Pulsating image degradation

Nitsuj

Nitsuj

Well-known member
I have been working with the GH1 to try and find out what this is and I can't seem to find the answer yet. I am wondering if anybody else has seen this. At times, it seems to depend on certain low light situations, I notice what looks like pulsating mud. It will degrade maybe twice per second at a constant rate. I turned off the O.I.S. thinking it was something about the lens stabilizer. That didn't seem to change it. I have turned off auto focus and that isn't it either. I am at a loss at this point what it could be. I get it in 720 and 1080p. I thought maybe it was my system during the encoding process however I viewed the direct .mts file using VLC (I know that has a lot of artifacts when viewing) but now I am at a loss as to what this could be and how to avoid it.

Anybody have any ideas?

Video example of what I am talking about

http://exposureroom.com/60839febfaa540f8964174db9c269468/
 
Last edited:
GOP size is 15 frames ==> 2 I-frames per second in 30p. One more evidence of how crappy the GH1's codec is. I usually see this when encoded with free/cheap DivX or Mpeg-2 codecs and low bitrate. In particular, I see the same pulsating with Mainconcept MPEG-2 encoder, what I do is I set it to insert I-frame every 3 frames, this helps. But you cannot tweak the built-in camera codec like this, so you are stuck. The only hope is for firmware upgrade.
 
Thanks Cranky for your input. I will look into that I-frame set. I was beginning to suspect as such.

How about anybody else with the GH1. You guys found a work around for this yet? It almost seems to me at this point the lighting needs to be flat and camera should be moving at all times. I really notice it when the camera is locked down.

I'm right now running through every possible setting on the camera and it is looking like Cranky has hit the head. I guess now is trying to find a way to mask it.
 
I haven't seen what you're referring to. Maybe I'm not understanding. Can you post a video or a short sequence of frames showing the anomaly?
 
I just finished up running through different shutter speeds at different ISO and it looks like shutter has a lot to do with how well you can see this. Seems like the higher the shutter the more visible it is. Anything at 60 or higher looks worse. However at ISO 400 I can see it from 30 on up. ISO 100 with shutter between 30 - 60 isn't really visible. Same goes at the other end with ISO 1600 30 - 60 not really visible. Sucks that the ISO sweet spot seems to be 400 where the problem is more visible in those shutter ranges.
 
Why would you shoot greater than 1/60 shutter? If you're shooting 1080 you should shoot with 1/50. If you're shooting 720p60 you should shoot 1/60. And if you're shooting 720p30 you should shoot 1/30 or 1/60, right? Do you still see the anomaly when you shoot at those shutter speeds?
 
Ah yes I am aware of the shutter speeds to use but it seems to really bring out the problem when raising the shutter speeds. It is still there at 1/50 @ 1080p. I first noticed it at 1/60 720p.
I recently discovered that certain film modes hide it better than others. For example Standard film mode seems to hide it better than Smooth however this is also depending on the lighting and gives me different results. Different results drives me crazy because it is hard for me to write down what not to do. I will figure this out if it kills me. ;)

I will put a vid together to show what I am talking about. It's pixels that are distorting 2 to 3 times a second and it is only at certain light levels. Seems to be at the edge of shadows. I did notice it outside as well on a previous test. It was what made me notice it. Since I have seen it outside I don't think it is the interference. I should get something together by tomorrow. I'm on Baby detail at the moment with my 3 week old and have to stop my testing. ;)
 
I'm pretty sure I've seen exactly what you're talking about and (at least in my case) it was nothing to do with the camera, it was down to the playback. As has been said, the file you end up with has P frames every half second or so and I frames in between. If the playback method you're using doesn't interpret these properly the I frames will look different to the P frames giving that regular pulsing. I saw this with VLC and also with the free included playback software. However, when I played it direct from the camera or when I converted it properly there was no pulsing.

So assuming you're getting the same effect as me, its not a problem with the original footage or the camera so don't worry.

The best way to find out is playback using the camera.
 
nathankw said:
I'm pretty sure I've seen exactly what you're talking about and (at least in my case) it was nothing to do with the camera, it was down to the playback. As has been said, the file you end up with has P frames every half second or so and I frames in between. If the playback method you're using doesn't interpret these properly the I frames will look different to the P frames giving that regular pulsing. I saw this with VLC and also with the free included playback software. However, when I played it direct from the camera or when I converted it properly there was no pulsing.

