AVS Forum banner

1 - 18 of 18 Posts

·
Registered
Joined
·
5,750 Posts
Discussion Starter #1
The above "RAW"s process the information from the sensor - this includes noise reduction and sharpening (edge enhancement) and also, if applied, compression. The result is the clips have halos, smearing and compression artifacts, just like those codecs we are all used to using, and baked-in color. They are not true RAW clips.

We have all shot RAW stills, and know that one great feature was that you could ignore white balance while shooting. This could be set with no cost in post. This not true for these video "RAWs".

As far as I know, Cinema RAW or RAW cng is a true RAW, with no additional processing as above (no sharpening, noise reduction, compression). All of the information (and noise) the sensor provides is captured. The idea is that you the "author' can decide how much sharpening or noise reduction you want and you can color and shape the look (eg, contrast, saturation) too, taking advantage of 12 or 14 bits. And doing this processing with software can be much more effective (better) than baking this stuff in in the camera.

The advantages of these pseudo RAWs are that the files are smaller than true RAW cng and they are better than other codecs, even regular ProRes. But they are not RAW, and they have artifacts and less latitude than true RAW.

So, what's the problem in shooting real RAW (RAW cng)? just big files and massive bitrates - a challenge for storage only. You cannot shoot 4K 12bit true RAW on any sd card that is available (including II). You need something bigger and faster - CFAst or SSDs.

How big? Well, less than 1/2 hour of 4K 12bit true RAW takes up one terrabyte of space!

Recent cameras permit storing video via a usb-c port onto those small external and relatively cheap ssd's (Samsung T5, SanDisk Extreme). This means in principle the camera can be small and obviates the need for the added bulk of an external recorder.

The original BMPCC was a tiny camera that shot true RAW, and as it was only HD, you could use sd cards, even fast I's. The 4K and 6K successors also shot in true RAW. Now they shoot in BRAW, but you need the large CFast cards to do that. The cameras are big and bulky (and ugly). And, they removed the ability to shoot in true RAW when BRAW became available. What is going on?

As a consequence, the original BMPCC is one of the few affordable cameras left that shoot in true RAW. I used it as my travel camera six years ago, shooting RAW in China, Florida and NYC. It was difficult to get the colors right, and it was slow in post, but not impossible even on a laptop. But now, DaVinci Resolve has evolved to be much more efficient and with better color management (and I know much more), plus computers are faster.

So, I took out the BMPCC recently to shoot true RAW to try to understand what the problem is that has led to proprietary processed flavors of "RAW" requiring bulky cameras or external recorders that are replacing true RAW.

Here is a video I produced. Much easier to work with than HEVC (H265) files and heavily-compressed H264 files. But, of course, it is only 1080. Color management: tell DaVinci it is a dng RAW file (with its expanded color gamut and linear gamma) and that you want it transformed and viewed in REC709. The rest is just playing with the look (get exposure and WB right). And, of course, with the expanded color space of RAW and all of the dynamic range of the sensor available you can also produce HDR videos. Note: no need for log profiles in the camera - you can apply a log profile in post.


This video is nothing special ( I cannot go anywhere), but you can really see the colors are rich and the resolution is with less artifacts, even with YouTube compression - the video is easy on the eyes, and natural-looking (forget "cinematic").

OK, so why am I looking at this now? Well, there is another exception - the Sigma fp. It shoots true 4K 12bit RAW onto an external SSD (Samsung T5) and it is as small as the original BMPCC with much better usability and battery life. And it is full-frame! Tiniest full-frame camera on the planet. In comparison, the BMPCC is only s16 - the sensor is half of the size of that of m43 cameras like the GH5.

Sounds worth trying, no?
 

·
Registered
Joined
·
5,750 Posts
Discussion Starter #3
Why can't you go anywhere now? You were going everywhere during the height of the pandemic. :confused:
All within the NYC-CT "bubble," which now has the lowest infection rates. By anywhere, I mean beyond the bubble. We cannot go to Europe, to China or to Hong Kong. US residents are currently global pariahs.
 

