Please select Into the mobile phone version | Continue to access the computer ver.
Dlog/RAW with X5S
12Next >
10951 42 2017-3-27
Uploading and Loding Picture ...(0/1)
o(^-^)o
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

On another thread (https://forum.dji.com/forum.php? ... D385%26typeid%3D385), there was a pretty extended discussion of how to expose DLOG. I was kinda surprised, because there wasn't a lot of distinction being made between DLOG for shooting ProRes and DLOG when shooting cDNG RAW. I made the comment that DLOG is a non-sequitor when discussing RAW because there is no gamma curve applied to RAW recorded data. (DLOG is a gamma curve for converting RAW to REC709)

I just ran a test with my camera, shooting a step wedge under various conditions. A step wedge is a pretty nice way of being able to see the gamma curve being applied to the image.

As I suspected, there is ABSOLUTELY NO difference when shooting RAW whether one applies the DLOG or STANDARD "look". Which begs the question, WHY does DJI alter and lock the ISO at 500 when recording cDNG? It's a nonsense setting to pick DLOG for cDNG.

Things are much more rational when shooting ProRes, because ProRes is a debayered REC709 image. The DLOG recorded footage does indeed show a compressed gamma curve, as one might expect when compared to the STANDARD gamma curve. The "compressed" gamma curve is what allows the expansion in dynamic range, up to the limitations of the sensor. It's an interesting side note to state that using the DLOG gamma curve is useful when the overall dynamic range is high. For low dynamic range scenes, I'd still prefer to use the "STANDARD" gamma curve as it will distribute all the exposure values over the whole exposure range of ISO. The DLOG gamma curve sacrifices shadow detail in order to increase hilight detail.
2017-3-27
Use props
Gybo102
First Officer
Flight distance : 2426230 ft
United States
Offline

DJI could you please explain the above? It is odd having a color profile for shooting RAW.
2017-3-28
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Gybo102 Posted at 2017-3-28 03:10
DJI could you please explain the above? It is odd having a color profile for shooting RAW.

By definition, there should be no color profile applied to RAW. A RAW file is recorded directly from the sensor, without any significant processing. To view a RAW file as REC709, REC2020, or any other viewable format. it has to be "debayered". Part of the debayering process is to apply a gamma curve. So, it doesn't fit conventional ideas of what RAW is to be able to apply a DLOG gamma when you shoot RAW. Yet, the current release of the I2 firmware allows selection of DLOG as a recording choice for RAW. If you pick DLOG, you lock the ISO to 500. It's nonsense in this context.

The entire intent of DLOG is as applied to a debayered recording format, like Prores or DNxHR. This is where a gamma conversion should be applied.\; and, it's correctly done when you select ProRes 422 in the Go4 application.
2017-3-28
Use props
Gybo102
First Officer
Flight distance : 2426230 ft
United States
Offline

raven4 Posted at 2017-3-28 04:06
By definition, there should be no color profile applied to RAW. A RAW file is recorded directly from the sensor, without any significant processing. To view a RAW file as REC709, REC2020, or any other viewable format. it has to be "debayered". Part of the debayering process is to apply a gamma curve. So, it doesn't fit conventional ideas of what RAW is to be able to apply a DLOG gamma when you shoot RAW. Yet, the current release of the I2 firmware allows selection of DLOG as a recording choice for RAW. If you pick DLOG, you lock the ISO to 500. It's nonsense in this context.

The entire intent of DLOG is as applied to a debayered recording format, like Prores or DNxHR. This is where a gamma conversion should be applied.\; and, it's correctly done when you select ProRes 422 in the Go4 application.

I understand what you are saying, and I agree...  RAW should be RAW and nothing else.
Its like me taking my olympus EM1MII and that camera being able to apply "vivid" or some other image mode when shooting RAW photos... should be right off the sensor with no processing.
So when shooting RAW video on the I2, im curious what the reason is for adding this?
2017-3-28
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Gybo102 Posted at 2017-3-28 04:52
I understand what you are saying, and I agree...  RAW should be RAW and nothing else.
Its like me taking my olympus EM1MII and that camera being able to apply "vivid" or some other image mode when shooting RAW photos... should be right off the sensor with no processing.
So when shooting RAW video on the I2, im curious what the reason is for adding this?

That's exactly my point. There's NO reason I can see for adding a DLOG option when recording cDNG. If DLOG is needed from a RAW file, application of a conversion LUT, thru something like Davinci Resolve, is the only sensible way to go.
2017-3-28
Use props
I 2 Fly
Second Officer
Flight distance : 8176 ft
South Africa
Offline

I also shot some stuff in DLOG and was not (very) impressed with the end result after correcting and grading.....  
2017-3-28
Use props
Gybo102
First Officer
Flight distance : 2426230 ft
United States
Offline

I 2 Fly Posted at 2017-3-28 06:46
I also shot some stuff in DLOG and was not (very) impressed with the end result after correcting and grading.....

I will admit, after the latest update, DLOG in ProRes has been improved some (minus the ISO lock at 500) As far as DLOG in RAW... I will not use it because again, im locked at ISO 500 and we should not be using a "color profile" or anything in RAW as the intended purpose is to get stripped down off the sensor data.
With good ND, and a good exposure (which is sometimes hard to get) the camera is giving me some good results..
Content in this video was shot before the last update was applied:


Im working on getting some new footage, but weather has been crappy lately.. only have a couple of shots. from what I have seen, a tad more noise in the image.. which really sucks.. but... either way, kind of off topic now.

