GH4 Settings for maximum DR (shootout tonight!!)

OK, so that is CineD with -2 contrast and iDynamic off for all the tests except the DR one, which will be -5 contrast and not-sure-yet iDynamic.

I'm a DR nuthead, and the guy I'm shooting this for is even worse...

If we are talking now about getting a usable picture with as much DR as possible then my recommendation is going to change also. Do not shoot with saturation all the way down to -5 !!! Adding saturation to deprived low bit codecs in post is one off the things that kills it the most! I wouldn't go down more then -2 with saturation. When shooting skin just be careful to underexpose a bit. Skin should be around 40 IRE when shooting flat with GH4.

As you've seen from my charts (probably) curves add a lot of noise, more so than lowering contrast. BUT I have used them with success, so it's not all bad. You will "peep" into the shadows a bit more, but if you use something like NetVideo, you can get away with it pretty easily. So I would then recommend:

Cine D
Contrast -3
Curves -2, +2
Saturation -2 (if there are no reds in the shot u can go higher with sat)
All else 0

But still, I don't see a reason why limit yourself... Just go for the max DR and the crush the blacks untill you feel comfortable with the noise level. This way you'll see what the sensor is actually capable of and then decide for your self how much of that is usable.

Now on the other cameras you also have to produce a picture profile that isn't maxed out flat, but is something between usable and flat...
 
^ I will use Flaat on the 5D3, D5200 and a7s. That is never overly desaturated. So yes, I will avoir -5 saturation on the GH4 too.

The problem with going for max DR in camera and then crushing what's unusable is that some settings create an awful light-response function: wiggly, very far from log. That's a nightmare to grade. I will go as flat as wide as I can without getting into those issues.
 
If we are talking now about getting a usable picture with as much DR as possible then my recommendation is going to change also. Do not shoot with saturation all the way down to -5 !!! Adding saturation to deprived low bit codecs in post is one off the things that kills it the most! I wouldn't go down more then -2 with saturation. When shooting skin just be careful to underexpose a bit. Skin should be around 40 IRE when shooting flat with GH4.

As you've seen from my charts (probably) curves add a lot of noise, more so than lowering contrast. BUT I have used them with success, so it's not all bad. You will "peep" into the shadows a bit more, but if you use something like NetVideo, you can get away with it pretty easily. So I would then recommend:

Cine D
Contrast -3
Curves -2, +2
Saturation -2 (if there are no reds in the shot u can go higher with sat)
All else 0

But still, I don't see a reason why limit yourself... Just go for the max DR and the crush the blacks untill you feel comfortable with the noise level. This way you'll see what the sensor is actually capable of and then decide for your self how much of that is usable.

Now on the other cameras you also have to produce a picture profile that isn't maxed out flat, but is something between usable and flat...


KINOKS, have you noted whether pulling highlights DOWN in curves adds any noise while recovering the blown out stuff? Say like -3 or -5, leaving everything else normal?
 
KINOKS, have you noted whether pulling highlights DOWN in curves adds any noise while recovering the blown out stuff? Say like -3 or -5, leaving everything else normal?

I don't know to be honest, but I presume it doesn't recover much but I am under the impression it saves some... What I know it does is it stretches down the highlights (especially the last stop) and compresses the midtones (in my test there were 4 stops between 25% and 50% when using curves, that's one stop more than just dialling down contrast) - but I don't know if the compression of midtones is a function of shadows or highlights curve or both. Probably both... I also don't know how GH4 is allocating bit data to the curve. If it's allocating the most to midtones than this is a good thing. I know it doesn't allocate a lot to the highs, as there are lots of problems with encoding in the upper part of the gamma. This is also however a characteristic sign of 8 bit compression that wasn't designed for a flat log like gamma. Canon C series have this much more controlled than let's GH4.

However I think that most of the DR is "saved" by lifting the shadows and then using noise reduction (primarily in the shadows) and not pulling down the highs. I presume that pulling down do highs gives a more pleasing highlight rolloff and it definitely pushes down middle grey.

I will test if highlights curve gives us an advantage. If it only stretches the last stop and a half it makes no sense to use it as that information doesn't need all that bitrate that you are effectively allocating it by using the higs curve.
 
Last edited:
^ I will use Flaat on the 5D3, D5200 and a7s. That is never overly desaturated. So yes, I will avoir -5 saturation on the GH4 too.

The problem with going for max DR in camera and then crushing what's unusable is that some settings create an awful light-response function: wiggly, very far from log. That's a nightmare to grade. I will go as flat as wide as I can without getting into those issues.

I did my own testing with the GH4 and found that getting the color right was more important to me than the DR. I use natural pretty exclusively. The DR doesn't seem to be lower to my eye (all I really care about) than cine-D and the colors are good out of the box and gradable.

Might I suggest repeating your test with one of the standard profiles, then repeating it with a flat one? I know it's more work, but it would mitigate people accusing you of not having the right settings. Thanks for doing this test; looking forward to the results!
 
I'll have to talk about that with the GH4 owner. If his preferred settings are like what people are suggesting here (cineD with contrast -2), that's what I will test (with -5 for the DR test). If he likes something different (like what you mentioned) then we'll have to see. Limited time (around two hours) and many things to test (8 cameras, some of them shooting RAW and standard, may shooting several frame rates; and tests for sharpness, DR, aliasing/moire, color, etc). I'll have everything ready before we start shooting but it's going to be tight.
 
Make sure you lower sharpness and noise reduction to a minimum (-5). This is absolutely necessary to make the GH4 not look like cheap video. ;)

And don't ever use iDynamic, iResolution as well as highlight/shadow adjustments. These settings will introduce a lot of blocky noise and make skin look like crap.

