Will 24 Mbps make a difference?

W

weaselander

New member
Please keep in mind that I am an Engineer by training who is looking to buy a GH1.

It seems to me like this MUD problem is due to the high compression ratio when shooting at 1080P with only 17Mbps. Some have mentioned that 24 Mbps will help, and I sure it will, but to what degree? This change would allow the transfer of ~40% more information, but I don't know how that translates into real world results.

However, users have said that the 720P has far less mud and the camera allows you to shoot in 13Mbps, or a roughly 30% lower transfer speed. So if someone could take similar footage with both the 13 and 17Mbps in 720P and compare the mud, I think we could get a pretty good idea at the level of improvement if Panasonic gave us 24Mbps for 1080P.

What do you think?
 
The number of pixels has a lot to do with how the codec responds from what I understand, so I don't think treating it by percentage works. Another thing to keep in mind is that some people have said it may relate to the transfer speed with the memory device, though I'm not sure how much faith to put in that statement. Seems unlikely to me, but if that's the case it's even less a matter of compression issues. Someone with more knowledge on the codec in question could probably provide a much better response.
 
Nah, I don't think it's to do with memory bandwidth or write speeds either. As someone else pointed out earlier, 17Mbps works out at 2MB write speeds - I have a 30MB write speed Sandisk (peak write speed that is) and still get mud.

I think it's to do with the lack of B frames, an encoding issue pure and simple.
 
1080 is 2.5x the amount of pixels/data than 720 squeezed into the same 17Mbps data stream.
 
Guys, HMC150 footage looks *fantastic* in PH mode, and HMC70 footage looks really, really good at 13mbps. It's not so much the megabits, as it is the lack of b-frames that seems to be impairing the efficiency. More megabits will help, but basically this thing needs a proper fully-implemented codec chip like the HMC150 has.
 
I wonder what the reasons are for the GH1's woes?

Is it the codec implementation in software? They didn't have time to complete all the features?

Is it the codec chip? If so, why did Panasonic do this, knowing it would cripple their flagship in this way.

Or is it to do with the design of the sensor? It's a pretty different system from that found in a dedicated video camera, for sure.

I doubt they can just drag and drop software & hardware implementations designed for video cameras into the GH1. Give them time and hopefully the GH1 will be a more complete article come August.
 
What is "mud" in our case? It's a sign of total overcompression. Some blocks are compressed too strong so they totally loose their texture while neighbor blocks can still have some details. It looks like raindrops on watercolor painting.

Will thicker bitrate make the situation better? Of course! CABAC, b-frames are the tricks that will help, too. But they are not universal. Why? Cause they are the methods for finding more redundancy in video stream. When there _are_ some redundancy in video they will surely save some bitrate for complicated parts. But if video is noisy, shaky, flashy and crappy then camera coder will have to create redundancy by itself - blurring and washing out video. Thick bitrate is the best way to prevent it.
 
I am pretty confident 24mbps will do nothing. I've seen 9mbps avchd that looks great and doesn't break up nearly as much. My canon Xh-a1 even though 25mbps never breaks up like that and is using mpeg2.
 
For my Canon SX1 I'm using cheapest SDHC Class6 cards - and all are fast enough.

SX1 makes 43 Mbps (!) H264 - same like Canon 5DMkII - with 1920x1080. This are 5.4 MBps - EACH Class6 (guaranteed 6 MBps) are enough!

If someone has to much money: buy a new Class10 card :beer:

b9njq7pwz88m615yw.jpg
 
ryansheffer said:
I am pretty confident 24mbps will do nothing. I've seen 9mbps avchd that looks great and doesn't break up nearly as much.
Click to expand...

Totally co-sign that. Clips from other cams I've seen/used @ 17mbps and less don't have issues @ 1080 that the GH1 seems to exhibit :(

I'm still getting a couple for a web series cause I think the 720-60p is plenty good and I can slomo anytime I feel like.
Hopefully Panny will address the 1080 issues for you guys that need it :beer:
 
Edit.. Well at first I said that I disagreed with Barry, but I think we are saying the same thing.