·
Registered
Joined
·
33,386 Posts
All within the NYC-CT "bubble," which now has the lowest infection rates. By anywhere, I mean beyond the bubble. We cannot go to Europe, to China or to Hong Kong. US residents are currently global pariahs.
How ironic, the country that started it all now considers us pariahs. :rolleyes:
 

·
Registered
Joined
·
4,614 Posts
What is going on?
What's going on is very old news. Red's Jim Jannard sued Sony for patent violation on internal raw recording, for monetary damages as well as the recall and destruction of all F5/F55/F65 cameras. Well oops, it worked both ways because Sony counter-sued claiming Red violated a bunch of Sony patents. Red and Sony eventually agreed to drop their lawsuits against each other but for the others including Canon, BMD, Arri were left to develop their own non-patent infringing raw solutions that for most involved external raw recorders. BMD came up with their internal solution BRAW which involved a partial debayer in-camera. But everybody now uses raw compression ranging from 3:1 to 12:1. Only Red maintains a non-compressed raw option in addition to compressed redcode rd3. Only Sony and Red have 16 bit linear raw. The dynamic range of 12 bit linear raw is only 12:1, so BMD and Canon 12 bit raw are transformed after the sensor to log raw, post A/D to maintain the full DR of the sensor. Raw compression is DCT but not GOP. H.264/H.265 also use DCT but additionally 12-50+ frame GOP meaning that only 1 new frame is recorded from each 15 (h.265), the other 14 record only the changes from the previous frames. XAVC-I, AVC-I and prores can optionally record GOP or all-I but at correspondingly larger file sizes, which for RGB is less efficient than un-debayered raw.

Imaging sensors are inherently linear devices, voltage output is directly proportional to light intensity. An f-stop is a doubling of light intensity on the sensor. Each bit of a 12 bit A/D is also a doubling of the data value, 1-2-4-8-16-32-64-128-256-512-1024-2048, all bits high summed together equal 4095; along with 0 gives 4096 discrete values than can be represented in 12 bits, each of the 12 bits representing a doubling of the data value corresponding to the doubling of light intensity with each f-stop. "True raw" means recording the sensor A/D. Fake raw as you call it means a log transfer is applied to the sensor data after the A/D but before compression/recording to the codec, i.e. baked in. Thus if the Sigma FP is, as you say shooting true 12 bit raw, then it is sensor raw and not log raw. And if it's sensor raw, 12bits can only record 12 stops. (see math above.)

What does it all mean? I could have saved explaining it if I had just read Sigma's firmware 2.0 update states external raw support is Prores raw with the Atomos Ninja V or BRAW with the BMD Video Assist 12G, both of which according to you are fake raw anyway. The FP is either true raw but only 12 stops DR, or fake raw with a log transfer and plagued with halos, smearing and compression artifacts, according to you. I have to also add here, the folly of any judgment about the Sigma FP's raw as a takeaway from observing through the lens of YouTube compressed AVC1 codec.
 

·
Registered
Joined
·
5,750 Posts
Discussion Starter #6
What's going on is very old news. Red's Jim Jannard sued Sony for patent violation on internal raw recording, for monetary damages as well as the recall and destruction of all F5/F55/F65 cameras. Well oops, it worked both ways because Sony counter-sued claiming Red violated a bunch of Sony patents. Red and Sony eventually agreed to drop their lawsuits against each other but for the others including Canon, BMD, Arri were left to develop their own non-patent infringing raw solutions that for most involved external raw recorders. BMD came up with their internal solution BRAW which involved a partial debayer in-camera. But everybody now uses raw compression ranging from 3:1 to 12:1. Only Red maintains a non-compressed raw option in addition to compressed redcode rd3. Only Sony and Red have 16 bit linear raw. The dynamic range of 12 bit linear raw is only 12:1, so BMD and Canon 12 bit raw are transformed after the sensor to log raw, post A/D to maintain the full DR of the sensor. Raw compression is DCT but not GOP. H.264/H.265 also use DCT but additionally 12-50+ frame GOP meaning that only 1 new frame is recorded from each 15 (h.265), the other 14 record only the changes from the previous frames. XAVC-I, AVC-I and prores can optionally record GOP or all-I but at correspondingly larger file sizes, which for RGB is less efficient than un-debayered raw.