Back on topic... DJI, why DLOG in RAW?
2017-3-28
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

I'll just post a contrasting opinion here, with a little info thrown in. Your question essentially is why is D-Log applied to CDNG, followed by "it's wrong to apply a gamma curve" to CDNG.

Whenever you view a raw image in a RAW processor like ACR, lightroom et al...you aren't looking at the "raw" image. You are looking at profile that has been applied to the RAW based on metadata tags embedded in the file. That profile is a Gamma Curve, just like the log profile DJI is tagging into CDNG files shot in "log". So in that sense, there's really nothing to see here...the raw file is the raw file...they are exactly the same "under the hood", you are just being shown a different, post-debayer preview when you open the file in the viewer.  There is no loss of shadow detail (I wish this one would go away, because it's not what happens in LOG BTW), no extended highlights in the RAW file...but simply a curve applied that is distributing the tones differently than "normal".

But your question as to "why" is a better one. DJI has a history of strange moves in this category. Their profiles have rarely made any sense, actually, until the recent implementation of D-log, which is an actual Log gamma compared to what they had before. I'm not sure why they've applied it (Log to CDNG) in this case, but it certainly isn't without precedent. Canon applies Log gamma's to it's RAW CDNG output (actually I think they bake theirs in), and I'm pretty sure RED does as well. But I will give you one situation where to me it makes sense. Say I'm shooting most of my footage in D-log ProRes, but I have a few special shots I want to shoot in RAW. With a "normal" gamma applied to the RAW image, I have a completely different workflow in terms of grading, but with the Log applied, it's much more similar. Last week my I2 inadvertently changed my shot settings to RAW on me mid flight (I probably just accidentally hit the button)...so I had some clips in RAW and some in ProRes of the same subject. At first I was perplexed by how I would approach log in ACR...but once I got it going, it was very easy...and resulted in an extremely similar final image to what I had going in FCPX with the ProRes. D-Log grades very easily without a LUT, and so if you are used to grading with it, it really isn't a big deal to do it in a RAW processor.
I'm not a Resolve user, so I'm surmising, but say you have a custom LUT (look) you want to use on both your ProRes D-Log and RAW files. It seems to me having the RAW in Log would be a benefit in that situation.

You certainly have the option of shooting RAW in "standard" with the X5s, so maybe one way to look at it is that the LOG RAW option is an added feature, that some folks may actually use. I can see myself using it.
2017-3-28
Use props
Gybo102
First Officer
Flight distance : 2426230 ft
United States
Offline

Barry Goyette Posted at 2017-3-28 07:36
I'll just post a contrasting opinion here, with a little info thrown in. Your question essentially is why is D-Log applied to CDNG, followed by "it's wrong to apply a gamma curve" to CDNG.

Whenever you view a raw image in a RAW processor like ACR, lightroom et al...you aren't looking at the "raw" image. You are looking at profile that has been applied to the RAW based on metadata tags embedded in the file. That profile is a Gamma Curve, just like the log profile DJI is tagging into CDNG files shot in "log". So in that sense, there's really nothing to see here...the raw file is the raw file...they are exactly the same "under the hood", you are just being shown a different, post-debayer preview when you open the file in the viewer.  There is no loss of shadow detail (I wish this one would go away, because it's not what happens in LOG BTW), no extended highlights in the RAW file...but simply a curve applied that is distributing the tones differently than "normal".

Well put... that is one way to take it. Agreed, it does give "options" in post, and like you mentioned, can even come in handy to have. I still would like to hear from DJI (mainly "DJI-CAO") Even if something can be beneficial, I feel its good to understand the thinking behind the decision, to better help me make a decision of weather to use or not.

.... Im STILL not a fan of locked ISO. Im sure im not alone on this one.
2017-3-28
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

Gybo102 Posted at 2017-3-28 07:50
Well put... that is one way to take it. Agreed, it does give "options" in post, and like you mentioned, can even come in handy to have. I still would like to hear from DJI (mainly "DJI-CAO") Even if something can be beneficial, I feel its good to understand the thinking behind the decision, to better help me make a decision of weather to use or not.

.... Im STILL not a fan of locked ISO. Im sure im not alone on this one.

