Dynamic range on RED
- Thread starter casey warren
- Start date
glenn chan
Well-known member
Simon:
Yes, but Glenn, they are set up that way for a reason. It would be impossible to set those cameras up for maximum range and have a picture that could be edited or output to a television without needing post grading to avoid a very flat low saturation look.
The SI and (to some degree, the DVX100) have a higher than normal dynamic range but they boost the saturation so saturation looks about right. SI shots can be seen on their website. Contrast may look a little lacking though.
If you wanted to fix that, you could apply a LUT in the camera to do in-camera color grading (so to speak)... then you wouldn't have to grade in post. Although for digital cinema use, I wouldn't see why you'd necessarily need to do that.
With some other cameras, you can adjust the camera's LUT for good control over the camera's transfer function / LUT (i.e. Sony F900's CVP file editor), which determines dynamic range and can let you grade in-camera. Other cameras don't allow this.
2- You really can keep all of the dynamic range captured. A simple method is using contrast functions that have a s-shaped curve (if you plot its transfer function). There are some more advanced functions that keep hue/saturation more constant (i.e. luminosity composite mode in Photoshop).
Then there's tricks you can do that work spatially (i.e. horizontally, vertically)... shadow/highlight in PS, power windows / masking color corrections, etc.
Yes, but Glenn, they are set up that way for a reason. It would be impossible to set those cameras up for maximum range and have a picture that could be edited or output to a television without needing post grading to avoid a very flat low saturation look.
The SI and (to some degree, the DVX100) have a higher than normal dynamic range but they boost the saturation so saturation looks about right. SI shots can be seen on their website. Contrast may look a little lacking though.
If you wanted to fix that, you could apply a LUT in the camera to do in-camera color grading (so to speak)... then you wouldn't have to grade in post. Although for digital cinema use, I wouldn't see why you'd necessarily need to do that.
With some other cameras, you can adjust the camera's LUT for good control over the camera's transfer function / LUT (i.e. Sony F900's CVP file editor), which determines dynamic range and can let you grade in-camera. Other cameras don't allow this.
2- You really can keep all of the dynamic range captured. A simple method is using contrast functions that have a s-shaped curve (if you plot its transfer function). There are some more advanced functions that keep hue/saturation more constant (i.e. luminosity composite mode in Photoshop).
Then there's tricks you can do that work spatially (i.e. horizontally, vertically)... shadow/highlight in PS, power windows / masking color corrections, etc.
pretopost
Well-known member
Gibby said:
Are you seriously saying that most people don't grade there footage in some respect? I do a lot low end budget work but we manage to grade and correct 90% of our footage before it goes out to get completed. It might not be on a 'Q' but it still gets processed and moved...
Or is it because it is RAW?
ZaneIsNumber1
Well-known member
pretopost said:
RAW takes a bit more work to get looking good, compared with regular video that should come out looking pretty decent to begin with, assuming it was shot properly. Any grading is a bonus for video in most situations and generally consists of minor tweaks, but with RAW, grading would be a necessity, to avoid delivering a washed out looking image.
glenn chan
Well-known member
If you're shooting news, then you probably want to shoot close to the Rec. 709 transfer function but with video knee (like every other ENG camera out there). It might be nice if the RED camera had a dynamic knee like the Sony cameras do (maybe)... which will allow slightly better knee optimization for whatever scene it is. Even better would be a contrast function that keeps hue and saturation fairly constant (i.e. R:G:B ratio the same)... but that takes (a lot of) extra processing.
The RED camera should allow customizable 1-D LUTs to define whatever transfer function you want (this is what Graeme indicated on dvinfo I believe; a lot of cameras have a 1-D LUT anyways). A reasonable workaround to dynamic knee would be to implement something like the scene file on the DVX100... allowing the shooter to quickly switch between user-defined LUTs. This would be the simple uncomplicated solution (not like the constant hue+saturation knee, which would require something like a 3-D LUT or just extra processing).
2- If RED wanted to target ENG, it could have strengths in a data workflow and acquisition.
Buffering the recording allows the shooter to capture events they don't have time to hit the record button for.
Being able to edit directly off the media means zero ingest time. This would have to work in conjunction with the editing system.
Get the ergonomics/controls of the camera sensible, so it's easy to have focus + good audio + good exposure, etc.
For ENG, I'd aim for those three things before worrying about how good the knee / in-camera grading is. Anyways, I'm really digressing. The original point was... in-camera grading (or lack thereof) is not a big deal for ENG use. Implementing a scene file (with a wheel on the camera body to easily switch things) + user-definable LUTs will likely be good enough.
