Strobing in Flat/Underexposed Areas

Shawn Storm

Active member
Hi Peeps,

I applied the hack about a week ago and have been taking a fair bit of footage since then. One thing I have noticed is that flat area & underexposed areas of the footage seem to have a "strobing" effect where it alternates between somewhat sharp & blurry over and over. In pre-hacked footage this isn't present and just displays as blurry (or "blobby" may be more accurate).

I am only getting this behavior with 720p AVCHD footage, and it doesn't seem to be an issue with MPEG (which is also hacked). I mostly shoot with 720p AVCHD, so I would like to figure this out. It isn't horrible (such as the similar but worse strobing that happens when shooting below 60 shutter speed), but it is noticeable and somewhat annoying to me.

I posted an example on Vimeo, but the strobing isn't as noticeable there (I assume their processing does something that reduces the effect somehow), so instead here are two screen shots of two frames from the video. Look closely at the blue shirt and imagine the video footage going back and forth between each frame (about every second) and you will get an idea of what I am talking about:

shirt1.jpg


shirt2.jpg



Here are my current hacked settings:

MJPEG:
E1 Quality 256
E1 Table 5
E2 Quality 220
E2 Table 5
E3 Quality 200
E3 Table 5
E4 Quality 184
E4 Table 5

AVCHD:
Changed FHD/SH Video Bitrate to 38,000,000
Changed H Video Bitrate to 38,000,000
Changed Overall Bitrate to 40,000,000

I am using a SanDisk 8gig Class 10 card.

Any help would be appreciated. :)

-shawn-
 
Last edited:
Hi

I am having this problem too. However I tried the setting below and it works well. However a word of caution. When I first tried it I left out the patches for testers. When I tested it created really bad pulsing on some flat shadow areas on a road. I then redid the settings but with the patches for testers enabled and it seemed to clear it up a bit. It's not perfect however and I'd be interested in what other people have to say regarding this. Overall when you up the bitrate, high detail and fast moving stuff becomes a lot more usable, but if you have areas of low detail ( road in foreground out of focus) or slightly underexposed areas of a frame then the pulsing starts to happen. In fact it is because the picture is overall better after a hack that the pulsing becomes more obvious. It must be based on the i frame way that AVCHD handles compression. This is why MJPEG is better because it sees every frame as a separate item. This is why it's a lot bigger data wise than AVCHD.


40Mbps AVCHD High Reliability Patch

In short, this patch combines the average bitrate of the C settings with the reliability of the B settings.

40Mbps AVCHD High Reliability Patch:

Patches for End Users
. Wrapper
.. Native 24p/25p: unchecked
. AVCHD Compression
.. Video Bitrate FHD/SH: 41000000
.. Video Bitrate H: 24000000
.. Video Bitrate L: 16000000
.. Overall Bitrate: 43000000
. AVCHD GOP Length: unchanged

Patches for Testers
. AVCHD Compression
.. Preset bitrate: 180
.. Preset bitrate 2: 180


Also I have used 5DtoRGB and this has cleaned up footage quite nicely. However it's not really much good for longform as there's no batcvh support and it loses the audio. But picture wise it does a better job than FCP and Neoscene in my opinion.

I would really like to be able to control this pulsing so if anyone has any ideas I'd be grateful

all the best

Richard

P.S. I have uploaded a video to Vimeo that shows the problem. It needs to be downloaded to see it properly as Vimeo has added a lot of artefacts to the one you view in the player.

http://vimeo.com/15546087
 
Last edited:
Hi Richard,

Thanks for the feedback.

Honestly I am starting to use the MPEG footage more because it is super clean and seems slightly better than the AVCHD footage whenever I do comparisons. The main issue is that my MPEG settings don't allow me to play the footage in camera which kind of bugs me. I have also heard that AVCHD is sometimes better depending on conditions of shooting. Of course perhaps that logic only applies to the pre-hacked GH1??

So perhaps this is opening up a can of worms, but if I am only shooting 720p and don't care about the 60fps then is there any reason why hacked AVCHD would be better than hacked MPEG?? The only reasons that I am certain about at this moment are that AVCHD files are smaller and don't have size limits & that I can't play the MPEG files in camera.

Back to the issue in question, I have done some more testing and haven't come up with any answers. I just did some footage with my 20mm Pancake lens and damn it looks nice (all my previous testing has been with the 14-140 kit lens) and haven't been able to replicate the issue either (although I haven't yet tried to replicate the issue by purposely shooting flat/underexposed stuff). Is it possible that the lens being used could impact this issue somehow? Hmm.

-shawn-
 
Also noise reduction and sharpness settings can alter the outcome. Try the settings I have posted. They seem very good and did clear up the pulsing slightly. You have nothing to lose and non pulsing footage to gain.
From what I can gather this is an issue with P frames and something that a lot of people are aware of. It won't happen or at least it will be hard to discern if the camera is not locked off. If the camera is on a tripod however it will occur. 5DtoRGB slightly masks it but doesn't entirely get rid of it.