The locking of the ISO is a complicated issue. Many logs are locked to some extent (you can alter ISO, but it affects your distribution of highlights and shadows..so really nothing changes). In RAW, again...this is just a "front" ISO...the actual RAW file is at base ISO (100)....the profile applied in ACR adds gain to make the image fit an ISO500. You could simply expose your image up to 2.3 stops brighter, (and lose highlight range) or underexpose  and add gain in post (and add noise to your shadows --one benefit of RAW is that you can control this noise). I still argue that the new DLOG500 is superior in every way to the old adjustable DLOG, I just hope that when they start  adding back in ISO adjustability per CAO's comments, that they don't screw it up, because it's finally a useful gamma. (I'm shooting everything with it.)
2017-3-28
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Barry....Thanx for your input. Yeah, I realize that any viewable images I get from ACR or Resolve are debayered images, or else I wouldn't be able to view them in an understandable image. But, here's the crux of my post: I shot two sets of RAW (cDNG) images of a 21 step gray scale wedge. One set was with DLOG turned on and one set was with STANDARD. I adjusted exposure to give me identical exposures. Running both RAW files thru RESOLVE (Resolve debayers and applies BMD Film Gamma to both files, equally), and looking at the resulting Waveform Monitor, I see an identical WFM display for both DLOG and STANDARD. This is telling me there is NO difference in either setting except that the DLOG setting locks the ISO to 500. If the DLOG setting was applying some kind of gamma curve to the RAW recording, I would expect to see a different WFM display when debayered in Resolve. I can confirm that this is true when I shoot DLOG and STANDARD with ProRes. But, the WFM displays are identical in RAW.
If it's instructive, I can upload screengrabs of the WFM display of the debayered grayscale wedge clips. It's quite representative of the gamma curve applied by using BMD Film gamma plus whatever gamma curve DJI may (or not) be applying.

2017-3-28
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

raven4 Posted at 2017-3-28 08:08
Barry....Thanx for your input. Yeah, I realize that any viewable images I get from ACR or Resolve are debayered images, or else I wouldn't be able to view them in an understandable image. But, here's the crux of my post: I shot two sets of RAW (cDNG) images of a 21 step gray scale wedge. One set was with DLOG turned on and one set was with STANDARD. I adjusted exposure to give me identical exposures. Running both RAW files thru RESOLVE (Resolve debayers and applies BMD Film Gamma to both files, equally), and looking at the resulting Waveform Monitor, I see an identical WFM display for both DLOG and STANDARD. This is telling me there is NO difference in either setting except that the DLOG setting locks the ISO to 500. If the DLOG setting was applying some kind of gamma curve to the RAW recording, I would expect to see a different WFM display when debayered in Resolve. I can confirm that this is true when I shoot DLOG and STANDARD with ProRes. But, the WFM displays are identical in RAW.

maybe you can clarify a couple of things...I'm a bit confused about your points..

I shot two sets of RAW (cDNG) images of a 21 step gray scale wedge.
--Is this a transmissive scale with each step showing 1 stop, or a reflective scale?

I adjusted exposure to give me identical exposures.
--In camera or in post?

I see an identical WFM display for both DLOG and STANDARD. This is telling me there is NO difference in either setting except that the DLOG setting locks the ISO to 500. If the DLOG setting was applying some kind of gamma curve to the RAW recording, I would expect to see a different WFM display when debayered in Resolve.
--isn't this what I was saying above?
2017-3-28
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Barry Goyette Posted at 2017-3-28 08:19
maybe you can clarify a couple of things...I'm a bit confused about your points..

I shot two sets of RAW (cDNG) images of a 21 step gray scale wedge.

The 21 step BW wedge is a reflective scale available from Alistair Chapman at XDCAM-USER.com. http://www.xdcam-user.com/2015/1 ... ay/xdcam-greyscale/
While each step is not an even 1-stop step, each step is an equal step size. Exposure was normalized in camera, not in post. And, I should note that as a reflective object, it's pretty important that it gets evenly illuminated for the test. But, since the illumination is not changed from one take to the next, one would not expect the WFM to show a change due to illumination.

Yeah, that's what you said above....except that I note that if DJI was applying some kind of DLOG gamma curve to their data before it was recorded, it would be added to the gamma curve you're referring to; and, would show up as a noticeable difference on the debayered WFM.
Being redundant, I'll say again that I repeated the test with ProRes, DLOG vs Standard, and the expected compression of the WFM display is quite obvious with DLOG.

2017-3-28
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

raven4 Posted at 2017-3-28 08:37
The 21 step BW wedge is a reflective scale available from Alistair Chapman at XDCAM-USER.com. While each step is not an even 1-stop step, each step is an equal step size. Exposure was normalized in camera, not in post. And, I should note that as a reflective object, it's pretty important that it gets evenly illuminated for the test. But, since the illumination is not changed from one take to the next, one would not expect the WFM to show a change due to illumination.

Yeah, that's what you said above....except that I note that if DJI was applying some kind of DLOG gamma curve to their data before it was recorded, it would be added to the gamma curve you're referring to; and, would show up as a noticeable difference on the debayered WFM.

sorry I am having a really hard time following. talking about charts is really hard without the accompanying visual and a very specific methodology.

first... A reflective chart is at best, only capable of showing approximately 7 stops. So whatever you're seeing on your waveform is only representative of the very middle of the range of the X5s, unless resolve is applying a rec.709 curve to the image, which would compress the shadows under 4 stops below middle grey and highlights 2.5 stops above. Again...when you say they are identical -- they should be, but...

So you say you "normalized" the exposure in camera? Meaning you exposed the two shots differently? If this is the case, and you got an identical waveform, then this is implying the opposite of your conclusion...DJI must be adding some gain to the raw DLOG for them to look the same if you exposed differently. However...there is no way that the highlight end of the chart could look the same if you exposed them 2.3 stops apart (this is where the limitations of a reflective chart comes into play, because you're not seeing anything beyond its maximum 7 stops of range.)

