RT2000 and MPEG2. Please help.

[jack99]

According to Matrox, graphics and titles on the RT2000 are 4:2:2 quality, but video captured is 4:1:1 quality? Am I reading this right? Or is all video, regardless if it's a graphic created in Premiere or Photoshop, or NTSC video, captured at 4:2:2 quality. I was under the impression that 4:2:2 quality was only for D-9(digital S) or Digital Betacam (50 Mbits)
Is video captured at MPEG2 4:2:2 25Mb per stream better than DV video 4:1:1 at 25Mb per stream? RT2000 captures video at 3.6 MB/sec while Pinnacle DC1000 captures up to 6 MB/sec (both S-video connections and both specs say 25Mb per stream) Neither capture at 50 Mb per stream like Digital S or DVCPro50 quality. So what's with the 3.6 MB vs 6 MB per second? (I hope I'm clear with the Mega bits and Mega bytes) I really want to get the RT2000 but I just want to know what I'm getting. I'm sure this all makes your head hurt but please help.

[DrMordrid]

First remember that both DV and MPEG-2 are DCT based codecs that can generate the same kinds of recompression artifacts as MJPeg given the right conditions.

The colorspace used is not propriatory to any given video format. 4:2:2 is used in many formats from MPEG to MJpeg and many others between. That isn't to say you can't encode these with less than 4:2:2 either. For example: the PICVideo MJPeg codec allows for 1:1:1, 4:1:1 and 4:2:2.

NTSC DV by its very nature is 4:1:1 (PAL DV is 4:2:0) so there isn't any escaping a reduced colorspace with it. On the other hand an advantage for DV is that it has built in basic error correction schemes that minimize the visibility of the above mentioned recompression errors to the point of insignificance out to 3-4 generations.

In theory DV's main weakness is due to it's 4:1:1 colorspace and arises when doing keys. It sometimes shows as the key edges being a tad "rougher" and not as well gradated as the edges in 4:2:2 video. In actual practice a properly lit and staged key isn't usually a problem.

While MPEG-2 captures and outputs at 4:2:2 its weakness is the lack of error correction in the MPEG formats. This of course means it can generate more rendering artifacts than you would see with DV footage, often in the FIRST generation of recompression. Even so because of its 4:2:2 colorspace MPEG-2 can be used to advantage for keying if high quality settings are used.

The RT-2000 can, using either codec, perform bitmap overlays, DVE's, Organic wipes or Page Curls done in a 32 bit colorspace, so the final quality is quite high.

Also, the RT-2000 can perform various fixed or variable (VBR) MPEG-2 datarates. Your choice.

As far as the RT-2000 vs. DC-1000 data rates you're comparing apples & oranges. The 3.6 mB/s rate for the RT-2000 is its DV data rate and applies to ALL DV devices.

The 6 mB/s rate you quote for the DC-1000 applies to MPEG-2. The DC-1000 cannot do DV at all.

As far as 50 mb/s you quote for DigitalS, DVCPRO etc., these are propriatory DV tape formats so support from other manufacturers will be spotty.

Dr. Mordrid



[This message has been edited by DrMordrid (edited 12 February 2000).]

[DrMordrid]

Actually I frame only should give better quality than IP. Higher compression comes at a cost....

I frame is the encoding technique used by MJPeg (and other) .AVI files wherein each captured frame is followed by another captured and fully detailed frame.

IP frame is where each I frame is separated from the next by one or more P, or "predicted", frames which are generated by calculating an "average" (vastly simplified) of the two involved I frames.

In short, half or more of the frames in IP video are "faked".

Given the same data rate and codec I frame only video should be of higher quality.

I frame only is also better for editing since you can do frame accurate cuts. Not so with IP.

With IP and its variants you can only cut acurately to within 2 or more frames, depending on the exact GOP structure used. With extended GOP structures this could cause serious editing problems.

Also the RT-2000's MPEG-2 data rate can be manually adjusted up to 25 mbits/s or allowed to "float" as VBR encoding.

Dr. Mordrid

[andrewh]

I think you will find that the MPEG2 in DC1000 and RT2000 is 25Mb/s in both cases. DC1000 does IP encoding whereas RT2000 does I frame only. At first thought one would expect that the greater compression capabaility of IP encoding would give better quality than I frame only at the same bit rate.

However, life is never simple! They both use the same Codec (DVexpress25) and have exactly the same amount of processing power available per frame. Now, the big question: Does the extra processing power required to generate P frames degrade the quality of the image to a smaller or greater extent with respect to the I frame only encoding that needs more bits? Hmmm... I don't know and I do not expect C-Cube are about to tell you the answer.

Andrew