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 :)

Comparing 3K and 2.5K IQ

Les Dittert

Les Dittert

Well-known member
Joined
Sep 29, 2011
Messages
1,543
Reaction score
0
Points
0
Location
Los Angeles
As someone who is always pursuing the maximum detail possible from my 3K Scarlet footage, I pre-ordered a 2.5 K camera today.
I am keen on seeing fine detail, noise, etc.

This is indeed an interesting time for filmmakers!

edit: I'm not planning to keep the 2.5k , only getting it for a reality check test.
 
Last edited:
What are the sensor-size and resolution comparisons between the two sensors? E.g. what is the frame/sensor size of MX at 2k (or 3k) in millimetres compared to the 2.5k camera sensor? I'd imagine that would cause differences in the image characteristics (pixel density and pixel size play a pretty fundamental role in noise/grain and fine detail departments, right)? Also, I think it might be CCD while the other is CMOS (BMD said it was CMOS, but in one of the interviews, the CEO said, "we aren't even using all of the CCD"... wasn't sure if it was a verbal typo or not.... by the way, I'm coining the phrase "verbal typo" as of right now.)

Off topic, but if you do end up selling your 2.5k test machine relatively quickly, I'll buy it for $2.5k (solely because I want to say I got 2.5k for $2.5k now that 3k for $3k is dead.)
 
6.5 micron vs 5.5 micron size.
With good amounts of light, this won't matter. It's going to be all about image processing and that's what I want optimal.

An idea for RED : Hearing the news of the 512 SSD's being fast, let the scarlet shoot uncompressed to it (new RC setting =ZERO ) , avoiding any wavelet smear.
Chews up SDD, but cleaner for those difficult shots. You saw it here first ;)
 
Les Dittert said:
Indeed. But it's 2.5 K at 30 better not look better than my 3K RED at 30.
Now do you see where I'm coming from ?
Click to expand...


No, because pixel resolution is not the sole determining characteristic of image characteristics. Far more important is the size of the sensor, the size of the pixels on that sensor, the optical system, the color science - in fact, everything other than the number of pixels.

I know for many on RedUser (not necessarily you) "resolution is everything." Except that it's not.
 
Les Dittert said:
6.5 micron vs 5.5 micron size.
With good amounts of light, this won't matter. It's going to be all about image processing and that's what I want optimal.
Click to expand...

I'd imagine Scarlet would have more processing power solely for offloading 4k, although the 2.5 does have ProRes/DNxHD compression built-in which wouldn't be processor-free.

Anyone know how CinemaDNG stacks up to R3DRAW (on paper at least)? Actually, is CinemaDNG compressed much if at all (the 1080p files are double that of even 5k r3d RAW)? I have some Ikonoskop footage somewhere (which is cDNG) and it looks a lot like 16mm filmscan, but that's (presumably) because of the sensor rather than the codec/format... it certainly lacks resolution though.

I get the feeling that it's going to be very difficult to objectively compare the two (MX at 2-3k and BMD at 2.5k). I mean look at all the hoops and stipulations that have to be made for the comparison -- MX restricted to 2-3k, at what compression ratio, running at 30fps (when really you'd only drop down to those resolutions for the higher frame rates, anyway)... seems kind of forced.

As for RC1:1; I didn't think it was the read/write speed of the REDmag that prevented it, but the actually throughput of the data (at 4k, anyway)... Again, I think CinemaDNG's compression scheme isn't as aggressive as RC (which itself isn't as aggressive as going to ProRes/DNxHD), and at 2.5k it's being dumped to over-the-counter/regular SSDs.
 
Last edited:
Mike P. said:
Anyone know how CinemaDNG stacks up to R3DRAW (on paper at least)? Actually, is CinemaDNG compressed much if at all (the 1080p files are double that of even 5k r3d RAW)?
Click to expand...

Well, as a format it's a lot more open than RC, for one thing.......

CinemaDNG only allows lossless compression, but is usually passed uncompressed, whether it's in single frame sequence form or wrapped into an MXF wrapper. An HD sized frame is approximately 3 - 3.5 MB, which comes out to about 85MB per second (at 24 fps) , 5GB per minute, or 300 GB per hour. That is about double the size of Redcode 42 at 4K, or Redcode 10:1 at 5K. Utilities are available to convert Cinema DNG files to, say, Cineform RAW, which would greatly reduce that file size, but Cinema DNG itself does not support lossy compression of any type at this point in time.
 
Apparently the BMD cinema DNG "RAW" footage has white balance baked in, according to their website.
That's not what I understand by RAW - raw data from the sensor?
 
FWIW, I'm shooting with the Epic at 2.5K already – actually, 2400x1200, since that's nice and clean within the Super16mm image circle.