As for what Resolve is doing, I don't use the program and certainly have no info regarding how it reads these raw files, and applies its own profiles...without seeing what you're doing, it's hard to comment...but....

In theory, if you expose RAW as "standard" at ISO 100, then you'd have more info in the shadows, and less noise, but lose 2+ stops of highlight compared to shooting log at 500. The only way to test the difference between the two gamma's (standard and Log) would be to use identical exposures and then attempt to normalize them in post, ideally using a scene that encompasses beyond the available DR of the camera. It is possible that DJI is baking in gain to their RAW file...meaning that 500 is a true 500 (and not a profile applied in post) and that there is more (or less) noise because of this. IDK.
2017-3-28
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Barry Goyette Posted at 2017-3-28 09:18
sorry I am having a really hard time following. talking about charts is really hard without the accompanying visual and a very specific methodology.

first... A reflective chart is at best, only capable of showing approximately 7 stops. So whatever you're seeing on your waveform is only representative of the very middle of the range of the X5s, unless resolve is applying a rec.709 curve to the image, which would compress the shadows under 4 stops below middle grey and highlights 2.5 stops above. Again...when you say they are identical -- they should be, but...

Barry, Thanx for the feedback. Yeah I see your point about in camera exposure. I'll go back and look at that again. I had an inkling that normalizing in camera would not be right. So, thanks for that. Also, yeah, I recognize the serious implications/limitations of using a printed reflectance chart. Sadly, a nice transmissive test object, like a Chromadumonde step wedge is not part of my equipment inventory. Clearly, printed reflectance step chart will never, ever give the full range of exposure. But, It's all i got. Wouldn't it be possible to change the exposure of the chart to capture both the hilites, then stop down to capture the shadows, using a WFM as a reference? Obviously, the same aperture settings would have to be reproduced for DLOG and STANDARD. Thanx again for your feedback...much appreciated.
-Bill
2017-3-28
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

Well, short of a $3500 xyla chart, you sort of have to get creative. If you are trying to measure "actual" dynamic range in a single exposure, without a transmissive chart, you need a light meter, and it can be as simple as pointing a light obliquely at a long white or neutral colored wall and marking off the stops based on illumination with tape. Put the light very close to the wall, or you're gonna need a really long wall :-).  You want to see how creative you can get with this...watch this  

Generally, though, the method is to photograph a single 'patch" with a defined ramp of separate exposures. You would then plot the WF readings on a graph to see what the gamma curve is doing. Again, the resolve workflow you are using, I'm not sure what it's doing, so that may impact things. I think raw therapee allows you the option to view the raw image without a profile. This might be an easy way to compare the actual files.

Another approach would be to use a stouffer chart...which are designed for darkroom work. Stouffer makes an inexpensive 41 step transmissive chart that encompasses 13 stops...which is borderline, because you ideally need to clip the first two stops...however due to it's 1/3 stop gradation, you could cheat a little and show 12+ stops...or you could buy two (4110 --they are only $38 a piece or you can buy a calibrated, certified for $178) and stack them. which would give you 2/3 stop increments with a 26 stop range.



2017-3-29
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Good info, thanx. I repeated my test without changing in-cam exposure, however, I am seeing a bunch of fallacies with my approach. So, maybe one of the methods you've described will be my next attempt. P.S. Like Raw Therapee, Rawdigger is also a good diagnostic viewer.
2017-3-29
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

raven4 Posted at 2017-3-29 10:36
Good info, thanx. I repeated my test without changing in-cam exposure, however, I am seeing a bunch of fallacies with my approach. So, maybe one of the methods you've described will be my next attempt. P.S. Like Raw Therapee, Rawdigger is also a good diagnostic viewer.

Good luck with it. I'm curious to see what you find. Your Resolve test does sort of indicate that DJI may be adding gain to the D-Log Raw (versus a metadata profile). It would be good to know if that's what is really happening, not that it's necessarily a bad thing. Canon does it on the C300 mark II (gaining the analog circuits before encoding the raw data in LOG) and it can be remarkably effective at controlling noise at higher ISOs. Who knows what DJI DLOG500's secret sauce is and whether it's a good or a bad thing.
2017-3-29
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Barry Goyette Posted at 2017-3-29 12:50
Good luck with it. I'm curious to see what you find. Your Resolve test does sort of indicate that DJI may be adding gain to the D-Log Raw (versus a metadata profile). It would be good to know if that's what is really happening, not that it's necessarily a bad thing. Canon does it on the C300 mark II (gaining the analog circuits before encoding the raw data in LOG) and it can be remarkably effective at controlling noise at higher ISOs. Who knows what DJI DLOG500's secret sauce is and whether it's a good or a bad thing.

What I'm finding seems to confirm your thinking. First, looking at the RGB parade that's available in Go4, before uploading to Resolve, the parade looks more compressed with Dlog than without. So, even with some added gain, there is an apparent redistribution of hi's and shadows towards the middle. Then, uploading to Resolve does some range changing, depending on what gamma and gamut settings I pick, whether it's Rec709, Blackmagic, Arri, or whatever.
BTW, Ben Turley claims to have reverse engineered the DLOG gamma curve...if this is right, I'm not sure I like the knee at 95%...