It's still not perfect but I did some tests on the same scene and these settings were better.

all the best

Richard

PS Scratch the noise reduction setting. I thought it did something bt it didn't fix the problem
 
Last edited:
Hi Again Richard,

I took your advice and applied Lpowell's settings ("40Mbps AVCHD High Reliability Patch") and also went ahead and applied his "Stable 35Mbps low-light MPEG patch with in-camera playback". Here are my findings:

MPEG Results:
*The settings seem stable and the MPEG file does indeed play back in camera just fine.
*The quality is noticeably less than the settings I was using before.
*The "static noise" that drives me crazy is much worse with the new settings (it was barely noticeable with my previous settings).
*The file size is much smaller (approx. half the size) as before.

AVCHD Results:
*Strobing issue is still present (and actually seems a little worse).
*Detail seems to be slightly less than with previous settings.
*More noticeable compression than with previous settings.



Overall thoughts:
It is nice to be able to play the MPEG back in camera and it is also nice having the smaller files so that my little 8gig card goes further, HOWEVER I was really getting used to the super clean footage from my previous MPEG settings (which appeared to be yielding around 80Mbps) and I honestly miss it. I think I may go back to my previous MPEG settings and just deal with the larger files and inability to play back in camera.

I had a similar experience with AVCHD. My previous settings actually seemed to look better and had the same amount of "strobing" and possibly even a little less! I am starting to feel like I actually had pretty decent settings before and I may go back to them. The footage I have tested is fairly high detail, but also a fair bit of underexposed and flat area. I took a screenshot which is below. The bitrate maxed out at around 23Mbps (using MediaInfo to analyze), so I am thinking that this "AVCHD High Reliability" patch may do good with properly exposed highly detailed scenes, but perhaps doesn't do so great with scenes that are lower bitrate (less detail and less exposure).

I am going to do a little more testing, but my gut feeling at the moment is to go back to my previous settings (which are outlined in my first post).

Also - I downloaded the 5DtoRGB program and tried it out, but I got a message saying that I didn't have any ProRes codecs installed. Where can I find a good ProRes codec? :)

tester.jpg


-shawn-
 
Re: Strobing & Quality Settings

Re: Strobing & Quality Settings

Hi Again,

So I decided to go back to my original settings (in my first post above) and do some additional comparisons in the process. This comparison is between my "original settings" above, compared to LPowell's settings that I mention in my previous posts.

This time I shot a very high detailed scene and moved the camera around a lot. I was attempting to create the highest bitrate scene possible. Here are my observations:

AVCHD:
*I like my original settings more. The details seems slightly better, especially in dark underexposed areas. When comparing bitrates, they were almost the same (my previous settings was 34Mbps and the LPowell settings was 36.6 Mbps).

MPEG:
*The difference is staggering (and the file size is too!). My original settings yield about 80Mbps and it looks great.
*The static noise using LPowell's settings is very noticeable. The static noise with my original settings is barely noticeable (about the same level as the AVCHD version).

Conclusion:
I am going to stick with my original settings for the time being. The MPEG files are HUGE and can't be viewed in camera, but they are beautiful and I will most likely use MPEG for short clips that are underexposed and/or highly detailed static scenes. I will probably also get a larger memory card (or a couple). The AVCHD files also seem better and more versatile in both high detail and low detail/low exposure scenes and as I said, the strobing effect is about the same if not better.

Here is a screenshot from my most recent test mentioned in this post:
grass.jpg


-Shawn-
 
I am too starting to wonder about these AVCHD settings. The biggest single problem is in underexposed areas. Now people will say expose properly but you can't expose sky and shadows together. One has to usually give and to be honest I have always known that you can't get back blown hilights but you can push shadows if needed. Therefore I almost always expose for sky and then under expose a notch, knowing I will be able to push some light into the shadows if needed.

But the problem is AVCHD doesn't like underexposed areas. I have a patch of grass in the foreground of a shot and it falls into shadow. It looks blurred and muddy in comparison to the lit grass just next to it. So detail is good with these settings but mud is apparent in shadows and underexposed areas.

I like MJPEG to be honest. At least you know what your'e getting won't be affected by this and all you then have to do is deal with scaling up later. Yes detail is lacking because of the smaller frame size but that's the compromise.

More testing needed I think
 
But the problem is AVCHD doesn't like underexposed areas. I have a patch of grass in the foreground of a shot and it falls into shadow. It looks blurred and muddy in comparison to the lit grass just next to it. So detail is good with these settings but mud is apparent in shadows and underexposed areas.