Quality at the 1920x1080p deliverable is very high.
 
nigelsmith said:
Apparently the BMD cinema DNG "RAW" footage has white balance baked in, according to their website.
That's not what I understand by RAW - raw data from the sensor?
Click to expand...

RAW is defined as "minimally" processed, not unprocessed:
http://en.wikipedia.org/wiki/Raw_image_format

I hear mixed things regarding whether the Alexa bakes in white balance into ARRIRAW, but this is from ARRI's website:

D-21 ARRIRAW has the white balance baked in, since back in 2002 we felt that it would be closer to the way film negative behaves. The D-21 EI setting is not baked in. The 2880 x 2160 pixel images (4:3 aspect ratio) are transported as 12 bit linear data.

The ALEXA ARRIRAW file format has neither the white balance nor the EI setting baked in. However, both white balance and EI that are set in the ALEXA are stored in the ARRIRAW header as metadata. This metadata also includes other information like the green/magenta shift value (CC) that has been set in the camera. The 2880 x 1620 pixel images (16:9 aspect ratio) are transported as 12 bit log data, since this is a better way to transport ALEXA's wide dynamic range.
Click to expand...

Since there is an analog-to-digital conversion step in order to store the sensor's output, some processing can occur in this step without necessarily debayering to RGB.
 
My interest starts with resolution. My hope is we can still squeeze some more detail out of the r3d. Advanced motion debayer.
The 2.5k will not stress the lens sharpness as much because of it's slightly coarser pitch ( so less lp/mm are needed to resolve to the sensor limit ) but modern lens's are not bothered by such low resolutions much anyhow.
Obviously I'll test DR, noise, all that.
I never said resolution is everything. Resolution is important to me. If I want creamy/organic/insert-artsy-fartsy-word-here , it can happen in post.

The R3d's may cost some high frequency detail, hence my suggestion to bypass it with direct to SSD with the higher bandwidth 512's coming soon.
Brute force three thousand megabits to a SSD is not out of the question.
For the same reason I like having AMD around to 'encourage' Intel, this is all good for the cinematographer.


M Most said:
No, because pixel resolution is not the sole determining characteristic of image characteristics. Far more important is the size of the sensor, the size of the pixels on that sensor, the optical system, the color science - in fact, everything other than the number of pixels.

I know for many on RedUser (not necessarily you) "resolution is everything." Except that it's not.
Click to expand...
 
I was reading thru the Cinema DNG format specification yesterday and there were several interesting details.
DNG can be uncompressed or mathematically lossless Hufner JPEG compressed. (2.5/1)
The image file definition can be tagged for TIFF compatibility
DNG clip files can be stored as single frame image sequences in a separate file folder, various metadata elements are each stored in their own file folders.
DNG clip files can be stored as "video essence" within an MXF format wrapper.
Metadata for image luts can be stored within the DNG image files, but this is not recommended as it overrides post metadata changes. Adobe recommends storing metadata in separate XMF or sidecar files.
DNG supports any sensor resolution and both single value per pixel color filter array sensors and 3 value per pixel multi-sensor arrays.
Sample bit depths from 8 to 32 bpp are supported.
Base metadata file definitions for white point reference and exposure index are allowed but not required.
Most other metadata related to defining a finished color image are not allowed since this is a raw file format. Among things that are not defined in a DNG file are color space, any form of RGB or YUV image encoding, gamma curves or sub sampling.
 
The fact that you have access to all of that information is a good testament to the value of open formats, and why they are greatly preferred by both standards setting organizations and archivists.
 
David Mullen ASC said:
Since there is an analog-to-digital conversion step in order to store the sensor's output, some processing can occur in this step without necessarily debayering to RGB.
Click to expand...

Although it's rarely talked about, I believe that even Red does some mathematical "manipulation" of the sensor data prior to compression both to reduce the data payload and make the compression more efficient.
 
I'm excited to hear/read/see your tests. I'm not a fanboy. I'm impressed by all the new technology out there to let me do the things I want to do.
 
Les Dittert said:
As someone who is always pursuing the maximum detail possible from my 3K Scarlet footage, I pre-ordered a 2.5 K camera today.
I am keen on seeing fine detail, noise, etc.

This is indeed an interesting time for filmmakers!

edit: I'm not planning to keep the 2.5k , only getting it for a reality check test.
Click to expand...



Les Dittert said:
I like the higher frame rates of 3K.
Click to expand...

Les Dittert said:
Indeed. But it's 2.5 K at 30 better not look better than my 3K RED at 30.
Now do you see where I'm coming from ?
Click to expand...

Can't you just look at your 3k footage? Are you happy with it? Do you need confirmation that you are happy?

Not that there's anything wrong with ordering the 2.5k camera. Wouldn't mind having one myself.
 
You must log in or register to reply here.
Back
Top