Personal View site logo
GH3 rumors topic
  • 869 Replies sorted by
  • On the GH3 and the 50mbs and 72mbps I-frame. The way I read it is if the I-frame codec is H264 then that would be equivalent to prorez 144mbps as H264compression is said to be twice as efficient as DCT... does that sound right Vitaly?

  • @berniez

    So 8 bits gives you 8 stops... 10 bits gives you 10 stops.

    I know what you are getting at but I think this is relating two separate things. You can put any number of stops of luminance into however many bits you want. The only result that comes from using 8bits instead of 10,11,12 ... is that of greater quantization errors in the A/D conversion which perceptually looks like posterisation/banding etc.

    Sure it is semantic but at least I think you were referring to 'usable' DR.

    You are never going to get 12 stops dynamic range out of an 8 bit video system.

    We can and do get a usable equivalent of 12stops of dynamic range out of 8bit systems and sometimes it looks great like blu-ray video on a plasma screen.

    As endusers I'm not sure we have the intimate knowledge of the sensor to know how much more information is lost or retained by using 8bits vs 10bits vs 12bits for 8stops of light and what the perceptual differences are once the data is shoehorned into an 8bit container. Someone could do a synthetic test to see if it matches up with what camera makers actually are doing right now.

    I see on some forums that when it comes to the OMD most of the users are raving about it's jpeg images more than about it's RAW. I think that tells you something about how people perceptually rate the usability of an image. I can agree that it does look like the best jpeg to come out of a camera.

    Last point. In the GH2 and its direct competition there are two very significant steps of data conversion resulting in compression, distortion and loss. Data is mangled once at the sensor in an A/D conversion and then secondly at the AVCHD encoding stage. Of the two manglings I think the AVCHD stage is by far the worst. (edit) I overstated in that point. More correct that the AVCHD stage is the one they can more easily improve to our benefit. Hacks demonstrated this. The sensor remains an important target for improvement.(/edit)

  • I don't know that much about sensors. What I've described is just the video signal. Sure you can compress lots of latitude into a video signal like what happens with a raw workflow, but you are making choices and throwing away information in through this process. The misunderstanding is that some expect that you can recover the lost information... you can't. An 8 bit system is all you need for a fine end product but you must make your choices 'in camera'. The C300 is a fine example but you must make the choices shot by shot. I actually think this is a good thing because you are doing the creative act on set and your choices are more vital... both artistically and creatively. Having said that, my ideal is a 10 bit system with 2 stops of play and manageable data. Not too far in the future though the amount of data you collect will be irrelevant and we would have lost another skill to technology. I really want the GH3, I have GH1 running a 50mbs and have shot many broadcast shows with it... even did a video for the gurs from Black Sabbath/Deep Purple/Iron Maiden/Mettalica. I wish BMCC had M43 mount... that would've been game over for me.

  • This DOES not apply to foveon sensors @berniez? since there is no sum of color per pixel (mixing colors in bayer to make one pixel)

    Could you explain how does foven sensor will apply this maths?

    please use this:

    http://personal-view.com/talks/discussion/4504/understanding-dynamic-range-type-of-sensors-and-film.#Item_1

  • Thank you @berniez !!!!!!!, it has more sence now. hmmm....but upps , ive allredy made topic. Please vitalyi dont erase it.

    @berniez please feel free to share your info in topic.

  • OK, more detail. The video signal is linear and digital. Every time we add a bit we double the number of levels available. 8 bits is 256 levels. Of course luminance is made up of red+green+blue. So the best possible outcome with 8 bits is 256 levels of R + 256 levels of blue (Giving one extra doubling ie: 9 stops) + 256 levels of green which is an extra 1/2 stop. So the best possible outcome with 8 bits is 9.5 stops if you recorded RGB. But video doesn't record RGB, it records YUV which is luminance (Y) plus two color difference signals (U and V). So you begin with 256 levels of luminance plus or minus some color signal (it gets complicated because luminance is 68%G 22%R 10%B) So, and I'm talking video system here not raw recording, an 8 bit system gets you about 9 stops, a 10 bit system gets you 11 stops. Whatever processing you do before this cannot change this limitation on the video system signal... this is a mathematical certainty!

  • @berniez.

    Could you please explain more detail about the Dynamic range and the bits limit on a sensor. Also, what does vitaly means with only linear case or near linear. I want more detailed explanation, if can`t be done here, please make new thread about Dynamic range from basic to advance information.

    Thanks in advance.

  • @berniez

    It is more complicated as you are telling us about linear case. Only raw is stored in linear (or, to be more correct - near linear) form.

  • I just wanted to say something about dynamic range because it seems that it is misunderstood. The amount of dynamic range in a digital system is directly proportional to the number of bits or "doublings" you use. So 8 bits gives you 8 stops... 10 bits gives you 10 stops. You are never going to get 12 stops dynamic range out of an 8 bit video system. RAW gets 12 or 13 stops because 12 bits is recorded each for RGB (adding the 3 gives you another stop for luminance)

  • @apefos

    No need to repost stuff that had been posted already.

    Check first topic page from time to time also.

  • If these specs are true the GH3 will be a killer camera. The big question now is: how many fstops dynamic range?