So assuming you're getting the same effect as me, its not a problem with the original footage or the camera so don't worry.

The best way to find out is playback using the camera.
Click to expand...

I'm putting together some test video together to check out this theory. I am viewing it first within FCP after a log and transfer. Then after removing pulldown I am viewing the QT file in quicktime. I am wondering if this is true because in VLC I get all kinds of artifacts and to me it is useless. I hope you are right because it may not be a problem for web viewing but I could only imagine what it would look like blown up on a silver screen or even an HD tv for that matter. (I don't watch TV so I don't have an HD set) I'll throw this up on exposureroom or something and see what it looks like there. If it isn't seen there then that is another clue that your theory is correct which I can only hope. ;) Thanks for the input.
 
I have uploaded a video example of what I am talking about. Maybe somebody can shed some light on this. It doesn't happen all the time, just in certain situations.

http://exposureroom.com/60839febfaa540f8964174db9c269468/

I also found it on Philip Blooms beach video. It is in the clouds in certain shots. It is faint but I can see it. If I can figure out a good way to avoid it and not be a surprise to me then I will be golden. You really can't see it in the LCD on the GH1.

Could it be related to the stock lens? I don't recall seeing it before but then again I wasn't really looking for it until I found it.
 
Wow, I've never seen anything near that in anything I've shot so far, even in my tests where I really pushed the camera. And I have no idea what could cause that.
 
I hope somebody does. I would like to get to the bottom of it. It isn't in everything. I have gotten a lot of great stuff out of it. What bothers me is not knowing exactly what causes it and how to avoid it.
 
If it's every 12 or 30 frames, then it's probably the i-frame kicking in. Is it always the first frame and every multiple of 30 from then on?

If so, it's yet another long-GoP issue. You're dealing with scenes there with extreme amounts of detail, so the i-frame probably doesn't have enough of a bit budget to encode all that detail, so when the i-frame comes up it goes "muddy", then the delta frames fill in the gaps for the remaining 29 frames of the group.

How to avoid it? Lessen the amount of detail (shallower DOF, or less harsh lighting maybe)... or don't shoot strictly static scenes where it's possible to notice the pulse...

I hate long-GoP. Have I ever said that out loud before? :)
 
Yep, this is suboptimal long-GOP encoding. I get similar results with MainConcept MPEG-2 encoder bundled with Vegas. On the other hand, you don't see this on big production DVDs, because they use much more sophisticated encoders, multiple passes, and they filter out hi-freq details.

The problem is not in long-GOP per se, the problem is in implementation.

> I hate long-GoP. Have I ever said that out loud before?

Some members of this board prefer the 150 to 170 for smoother and more natural looks, because DVCPRO HD on the 170 shows noticeable macroblocking at times (and they posted frame grabs, so I saw it myself). Long-GOP is not evil, one just have to cook it right. As a computer programmer, I like long-GOP because it is a smart way to squeeze as much quality as possible from limited bitrate.
 
As a computer programmer myself, I can appreciate the genius behind the technology. But as a video user, long-GoP has yet to be implemented in any codec with substantial enough bandwidth in order to make it a viable and valid alternative to intraframe for reliable professional use. PH mode is the closest I've found yet, and given the price point of AVCCAM products I find it a reasonable choice for that, but I'm not giving up my 170 or AVC-Intra any time soon.

If they'd keep the same bandwidth, but add GoP encoding, they'd improve video quality. But they never do -- they always take advantage of the GoP efficiencies to enormously shrink the bandwidth, but leaving behind the long-GoP issues that always crop up.
 
Barry,

I'd be curious to get your take on the Long-GOP results from the Convergent Designs Nano Flash unit. It seems from footage I've both seen and shot, that their implementation of 100Mbps Long-GOP is as clean as the AVC-Intra footage I've seen (though I can't say I've shot any). According to CD, 100Mbps is the sweet spot for Long-GOP, with little benefit beyond that number, even though you can go to 160Mbps on the Nano.

Granted, this is using the XDCAM 422 codec, but we're still talking about Long-GOP.

So as to the OP's issue, perhaps a combination of bitrate and implementation is the issue. Or, as I've not seen any behavior like this in my GH1, could it be a faulty camera/sensor?
 
You must log in or register to reply here.
Back
Top