DJIGamma.JPG
2017-3-29
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

raven4 Posted at 2017-3-29 13:04
What I'm finding seems to confirm your thinking. First, looking at the RGB parade that's available in Go4, before uploading to Resolve, the parade looks more compressed with Dlog than without. So, even with some added gain, there is an apparent redistribution of hi's and shadows towards the middle. Then, uploading to Resolve does some range changing, depending on what gamma and gamut settings I pick, whether it's Rec709, Blackmagic, Arri, or whatever.
BTW, Ben Turley claims to have reverse engineered the DLOG gamma curve...if this is right, I'm not sure I like the knee at 95%...

As Far as I know...the Waveform only shows the H264 video stream and really wouldn't have anything to do with what the raw recording looks like, so I wouldn't count on that to tell you much. If I get what i think you're after, which is answering the question as to whether DJI is encoding the raw data differently...ie baking in the Log Gamma...then your best approach would be to compare the two test clips in something that isn't applying any profile, and is ignoring the metadata. I don't know if Resolve does this or not. So you might try RawTherapee, or the program you mentioned and see what that shows you. Once you know that (same or different)...if it's different, then you would attempt to grade both to an identical output and see if there were any differences in DR, color depth, noise etc.

Ben certainly knows his sh*t. I'd love to see that post if you have a link you can share. That "knee" is a little odd, but not unheard of especially in an 8 bit curve, which d-log seems optimized for. Arri does similar compression in its last  highlight stops, which is responsible for the Alexa Look as well as partly responsible for its superior noise in the shadows compared to other cameras (as it allows you to increase exposure by .5-1stop). C300 mark II also has a weird flatter upper section in C-Log2 that extends beyond it's 15 stops, and we see that same thing here. (although it would appear to be encompassing the last stop of DJI's 13 stops of DR). Generally we tend to compress those upper stops anyway in grading, so its allowing for a little more wiggle room in the rest of the curve. Doesn't concern me.
2017-3-30
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Barry Goyette Posted at 2017-3-30 08:04
As Far as I know...the Waveform only shows the H264 video stream and really wouldn't have anything to do with what the raw recording looks like, so I wouldn't count on that to tell you much. If I get what i think you're after, which is answering the question as to whether DJI is encoding the raw data differently...ie baking in the Log Gamma...then your best approach would be to compare the two test clips in something that isn't applying any profile, and is ignoring the metadata. I don't know if Resolve does this or not. So you might try RawTherapee, or the program you mentioned and see what that shows you. Once you know that (same or different)...if it's different, then you would attempt to grade both to an identical output and see if there were any differences in DR, color depth, noise etc.

Ben certainly knows his sh*t. I'd love to see that post if you have a link you can share. That "knee" is a little odd, but not unheard of especially in an 8 bit curve, which d-log seems optimized for. Arri does similar compression in its last  highlight stops, which is responsible for the Alexa Look as well as partly responsible for its superior noise in the shadows compared to other cameras (as it allows you to increase exposure by .5-1stop). C300 mark II also has a weird flatter upper section in C-Log2 that extends beyond it's 15 stops, and we see that same thing here. (although it would appear to be encompassing the last stop of DJI's 13 stops of DR). Generally we tend to compress those upper stops anyway in grading, so its allowing for a little more wiggle room in the rest of the curve. Doesn't concern me.

Sheesh....I didn't think of that, so thanx. I should've known that in cam RGB parade wasn't the RAW signal, doh! I did compare a couple of images in RAWDIGGER. It looked like the DLOG histogram wasn't the same as the STD histogram, but, all in all, very subjective and not conclusive.

Ultimately, what I'm after is a Davinci Resolve workflow that maximizes the DR. It's really easy to clip highlites in any LUT application I've found. I ususally have to apply contrast and offset corrections BEFORE applying any Dlog-to-Rec709 LUT. Honestly, once I start manipulating the footage in Resolve, I don't see much difference in STD vs. DLOG.

The gamma curve from Ben Turley is straight off his free software LUTCALC. I set a generic camera and used Dlog Gamma, REC709 gamut for both In and OUT settings. It's interesting to set REC709 Gama for the output, as it then compares the DJI gamma to the REC709 gamma.
2017-3-30
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

raven4 Posted at 2017-3-30 12:45
Sheesh....I didn't think of that, so thanx. I should've known that in cam RGB parade wasn't the RAW signal, doh! I did compare a couple of images in RAWDIGGER. It looked like the DLOG histogram wasn't the same as the STD histogram, but, all in all, very subjective and not conclusive.

Ultimately, what I'm after is a Davinci Resolve workflow that maximizes the DR. It's really easy to clip highlites in any LUT application I've found. I ususally have to apply contrast and offset corrections BEFORE applying any Dlog-to-Rec709 LUT. Honestly, once I start manipulating the footage in Resolve, I don't see much difference in STD vs. DLOG.

