What a difference -1 makes....

[peej1977]

So, I know this isn't a particularly new topic but there seems to be a ton of "Best Mavic video settings" tutorials around which often seem to contradict each other. Then a new firmware update appears and seems to nullify those.

Since I got my Mavic I've been playing around the settings, some recommend shooting in D-Log, some in D-Cinelike and there seems to be arguments for using either. I was finding that alot of my 4K footage was blurry yet anything shot in 2.7K seemed fine, so I ran a little experiment and I'm shocked how much difference it makes.

Again, I realise this is probably old news but it surprised me alot since I consider myself a fairly tech able guy,.

Both the following pics are screen grabs of 4K footage shot in D-Cinelike with the Contrast and Saturation set to -1. Only difference between them (aside from the 30 mins or so apart of taking the footage) is that the first pic was from footage shot with sharpness set to -1, the second with it set to 0. I've seen a number of recommendations that -1 for sharpness is good due to to the Mavic's tendency to over sharpen but it seems for me at least not to be the case. Shooting in D-Log with the same settings doesn't seem to make any difference by the way.

vlcsnap-error580 by Paul Jones, on Flickr

vlcsnap-error225 by Paul Jones, on Flickr

One thing I will clarify, is that in both cases the drone was hovering and I used centre focus and confirmed it was "in focus".

Conversely, the same settings in 2.7K cause the footage to be massively over sharpened and I have to use -1 on the sharpness for 2.7K.

Again, sorry for the waffle, I was just taken aback at how much difference the -1 made compared to 0.

Anyway, I'll wait to be told this is old news and it's been posted a million times haha


Paul.

 

[CyberNate]

So, I know this isn't a particularly new topic but there seems to be a ton of "Best Mavic video settings" tutorials around which often seem to contradict each other. Then a new firmware update appears and seems to nullify those.

Since I got my Mavic I've been playing around the settings, some recommend shooting in D-Log, some in D-Cinelike and there seems to be arguments for using either. I was finding that alot of my 4K footage was blurry yet anything shot in 2.7K seemed fine, so I ran a little experiment and I'm shocked how much difference it makes.

Again, I realise this is probably old news but it surprised me alot since I consider myself a fairly tech able guy,.

Both the following pics are screen grabs of 4K footage shot in D-Cinelike with the Contrast and Saturation set to -1. Only difference between them (aside from the 30 mins or so apart of taking the footage) is that the first pic was from footage shot with sharpness set to -1, the second with it set to 0. I've seen a number of recommendations that -1 for sharpness is good due to to the Mavic's tendency to over sharpen but it seems for me at least not to be the case. Shooting in D-Log with the same settings doesn't seem to make any difference by the way.

vlcsnap-error580 by Paul Jones, on Flickr

vlcsnap-error225 by Paul Jones, on Flickr

One thing I will clarify, is that in both cases the drone was hovering and I used centre focus and confirmed it was "in focus".

Conversely, the same settings in 2.7K cause the footage to be massively over sharpened and I have to use -1 on the sharpness for 2.7K.

Again, sorry for the waffle, I was just taken aback at how much difference the -1 made compared to 0.

Anyway, I'll wait to be told this is old news and it's been posted a million times haha


Paul.

You need to use +1 sharpness to turn off the horrible noise filter that makes trees look like water color paintings. Please check the existing threads that discuss this in detail. Yes, it will over sharpen, but you can soften in post processing. 2.7K looks pretty good with +1 sharpening.

 

[Ender]

The ugly part is that these behavious change from FW to FW, so forget discussion w/o FW versions.
-1 was a MUST in early FW versions but things change. Those stupidos should simply offer a mode w/o sharpening and w/o denoising.

Ender

 

[jeremybirn]

It looks like the big, important differences there are the sun and the exposure, not the sharpness tweak.

The top image is lit by soft light, as if the sun were hidden behind a cloud. You can see very soft shadow areas under the trees, but none of the shadows are very clearly defined. The second image is lit by direct sunlight, with sharp, well-defined shadows under every tree, and a lot more contrast as a result of that.

The exposure is also very different between them. Maybe you were using auto exposure and the drone chose the shutter speeds for you, but the first shot is over-exposed, which you can especially see in the sky. Instead of a blue sky with white clouds like the bottom image, in the top image even the blue parts of the sky look pretty white. If you're using center-weighted auto-exposure, the camera would adjust the exposure for you when the sun goes behind a cloud, and could end up overexposing the sky in those cases because of that. But with different lighting and different exposure, those are the big differences between the images.

