Personal View site logo
GH2 in 24p: noticeable "strobe effect" with any motion in frame?
  • 259 Replies sorted by
  • The screen size matters, too. I currently sit 1.5 meters away from my Vieira 42" HD plugged into my PC via HDMI (for working) but it's interesting how small juddering movements become noticeable when I watch video from this close.

  • Hi All!

    I just saw @davidhjlindberg nicely shot shorty "The Final Dance". I downloaded the video and watched on my edit. It's also good example of strobing shots. I'm not sure what shutter speeds David used in this, and my opinion is that specially in war scenes strobing works great - aka style of "Saving Private Ryan".

    Go and see and it

    But as @rajamalik commented that there is no strobing effect in latest hacks and many of You are saying that there is no eye-bothering strobing effects in material - and - testers don't see any problem how hacked GH2 captures this "cinematic motion": You must be right after all. Probably I should just forget it and shoot. = )

    So this is my last try. =D @bwhitz @Roberto @mo7ies @LPowell @driftwood I did a test shootings with sedna A, crossfire variation 1 and Lpowell's latest Flow Motion v2.02. I did not have OIS on and I used Kit Lens 14-42 and Voigtländers NOKTON 40mm. I shot using same settings that I used shooting those flowers earlier on. And I see there is differences. LPowells hack seems to have more motion blur during handheld camera movement than others. After I exported 422(HQ) from NLE I conformed all three videos from 23.98fps to 25fps and compared again. LPowell's version passed without any strobing.

    I like most SEDNA A in picture quality, but Flow Motion is better capturing motion. In a slower camera movements, sedna did not do almost any motion blur. When you make a zip-pan blur is there, but in a slower movements its notn

  • @MisterJ
    How is it possible that "Sedna Intra ' imitates motion worse than" Flow Motion " ?

  • Possibly because it 'is' Intra. Flow motion and all longer GOP settings use P and B frames which maybe making the blend from frame to frame less obvious. Pure theory mind.

  • @itimjim can be on right track. But here my skills ends. I'm just a novice considering hacks.

  • @Misterj Thanks for the props on Flow Motion v2. One of my goals was to see if I could improve frame-to-frame motion picture continuity by boosting B-frame image quality to match the quantization quality of the I-frames. In an FM2 1080p video, there are two B-frames between each pair of I-frames. The macroblocks in each B-frame are formed from reference blocks taken from both previous and following I-frames, averaged together. This inter-frame compression process is similar to techniques used to generate "tweened" frames between the key frames in animated videos.

    In FM2, the theory is that high-quality B-frames can effectively form a bridge between consecutive I-frames, by blending object details from both I-frames together. (Of course, B-frames are not simply averaged pixels, the details in each frame are individually refined to closely approximate the details captured by the image sensor for that frame.) My hunch is that when viewing moving objects in a video, your eyes make every effort to integrate the consecutive frames into a continuous stream, and that random frame-to-frame discrepancies are perceived as undesirable distractions. Hence, the disturbing judder we see in some videos, that mysteriously vanishes when you step through the video frame-by-frame.

    In an intra-frame motion picture, whether film or video, each frame is individually captured as a self-contained still image, with no inherent connectivity to any other frame. Frame-to-frame continuity is produced only by the coincidence of having captured consecutive frames at very short time intervals. At 24p, the frame rate interval is just barely short enough to maintain the illusion of continuity, and it can easily be dispelled by any perceived irregularity. I personally find the mismatch between the 60fps refresh rate of LCD monitors to clash noticeably with a 24p video cadence, whether interlaced or progressively displayed. And that's the main reason why in practice, I prefer 30p to 24p.

  • LOL... This is a topic on every forum on every video camera or DSLR capable of video. There's nothing magical or unmagical about the frames.
    They are 24FPS.

    Now having said that, if anything, the GH2 now with the capability of producing all i-frames (thanks to Vitaliy and crew) now produces clean 24P motion.

    The motion is better than my Sony EX1 which is long GOP. Of course, the EX1 motion is fine when running SDI out to intraframe..

  • After reading all of this, even myself have recorded on 24p 1/50th shutter speed, the motion blur didnt look smooth as you would get with a perfect synced shutter speed with the recording framerate, and it does look abit unattractive.

    You can it is most noticable at 29secs in this video:

    Therefore, hasnt anyone found a way to at least implement a 1/48 Shutter Speed setting? like 1 step/level/click away from 1/50???

  • @Roberto:

    "@svart

    when I went frame by frame, I could see strobing as slight changes in color.
    

