This is probably a dumb question, but... Panasonic says the GH4 can output 4k 30p 10 bit 4:2:2 via the micro HDMI, ATOMOS SHOGUN just posted a video claiming that they can record it from the micro HDMI and Convergent Designs says the HDMI 1.4 spec (used by SHOGUN? and Odyssey 7Q+) doesn't support 4k 10 bit 4:2:2. How would you tell that the ATOMOS SHOGUN recorded file (on PRORES HQ I believe) was actually recorded with 10 bits of data and actually recorded with 4:2:2 color data? Is there something that would show up in an NLE that would confirm it? I just find it interesting that there actually is controversy as to whether this could work and wondering how it could be confirmed objectively?
GH4 Dumb Question re: 10 bit 4:2:2 Output
- Thread starter BobPilz
- Start date
no -- that does not help. you can allways pack 8bit data in 10bit file! analysing a line of video data from a unsharp shot or some slightly gradation in color (e.g. sky) would be more usefull. but it's not that trivial as you may find compression artifacts and various other image processing effects in this data too.
Bern Caughey
Veteran
FYI supposedly the 7Q+ will record 10-bit over HDMI.
http://www.dvxuser.com/V6/showthrea...ipping&p=1986504133&viewfull=1#post1986504133
http://www.dvxuser.com/V6/showthrea...ipping&p=1986504133&viewfull=1#post1986504133
Mitch Gross
Veteran
Yes it will. We have expanded the spec of our device to include it.
That's a very positive development that everyone's now in agreement that 4K 10 bit 4:2:2 over HDMI is workable. wwjd, I'm with you. I'd also like to see some comparisons. I saw one earlier this year on the web that suggested there is a useful improvement, but hopefully, now that the Shogun and 7Q+ are shipping, we'll see some fresh apples-to-apples comparisons soon.
From a theoretical standpoint I'm still unclear how you would check a video file to determine that the data really encompassed 10 bits and 4:2:2 color information. I suppose I'd try a blue sky gradient and see if a high degree of grading produced less posterization with 10 bit vs 8 bit and maybe checking hard edged colors to see if the edges were cleaner with 4:2:2 vs 4:2:0 recordings.
From a theoretical standpoint I'm still unclear how you would check a video file to determine that the data really encompassed 10 bits and 4:2:2 color information. I suppose I'd try a blue sky gradient and see if a high degree of grading produced less posterization with 10 bit vs 8 bit and maybe checking hard edged colors to see if the edges were cleaner with 4:2:2 vs 4:2:0 recordings.
GH4 100% sensor readout is unperfect even in RAW photos so it will be very hard to see the difference of 4k 4:2:0 or 4:2:2. 10/8 bit is easier to see.
I tested GH4 4k video vs JPG vs RAW:
http://www.dpreview.com/forums/post/54879969
I tested GH4 4k video vs JPG vs RAW:
http://www.dpreview.com/forums/post/54879969
which software/algorithms did you use for RAW processing? did you use mechanical lenses or electronic coupled lumix ones?
Raw was made with Adobe Camera Raw. The lens is not affecting sensor color quality. It is the debayering, there are different amount of sub pixels for 3 colors (RGB).
that should give quite represetative results. i don't use adobe products, but on linux most applications let you choose between different debayering algorithms. there are fast ones around and very time consuming high quality alternatives (e.g. "AMaZE"). they all lead to slightly different results.
yes -- but lumix cameras use this information from electronic lenses to do a lot of correction on the images resp. adding metadata to the files. if you wante a clean pixel corresponing RAW and JPG pair, you have to use legacy glas on mechanical adapters, to avoid all this influences.
the GH4 bayer pattern is a very common one. it uses a slightly different order of color fields in readout compared to previous models, but otherwise it doesn't show any uncommon attributes.
Last edited:
That is the point. GH4 sensor is OK but when using 100% bayer sensor the result is far from perfect. If down sampling the sensor like in Samsung NX1 (6k to 3.8k) there may be very good color resolution and pureness if the scaler is good.
My tests just shows that JPG photo (I think it is 8 bit 4:2:2) and 4k video 4:2:0 are very close in color resolution and look.
for use in video it does not matter, because the chroma information is not needed in full resolution in most cases. higher pixel counts and down sampling may have some benefits, to combine maximized image resolution for stills with acceptable video quality. but it does not come without drawbacks. the noise ratio resp. usable sensivity does not like this kind of pixel inflation.
yes -- stills from GH4 4K video recordings look unexpectedly good. (take a look at http://bellard.org/bpg/ to see how the next generation of video compression will influence still photography again)
i don't think the sensor is the limitting factor. it has so much dynamic range in comparison to previous models, that flat 8bit recording profiles can not be seen as very adequate anymore. but im also very sceptical, if 10bit HDMI output will give us any better results. we don't know, if the internal processing is optimized for it. there is a lot of internal image processing going on (the shift of noise from chorma to luma is a very obvious indicator for this kind of hidden magic) that can not be compensated later on the signal processing path. all this inefficent hacks to get more digital range out of 8bit 4k recordings during the down scaling process are typical for this misconception. you never get more precission out of a calculation using a limitted precission of input values. so i'm not very optimistic about the real benefits in image quality of this external recording solutions. getting raw video data or a more advanced internal recording would be much more promising. but the actual quality of GH4 4K files aren't that bad at all. if you use them as source for nowadays common FHD production, they are more than sufficient in most respects. external recording will be of great value to bypass synchonisation issues and the 30min limit of the european product -- even though a quite expensive workaround for simple features/issues.
Last edited:
tflak
Veteran
When you drop 4K into a 1080 timeline in your NLE (scaled to fit), you get the benefit of 10-bit and 4:4:4 color sampling for color correction anyway. So unless you're editing and delivering in 4K there may be little benefit to using an external recorder in that regard.