The 5D lacks B-frames too and does not demonstrate the same problems. Allow me to explain what I believe is happening:

When you pan your GH1 a lot is changing in frame all at once. Therefore, each frame now must carry much more information to keep up the integrity of the picture. Using B-frames you can reconstruct data based on algorithms. With B-frames you have more information using the same bit rate.

Since the GH1 has no B-frames, you don't have that extra headroom. So our problem is that we are not packing enough data into the signal, how do we get more? By bumping up the bit rate! This is exactly what the 5D does, and the signal is flawless. Well.. relatively speaking :)

--

This is really all speculation.... It's a lot more complicated than it appears. The problem can really be a combination of factors, and possibly not even related to bit rate at all! We have no way to tell how the GH1 h.264 implementation works.

Personally, I'm convinced it's a hardware limitation. There is only so much you can fit into such a small and inexpensive body. The camera is a fraction of what the 5D costs, and you are really paying up for the extra horsepower & electronics in the Canon camera. Based on specifications it's a much faster system which makes video more robust.

This leaves a GH1 firmware update less likely but still possible. There may be something they can do as far as increasing the efficiency of the system. Just a little bit of improvement could make all the difference.
 
Last edited:
Just wanted to add that the lack of B-frames is not necessarily a bad thing. In fact, for our purpose it can be quite good.

This really comes down to fundamentals. The h.264 codec was designed to be a delivery format. This is bad for us, delivery formats are heavily compressed and optimized for small file sizes while keeping up a detailed signal. Trouble is, this destroys much of the information we need to do proper color work, keying, etc...

B-frames are used to reduce the amount of actual data packed into the signal and increase the compression ratio. By not including B-frames you reduce the compression ratio, thus getting a better source.

Using math to reconstruct information with B-frames is much less accurate than the original source from the sensor. By not including B-frames you depend less on the math predicting data and more on factual image data.

At least, if you have high enough bandwidth to get it across...
 
If they lock it down to 17Mbps and get rid of the VBR then I think it will help. From my understanding the VBR has it averaging around 8-10Mbps - I assume on the fast pans people are doing this drops down to compensate for sustained write speeds and we get the results some people are getting. But this is only in 1080p mode from what I understand as well. So if you are going to do a fast pan then drop down to 720p.
 
Nitsuj said:
If they lock it down to 17Mbps and get rid of the VBR then I think it will help. From my understanding the VBR has it averaging around 8-10Mbps - I assume on the fast pans people are doing this drops down to compensate for sustained write speeds and we get the results some people are getting.
Click to expand...
This is actually the reverse of how Variable BitRate works. It's main purpose is not keep sustained write speeds but to conserve file size. A VBR encoding drops down the bitrate when image content is simple - like a rose against a clear blue sky - and ups the bitrate when you tilt down on a field full of roses. I doubt VBR is causing the 'mud'.

Read about VBR here: http://en.wikipedia.org/wiki/Variable_bitrate and about H.264/AVC in specific here: http://www.drunkenblog.com/drunkenblog-archives/000312.html (scroll down a long way for tech info)

To very very shortly answer what B-frames are; they 'predict' coming changes in the image which increase quality without adding much to file size. However they also add CPU cycles to the encoding process so leaving them out in GH1 might be dictated by hardware limitations as several have suggested already.
On the other hand by upping the bitrate it might be possible to decrease the intervals between I-frames (the 'whole' frames in the stream) thus maybe making 'mud' less likely to occur. Tweaking interframe compression for optimum file size and image quality is a mix between tech know-how, art and voodoo so saying exactly what Panasonic should/could do to reduce mud is impossible for us outsiders.

Side note; some people has made a art form out of breaking interframe compression aka "data moshing". Example: http://www.youtube.com/watch?v=ZMZu0FO8G8U&fmt=18
 
You must log in or register to reply here.
Back
Top