C-Cube DVxplore + Ulead VideoStudio 4

[Jerry Jones]

http://www.ulead.com/vs/news0507.htm

It's been 8 months since this news
release... and so far... I am unaware
of anything with which VideoStudio 4
has been bundled that utilizes the
C-Cube DVxplore chip.

I was wondering if Matrox had anything
quietly cooking on the burner where that
is concerned?

[Raptor]

They seem to release a new low end card every year, so maybe MJPEG isn't eternal in their hearts?


------------------

<<=============== RAPTOR ===============>>

[Jerry Jones]

I heard a rumor from somebody once... I
could swear... that Matrox would probably
have one of the low-end C-Cube chips in
production by year's end... and as I recall
it wasn't the RT2000. It was something
that was supposed to be for the really
low-budget types. Anybody heard anything
on that front?

[Grigory]

Why do you name dvxplore low-end chip? For DVD mpeg2 it is the same as the others. Same algorithm. It is simply not for prfessional production use.

Grigory

[Jerry Jones]

>Grigory: Why do you name
>dvxplore low-end chip?

Grigory,

I didn't name it----Ulead did.
If you read their news release
at the link I provided...

http://www.ulead.com/vs/news0507.htm

...please note Ulead specifically
vows support for the DVxplore chip...
which--as you point out--is a lower end
C-Cube chip.

[Jerry Jones]

Oh... I may have misunderstood your
question, Grigory. Did you mean to ask:
Why did I use the term "low end" to
describe the DVxplore? It isn't because
I believe the chip isn't capable of doing
a good job. It's because this chip is
always described---in the literature---as
being one of the less expensive C-Cube
chips. Nothing critical was intended by
my use of the term... assuming that's what
you were asking.

[Grigory]

Jerry,

You did not use the term low end, the question was for all persons who do use.

However, you last reply makes me asking again:
Why do you think that dvxplore cannot do its job well?
As far as I uderstand the end of chip is determined by its inability to do some jobs that others can do.

Concerning the low end job that it can do - 4:2:0 mpeg at DVD level, the specifications point out that all c-cube chips are the same by algorithm.

So, isn't capable of doing
a good job sounds like you know that the chip cannot do its job well. Does it mean that any other, based on the same algorith c-cube chip cannot do well this low end job?

You know, the quality at 5-8 mbps data rate depends on algorithm, but not on extras like 50 mbps 4:2:2 mpeg support, speed, other goodies that affect price.

Software encoder will give you the same quality on 386-40 and PIII-800, but the price of solution is quite different.
What do you think?

Grigory

[Jerry Jones]

We have a language barrier going on
here, I think.

I'm finding it difficult to understand
your posts, Grigory. Sorry.

I'm reading your last post.

It suggests you may have misread my
prior post.

Bottom line: I think the DVxplore chip
*can* do a reasonably good job.

[Grigory]

Well, I take the answer

Now, second step. Assume somebody takes high quality clip in DV format or better. Then, he makes typical mpeg2 file with say 6 mbps data rate by:

1. LSX software encoder
2. Vitec software encoder
3. bbmpeg software encoder
4. pixeltools software encoder
5. c-cube xplore hardware encoder
6. c-cube xpress hardware encoder
...

Is there anywhere an information what solution is better by the quality of movie?

This is important for me and other home users. I have tried #1-4, they are very different in speed, comparable in quality, if you adjust compression modes for each encoder carefully. Some work better with one clip, and another are best for the other clips.

I have no idea how two different chips from c-cube are rated, as well as where can we place them in a quality order?

Do you have any ideas or links?

I do not need to use 4:2:2 mpeg2 for my 4:2:0 source video, and do not need to go over 8000 mbps level too. So, is there any image quality difference between c-cube chips in that area?

Who has any info about this?

Grigory

[DrMordrid]

I had the opportunity recently to try the new version of the Ligos encoder encoding DV clips at 9 kb/s NTSC.