I think I mentioned it in another thread, but using stock LUT's in non-controlled lighting conditions is generally a bit of a fools errand -- as you're usually trying to hold shadows and highlights that the LUT naturally wants to compress -- and certainly with a LOG like DLog that is really easily handled with color, exposure and saturation controls -- I'm finding it's best to ignore the LUT workflow and just start grading. A LUT workflow should involve a monitor LUT at time of capture (which is currently impossible the way Dlog is implemented across codecs), precise correction "in Camera", with various controls and...uh...lighting :-)  -- then what you see in resolve would match what you saw "in-camera". All the rest of the DR is there for secondaries. The reality of the drone world is that you're trying to hold the corona around the sun and then cramming all the rest of the DR into a nice 7 stop box with a LUT. Not gonna happen, and trying to fit 13 stops of DR into rec 709 is always going to involve a huge amount of compromising. Most likely both RAWs are going to have the same amount of DR...The LOG version, if it's any different at all, should have more gradations in the tones below middle grey, and a few less in the tones above middle grey, but I'll bet they aren't that different at all.
2017-3-30
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Barry Goyette Posted at 2017-3-30 13:22
I think I mentioned it in another thread, but using stock LUT's in non-controlled lighting conditions is generally a bit of a fools errand -- as you're usually trying to hold shadows and highlights that the LUT naturally wants to compress -- and certainly with a LOG like DLog that is really easily handled with color, exposure and saturation controls -- I'm finding it's best to ignore the LUT workflow and just start grading. A LUT workflow should involve a monitor LUT at time of capture (which is currently impossible the way Dlog is implemented across codecs), precise correction "in Camera", with various controls and...uh...lighting :-)  -- then what you see in resolve would match what you saw "in-camera". All the rest of the DR is there for secondaries. The reality of the drone world is that you're trying to hold the corona around the sun and then cramming all the rest of the DR into a nice 7 stop box with a LUT. Not gonna happen, and trying to fit 13 stops of DR into rec 709 is always going to involve a huge amount of compromising. Most likely both RAWs are going to have the same amount of DR...The LOG version, if it's any different at all, should have more gradations in the tones below middle grey, and a few less in the tones above middle grey, but I'll bet they aren't that different at all.

"...but using stock LUT's in non-controlled lighting conditions is generally a bit of a fools errand". Yup, exactly the conclusion I've come to. Here's the REC709 dilemma....with a portion of the acquired imaging usually encompassing sky and some cloud structure, yet still not pushing the area below the horizon into the mush, there's the challenge. Short of using grad filters, the great white hope is DR. Since shots from 200 ft altitude don't lend themselves to key or fill lights, it's a big problem to bring light to the landscape, but, retain the cloud structure. Mostly, I can only get this done in post with judicious masking....as long as I haven't blown those hilites out. I'm finding that my original concept to ETTR is wrong. With the X5s, it's best to keep exposures right on, exposing clouds at 90%IRE, which is where I presume the zebra function is. But, even there, DJI has not published any data, to my knowledge, when the zebras come on. Bit of a rant, here. After promoting the I2 as a camera for film makers, the current firmware/application useability is pretty toy like.
2017-3-30
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

raven4 Posted at 2017-3-30 13:29
"...but using stock LUT's in non-controlled lighting conditions is generally a bit of a fools errand". Yup, exactly the conclusion I've come to. Here's the REC709 dilemma....with a portion of the acquired imaging usually encompassing sky and some cloud structure, yet still not pushing the area below the horizon into the mush, there's the challenge. Short of using grad filters, the great white hope is DR. Since shots from 200 ft altitude don't lend themselves to key or fill lights, it's a big problem to bring light to the landscape, but, retain the cloud structure. Mostly, I can only get this done in post with judicious masking....as long as I haven't blown those hilites out. I'm finding that my original concept to ETTR is wrong. With the X5s, it's best to keep exposures right on, exposing clouds at 90%IRE, which is where I presume the zebra function is. But, even there, DJI has not published any data, to my knowledge, when the zebras come on. Bit of a rant, here. After promoting the I2 as a camera for film makers, the current firmware/application useability is pretty toy like.

Yup 95 ire is max on D-log. I usually push my clouds right up against that and then back off a half stop, That keeps the landscape pretty open. If its a backlit cloud or sun, there isn't going to be enough DR to pull the landscape up in post, so you either let the sky go, or ad a grad (which are largely unknown in filters of this size). As we move into HDR land with televisions and computers being able to display more than 7 stops, this will ease somewhat, but the need to not bury your actual subject matter down in the shadows will always be important, even in HDR.
2017-3-30
Use props
Barry Goyette
Second Officer
Flight distance : 14928 ft
  • >>>
United States
Offline

Also...curious where you came up with the D-log curve in LutCalc. I'd been aware of the App, but never used it. Happily playing with it and creating LUTS for my C300mkII, but I don't see DJI or D-log anywhere in the program. I do know that he had previously profiled the old D-log, but I don't see that, or anything in his notes about the new D-log. Just curious where you found it.

edit: Never mind. I looked through all the gamma menus and found it under the uncerimonious "LOG". Based on his notes in the NEWs section, this was added to the program back in May, and so this would be the "old" D-log...which is not the "new" d-log.

2017-4-1
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Barry Goyette Posted at 2017-4-1 07:38
Also...curious where you came up with the D-log curve in LutCalc. I'd been aware of the App, but never used it. Happily playing with it and creating LUTS for my C300mkII, but I don't see DJI or D-log anywhere in the program. I do know that he had previously profiled the old D-log, but I don't see that, or anything in his notes about the new D-log. Just curious where you found it.

edit: Never mind. I looked through all the gamma menus and found it under the uncerimonious "LOG". Based on his notes in the NEWs section, this was added to the program back in May, and so this would be the "old" D-log...which is not the "new" d-log.

