GH2 AF100 with a hacked GH2

dcloud said:
anybody using mac here?
id like to know if native FCP transcoding is similar to 5DtoRGB's
Click to expand...

Miles apart.

Native FCP Transcode looks like garbage in comparison to 5DtoRGB. It isn't just lifting the gamma, it's smoothing a ton of chroma issues out, balancing the image.

I didn't believe it, but I went back and forth between FCP L&T and 5DtoRGB a few times, even adjusted contrast to match. It's not the same image.

Remember, not all compressors are created equal, and not all compression schemes either.

ProRes from the Ninja doesn't look like ProRes from my computer.
 
More findings!

Comparing AF100 and GH2 at ISO 200, "properly" exposed for maximum appearance. AF100 was black balanced. Both cameras at Tungsten preset WB. I should also mention that all the GH2 files went through 5DtoRGB.

Same Nikon 35mm lens on both cameras. Aperture went between 2.8 and 4 allowing the brightest parts of skin to barely hit 75-80 IRE on all film modes.

Comparing image quality, skin tone, compression artifacts, noise and posterizing.

I bounced a 1K tungsten lamp off of the ceiling on full flood and made sure that it hit the back wall slightly to create that smooth gradation that we all know and love to look at the posterizing. You can see it on the upper left portion of the image against the white wall.

All GH2 images were given minor tweaks in Saturation to match the AF100 as best as possible. All Curve images were given the exact same curve to see how the codecs respond to increasing the contrast in the image.

AF100:
Detail and V Detail at -6
Chroma -5
CineD gamma
Norm 2 matrix

GH2:
Quantum V2 patch
Nostalgic
Cinema
Smooth
All at -2 settings.

( To go back and forth between images for best comparisons, right click and create a new tab so you can tab back and forth to pixel peep lol ) Or just click on the link underneath the image as imageshack doesn't allow full rez postings ( or maybe that's DVX )

AF100

Clean:

af100clean.jpg


link: http://img716.imageshack.us/img716/7433/af100clean.jpg

Curved:

af100curve.jpg


link: http://img13.imageshack.us/img13/5527/af100curve.jpg

Clean Crop:

af100cleancrop.jpg


link: http://img815.imageshack.us/img815/5364/af100cleancrop.jpg

Curve Crop:

af100curvecrop.jpg


link: http://img546.imageshack.us/img546/7316/af100curvecrop.jpg



GH2

Smooth Clean:

gh2smoothclean.jpg


link: http://img855.imageshack.us/img855/4170/gh2smoothclean.jpg

Smooth Curve:

gh2smoothcurve.jpg


link: http://img694.imageshack.us/img694/3055/gh2smoothcurve.jpg

Smooth Clean Crop:

gh2smoothcleancrop.jpg


link: http://img221.imageshack.us/img221/2553/gh2smoothcleancrop.jpg

Smooth Curve Crop:

gh2smoothcurvecrop.jpg


link: http://img267.imageshack.us/img267/6555/gh2smoothcurvecrop.jpg


Nostalgic Clean:

gh2nostalgicclean.jpg


link: http://img829.imageshack.us/img829/2269/gh2nostalgicclean.jpg

Nostalgic Curve:

gh2nostalgiccurve.jpg


link: http://img196.imageshack.us/img196/135/gh2nostalgiccurve.jpg

Nostalgic Clean Crop:

gh2nostalgiccleancrop.jpg


link: http://img714.imageshack.us/img714/7761/gh2nostalgiccleancrop.jpg

Nostalgic Curve Crop:

gh2nostalgiccurvecrop.jpg


link: http://img542.imageshack.us/img542/9629/gh2nostalgiccurvecrop.jpg


Cinema Clean:

gh2cinemaclean.jpg


link: http://img32.imageshack.us/img32/2732/gh2cinemaclean.jpg

Cinema Curve:

gh2cinemacurve.jpg


link: http://img834.imageshack.us/img834/3018/gh2cinemacurve.jpg

Cinema Clean Crop:

gh2cinemacleancrop.jpg


link: http://img220.imageshack.us/img220/6003/gh2cinemacleancrop.jpg

Cinema Curve Crop:

gh2cinemacurvecrop.jpg


link: http://img151.imageshack.us/img151/2272/gh2cinemacurvecrop.jpg
 
Last edited:
Kholi said:
Miles apart.

Native FCP Transcode looks like garbage in comparison to 5DtoRGB. It isn't just lifting the gamma, it's smoothing a ton of chroma issues out, balancing the image.

I didn't believe it, but I went back and forth between FCP L&T and 5DtoRGB a few times, even adjusted contrast to match. It's not the same image.

Remember, not all compressors are created equal, and not all compression schemes either.

ProRes from the Ninja doesn't look like ProRes from my computer.
Click to expand...
wow that could also be another factor on these cameras tests on why af100 looks video-y to some.
 
PaPa said:
More findings!

Comparing AF100 and GH2 at ISO 200, "properly" exposed for maximum appearance. AF100 was black balanced. Both cameras at Tungsten preset WB.

Comparing image quality, skin tone, compression artifacts, noise and posterizing.
Click to expand...

Nice. 200 ISO indoors is lot of darn light!

Post uuupp.

dcloud said:
wow that could also be another factor on these cameras tests on why af100 looks video-y to some.
Click to expand...

I would be interested in seeing some AF100 5DtoRGB footage, but my issue with AF100 footage was never really color (although it's not pleasant there, either), it's been how it resolves detail, or the lack there of. I dunno what the camera's doing to the image, but it is rarely ever pleasant to look at. However, again, I'd be interested in seeing what happens when you 5DtoRGB the Af100 footage.
 