Imaging sensors are inherently linear devices, voltage output is directly proportional to light intensity. An f-stop is a doubling of light intensity on the sensor. Each bit of a 12 bit A/D is also a doubling of the data value, 1-2-4-8-16-32-64-128-256-512-1024-2048, all bits high summed together equal 4095; along with 0 gives 4096 discrete values than can be represented in 12 bits, each of the 12 bits representing a doubling of the data value corresponding to the doubling of light intensity with each f-stop. "True raw" means recording the sensor A/D. Fake raw as you call it means a log transfer is applied to the sensor data after the A/D but before compression/recording to the codec, i.e. baked in. Thus if the Sigma FP is, as you say shooting true 12 bit raw, then it is sensor raw and not log raw. And if it's sensor raw, 12bits can only record 12 stops. (see math above.)

What does it all mean? I could have saved explaining it if I had just read Sigma's firmware 2.0 update states external raw support is Prores raw with the Atomos Ninja V or BRAW with the BMD Video Assist 12G, both of which according to you are fake raw anyway. The FP is either true raw but only 12 stops DR, or fake raw with a log transfer and plagued with halos, smearing and compression artifacts, according to you. I have to also add here, the folly of any judgment about the Sigma FP's raw as a takeaway from observing through the lens of YouTube compressed AVC1 codec.
Thanks Tom, good info. The Sigma fp is true 12bit dng RAW. And they claim "only" 12.5 stops of dr, sort of consistent with this. There is no compression at all, so the bitrates are huge. However, a program slimraw losslessly batch compresses dng RAW on the computer, basically doing a better job than what RED does in camera that no one else can do in camera for free. I got a reducion, lossless, to 38% of the originals.

In Resolve, if in managed color you choose linear for the input gamma (and REC2020 for the input gamut) and select REC709 for output and timeline the Sigma fp "clips" look fine, further indicating that is is "pure" RAW.

Here is a quick example using those settings:

 

·
Registered
Joined
·
1,218 Posts
The issue with Red does not involve RAW itself, rather compressed RAW. If you record RAW directly without compression, or with compression ratios below a certain number, it is OK from an IP point of view. I think it only affects 4K as well, HD is ok.

All these other so called "RAW" formats are attempts to work around the Red patents basically.

The halo effect is not due to computational sharpening btw, it is an artifact of the debeyering process. The extent to which it happens depends on whether you weight debeyering towards resolution or color fidelity at edges. With a Beyer filter you get one or the other, but not both.
 

·
Registered
Joined
·
5,750 Posts
Discussion Starter #8
The issue with Red does not involve RAW itself, rather compressed RAW. If you record RAW directly without compression, or with compression ratios below a certain number, it is OK from an IP point of view. I think it only affects 4K as well, HD is ok.

All these other so called "RAW" formats are attempts to work around the Red patents basically.

The halo effect is not due to computational sharpening btw, it is an artifact of the debeyering process. The extent to which it happens depends on whether you weight debeyering towards resolution or color fidelity at edges. With a Beyer filter you get one or the other, but not both.
Yes, no lossless compressing of true RAW from the sensor in camera in 4K without a payment. Other manufacturers chose partial debayering to get around that. But the Sigma fp does RAW without any compression.

However, just as debayering or noise reduction in the computer can be more effective than in the camera, you can compress losslessly true RAW on the computer even more effectively The program slimraw does that, and reduces the RAW clips to from 37-46% of the original sizes, again, with no IQ penalty. Essentially you just use slimraw to copy/compress the files from the ssd to your storage disk (one step) and then wipe your camera ssd. It is very quick and effective.