The results were excellent and almost as good as the DV clip. Playback was smooth and even areas with a fast moving object, my 21 month old son Erik doing about warp 5, were smooth.

Dr. Mordrid

[Grigory]

I got very good result with LSX too.

I used variable bitrate with 4 or 3 constant quantization value. This appeared to be the est mode, although you have to take care about file size by yourself. Average bitrate was about 7 mbps.
One clip contained Brugge (Belgium) city scenes with a lot of details and camera pans (especially when my 14 year old son was an operator).
Another one was nord sea beach in Beligium - huge sand space with only few details, but large areas of nearly constant color.
Both scenes were hard to compress. First one can give some mosquitoes around sharp edges of buildings. Beach scene was subject to distortions of sand color and color banding.

I tried to use constant bitrate, and several variants of variable bitrate setting. Constant quantization was the best. For these two scenes, LXS appeared to be the best of four encoders.
I play mpeg2 on H+ decoder, so I kept interlacing ON. Because LSX cannot compress by fields, it cannot set correct field order in decoder.
So, it is usable with DV footage only if I recompress all to DV with reversed field order. Straightforward recompression produces high frequency components that are not visible, but can increase average datarate by 10 percent in mpeg2 movie. I made such recompression with some filtering -used to make a sum of tree pixels in line with 2, 20, and 2 coefficients and divider by 24. This is equivalent to weak low-pass filtering.

LSX has to compress interlacing artefacts that appear in full frames with motion. This makes its job harder than for other encoders. If I turned ON deinterlacing, the bitrate dropped very noticeably, and the quality of individual frame even impoved. However, deinterlacing removes entire field from a frame which is not acceptable.

The motion estimation of LSX encoder works well (compared with bbmpeg or pixeltools) only in very slow mode. Otherwize you get noticeable noise on fragments with motion. Note - every indiviual frame is good, but they are changing slightly even on still objects, which looks like noise.
Very slow mode takes 1.6 second per frame to compress on celeron 450.
Slow compression mode may also give very good quality and takes a lot less time to compress. However, VBR compression produces up to 20% more data for this mode, which is an indication that motion compensation works less efficient.

I am interested in hardware solution quality comparisons. Actually, I can live with 30xrealtime actual LSX encoder speed (120x for bbmpeg), but 1-2xrealtime is always better.

Grigory

[dragmore]

I recently tested Pixel Tools Expert DVD aginst Ligos LSX 3.0, though the expert dvd was a little bit better in quality my pc used over 50 minutes to compress the file.
LSX used about 55 seconds in slow mode.
The pal clip was about 20 sec's long and if i had enough experince with mpeg encoding i'm sure i could tune the paramenters in lsx to compete with expertDVD.

The speed increase is probably because i've got a dual celery 450 setup in win2k.

I havent tried the Panasonic MpegX encoder. It's supposed to be the best in encoding quality.

Merry Christmas guys.

Thought i'd share my comments

--Dragmore

[Grigory]

For LSX and hard scenes, slow and very slow modes give you different quality and compression time.
Pixeltools uses probably the same compression technique as free bbmpeg. Both are similar to free encoding program from MPEG group.
Exprert dvd has well developed tuning abilities, but works very slow.
With the scenes I tried, pixeltools encoder quality appeared to be lower than lxs in very slow mode, 12-2 GOP structure and variable bitrate.

Grigory

[Jerry Jones]

...which brings us back to my
original question:

WHEN ARE WE GOING TO SEE A DVXPLORE
CHIP PRODUCT???? HMMMM????

[DrMordrid]

Certainly not from ATI that's for sure. I got this in the mail from them after asking about the "Video Wonder";

"ATI constantly reviews its product lineup to ensure that our product mix offers an excellent selection of products, features and price points. During our recent review, we have decided to postpone the introduction of ATI-VIDEO WONDER to the retail market. We will be re-evaluating this decision at the end of January."

Uh-huh....

Dr. Mordrid



[This message has been edited by DrMordrid (edited 30 December 1999).]