Panasonic 24 Mbit AVCHD is a lot better than you might think...

AF100 CODEC Stress Test (http://crews.tv/blog/2010/10/22/af100-images-pushed-to-the-limit/)

Guy thanks for the test.

Nice to see in the basic test how well the native codec stands up. One point I have to make is it is easy to see the difference between 4.2.0 and 4.2.2 in all of my testing. Surprised you could not see the difference in the color spacing. I know people who have purchased the Nano just for the color spacing.

Either way this camera is starting to shake it up which is great. Thanks for taking the time to do the test and share with us.

...I wish I could claim credit, but It's not my test. :-)

It was done by the Crews.TV guys and I thought it was worth posting, as many people seem to think that CODECs are all about bit-rate, and this test shows how close 24-Mbit can be to 100-Mbit sometimes.

What I would like to see next are some motion tests, which is where I think you will see better detail with the AVC-Intra 100-Mbit CODEC.

All in all it is a pretty impressive demo of how good the Panasonic AVCHD CODEC can be. ( I've been shooting with the Panasonic HMC-150 camera for the past couple of years which uses the same AVCHD CODEC )

I also use this codec in a Panasonic HMC40 and it is simply amazing. I do have to agree about the chroma but really it isn't as big of a deal as it once was. Outside of chroma subsampling I have also found 24mbit AVCHD to be at least as good as 50 mbit mpeg2. Remember I said not including the chroma subsampling of course.

I do however feel I need to point out that this wasn't exactly the best test for this sort of thing. These type of formats can look great when you are shooting a scene without a lot of movement. As soon as you add complexity to your scene however that is when a codec is really taxed. AVC-Intra is designed to look about the same regardless of how simple or complex the scene is. In these tests only still life like images were used with very little to no movement. The shots themselves are also pretty simple without a lot of complex changes such as tree leaves or a high level of noise. To really do an accurate test the people who tested this would have to use footage that is typically very hard to compress. These shots would look good with even HDV.

The part of the test I will commend is the use of difference maps to get an idea of what compression is doing to the images.

But a big problem is that the comparison is between AVC-HD and AVC-Intra, and only on fairly static images at that. It Both AVC-Intra and AVC-HD are (unsurprisingly!) based on AVC - and it is highly likely that for STILL images they will therefore each take a similar approach to how they compress. In other words, slight differences between them only tell you how they differ - not how good OVERALL either of them is.The difference map approach is far more revealing when the tested codec is being compared directly with uncompressed.

It would be far more revealing to see similar comparisons on high motion scenes - the main differences between AVC-HD and AVC-Intra are the long-GOP nature of the former, bit depth, and colour sampling. And static scenes won't tell you anything about the long-GOP/I-frame only aspects.

But more interesting still to see AVC-HD compared directly with uncompressed frames, ideally on scenes with both a lot of detail AND movement.

Using a 10-bit codec wont add anything to a 8bit hdsdi output, so the AF100 don`t do the avc-i codec justice...

There's not a whole lot of stress in that "stress test," they're not testing chroma, DR, or highlights, maybe shadows, and certainly not encoding. I'd wager that an iphone jpeg still would be comparable to those stills.

I'm not sure it's responsible to draw conclusions from this. Is there any reason to believe the camera sports a different encoder than the gh1?

The implementation of the codec on the AF100 includes B-frames, both the GH1 nor the GH2 use I and P only.

Where did you read that? I heard that at least in the 1080 24p mode it does indeed have B frames in the new GH2 and even Vitaliy Kiselev says that. The AF100's AVCHD codec is still slightly more advanced since it's closer to the HMC40's.

Jan Crittenden posts answers to queries in another forum. But you are correct in addressing my GH2 error -- it also uses B frames.

I think one answer to why the AVCHD 24 looks so good is how much image detail is the camera able to resolve anyway? If the lens isn't excellent, the camera is somewhat noisy, etc. , the higher rate codec can wind up recording mostly noise, aliasing, etc.

BTW, I actually think some of the "crappy" codecs can look quite good. For example, I've never seen HDV from a Canon XL-H1 break up. IMHO it's a fantastic implemtation of HDV.

True, Canon XL-H1 HDV is fantastic. I love the quality of the image and codec.