    I know you were trying to replicate this, but any chance you can post stills of this? Or was the colour change only perceived in motion? "

    Sorry, I've been traveling a lot for work lately. I'll try to replicate and post frames once I have some time. I could see this problem when going frame by frame. Some say that it's possibly the noise changing shape/grain but it was pretty clear that I saw yellow-ish tint wave through the normal blue/red noise. I also saw it in both long and short GOP patches. It was far less apparent in I-frame only patches and was more apparent when panning. In either case I only saw it when the majority of the frame was slightly underexposed and had large areas of single colors. When I exposed normally, or overexposed I didn't see it much at all. My test was looking at a black comforter illuminated by a tungsten spot light creating a grey-ish look.

    Also, this means I'm not convinced the strobing that some people are seeing is actually strobing. Most of the testing and evidence has shown that the camera has the correct amount of motion blur for 24P. I think it's an artifact of the compression process where some of the I, B or P frames are not encoded the same way or have something strange about others around them. It makes me think that the prediction stuff is not working quite right in the GH2, especially since I-frame only patches seem to look much better than others in this regard.

  • Wow...The thread came back to life....LOL!!

    I am one of the ones that had complained earlier on this thread about excessive strobing (even using the latest hacks) I will be happy to post some of my earlier clips showing small sections of the original MTS files for anyone who says they have never seen this...or its totally normal, or curiously that no one ever posts any examples etc...

    For me tho I have noticed it mainly panning the camera when there is higher contrast...black letters on a white or yellow background, trees against a blue sky...that kind of thing.

    Anyway for me the Driftwood all Intra or Flowmotion 2 hacks produce such a beautiful image that I am very happy with, so I just use workarounds for 24P...(shoot on overcast days, dont pan fast....or not at all) keep human motion so its not too extreme) etc..

    I really think if you plan to pan the camera a lot or are shooting sports its probably better to use 30P frame rate...as the problem is not really there, I have done countless tests on this BTW...at all relevant shutter speeds in 24P, and using several hacks as well. Also as LPowell pointed out 30P probably has a more fluid feeling on 60hz monitors, so its less troublesome overall and thats a fact.

    Bad juddering with the camera in 24P almost looks like the object is jumping back and forth slightly as the camera pans past (even quite slowly), its easy to see and very easy in higher contrast footage.

    But for me...I have more or less put it to rest, I know what the weaknesses are with 24P and I just work around them. I have never used a Red, but I have studied a lot of Blu Rays that were shot at 24P on the Red...and yes it is there too, but I would say its definitely less. As I mentioned earlier I saw it on Ghost Rider spirit of vengeance when the camera was panned over hi contrast rocks (similar to the GH2, but not quite as bad)....at 1hour 6mins....and other films too.

    Personally tho...I have ended up using the all I frame hacks for 24P, mainly cause I love the detail and Flowmotion for more movement on 24P...and 30P for everything else. You just handle the camera very very carefully in 24P (use shoulder rig) and so on. Cheers

  • Thanks for the info and great tips @LPowell and @Astro!

    That is exactly so as Astro mentioned. Intra-hacks are great in detail and detail can take you a long way. Myself I shoot a lot of static shots and there detail is the key feature. But it's good to have different kinds of approaches what hack is trying to do. And there I think LPowell has done a great task.

    And @Driftwood too. Speaking of witch I've done some testing with Driftwood's latest Cluster 6 and it's really promising. Best thing in that hack is that you get really great quality 2k mjpeg so it is easy to jump from 24H to mjpeg 30fps when you are shooting handheld(I always use shoulder rig). With that setting I can shoot in situation where I need to pan from person to another and so on. And that 24H looks just great with modest datarate. Great setting for documentary filmmaker or such. As Flowmotion is too.

    Well monitoring material is tricky thing. I'm on my summer vacation and don't have the best monitoring with best videocards at the moment. So I've looked material via my home-edit and then used my Home tv-screen witch has 24p -support. In home-edit I have only 60hz monitors, but tv-screen gets it quite close.

    Cheers, misterj

  • No one here seems to be thinking about GOP1 - shoot intraframe, it looks so so so much better for motion.

  • @bannedindv Exactly. Motion will always look better in Intra - my Intra settings are set pretty high to keep the IQ up too. Yes its costly in bitrate, but what do you want? Shit?

    Optical distortion has been with us since the very first day a photographer shot movement. With the GH2 we are without all the neccesary shutter speeds to perhaps compensate motion unlike higher end cameras but Intra will render the best possible motion. Period. 'Tweened' is in my opinion the wrong analogy being used for predictive frames and motion vectors - this is an animator's term of working in pairs of frames. In AVCHD there is no user control with these 'tweening' prediction frames - we are reliant on prediction of the h264 alogorithms. Tweening in animation is eventually rendered to full frame. On the contrary, i frame only is constant full frames - it records what it sees. Just try and get your shutter speeds and panning adapted to the frame rate. That's all.