Also, under CAMERA, you need to select "generic".
It's a great app, isn't it.
2017-4-1
Use props
DJI-CAO
lvl.3
Flight distance : 228714 ft
  • >>>
China
Offline

Well, Inspire 2 makes the mysterious digital filming open. The answer is, to your surprise, DLOG affects on the footage shot by CinemaDNG. Precisely speaking, the exposure is changed when applying DLOG.

I agree with you that CinemaDNG is a linear RAW independent to color and film tones. The key is the meta data which inform the workflow how to handle the RAW. Two among them are very important.

1. the exposure bias of 18% gray.
As I mentioned in another post, the exposure, to the camera engineering point of view, is only responsible for 18% gray.

2. the transfer matrix from sensor RGB to XYZ
XYZ is an absolute color space and this matrix tells what is the sensor RGB.

The post processing software, like Adobe Premiere and Davinci Resolve, might colorize the footage based on their own film curves by default. But these two information guarantee the exposure and color be aligned among different workflows. Please keep in mind this is "by default".

Let's go back to your question and take an example to show the magic of CinemaDNG.

When shooting DLOG with a standard AE, the camera will add a meta data, saying 18% gray is biased by 2.3 ev. So the linear raw is 2.3ev under exposed than shoot in Rec709.

We don't tell the workflow to use what specific film curve because it's not our job. Neither is the workflow. It's the job of you.

Shooting in CinemaDNG can theoretically approach to the color space of any cinema cameras, e.g., ARRI, RED, SONY, PANA, Canon, ... whatever you want.

In Davinci Resolve, the following steps show how to colorize CinemaDNG to ARRI gamut.

1. Shoot the CinemaDNG footage by DLOG;
2. Import to Davinci Resolve, choose linear and tick the highlight recovery in the camera raw settings;
3. In the input 3DLUT, select ARRI RGB, e.g., LCC (light-contrast color??);
4. enjoy

The behind color workflow, I guess, is
sensor RGB->XYZ->ARRI LCC RGB

These steps were summarized to a tutorial. I think you will get it soon in the forum.

Apple ProRes, as an intermediate codec, is another story of workflow, e.g.,
sensor RGB ->DJI Cinema RGB (in camera)->workflow RGB (in workflow)

For example, Adobe Camera RAW uses ProPhoto RGB for photography. I don't know what is used for cinema. Currently, we're still working with leading softwares to adapt to their workflow RGB.

Remember Inspire 2 camera covers a wide customer requirements from photography, pro-sumer video and cinema. GH4 users are quite familiar with the Rec709 shooting and grading. ARRI users definitely prefer LOG. So we have to create a new type of product adapt to different customers. It's our fault that the current options are confusing and even without any tutorials. In the future, we will improve it and open more options of looks to help customer better understand what they need.
2017-4-2
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

DJI-CAO Posted at 2017-4-2 05:58
Well, Inspire 2 makes the mysterious digital filming open. The answer is, to your surprise, DLOG affects on the footage shot by CinemaDNG. Precisely speaking, the exposure is changed when applying DLOG.

I agree with you that CinemaDNG is a linear RAW independent to color and film tones. The key is the meta data which inform the workflow how to handle the RAW. Two among them are very important.

DJI-CAO
Thanks so very much, for your input. This is the kind of info I was looking for. It all makes much more sense, to me, now.
Best regards,
-Bill R
2017-4-2
Use props
fans43da2ba5
Flight distance : 3186319 ft
  • >>>
Russia
Offline

DJI-CAO Posted at 2017-4-2 05:58
Well, Inspire 2 makes the mysterious digital filming open. The answer is, to your surprise, DLOG affects on the footage shot by CinemaDNG. Precisely speaking, the exposure is changed when applying DLOG.

I agree with you that CinemaDNG is a linear RAW independent to color and film tones. The key is the meta data which inform the workflow how to handle the RAW. Two among them are very important.

Dear CAO,
sorry if I did not catch something, but after your words we want to rec in dlog iso 500 always. but we can see pics in dji site -
https://yadi.sk/i/01mdtHX33HQS2Q
by that pictures we can see that working iso is 100 not 500. Could you clarify me by simple words))

Kind regards,
Dmitry
2017-4-27
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

Recording in cDNG, there are two selectable modes, STANDARD and DLOG. STANDARD allows selecting any ISO from 100 and up. Increasing ISO also increases s/n, so one must be careful how much to raise ISO. DLOG modes locks the ISO to 500. Doing this changes where middle gray is placed on the gamma curve, trading off s/n in the shadows for retaining detail in the hilites. In practical application, I would choose STANDARD at ISO 100 or 200 for low contrast scenes and DLOG for high contrast scenes. You can tell what kind of contrast a scene has by looking at the onboard histogram and noting how much of the exposure range is filled with values.
2017-4-27
Use props
aihome
lvl.1
United States
Offline

Been testing this out and noticed that when shot in D-Log the ISO metadata in Resolve 14 is brought in as 100 regardless. While no metadata in ProRes . I wonder what this says or means if anyone has an opinion.
DJI-CAO what do you say.