In any case, the Sigma fp provides true RAW without a bulky external recorder in a compact form, and the RAW is pristine. This means that in the future if there are advances in debayering, one can apply those advances to your fp RAW files. They are archival.
 

·
Registered
Joined
·
4,614 Posts
CinemaDNG is an open source Adobe format, and not just for raw. In the use case of the Sigma fp, raw video is recorded in a 8-10-12 bit DNG format from a 14 bit image sensor. That transformation is a rescaling process but it is not the only one. Rescaling from 6000 x 3376 readout for 16:9 UHD is another. In fact, it is the first instance I have ever heard of any sensor raw being anything but a 1:1 sensor readout. How can you debayer a CFA that's been resized with pixels that no longer exist and call that raw? But reverse mortgage is just a loan, and CinemaDNG is just a format, a container for anything that fits under ISO 12234-2. Sigma lists the color space as sRGB or or Adobe RGB. Those are applied with a gamma transfer of 2.2 - 2.4 power. Nothing unusual there, smaller color gamut than SLog3 or Rec.2020 but not materially different than other raw formats; Canon C-Log, Arri Log-C, BRAW Film log etc. Sigma fp also has an assortment of cinema color modes. While the feature can be switched off, the possibility (if it exists) of baking it into CinemaDNG is also inconsistent with raw. Sigma claims a DR of 12.5 stops. 2^12.5 power = 5792 but 12 bit only contains 4096 code values. There is no room in the bucket for the last 1/2 stop unless you apply a gamma transfer just like everybody else does except Sony and Red with 16 bit linear. But by applying a gamma curve, even 10 bit can hold 14 stops. But if your raw is recorded from linear voltages off the sensor A/D, that is a clearer definition of what raw really is than otherwise processed signals that include rescaling, remapping and grayscale gamma transformations. Sigma also makes no reference to CineDNG raw as being uncompressed; they just call it "lossless." BMD calls BRAW lossless as well. CinemaDNG is old and obsolete; good that it works for you but Prores Raw with the Atomos Ninja V or BRAW with the Video Assist 12G are going to be much better assuming you don't need HFR.
 

·
Registered
Joined
·
5,750 Posts
Discussion Starter #11 (Edited)
CinemaDNG is an open source Adobe format, and not just for raw. In the use case of the Sigma fp, raw video is recorded in a 8-10-12 bit DNG format from a 14 bit image sensor. That transformation is a rescaling process but it is not the only one. Rescaling from 6000 x 3376 readout for 16:9 UHD is another. In fact, it is the first instance I have ever heard of any sensor raw being anything but a 1:1 sensor readout. How can you debayer a CFA that's been resized with pixels that no longer exist and call that raw? But reverse mortgage is just a loan, and CinemaDNG is just a format, a container for anything that fits under ISO 12234-2. Sigma lists the color space as sRGB or or Adobe RGB. Those are applied with a gamma transfer of 2.2 - 2.4 power. Nothing unusual there, smaller color gamut than SLog3 or Rec.2020 but not materially different than other raw formats; Canon C-Log, Arri Log-C, BRAW Film log etc. Sigma fp also has an assortment of cinema color modes. While the feature can be switched off, the possibility (if it exists) of baking it into CinemaDNG is also inconsistent with raw. Sigma claims a DR of 12.5 stops. 2^12.5 power = 5792 but 12 bit only contains 4096 code values. There is no room in the bucket for the last 1/2 stop unless you apply a gamma transfer just like everybody else does except Sony and Red with 16 bit linear. But by applying a gamma curve, even 10 bit can hold 14 stops. But if your raw is recorded from linear voltages off the sensor A/D, that is a clearer definition of what raw really is than otherwise processed signals that include rescaling, remapping and grayscale gamma transformations. Sigma also makes no reference to CineDNG raw as being uncompressed; they just call it "lossless." BMD calls BRAW lossless as well. CinemaDNG is old and obsolete; good that it works for you but Prores Raw with the Atomos Ninja V or BRAW with the Video Assist 12G are going to be much better assuming you don't need HFR.
1. The picture modes are not baked in. Does not matter what mode you shoot in. This has been tested. WB also not, though I think it is in the metadata.