Using cinelikeD using contrast -2 is the best compromise for noise, dynamic range and good looking skin. Make sure you set saturation to -2 or even lower to prevent blown out reds.
 
Last edited:
Make sure you lower sharpness and noise reduction to a minimum (-5). This is absolutely necessary to make the GH4 not look like cheap video. ;)

And don't ever use iDynamic, iResolution as well as highlight/shadow adjustments. These settings will introduce a lot of blocky noise and make skin look like crap.

Using cinelikeD using contrast -2 is the best compromise for noise, dynamic range and good looking skin. Make sure you set saturation to -2 or even lower to prevent blown out reds.

Agreed about the sharpness. Turn that thing off. Film has no such thing as sharpness and neither is there such a thing when one shoots RAW. Sharpness is a byproduct of video cameras and a way to make the RAW data from the sensor look more punchy. Sensors inherently are always going to be soft just due to debayering and low pass filtering. For those that think -5 sharpness looks too soft that is actually what your sensor is seeing. Well at least much closer to what the sensor is actually seeing. I'm still not sure if -5 sharpness equals completely turned off.

I also believe sharpness negatively affects motion cadence because when objects move they should have a certain level of motion blur. By adding edge enhancement you are effectively telling the camera to remove the more subtle amounts of edge motion blur. Motion blur is all about the subtle gradation from your subject to the background. By messing with that you are messing with the natural movement we expect to see. Our eyes do not sharpen the edges of what we see moving and add ringing black or white halos on those edges. I really do think this is why some people feel as though digital 24p doesn't always look the same as 24p from other more professional cameras. I think it is the use of too much edge enhancement.
 
because it is better that we go for emulating 135 year old blurry technology instead of embracing the much improved clarity and detail that we have in our new cameras from 2014? :D
 
because it is better that we go for emulating 135 year old blurry technology instead of embracing the much improved clarity and detail that we have in our new cameras from 2014? :D

It isn't about that at all. Sensors are sensors and have to build an image with the way they do. Edge enhancement is just that, edge enhancement and not what the sensors actually capture. In fact it is the opposite. We are trying to emulate higher detail by using sharpness and not using the natural raw detail that our 2014 cameras are actually capable of. Sharpness is an optical illusion of detail and not the real detail resolving capability of our 2014 cameras.
 
k, in my tests -4 and -5 sharp made it blurrier than -3 and -2. why make it fuzzy? I don't want sharpness haloing but I don't want extra softening either. But, yeah, back it off so it is not adding fake sharp.
 
Make sure you lower sharpness and noise reduction to a minimum (-5). This is absolutely necessary to make the GH4 not look like cheap video. ;)

And don't ever use iDynamic, iResolution as well as highlight/shadow adjustments. These settings will introduce a lot of blocky noise and make skin look like crap.

Using cinelikeD using contrast -2 is the best compromise for noise, dynamic range and good looking skin. Make sure you set saturation to -2 or even lower to prevent blown out reds.

Are we sure the NR in this camera is a bad thing? Never got a difinitive answer on that and don't have my GH4 with me to test.
 
k, in my tests -4 and -5 sharp made it blurrier than -3 and -2. why make it fuzzy? I don't want sharpness haloing but I don't want extra softening either. But, yeah, back it off so it is not adding fake sharp.

The camera cpu doesn't go out of its way to add a blur filter to the video. This is actually what the sensor is able to resolve. There is added sharpness even at -3 and -2. Sure you may not notice it as much but it is still there. Look it is up to you to determine what level of sharpness you find acceptable. The fact is however that if you want the cleanest most RAW image the sensor can produce then you need to go as low as possible with the sharpness. That is the natural look of the sensor.
 
I will shoot with minimum sharpness, because it's easy to add sharpness in post but taking artifacts away is not. In any case, this thing is probably already "too sharp" to begin with!! I have no doubt it will be the winner in the resolution test, what I'm worried about is DR.
 
Please test another profile than cinelike D (or as well as). People always seem to harp on about it, but all the videos I have seen with cine d the colour tones are goofy. As you will well know, a flat image doesn't always mean more range. If it clips at the same point there is no extra latitude and if the master pedestal doesn't lift middle grey then it doesn't alow you to under expose either. You are just raising the blacks to crush them again. I guess you could call it fake flattening. Thats what cinelike D looks like to me.
 
Last edited:
settings for low light and low noise

settings for low light and low noise

Sorry to barge into this thread with a different (but related) topic. I am shooting soon in a low light nighttime setting and wanted to see if I could generate a consensus, or at least veer towards a consensus, on the settings that I should use. I don't mind grading the footage in post, and am well versed in that, but am thinking I should get as close as I can to usable out of the camera footage so as not to jump up the noise in post.

I have pretty fast lenses (one is f1.7, the other f1.4) so feel like I am covered well in that department.

I am planning on shooting with these settings -

Cinelike-D

all -5 except saturation at -2 and hue at 0.

Thanks to all for taking the time to help out a new owner with thoughts on best approach.

Sam
 
Nighttime typically presents less challenges to dynamic range then direct Sunlight.

Assuming your pushing the ISO to 1600, or so, consider raising your contrast to at least -3 or even as high as '0'. You won't capture as much dynamic range as -5 but the output will be cleaner.

Also raise your Noise Reduction setting into a similar range.
 
Thanks for the feedback.

I'll be trying my best to keep ISO to 800 at highest as in my experience anything above is quite noisy.
Also since I did not point it out before I'll be keeping master pedestal at 0 and shooting 0-255. That is the plan anyway.
 
Back
Top