If you want to see the results of a much more subtle little tweak like a softness change, try shooting images with the same lighting and same exposure both ways, and then look very closely at a full-resolution shot. It won't be nearly as dramatic a change as when you have different light and different exposure, but at least then you'll be able to see what that setting does all by itself.

 

[Viciam]

before firmware update to .900 i used -1 sharpness. after i had to use +1

 

[RedDevils88]

Sharpness has to be +1 or at least 0, - 1 kills the video

 

[Viciam]

Yes but it never used to be before the update to .900. On .800 -1 sharpness was the ideal setting.

 

[Michael W.]

Yes but it never used to be before the update to .900. On .800 -1 sharpness was the ideal setting.

WRONG

This has been covered ad nauseam in other threads. Going all the way back to the .200 firmware, -1 on sharpness was causing watercolor effects.

 

[Ender]

WRONG

This has been covered ad nauseam in other threads. Going all the way back to the .200 firmware, -1 on sharpness was causing watercolor effects.

I strongly disagree, but then perception is subjective but i did comparing shots with a Leica S wich is beyond doubt for me ;-)

Ender


P.S. i did that when i measured the Details comparison 2,7k vs 4k (that came out to the conclusion that 4k is useless). As i was measuring anyways i also did those comparing shots and worked far more init, just posted a subset here to make my point about 2,7k.
(A decision that i had to do scientific while most talented video-guys came to the conclusion by gut feeling and tehy were right ;-) As i am not talented i had to use science).


P.S. the watercolor effects are also way stronger on 4k as the compression is too harsh to deliver Details it often decides to mush the image. in 2,7k that happens much less as compression to pixel ratio is more favourable.

 

[Michael W.]

I strongly disagree, but then perception is subjective but i did comparing shots with a Leica S wich is beyond doubt for me ;-)

Ender

What part do you disagree with? That firmware before 0.900 could be used with a -1 sharpness without seeing the same affect as with firmware 0.900? Or, that -1 sharpness causes watercolor effects?

You do realize that 0.900 firmware just came out recently and here have been threads for months showing the watercolor effects that are solved by using sharpness greater than 0? Have you viewed those threads and seen the resulting photos? Just FYI, this issue doesn't exist with RAW images, only compressed images (still or video).

 

[Ender]

The latter, -1 is not always worse ;-)

 

[Viciam]

WRONG

This has been covered ad nauseam in other threads. Going all the way back to the .200 firmware, -1 on sharpness was causing watercolor effects.

Err not wrong because this is what was happening with my on .800 and .900 respectively

 

[Ender]

Of course they do not exist with raw images as they derive from very bad / harsh denosing, unsuitable sharpness and general carelessness concerning image information contents at DJI.
To add to this BS they tweak it every now and then. Maybe the recent hacks oüen up a way to tweak JPG parameters to good use where you dont want to toy with raw and to remove denoising & sharpening from video and give us back the higher bitrate for 2,7k that existed in early & pre release Firmwares.
But they have to sell that P4P somehow ;-) ( It HAS better sensor & optics, just not so dramatic as they want us to believe)

Ender

 

[fnf]

I've done a fair amount of research on the best video settings for the Mavic and concluded that 2.7K@30fps D-Log +1, 0, 0 is best in terms of sharpness. You can find proof of it in my post: Mavic Pro: why DJI doesn't fix this huge flickering/compression problem? .

There is some flickering issue with very complex scenes (i.e., lots of fine details and high contrast) that contain nuanced changes such as slow-moving shadows created by clouds in a bright sunny day or slow-moving ocean waves. As long as you are aware of that and reduce the sharpness intentionally you should find that +1 sharpness is the best in terms of giving you details which are unrecoverable in post.

Personally I do not shoot in bright sunny days often as only sunrise/sunset tend to produce outstanding footages so I only need to be aware of this when shooting near the sea.

 

[grizzard]

I've done a fair amount of research on the best video settings for the Mavic and concluded that 2.7K@30fps D-Log +1, 0, 0 is best in terms of sharpness. You can find proof of it in my post: Mavic Pro: why DJI doesn't fix this huge flickering/compression problem? .