2. Nikon line skips in its full-frame implentation of ProResRAW. So I guess you missed that. The artifact from that can be moire. The guess is the fp also does that.

3. Those gamut modes are for stills only. Actual measurement of the sensor gamut reveals it is closer to REC2020 than the one in Sony's latest cinema Venice camera. Both the fp and the Venice were used to shoot the latest Avatar.

This is the video interview with the videographer involved in shooting Avatar, with the fp and Venice tests:


3. I do not understand your insinuation that fp is compressing (which would violate the patent) nor your conclusion that somehow BRAW and ProResRaw are "better" when there are tests showing that the noise reduction and sharpening in at least BRAW result in visible artifacts not seen in the fp RAW files. Is it the dynamic range via log transforms? Only one flavor of BRAW compression is claimed to be lossless, but even that is defined after noise reduction and sharpening, which are not lossless.

People who have actually used RAW have concluded that the properties of the fp RAW files are the same as those from the original BMPCC and look better than BRAW. Are you claiming that that camera did not produce RAW files?

I agree that straight recording of the linear voltages is the best definition of RAW. And that is the claim. My test was to show that telling Resolve in color management that the input gamma is in fact linear (and the gamut REC2020) results in perfectly nice REC709 video when the ouput is specified as REC709. So, it is linear.
 

·
Registered
Joined
·
4,614 Posts
1. The picture modes are not baked in. Does not matter what mode you shoot in. This has been tested. WB also not, though I think it is in the metadata.

2. Nikon line skips in its full-frame implentation of ProResRAW. So I guess you missed that. The artifact from that can be moire. The guess is the fp also does that.

3. Those gamut modes are for stills only. Actual measurement of the sensor gamut reveals it is closer to REC2020 than the one in Sony's latest cinema Venice camera.
For green, not red and blue, but I am grateful for the video link because it answers some questions but leaves one open, which I'll get to.


3. I do not understand your insinuation that fp is compressing (which would violate the patent) nor your conclusion that somehow BRAW and ProResRaw are "better" when there are tests showing that the noise reduction and sharpening in at least BRAW result in visible artifacts not seen in the fp RAW files. Is it the dynamic range via log transforms? Only one flavor of BRAW compression is claimed to be lossless, but even that is defined after noise reduction and sharpening, which are not lossless.
.

I didn't insinuate compression, I said "rescaled." On the conversion from 14 bit sensor to 8-10-12 bit raw the word length is truncated. But yes, 12 bit log raw is better than 12 bit linear because the latter is limited to 12 stops. Not only that, but the darkest stop is represented by just a single shade of gray. Just one, yet the brightest stop has 4096 shades (0FFFh). That's way more than you need at the bright end and not nearly enough at the dark end. Log transformation in-camera redistributes the shades to be more equitable. There is a reason for it. Not to be argumentative, I leave that to you but your video at about the 1:50 mark is swimming in shot noise on the trash container in the shadowy left side of the frame illustrates this consequence.

People who have actually used RAW have concluded that the properties of the fp RAW files are the same as those from the original BMPCC and look better than BRAW. Are you claiming that that camera did not produce RAW files?
Straw man argument. Post some sources for consideration, but yes I have seen some frame grabs that purport to show that CinemaDNG has harder edges, aliasing actually, for the reason Tugela stated, result of the debayering process.

