AVCHD versus DVCPRO HD at DVinfo.net

TingSern Wong · March 15th, 2008, 10:59 PM

I have a HVX202 (PAL version of HVX200) - recording on P2 cards. I know Panny has announced a pro version of the AVCHD recorder - AG-HSC1U, recording on SD/SDHC cards.

Has anybody done any comparison of the video quality - between DVCPRO HD (100 mbits/sec) output on HVX200 - On P2 cards versus this AVCHD output on SD/SDHC cards?

I am more interested in knowing if I can see any compression artifacts of AVCHD that is objectable? Can it be removed at the post editing stage if there are any objectable compression artifacts?

Ken Ross · March 16th, 2008, 08:45 AM

The latest version of some of the new AVCHD cams show far less compression than prior models. The SR12 I've been using shows no compression artifacts that I can see.

But with that said, I'd be shocked if the DVCPRO didn't produce a better picture. The cams are larger, the sensors are larger and you have more bandwidth. Of course you also have less recording time.

David Saraceno · March 16th, 2008, 10:47 AM

Quote from Jan at other forum:

"First, DVCPRO HD is not MPEG or anything like it. It is more like an MJPEG, but not really. It does start with a DCT and goes from there, but on a frame by frame basis. When you talk about the recorded full raster pixel map, you need to recognize that if there is any lacking in the signal or in the algorithm, it will show. You really need a codec that will support the full raster or just expect that sooner or later you will find that the compression engine will come back to haunt you.

That is if you were to look at the HDCAM, DVCPRO HD, XDCAM HD algorithms you would see that the size of the record is not a direct match up to the HD format either in 720P or 1080i/p. This is not because of the recording is being limited to a less than pixel for pixel rendition of HD, it is due to the pre-filtering done in the codec for optimizing the compression. There are high frequencies that can be rolled off and not be missed and if you don't you end up with more data to compress. When you play it back out of the uncompressed output you have the full raster for display purposes.

For example, if you have a 1920 X1080 signal and you compress it to 25 Mbs, that is a lot of compression. HD starts at an uncompressed bandwidth 994Mbs in 8 bit. So there is a lot of compression required to get it to 25Mbs. However, you prefilter some of that off, the compression ratio falls. So take DVCPRO HD, without prefilter the compression ratio is 10:1 roughly, however with the prefilter it is 6.7:1. The second is vastly more desireable.

The problem that people have here is that the numbers mean something but they do not mean what people might think. At a full 1920 X 1080 image in contrast to a 1440 X 1080 image there is approximatyely 40% more signal to compress. More is not always better unless the algorithm can handle it.

That said, AVCHD for all intents and purposes is like the new HDV but more efficient. It is an interframe codec and thus no frame stands independently resolution-wise, just like all of the other long GOP formts, HDV, XDCAM HD, XDCAM EX.

DVCPRO HD is an intra-frame codec, so that all of the compression that is going to happen is going to happen within each frame and it doesn't need to consider the next frame or the previous frame."

March 15th, 2008, 10:59 PM	#1
TingSern Wong Major Player Join Date: Jan 2004 Location: Singapore, Rep of SINGAPORE Posts: 749	AVCHD versus DVCPRO HD I have a HVX202 (PAL version of HVX200) - recording on P2 cards. I know Panny has announced a pro version of the AVCHD recorder - AG-HSC1U, recording on SD/SDHC cards. Has anybody done any comparison of the video quality - between DVCPRO HD (100 mbits/sec) output on HVX200 - On P2 cards versus this AVCHD output on SD/SDHC cards? I am more interested in knowing if I can see any compression artifacts of AVCHD that is objectable? Can it be removed at the post editing stage if there are any objectable compression artifacts?

March 16th, 2008, 08:45 AM	#2
Ken Ross Trustee Join Date: Jan 2005 Location: New York Posts: 1,945	The latest version of some of the new AVCHD cams show far less compression than prior models. The SR12 I've been using shows no compression artifacts that I can see. But with that said, I'd be shocked if the DVCPRO didn't produce a better picture. The cams are larger, the sensors are larger and you have more bandwidth. Of course you also have less recording time.

March 16th, 2008, 10:47 AM	#3
David Saraceno Major Player Join Date: Jan 2006 Location: Inland Northwest Posts: 490	Quote from Jan at other forum: "First, DVCPRO HD is not MPEG or anything like it. It is more like an MJPEG, but not really. It does start with a DCT and goes from there, but on a frame by frame basis. When you talk about the recorded full raster pixel map, you need to recognize that if there is any lacking in the signal or in the algorithm, it will show. You really need a codec that will support the full raster or just expect that sooner or later you will find that the compression engine will come back to haunt you. That is if you were to look at the HDCAM, DVCPRO HD, XDCAM HD algorithms you would see that the size of the record is not a direct match up to the HD format either in 720P or 1080i/p. This is not because of the recording is being limited to a less than pixel for pixel rendition of HD, it is due to the pre-filtering done in the codec for optimizing the compression. There are high frequencies that can be rolled off and not be missed and if you don't you end up with more data to compress. When you play it back out of the uncompressed output you have the full raster for display purposes. For example, if you have a 1920 X1080 signal and you compress it to 25 Mbs, that is a lot of compression. HD starts at an uncompressed bandwidth 994Mbs in 8 bit. So there is a lot of compression required to get it to 25Mbs. However, you prefilter some of that off, the compression ratio falls. So take DVCPRO HD, without prefilter the compression ratio is 10:1 roughly, however with the prefilter it is 6.7:1. The second is vastly more desireable. The problem that people have here is that the numbers mean something but they do not mean what people might think. At a full 1920 X 1080 image in contrast to a 1440 X 1080 image there is approximatyely 40% more signal to compress. More is not always better unless the algorithm can handle it. That said, AVCHD for all intents and purposes is like the new HDV but more efficient. It is an interframe codec and thus no frame stands independently resolution-wise, just like all of the other long GOP formts, HDV, XDCAM HD, XDCAM EX. DVCPRO HD is an intra-frame codec, so that all of the compression that is going to happen is going to happen within each frame and it doesn't need to consider the next frame or the previous frame."