It appears as though nostalgic has the highest level of noise which helps break up the banding but only very very slightly.
 
I find it so sad that the GH2 gives a more resolved image than the AF-100 lol. I swear the cam is a little beast. Good test. I'm finding cinema mode is best for indoor/lowlight or situations where you want to intentionally under expose, smooth is good for outdoor because it's the best at perserving highlights (although I don't understand how it does this better than cinema because it seems to be more contrasty) and nostalgia is ok but the over warming of the image kills it sometimes for me. What are your thoughts on film modes?
 
PaPa said:
It appears as though nostalgic has the highest level of noise which helps break up the banding but only very very slightly.
Click to expand...

Man, thanks for doing something with an actual human in it. Greatly appreciated.

Which one do you prefer right now? Nostalgic or Smooth? It looks like Nostalgic's grabbing a tiny bit more in the highlights, and opening up the lows a smidge. Is it worth it in exchange for the noise?

I'm curious as to what either of the profiles do when you hit say 800 or 1000 ISO, as well.
 
You're welcome! Took about an hour lol.

I don't find tests of objects and flowers very useful, although they may be fun to watch. The only thing i care about are humans because quite frankly, when i watch something, that's the only real point of reference i have to tell how natural things look.

Man, i'm having a difficult time figuring out whicih one i like best.

Without further testing, i can almost say with 100% certainty, that if we raised the ISO, the added gain would definitely help dither the posterizing. But with the expense of much more added noise because of the low compression.

I don't know which way to think? Either both of these cameras are very noisy, or the stock af100 codec does a hell of a job compared to something of much higher bitrate. lol.

I am up in the air as to which mode i like hte best. I find that there is something about smooth and nostalgic that makes the lighter values too bright. Cinema flesh tones are generally, from my experience, darker and more rich while stil appearing smooth which we get a bit more here with the Cinema film mode, but at the expense of a little more posterizing and being weary of overexposing highlights.

hmmmm. much thinking to do.

And now that i look back on these images, I'm wishing that i took one of my little 150 fresnels and just spotted it on the wall somewhere to make a blown area as well, would have been good added information.
 
Also, weird, with 5DtoRGB Cinema mode looks a lot more pleasing, as well. Dunno why I didn't try all of this, just didn't even cross my mind that the program would change how you work with the film modes.

Good stuff.

From running around with the GH2 in Smooth, my conclusion was under-exposing anything important more than a quarter (maybe half) was not a good thing. Overexposing seemed to be the best way to go at it. If I didn't have go get ready to go out of town I'd go and do this stuff m'self. Grrrr~
 
Hopefully some find this useful and continue to look into these types of tests.
I want to try something similar with a lower bitrate so that i get more than 16 minutes out of a 16 gig card lol.
 
PaPa said:
Hopefully some find this useful and continue to look into these types of tests.
I want to try something similar with a lower bitrate so that i get more than 16 minutes out of a 16 gig card lol.
Click to expand...

Oh now that you mention the hack and bitrate, did you find a difference between the Kae patch and the Quantum patch? I noticed a subtle enough one in noise pattern to warrant the switch, but I had to go back to a previous location to test it out. Dirty test.

If I run across a patch that'll get you what you want I'll link you to it.
 
GREAT information here Papa and Kholi! This level of information is what makes DVXUser one of the best indie resources in the world! :) Thanks for sharing!
 
Oh and for those grabbing 5DtoRGB right now, Thomas Worth (Rarevision) said that he's removed the Gamma option from the program entirely, but I didn't get a word back if it's now defaulted to a 1.8.

I'm still using the prior version of 5DtoRGB, but will grab the newest one when I get to the office and see what it's doing.
 
Carp!

I was just about to ask, where this option might be...damned.

I have the following options in my brand new 5DtoRGB now:

Output Format:
- Uncompressed 10bit 422 (v210)
- Apple Prores 422 (and HQ, LT and Proxy)
- Avid DNxHD

Decoding Matrix:
- ITU-R BT.709
- ITU-R BT.601

Luminance Range:
- Full Range
- Broadcast Range

Postprocessing:
- None
- Technicolor Cinestyle

Start Timecode


Anything new for old version users? And any suggestions?

Thanks!

Edit: I forgot Decoding Matrix, updated the post.
 
Ok, besides the gamma my first findings are devastating for NeoScene (I still have it for my GH1 and that´s why I´m interested in its (dis)advantages:

Even though there´s no real visual difference detectable on 100% and in RGB, when zoomed-in there´s quite heavy jagged edging on the red channel.
That doesn´t occur both on the original MTS and the 5DtoRGB (uncompressed). I am not using the latest version of NeoScene as it would have cost me to upgrade.
In fact I couldn´t detect any difference between MTS and 5DtoRGB, on neither the red channel or RGB.

The reason I was testing the red channel was because of what´s on Rarevisions website. But I guess the GH2 simply has no issue with reds but Canons do?

So, without the gamma option I can just stick to my mts files. :undecided
 
Ha! I´ve found the old version I used quite a while ago in June. It´s v1.5.3 (32 Bit sadly) but it has the gamma setting.

And I tested but the file comes out not the tiniest bit flatter than the original. I am doing something wrong, I´m sure of it. But what?

I used: Uncompressed 10but 422, BT.709, Gamma 1,8

Is that version too old maybe and the gamma adjustment wasn´t properly implemented?
 
You must log in or register to reply here.
Back
Top