I like MJPEG to be honest. At least you know what your'e getting won't be affected by this and all you then have to do is deal with scaling up later. Yes detail is lacking because of the smaller frame size but that's the compromise.

I definitely agree. I did a bunch of testing this afternoon where I shot various scenes with both AVCHD and MJPEG and then compared them. In almost every circumstance the MJPEG image looked cleaner and gave more detail (especially in dark/flat areas). I have included the most drastic example below (for me it is light and day).

However, after carefully looking at lots of footage I am finding that I prefer the AVCHD in terms of motion in the frame. It seems smoother somehow, where the MPJEG is more jerky looking to me. In looking at individual frames I am thinking it is because the MJPEG frames are sharper which gives it a sort of fast shutter speed look. This is tricky because I love the crisp sharpness of the MJPEG, but I don't like the look of the corresponding motion. Hmm! :)

Here is a good example of MJPEG vs. AVCHD from today's tests:

AVCHD:
forest_avchd.jpg


MJPEG:
forest_mjpeg.jpg


-shawn-
 
I just looked through some of my pre-hack footage and was amazed to see the same strobing as mentioned in this thread, mostly in dark areas of the frame. I guess I just haven't noticed it before and it became obvious now that I am looking so much closer at the quality! It does seem like the hacked footage makes it more noticeable in certain situations, but it is interesting to note that it appears to also be present in non-hacked footage.

So I assume this has been discussed in this forum before? Anyone know of a good thread to check out? Is this just a standard glitch with the AVCHD codec that we have to live with??

-shawn-
 
Hey Shawn

stay tuned I have a very interesting "no scientific" test that really shows the flaws of AVCHD.

I am just encoding a movie so I'll post in a while
 
Re: Tripod Causing AVCHD Issue

Re: Tripod Causing AVCHD Issue

Hey Shawn

check out this thread. It's about the pulsing etc and is pretty interesting I think

http://www.dvxuser.com/V6/showthread.php?p=2123528#post2123528

I checked it out. I will be interested to see people's responses.

Honestly your footage doesn't show the issue as blatantly as some of the stuff I have witnessed from my own camera. The main spot that it is noticeable to me is the leather strap in the foreground, but its such a small area that I had to look really close. Here is a link to video footage of mine that I think shows it well:

http://www.lovespiral.com/shirt_strobe.mov

I enlarged the footage 200% so you could see it better (in the blue shirt).

If I get a chance I may try a similar test of doing tripod and then handheld just out of curiosity.

-shawn-
 
Last edited:
I am ultra critical (you wouldn't guess would you!). I want this camera to be as good as everyone says it is, but a lot of the time I just don't see it. I think underexposure is the key to the pulsing issue, but that shirt in your example is exposed fine. It's outside for cying out loud. And it looks dreadful.

Another thing is how do you expose for sky and shadows. You can't! You have to expose for the brightest part of the image or you lose it forever when you go to colour correct. So parts of an image in shadow are more than likely to be underexposed and that leads to the mud I am definitely seeing in grass that falls into shadow.

When you set your camera on a tripod you inevitably hilight this pulsing effect.

I really like MJPEG. I took some footage of a duck near our swimming pool ( very Sopranos) and it looked astonishing. There's grain yes but it looks more like film grain to me than video.

duck-01.jpeg

duck-02.jpeg

duck-03.jpeg

duck-04.jpeg


Only problem for me is it's 30 frames per sec and I work in PAL. I can re interpret the fotage to 25 p but the sound would obviously be slowed down slightly. This isn't really a huge problem for me as I mostly work in motion graphics.
 
Last edited:
Re: Ducks & 5DtoRGB

Re: Ducks & 5DtoRGB

Hi Richard,

Nice duck shots. What Mbps is your MPEG footage?

-shawn-
 
You have to expose for the brightest part of the image or you lose it forever when you go to colour correct. So parts of an image in shadow are more than likely to be underexposed and that leads to the mud I am definitely seeing in grass that falls into shadow.
With the GH1 I think it's crucial to achieve the best color balance you can obtain in-camera, rather than attempting to grade it extensively in post. The compressed frames it produces have at best 8-bit color depth and chroma is sampled at half the resolution as luma. When adjusted properly, the camera does an excellent job of balancing the 12-bit color depth of the raw Bayer Filter output of the image sensor, with far more resolution than is available in post.

If you do require the best results from color grading in post, I'd definitely recommend using MJPEG with the 4:2:2 color-depth patch. This doubles the vertical resolution of the chroma components, which can be especially noticeable with color keying. The MJPEG modes in both the 35Mbps MJPEG In-Camera Playback Patch and the 60Mbps MJPEG Class 6 SD Card Patch produce 4:2:2 color-depth images. Here are links to those patches:

http://dvxuser.com/V6/showthread.php?t=215624

http://dvxuser.com/V6/showthread.php?t=216238
 
Hi Richard,

Nice duck shots. What Mbps is your MPEG footage?