  • One thing that I have heard many requests for is an ultra wide video aspect ratio. It would be cool to have a 2.35 video aspect ratio even it is still only 2K video. I can’t imagine that it would be that hard to do with the GH3’s multi-aspect ratio sensor.

    I know it probably wouldn’t keep the exact same field of view as the other aspect ratios. However, if it was anywhere close it might be handy for in camera framing.

  • any rumors on better 3d quality and easy working 3d videomode ?

  • @danyyyel

    you're right. 14 stop RAW D800 stills are very close to eye vision. Retina have 6-7 stop "static" range, brain change exposure multiply this per 10. On extreme contrast scenes, human vision can go to 19-20 stop.

    Of course I agree about great improvement going to 12 stop on GH3 video mode. Talking about "wide dynamic range" at this level is for me a classic marketing lie...

  • If we get 12 stops of DR from the GH3, I'd be pretty happy. If we got that plus a 10-bit log format I'd be ecstatic. I'm not holding my breath on that last one though. I can't wait until we get official specs from Panasonic. The suspense is killing me!

  • I have seen many theories about that. Some saying 20 stop others stating about 14-16. Some saying that it is because our brain continuously changes exposure etc. For me the 14 stop are really close from what I can see. Just some example with the Alexa here http://blog.localheropost.com/main/tag/sample-footage I don't know if I can see much more than that. The one where she is back to the sea with the sun not completely behind ( it is not a sun set) and capturing all the detail from her black hair and the clouds and reflection of the sky behind is impressive. I am sure that 9 people out of ten would have to wear sun glases to be comfortable looking in this direction. The most importnat for me is that that it will start to get a pain in the ass because you will have to grade so much every shot to get a good image.

  • @danyyyel "At 14 stop, you have excellence, I would call it human vision."

    Human vision was measured at 20 stop with 1 000 000/1 contrast ratio.

    For me, a good DR is better than 15 stops, 55 000/1 CR we can definitively play with in post.

    That's for me the next challenge. Beating the 16 stop barrier in video mode. THIS is the real definition of "wide dynamic range". :-)

  • Yes, choice is the good thing. In fact we are getting close to what I call good enough camera. You know, the exact contrary of the good enough that red boss and his followers were mocking. A good enough camera, would have large enough sensor (M43 is good enough) to be able to isolate your subject without him living in like suspended in a blurry world were he has no contact. At least 12 stop DR, at least good 1080p resolution, good codec like a high bit prores in 10 bit, good usable 800/1600 iso. It looks very close to the BMC in some ways and the rumored gh3 in others.

    Once we will reach that, the rest of the tech will be only bonuses but nothing vital. There are some big technological boundaries to low budget people that have arrived or are coming at an affordable cost. The first thing was HD digital, which gave us good enough resolution to show on the big screen. Then DOF with the dslr revolution four years ago. Now with black magic and hopefully gh3 DR and Codec at affordable price again.

  • The GH3 with XLR would be a dream for people on a small budget. With 50Mbit, that's pretty good quality video. For years, I had a camcorder with XLR. It was good, but not great. Sony NEX-VG20 doesn't suit me in this price range. If I had a BMCC, I would probably use Prores on it, and RAW only for special occasion. Choice is good. Wonderful times.

  • I don't know about anyone else, but I can't really afford anything over a couple thousand dollars or so. That culls the field quite a bit for me. So I'm really happy when more cameras come out at lower price ranges. The extreme top of my affordability list would be the BMC. After that you have Canon's lower end offerings and Panasonic's GH series. Not such a large list (for me). Scarlets, Sony's FS series, and the like are great cams, but they might as well not exist since I'll never be able to afford one. But the great thing is, as that tech moves forward it trickles down into lower end cameras that I can afford. : )

  • Do you think we are reaching a camera overload? Once there was only the 5D and 7D. Then GH-1. Now we have too many Canon cameras, Black Magic, Scarlet, GoPros, tons of Sony cams from a pocket NEX 5 and 7, Kineraw, Digital Bolex, Nikon D800, etc. Can the market support so many "filmmaking" cameras? None of these are perfect cameras, but I think we are coming to a technology convergence were image quality, 1080p, 2k or 4k is not a real issue. If XLR is addressed in the GH-3 even issues of sync sound in camera won't be a work around anymore. Even if you have the extra cash, who will be buying all these different camera options?

  • @mpgxsvcd I was just going to tell you that. Dpreview by some strange coincidence stopped testing camera DR in Rawwwwwwwwwww, to only include Jpeg at the same time that Nikon/pentax/sony based exmor sensor started to trash Canon. Since then I have lost any respect for there testing.

  • @Vitaliy_Kiselev

    I am not sure if Dpreview uses RAW files for the tests where no DR tools are applied. It isn't clear what they use for those tests. However, aren't we talking about movie Dynamic range here anyway? Last time I checked there were not any RAW options for movies on the GH2. JPG is the closest thing to the movie formats.

    In any case the GH2 doesn't outperform ANY camera they have tested for DR. That is not a good thing.

    As for the dynamic enhancers. Yes if they work and are as good as doing it in post I would certainly compare them as well as comparing the RAW output. The fact is that the GH2 DR tool really doesn't work well. It does work for some other cameras.

    In all likelihood what Panasonic means by "Wide dynamic range" is that they have improved the I.Dynamic feature.

This topic is closed.
← All Discussions