Thanks!
2017-10-7
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

aihome Posted at 2017-10-7 10:26
Been testing this out and noticed that when shot in D-Log the ISO metadata in Resolve 14 is brought in as 100 regardless. While no metadata in ProRes . I wonder what this says or means if anyone has an opinion.
DJI-CAO what do you say.

Yeah, I noticed that. I've concluded, after many hours of working with DLog footage in Resolve, that the Metadata is incorrect. Apparently, DJI never updated the metadata and left it at default values that Resolve interprets as 100.
2017-10-7
Use props
fansa84fe8a4
First Officer
Flight distance : 3 ft
United States
Offline

For fun, I shot a "Sekonic Exposure Profile Target II" with the X5S using Normal (ISO100), Cinelike (ISO100), and D-Log (ISO 500).  These are all JPG from the camera into the Sekonic software at EV 0 stop, +/-5 stops, and +/-10 stops per their recommendation for their digital light-meter.  Target has 25 B&W patches.

Curves weren't much different for the Normal and Cinelike, but the D-Log shows maybe 2 stops more highlight definition at the shoulder.  JPG shadow was never black in D-Log unless post processed.

X5S Normal, Cinelike, and D-Log curves.

X5S Normal, Cinelike, and D-Log curves.
2017-10-7
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

yeah, that makes a lot of sense. a lot more hilite detail....like 2 stops worth, at the expense of shadow detail.
2017-10-7
Use props
aihome
lvl.1
United States
Offline

DJI-CAO Posted at 2017-4-2 05:58
Well, Inspire 2 makes the mysterious digital filming open. The answer is, to your surprise, DLOG affects on the footage shot by CinemaDNG. Precisely speaking, the exposure is changed when applying DLOG.

I agree with you that CinemaDNG is a linear RAW independent to color and film tones. The key is the meta data which inform the workflow how to handle the RAW. Two among them are very important.

Thanks for this great Info DJI-CAO. Would this be the tutorial you are talking about? is there a video available or in the works?

Thanks again
2017-10-8
Use props
aihome
lvl.1
United States
Offline

raven4 Posted at 2017-10-7 12:36
Yeah, I noticed that. I've concluded, after many hours of working with DLog footage in Resolve, that the Metadata is incorrect. Apparently, DJI never updated the metadata and left it at default values that Resolve interprets as 100.

Thanks man, not surprising that is a quirk on DJI's part and confirm not going crazy on my end
2017-10-8
Use props
aihome
lvl.1
United States
Offline

raven4 Posted at 2017-10-7 18:12
yeah, that makes a lot of sense. a lot more hilite detail....like 2 stops worth, at the expense of shadow detail.

Makes sense technically but maybe not good for aerial or low light work as the main content will be lost below a horizon line. I think this might have been mentioned before on this thread...

Lets see what the X7 brings and hopefully the X5S will not be forgotten.
2017-10-8
Use props
raven4
First Officer
Flight distance : 86194 ft
United States
Offline

aihome Posted at 2017-10-8 11:23
Makes sense technically but maybe not good for aerial or low light work as the main content will be lost below a horizon line. I think this might have been mentioned before on this thread...

Lets see what the X7 brings and hopefully the X5S will not be forgotten.

Looking at this data, the base sits at about 16 IRE, instead of zero IRE, below -5 stops,   Standard REC709 has a lower limit of 16 IRE. So you've, in effect lost nothing. In addition. the shadow detail is compromised up to -3 stops, gaining +2 stops on the high end. and much better detail above 0 IRE. All in all, you're sacrificing a bit of shadow detail for greatly improved midrange and hi range detail. That's the trade off of using log gamma. If there are times when one prefers shadow detail, don't use it. Being aware of when to use which gamma setting is a skill any good photographer knows how to use.
2017-10-9
Use props
FFAUS
New
Flight distance : 784816 ft
Australia
Offline

Im stuck on the fact that the ISO metadata recorded in the CinemaDNG file when in D-LOG is 100, when you have actually exposed the image for an ISO of 500, so this is the 2.5 stops that the footage looks to be under on play back. The footage is actually correctly exposed, it is simply the metadata recorded in the cinemaDNG file incorrectly and either premiere or resolve is rendering playback how it is told to by the metadata. If DJI were to correct this ISO metadata issue in a firmware update would we not see correctly exposed D-LOG CinemaDNG.

There may well not be any advantage of shooting in DLOG when recording to CinemaDNG format, but it would be nice if the camera recorded the metadata correctly. Should this not be a priority for DJI to fix?
2017-10-9
Use props
DJI-CAO
lvl.3
Flight distance : 228714 ft
  • >>>
Hong Kong
Offline

fans43da2ba5 Posted at 2017-4-27 02:04
Dear CAO,
sorry if I did not catch something, but after your words we want to rec in dlog iso 500 always. but we can see pics in dji site -
https://yadi.sk/i/01mdtHX33HQS2Q

That is sensor RAW test.

Please read the introduction of exposure, color and dlog.

The next FW will upgrade cinecore 2.1 and introduce EI mode with all ISO opened.

You will easily understand the EI mode based on my explanations.
2017-10-11
Use props
12Next >
Advanced
You need to log in before you can reply Login | Register now

Credit Rules