-shawn-

Hi

I used these settings for these particular shots. I also had sharpness halfway or between halfway and to the left. Made a big difference and I think a little sharpening in camera can help. Well as you can see the shots look like stills not frame grabs from video.

Q1=400, T1 = 4
Q2=300, T2 = 4
Q3=250, T3 = 4
Q4=188, T3 = 4

Also 4.2.2 color checked

http://www.dvxuser.com/V6/showpost.php?p=2038629&postcount=4

Have to say this produces the most gorgeous footage out of this camera I have ever seen. Unfortunately this will not survive the foliage of death pan or tripod on heavy detailed scenes even with a class 10 card. I have the default standard of Class 10 cards in my opinion a SanDisk Extreme Class 10 16 GB

I am now using the 60 Mbps MJPEG Class 6 SD Card Patch from lpowell (thanks for this lpowell) and find it to be not perhaps as detailed but still very presentable and more importantly a great deal more reliable.

http://dvxuser.com/V6/showthread.php?t=216238

I really loved the old settings but for reliability I had to downgrade. It's the kind of thing where you could shoot some similar stuff before a shoot to see if the high quality hack works, then use it on the day with a reliable version as a back up.
 
I really loved the old settings but for reliability I had to downgrade. It's the kind of thing where you could shoot some similar stuff before a shoot to see if the high quality hack works, then use it on the day with a reliable version as a back up.

I am very happy with my MPEG settings (which I outlined in my first post of this thread) and am getting around 80Mbps. Haven't had any issues yet and I have been shooting very complex foliage rich scenes. If I do encounter any issues I might try the 60Mbps patch you mention, but so far so good. The only thing that bothers me is not being able to play in camera and the large files sizes.
 
With the GH1 I think it's crucial to achieve the best color balance you can obtain in-camera, rather than attempting to grade it extensively in post.

I agree, however I have been personally struggling with finding that balance in camera.

I tend to shoot in scenes with a large range of brightness/darkness and the GH1 just doesn't seem to have the dynamic range to cover it all. I am then left with a decision of whether to clip the highlights or the dark areas and I usually prefer to clip the dark areas instead of having blown highlights. However, this usually results in slightly "underexposed" footage which the GH1 obviously struggles with.

I have been trying to follow the advice of people in this forum and "slightly overexpose", but that always seems to result in blown highlights for me so I usually end up going back to zero on the meter or slightly below.

I realize that extra lighting and carefully selected scenes can go a long way, but usually I am running around solo in the woods with just a tripod and so these aren't always options for me.

Thanks for the feedback,
-Shawn-
 
I agree, however I have been personally struggling with finding that balance in camera.

I tend to shoot in scenes with a large range of brightness/darkness and the GH1 just doesn't seem to have the dynamic range to cover it all. I am then left with a decision of whether to clip the highlights or the dark areas and I usually prefer to clip the dark areas instead of having blown highlights. However, this usually results in slightly "underexposed" footage which the GH1 obviously struggles with.

I have been trying to follow the advice of people in this forum and "slightly overexpose", but that always seems to result in blown highlights for me so I usually end up going back to zero on the meter or slightly below.

I realize that extra lighting and carefully selected scenes can go a long way, but usually I am running around solo in the woods with just a tripod and so these aren't always options for me.

Thanks for the feedback,
-Shawn-

I am rapidly approaching the conclusion that it is a flaw with the camera and we have to live with it, or shoot MJPEG and lose a little detail due to the smaller frame size. AVCHD just can't cut it, in certain circumstances where you have a large dynamic range.
 
I am rapidly approaching the conclusion that it is a flaw with the camera and we have to live with it, or shoot MJPEG and lose a little detail due to the smaller frame size...
Using a custom frame size, I've produced videos in MJPEG that when scaled up to 1920x1080 are nearly as sharp as those produced by the FHD AVCHD encoder. It does require preprocessing of the MJPEG files, but if you're already transcoding them to an intermediate format, it's no additional trouble.

The key technique is to use an anamorphic MJPEG frame size of 1920x810, similar to AVCAM format, but squeezed vertically rather than horizontally. Since it requires resizing only in one dimension, it retains sharper details than magnifying an entire 720p frame up to 1080p resolution.

Here is a comparison image magnified by 400%, using "nearest neighbor" scaling in Photoshop to preserve the shape of individual pixels. On the left is a section of an AVCHD frame shot at 1920x1080. On the right is a comparable section of an MJPEG frame, shot at 1920x810 and upsampled to 1920x1080.

ruler1920x1080compariso.jpg


For more details on how this custom MJPEG video mode is produced, check out the following post:

http://dvxuser.com/V6/showthread.php?t=225252
 
Back
Top