I agree that straight recording of the linear voltages is the best definition of RAW. And that is the claim. My test was to show that telling Resolve in color management that the input gamma is in fact linear (and the gamut REC2020) results in perfectly nice REC709 video when the ouput is specified as REC709. So, it is linear.
And I accept that conclusion, however I found it telling that when Pawel was asked a podcast question about the DR, he danced around it, not giving a direct answer, saying he did not know exactly, and answering instead about "sensitivity." He did not mention dual ISO and I didn't see it in the specs, but I'm sure he knows and understands, DR cannot be more than 12 stops in CinemaDNG, and if the claim of 12.5 is true, it is obtained using h.264. Pawel doesn't mention if he used crop mode for Avatar II. Wide DR is probably not needed for underwater cinematography, 6-8 stops is probably justified.

It's also abundantly obvious, that if abhorrent line skipping is indeed the process for full frame cropless raw, it comes with no free lunch. Vertical resolution is halved which means that to preserve the 16:9 aspect ratio, horizontal resolution must also be halved by column skipping or dithering, which as with downsampling is not a native raw processing function. Unless you have a better explanation and can post a source for it, the conclusion is APS-C sensor crop is the only possible true raw mode, a terrible loss for full frame format; better to stick with h.264. Comment?

It would be better if Sigma did their raw the right way, which is to sample the full sensor at 6k raw and deliver in 6k.

WB and ISO are btw not lossless functions of raw just because they are adjustable from the raw pane in your NLE. WB is an in-camera adjustment of red and blue gains. You can adjust it in-post from the recorded metadata in the raw pane but it is simply adjusting gain and noise with it. You still should want to WB in camera if possible. Source: Alister's June 29 blog http://www.xdcam-user.com/alisters-blog/
 

·
Registered
Joined
·
4,614 Posts
Alister Chapman said:
Now that raw video has become the latest must have camera function everyone and their dog is adding raw video to cameras. But because it is actually quite difficult to record good quality 12 bit log raw or 16 bit linear raw in a small enough file for low cost, compact and reliable media, they are adding much simpler and easier formats such as 10 bit log or 12 bit linear. Just so they can put a raw sticker on the box.
This is a must read!
 

·
Registered
Joined
·
5,750 Posts
Discussion Starter #14
For green, not red and blue, but I am grateful for the video link because it answers some questions but leaves one open, which I'll get to.


.
There's good info . But also some slant.

I didn't insinuate compression, I said "rescaled." On the conversion from 14 bit sensor to 8-10-12 bit raw the word length is truncated. But yes, 12 bit log raw is better than 12 bit linear because the latter is limited to 12 stops. Not only that, but the darkest stop is represented by just a single shade of gray. Just one, yet the brightest stop has 4096 shades (0FFFh). That's way more than you need at the bright end and not nearly enough at the dark end. Log transformation in-camera redistributes the shades to be more equitable. There is a reason for it. Not to be argumentative, I leave that to you but your video at about the 1:50 mark is swimming in shot noise on the trash container in the shadowy left side of the frame illustrates this consequence.



Straw man argument. Post some sources for consideration, but yes I have seen some frame grabs that purport to show that CinemaDNG has harder edges, aliasing actually, for the reason Tugela stated, result of the debayering process.



And I accept that conclusion, however I found it telling that when Pawel was asked a podcast question about the DR, he danced around it, not giving a direct answer, saying he did not know exactly, and answering instead about "sensitivity." He did not mention dual ISO and I didn't see it in the specs, but I'm sure he knows and understands, DR cannot be more than 12 stops in CinemaDNG, and if the claim of 12.5 is true, it is obtained using h.264. Pawel doesn't mention if he used crop mode for Avatar II. Wide DR is probably not needed for underwater cinematography, 6-8 stops is probably justified.

It's also abundantly obvious, that if abhorrent line skipping is indeed the process for full frame cropless raw, it comes with no free lunch. Vertical resolution is halved which means that to preserve the 16:9 aspect ratio, horizontal resolution must also be halved by column skipping or dithering, which as with downsampling is not a native raw processing function. Unless you have a better explanation and can post a source for it, the conclusion is APS-C sensor crop is the only possible true raw mode, a terrible loss for full frame format; better to stick with h.264. Comment?