  • @driftwood Yep agreed...I like the feel of all Intra best for 24P, so thats mainly what I shoot with, I just pan and adjust the shutter accordingly. As I mentioned earlier ..if there is a lot of motion and movement and it looks too juddery on 24P, then 29.97 works for me, other than that I really love the image quality of 24P.

    As far as 24P goes I shot some footage the other day for a music vid and it was hard to believe that type of quality can come from such a small and relatively cheap camera....just amazing and holds up very well in post.

    A question (even tho I have done many tests) is there really much difference between the contained bitrate All Intra settings (such as Quantam 100) and the higher bitrate ones (such as Crossfire etc...)?

    Plus strangely enough some of the longer GOP (cluster etc..) settings seem to produce a better image in HBR. (Although I have some more tests to do on that) Do you have any all Intra settings that you would recommend as being great in HBR as well?

  • In AVCHD there is no user control with these 'tweening' prediction frames

    If there really is just one nutshell we could put this stuttering 24p problem in, I like this one the best.

  • @driftwood Yes, but what You think about Cluster 6 compared lets say Sedna AQ1 in 24H mode? I think that Cluster has les strobing effect and still, it does not have any of that bothering 'tweening' problem. I've been testing these two hacks side by side last two days and think that Cluster 6 has advantages in motion = bit faster pans and tilts before strobing is visually there.

  • Wait for Canis Majoris Day AM...

  • Hi guys, I'm about to shoot with the GH2 using a jib. I'm planning to use Driftwood's Sedna AQ1 with Sandisk 64GB 95MB/s. I've read somewhere on these threads that shooting VMM 80% helps a lot with the judder issue when the camera pans. I'm trying to test this out but haven't noticed any improvements yet. I'm wondering if my post production workflow is wrong?

    I'm transferring the MTS files onto my macbook pro HD and converting the files with 5DtoRGB batch. 5DtoRGB reads the VVM 80% files as being 23.976 so following on from LPowell's note that the files are actually recorded at 29.97 fps (and don't need to be sped up in an NLE) I'm telling 5DtoRGB to encode them at 29.97. Is this the right workflow?

    Or should I import them into AE and interpret the frames as 29.97? If so, how do I export the MTS files from AE so that I can then transcode them with 5DtoRGB?

  • Motion artifacts are determined by sampling rates and the rate at which the image sequence is up-sampled during reconstruction. A very simple and easily implemented filter mitigates this annoying visual artifact

    Objects in nature tend to move in smooth trajectories. When motion is sampled by a video camera and reconstructed on a display that natural smooth motion can appear to be jerky and irregular.

    ....

    The moving object in world coordinates jumps across the frame of reference in discrete static steps. During the reconstruction the eye is smoothly pursuing the motion across the frame of reference. The pursuit eye movements match the average velocity of the moving object in nature. The reconstructed object, however, only moves relative to the scene at the sampling rate, i.e., when the image changes.

    from Judder-Induced Edge Flicker at Zero Spatial Contrast
    by James Larimer, Christine Feng, Jennifer Gille, Victor Cheung. Article first published online: 5 JUL 2012

    Download PDF -286KB

  • Well...what filter?

  • @Jim_Simon

    These results show that motion artifacts are determined by sampling rates and the rate at which the image sequence is up- sampled during reconstruction. A very simple and easily implemented filter mitigates this annoying visual artifact. Because high-resolution displays require severe data rates, finding the sweet spot for sampling and reconstruction will have important consequences for electronic displays and image communications systems.

    The paper certainly doesn't mention a filter in the way we're used to; perhaps they are referring to filtering more like an algorithm which might determine the up-sampling of a given piece of video. In any case, the paper's terms of reference do not include the actual building and deployment of the judder remedy they allude to.

  • @Jim_Simon Thanks for replying. I'm glad somebody has noticed. This put a lid on the mystery.

    (Quiet popping of champagne corks)

  • Hello, thank you everyone who has contributed to this topic. Could you please watch the example video:

    test from INHUMAN on Vimeo.

    (the password is personalview)

    and say that it is the same thing. This stutter kills me! (look at around 7th second of the clip). I used Cinema mode, 24p, 1/50 second, everything -2, (noise reduction 0), smooth, Flow motion v 2.02 patch, the lens was Nokton 25mm, 0.95.

    So what is the final conclusion on how to minimize this. Is it to set the shutter to 1/30 or to use HBR mode and 29fps.

  • Just watched this Vimeo test.

    Looks about right to me; taking into account speed of motion through frame, for a 50mm lens and guessing distance from subject.

    I am about to do a shoot with the GH2 so will see if I find any validity to this GH2 issue. I've had issues with RED when I first started using it as there is a hard line and a high resolution. It's not as forgiving and organic as film.

    Canon's and other softer acquisition devices will naturally even out this kind of issue, because the codecs are a bit of a mess.

    I'll know more after I shoot.