The RED camera should allow customizable 1-D LUTs to define whatever transfer function you want (this is what Graeme indicated on dvinfo I believe; a lot of cameras have a 1-D LUT anyways). A reasonable workaround to dynamic knee would be to implement something like the scene file on the DVX100... allowing the shooter to quickly switch between user-defined LUTs. This would be the simple uncomplicated solution (not like the constant hue+saturation knee, which would require something like a 3-D LUT or just extra processing).
2- If RED wanted to target ENG, it could have strengths in a data workflow and acquisition.
Buffering the recording allows the shooter to capture events they don't have time to hit the record button for.
Being able to edit directly off the media means zero ingest time. This would have to work in conjunction with the editing system.
Get the ergonomics/controls of the camera sensible, so it's easy to have focus + good audio + good exposure, etc.
For ENG, I'd aim for those three things before worrying about how good the knee / in-camera grading is. Anyways, I'm really digressing. The original point was... in-camera grading (or lack thereof) is not a big deal for ENG use. Implementing a scene file (with a wheel on the camera body to easily switch things) + user-definable LUTs will likely be good enough.
Last edited:
How about active cooling of sensor?
How about active cooling of sensor?
I think active cooling of the CMOS sensor could be implemented with a Pertier-effect chip (thermoelectric) to further reduce the noise especially for low-light situations and the heat could be transferred to the magnesium body without fans.
Because of the relatively high power consumption required, that could be optionally enabled if an additional power source (battery etc) is provided, but at least it could increase considerably the low light capabilities of the camera for special cases.
Just a thought.
How about active cooling of sensor?
I think active cooling of the CMOS sensor could be implemented with a Pertier-effect chip (thermoelectric) to further reduce the noise especially for low-light situations and the heat could be transferred to the magnesium body without fans.
Because of the relatively high power consumption required, that could be optionally enabled if an additional power source (battery etc) is provided, but at least it could increase considerably the low light capabilities of the camera for special cases.
Just a thought.
hiderocketdive
Member
ZaneIsNumber1 said:
Hmm...as someone who has never worked with RAW footages, I'd like to know what differences there are in grading RAW versus regular DV footages. Is it just working with the levels to keep it within the range of the final output format? Educate me please.... :dankk2:
hiderocketdive said:
well first of all you dont get the horrible blockiness from grading.. you can push colors WAY farther. Plus, you gain between 1-2 stops of latitude.
Graeme_Nattress
Red Team
I don't think it's as much as you gain a couple of stops DR, but that you don't throw away stops in the conversion to directly "viewable" image. Look at the Adromeda mod for the DVX - you get a lot more DR because you're recording RAW off the same sensor, because it doesn't throw away the DR in the process.
Graeme
Graeme
Barend Onneweer
Member
When you set your gain or exposure in a DV Camera, the raw data from the CCD is processed. Usually this includes adjusting gamma, clipping highlights etc. to create a decent looking picture. Some edge enhancement etc.
And the result is downsampled to 8-bit, and compressed.
So the sensor might be picking up a wide range of color values in 12-bit, but a lot of that detail (especially in the highlights) is lost when the signal is processed and clipped to fit into 8-bits.
If you have a DSLR try this: shoot the same scene twice. Once RAW and once as a jpg. Now open both into Photoshop CS. When you open the RAW file in Photoshop you can adjust the levels on the RAW file, and you'll find that you have a lot more detail to play around with than the jpg file. The jpg will look prettier at first - but the RAW file contains a lot more color data.
And the result is downsampled to 8-bit, and compressed.
So the sensor might be picking up a wide range of color values in 12-bit, but a lot of that detail (especially in the highlights) is lost when the signal is processed and clipped to fit into 8-bits.
If you have a DSLR try this: shoot the same scene twice. Once RAW and once as a jpg. Now open both into Photoshop CS. When you open the RAW file in Photoshop you can adjust the levels on the RAW file, and you'll find that you have a lot more detail to play around with than the jpg file. The jpg will look prettier at first - but the RAW file contains a lot more color data.
Graeme_Nattress
Red Team
Yes, the generally just crop off the top of the DR to make the rest of the image bright, thus loosing a couple of stops. That, and that they don't always have the greatest DR to start with! When you think about it, most video cameras are designed for a standard "news" like video image - bright and punchy - not too good for movie making.
Graeme
Graeme
casey warren
Active member
how much extra DR do you get when you use the Adromeda mod? Is it worth installing in a DVX or just spend the money on buying an HVX without the mod.
Whiterabbit
New member
The other day I was playing around with some high dynamic range images (HDRI) with my DSLR and it struck me that you could do the same thing with 2 video cameras side by side. I started investigating and found one approach using a single camera that alternates exposure from frame to frame to capture different parts of a scenes radiance map. Then they follow up with an offline process that motion-compensates and estimates the full radiance map at each frame called HDR stitching. From the examples, seems to work really well.