There is some flickering issue with very complex scenes (i.e., lots of fine details and high contrast) that contain nuanced changes such as slow-moving shadows created by clouds in a bright sunny day or slow-moving ocean waves. As long as you are aware of that and reduce the sharpness intentionally you should find that +1 sharpness is the best in terms of giving you details which are unrecoverable in post.

Personally I do not shoot in bright sunny days often as only sunrise/sunset tend to produce outstanding footages so I only need to be aware of this when shooting near the sea.

You are right on. 3.8K isn't bad either if you use a very slow yaw. +1 is mandatory. Do a -3 and then work up to ,+1 ... And then compare. It will be obvious. Snapshots I don't know, but for video, absolutely if you want to reduce the blurry areas.

 

[Ender]

I've done a fair amount of research on the best video settings for the Mavic and concluded that 2.7K@30fps D-Log +1, 0, 0 is best in terms of sharpness. You can find proof of it in my post: Mavic Pro: why DJI doesn't fix this huge flickering/compression problem? .

There is some flickering issue with very complex scenes (i.e., lots of fine details and high contrast) that contain nuanced changes such as slow-moving shadows created by clouds in a bright sunny day or slow-moving ocean waves. As long as you are aware of that and reduce the sharpness intentionally you should find that +1 sharpness is the best in terms of giving you details which are unrecoverable in post.

Personally I do not shoot in bright sunny days often as only sunrise/sunset tend to produce outstanding footages so I only need to be aware of this when shooting near the sea.

I trust that you found YOUR best setting but again you are making the mistake of assuming that its the same "best value" for all FW versions & workflows. Even if it were for you, the usage of D-Log on 8 Bit Data is considered a "near crime" for many professionals.
That doesnt diminish your studies or your results but should offer Motivation for others to think & check by themselves.

Ender

 

[fnf]

I trust that you found YOUR best setting but again you are making the mistake of assuming that its the same "best value" for all FW versions & workflows. Even if it were for you, the usage of D-Log on 8 Bit Data is considered a "near crime" for many professionals.
That doesnt diminish your studies or your results but should offer Motivation for others to think & check by themselves.

Ender

I do agree that the 8-bit colour depth doesn't make it very malleable in post, not to mention that not everyone is keen on doing so. That said, my post does give a few (subjective) reasons why I went with 0 colour and 0 contrast which is to not make the 8-bit depth worse than it already is. Some people went with negative colour and contrast on D-Log which in my opinion destroys what's left of the remaining few bits for tonal values and causes worse bandings.

Regarding D-Log vs others, I found that D-Log gives the best dynamic range according to my tests in high-contrast scenes (sunset with unlit foreground).

I haven't tried earlier firmwares than 0.600, the settings that I arrived at were on 0.700 firmware.

 

[Cut.Aussie]

Can I suggest another way of assessing detail in a two different photos.
Look at the jpg file size. The smaller size file has less detail, the larger file more detail because less data was compressed.

My personal favourite is 2.8k @ +1

 

[Ender]

I do agree that the 8-bit colour depth doesn't make it very malleable in post, not to mention that not everyone is keen on doing so. That said, my post does give a few (subjective) reasons why I went with 0 colour and 0 contrast which is to not make the 8-bit depth worse than it already is. Some people went with negative colour and contrast on D-Log which in my opinion destroys what's left of the remaining few bits for tonal values and causes worse bandings.

Regarding D-Log vs others, I found that D-Log gives the best dynamic range according to my tests in high-contrast scenes (sunset with unlit foreground).

I haven't tried earlier firmwares than 0.600, the settings that I arrived at were on 0.700 firmware.

I can see !
The D-Log was *absolutely* unusable at early versions and the Parameters changed more drastically.
Due to this or that reason many (see polls) people are still using .200 or .400 and have to find their own set of perfect parameters or stick to the ones they found early.

Can I suggest another way of assessing detail in a two different photos.
Look at the jpg file size. The smaller size file has less detail, the larger file more detail because less data was compressed.

My personal favourite is 2.8k @ +1

You are right in general, but sadly this is also true if the processing introduces sharpening & Aliasing Artifacts which it does :-(

Greetings,

Ender

 

[doccy]

Arghhh - I have been using 2.8 @-1 after seeing a very compelling video arguing for this a while back.

Will need to go and reassess now with the latest firmware - thanks for the heads up.

If anyone has seen some good testing in a video online that is up to date drop me a link please - save me doing all the work!