It would be better if Sigma did their raw the right way, which is to sample the full sensor at 6k raw and deliver in 6k.

WB and ISO are btw not lossless functions of raw just because they are adjustable from the raw pane in your NLE. WB is an in-camera adjustment of red and blue gains. You can adjust it in-post from the recorded metadata in the raw pane but it is simply adjusting gain and noise with it. You still should want to WB in camera if possible. Source: Alister's June 29 blog http://www.xdcam-user.com/alisters-blog/
There's good info here and in Chapman. But also some slant.

What it comes down to is 10 bit highly compressed versus 12bit RAW. Compression is the problem. The reasonable argument is that processing in camera might be using 14bits before compression, so that is good. Noise reduction etc with the high bit source could be good. But then any additional noise reduction in post is working with 10bit compressed, which is not good. All processing with raw clips is done on noncompresed 12bit, and the noise reduction in software can also be a lot better than what the camera can do. Neatvideo is praised as using very sophisticated algorthms. I use noise reduction as an example because of your one criticism of the video that there is noise in a dark area. You seem to claim that is due to limited dynamic range. But in fact there is zero noise reduction in the camera, unlike all h264 or BRAW clips, so you can't point to noise as a defect of the clips. I could apply noise reduction very effectively on the 10bit raw clips, unlike if that noise were in h264 clips.

On white balance, if you get it wrong with h264, you are again correcting a highly compressed 10bit clip. That will be worse than correcting an uncompressed 12bit clip. Thats the point about white balance. But its the same point as above - much more quality is retained working with high bits before compression. The camera does that, which is good, but any further alterations in post will be not so good.

A perfectly shot, perfectly graded and noise-reduced *in camera* clip as the final result will be better possibly than a graded 12bit raw clip. But that means you have much less latitude with h264 clips - you gotta get them exactly right to your taste in the camera. Bake in the look you want and stay with it. Or get artifacts of all kinds.
 

·
Registered
Joined
·
4,614 Posts
Alister Chapman said:
Linear raw tries to record the scene with the correct amount of data for each stop. So that means each stop requires twice as much data as the previous one. The problem is that if we use just 2 code values for the first stop we need 2048 code values to record stop 11. 12 bit data only has a maximum of 4096 code values, so by stop 11 we have run out of data. 14 stop linear raw actually need 16 bits. 12 bit linear raw, really, really needs to be well exposed as anything down in the lower end will lack code values so will appear coarse and grainy. But there is a huge amount of data up in the highlights of 12 bit raw. Bright scenes work really well with 12 bit linear, low key can be a disaster.
..
 
  • Like
Reactions: markr041

·
Registered
Joined
·
9 Posts
I am intrigued by the Sigma fp camera. The DNG video workflow is quite familiar to me since I've used Lightroom a lot with raw photos in the past. @markr041 Thank you for the extensive examples. Just to make sure, is http://www.slimraw.com/ that DNG raw compression software you are referring to?
 

·
Registered
Joined
·
5,750 Posts
Discussion Starter #18
BRAW vs DNG RAW Comparison

https://www.4kshooters.net/2018/11/03/blackmagic-raw-braw-vs-cinemadng-side-by-side-comparison/

Bottom line: BRAW clips had less noise and were softer compared with DNG RAW clips. As suspected, BRAW is reducing noise, and thus smearing, to enable more compression (since there is less information).

Noise reduction can always be done in post (and more effectively). You cannot in post, however, retrieve lost information.

I am still pursuing this as I have a camera that (Uniquely) can shoot DNG RAW, BRAW and ProRes RAW, the latter two requiring adding a bulky recorder externally. The cost is obvious; I am trying to figure out the returns, if any.
 
1 - 18 of 18 Posts
Top