It is not the codec, it is the onboard computer!

[SLoNiCK]

2 subj.

Panasonic knows how to make good AVCHD codecs, what is happening is that the onboard computer is insufficient to handle all the tasks that are thrown at it, and so it is programed to DELIBERATELY reduce the AVCHD quality when it is stressed. This allows it to put out something. Unless it reduced IQ there would be dropped frames when the computer fell behind.

It's not valid to treat camera processing CPU as general purpose one. It's not a powerful CPU with universal amount of mips, flops or any other "processing power units" to distribute between tasks (de-bayer, NR, CC, distortion fix, JPEG/AVC encoding...). It's usually some weak lo-power CPU sitting atop of DSP heap, managing data streams, driving UI and peripherial buses like flash cards, USB, AV in/out. While the main work is done by DSPs. Currently it's the only way to make SoC cheap, powersaving and fast for some strictly defined tasks. I guess distortion correction and AVCHD encoding are two different independent blocks and none of them can directly affect other one's performance. All IMHO.

 

[shzr]

I'm under the impression that higher bitrates require less compression. Thus, is it safe to say that if Pansonic increases the bitrate on the GH1 it will alleviate the CPU? Two birds w/ one stone?

 

[SLoNiCK]

Less compression doesn't mean less processing. For example, coder can simply skip calculation of higher order coefficients in DCT for small bitrate cause they won't fit defined bandwidth. Same with motion estimation, I guess. I tried to compress same footage with XviD (it was the fastest I could make). The results: 10Mbit/s - 4m 34s, 5Mbit/s - 4m 19s, 2Mbit/s - 4m 04s. Not a huge difference, but you can see the trend.