be, if any, had they shot Arri Raw instead.
Click to expand...
i tested and also asked same question to ARRI Tech.here is the answer
Raw uncompressed is untouched and there are some choices for de-bayering, sharpening, downscaling and compression.
In ProRes these steps are done in camera and there are few to no options to change the result.
ProRes will look slightly softer, not sure if you can call this "grain" structure.
On the Mon-out the images will always look the same. This is always an uncompressed HD feed. The difference can be seen when you pick a small segment that you blow up 500% and compare them.
Click to expand...
 
It's not that I don't believe them… but I'd still like to see the results myself. In previous tests CML posted raw frame grabs in their native format (wish they did that here, even though, as Phil says, it'll be moot come Dragon comparison time).

To be honest, out of all the frames I've seen, Dragon's noise floor doesn't seem to be all that much cleaner than MXs. Take it with a grain of salt, but I think the 6k/smaller photosites are perceived as less noisy, even though it's pretty much the same. Doesn't really bother me, since the end result is the same (you can push the ISO higher before it looks too dirty), but that's what I've noticed… It's like the difference between Epic 5k and Scarlet 4k -- they are the exact same chip, but Scarlet has a tendency to look noisier because of the crop-factor/amount of pixels being lower.
 
Last edited:
Phil Holland said:
The test above is showcasing Epic with Mysterium-X, not Dragon.
Click to expand...

Yes, but Kemalettin asked about Dragon, and I did not refer to the test in my answer. :)
 
Mike P. said:
It's not that I don't believe them… but I'd still like to see the results myself. In previous tests CML posted raw frame grabs in their native format (wish they did that here, even though, as Phil says, it'll be moot come Dragon comparison time).

To be honest, out of all the frames I've seen, Dragon's noise floor doesn't seem to be all that much cleaner than MXs. Take it with a grain of salt, but I think the 6k/smaller photosites are perceived as less noisy, even though it's pretty much the same. Doesn't really bother me, since the end result is the same (you can push the ISO higher before it looks too dirty), but that's what I've noticed… It's like the difference between Epic 5k and Scarlet 4k -- they are the exact same chip, but Scarlet has a tendency to look noisier because of the crop-factor/amount of pixels being lower.
Click to expand...
i agree with you on Scarlet.Recently worked with my scarlet on commercial it was 1 hour stuff ready shoot leave type of shooting.finished in 50mins.Later i checked the files they had too much grain.three times more than usual.Dunno what went work i knew DP underexposed lot but seemed excessive to me.
Speaking of Dragon some shots looks same with MX.Some of them looks way better than anything i have ever seen when it comes to grain.
Like this one (dragon framegrab 12800 iso)
yIwk9kh.jpg

looks cleaner than 3200 ISO MX.


Also we have Alexa too.did test with raw vs prores looks same to me.Do your own test if have chance to do.
 
To be fair, grain looks far less apparent when it's not in motion. On the flip side, properly exposed MX footage still looks pretty clean at ISO12800 (which has been touched on in this thread by me, phil, and others.) It only gets noisy when you've hit the noise floor and are trying to pull image detail out of that (for example, a candle lit room that is clearly dark, shouldn't be pushed to show what's in the shadows… there's no light/image there to boost.)
 
Mike P. said:
To be fair, grain looks far less apparent when it's not in motion. On the flip side, properly exposed MX footage still looks pretty clean at ISO12800 (which has been touched on in this thread by me, phil, and others.) It only gets noisy when you've hit the noise floor and are trying to pull image detail out of that (for example, a candle lit room that is clearly dark, shouldn't be pushed to show what's in the shadows… there's no light/image there to boost.)
Click to expand...

I remember back in the early days of pushing the Genesis sitting in a screening room and trying to come up with a rating system for how "excited" the noise is and how we could logically rank that :)


Kemalettin Sert said:
Like this one (dragon framegrab 12800 iso)
looks cleaner than 3200 ISO MX.
Click to expand...

The reason this looks clean is due to the exposure rating. Nothing is hitting the noise floor. In Dragon's case if there's nothing at all in the noise floor you can do almost anything with the footage and maintain decent results because the data falls within the workable latitude. Which is massive on Dragon.
 
People really should keep in mind that post-processing a file to 12800 ISO is not at all
the same as metering and shooting a scene at 12800 ISO.
 
Phil Holland said:
The reason this looks clean is due to the exposure rating. Nothing is hitting the noise floor. In Dragon's case if there's nothing at all in the noise floor you can do almost anything with the footage and maintain decent results because the data falls within the workable latitude. Which is massive on Dragon.
Click to expand...

I don't know why, but this comment makes so much sense to me. As always Phil, you are a master at reducing things to it's essence.
 
The film vs epic , etc test is hard to do a true evaluation on, since they did overexpose the Epic.

You can see that the Epic image of the girl in front of the fake seascape has almost zero noise in the image. But the whites are blown out with no detail.
It's all about noise floor folks. If they would have cut the light by a stop or two to the Epic, there would be plenty of detail in the seascape, and only a little more noise in the blacks.
So the test is completely invalid, obviously shot by a digital camera hater or someone with little understanding of exposing anything. Sorry if I seem a bit brash, but this kind of bad PR really pisses me off to the Nth degree.
Looking at the film scan, the noise is horrid and overpowering, as expected, even from the low rez 1920 image ( they didn't say it's a crop from the 4k )
I have demonstrated in an old thread that the MX sensor clearly blows away even ISO 50 Kodak Vision III 35mm negatives. I invite anyone in Hollywood to do a shootout and prove me wrong.

edit: Phil said the same thing as me , but more politely .... ;)
 
Patrick Tresch said:
I think it goes down to physics. Larger photosites means better light capture, less noise. Red is challenging those sensors that have fewer pixels with the Dragon. But if Arri stays with less pixels policy they will always perform better (if they push their sensor technology) with less resolution/photosites.

What do you think?

Regarding the camera as a tool, I prefer the Epic.

Pat
Click to expand...

More pixels means noise reduction in post will be far more effective and leave less artifacts
 
Mark Toia said:
Red prefers you shoot at 800 because it's like giving you a stop up your sleeve for over exposure. In other words it saves you if you are a bit loose with your exposures. Saves clipping of highlights in other words.
I shoot exclusively at 320 and on the Dragon 250 . But mind you I do have one eye firmly planted on the histogram making sure I never clip. Sometimes I do because I pan into the sun or do something stupid like that.
Click to expand...

Actually I find highlights like backlit hair look better - they roll-off better - if you expose for ISO 800 than 320 - this improvement means more visually to me than absolutely absent noise, and most noise can be reduced with clever application of Neat Video. But, you know, it's all subjective.
 
My wish for 2014:

Let somehow people understand that the only thing that affects "exposure" and noise, is the number of photons hitting the sensor.

ISO (on RED cameras) does not alter the noise or "cleanness"/Fidelity of an image. You can rate it at "any" ISO and get identical results when you develop from "any" ISO.

You cannot change the amount of light you used in post, though…

There is no magic involved. Just physics...
 
You must log in or register to reply here.
Back
Top