testing of video data compression quality of home DVD record

[Andrey /MAG/]

If anybody is interesting several times ago project named "Testing of video data compression quality of home DVD recorders" was started. Adress: http://mag-page.narod.ru/index_en.html (quick mirror: http://mag.irk.ru/index_en.html).

At this moment you can compare results of two home DVD recorders (Panasonic DMR-E100H and SONY RDR-HX900) with the testing results of software encoder Canopus ProCoder.

Test video data using for recording is HD trailer of cartoon Shark Tale. Mark method consists of visual analisys and analisys of bitrate diagrams. You can go there and compare the results of testing.

The project is open. You can take part in the testing and post the results to publishing too. All details are described there (but most part in russian).
May be someone of DVD recorders users makes such testing and post results for publishing. It wil be interesting to compare quality of old and new models of recorders.

There are two pages in this site are translated into english: Main page and page of comparison. Others pages stays in russian.
If anyone can help with translating you are welcome.
Thanks.

[lordsmurf]

No. Cartoons make for terrible encoder tests.
Not to mention your site timed out after 2 minutes.

[Andrey /MAG/]

lordsmurf
Sorry for inconvenience.
This problem are fixed.
Please try any more.

[trhouse]

Lordsmurf,

I have been thinking about uploading some clips in .vob form as suggested by vhelp. Still working on the finding all freeware to do this so others can repeat the test at no cost. The clip was to be the videoclip from the THX test patterns , but if cartoons are not a good test, what would you suggest? It would be good if whatever it is, is something most people could easily find.

Andreys site is interesting. Once the "how to do the test" part is translated from the Russian, it may prove to be a good contribution. The site does seem slow loading however.

[lordsmurf]

Good tests are live action shows. Explosions. Water. People. Movement. Fire. Dark. Sunlight.

Cartoons are just big blobs of solid colors that move around. No good. No detail. The good thing about cartoons is they are less demanding on an encoder. And if done correctly, can survive with slightly lesser bitrate. But for a test, bad. Gives false results. Most people will have uses dominated by live action, not cartoons.

[Andrey /MAG/]

It is hard to me to explain in engish but I will try.

Why cartoons?
Not cartoons but HD trailer of cartoon Shark Tale (resolution 1288x688). It is not drawing but 3D rendering. The quality of this trailer is fine.
There are a lot of movements, transition effects and details there. All scenes of trailer (103 scenes) shows actions under water. Dynamic scenes are interleaving by static scenes. It is the better way to test quality of bitrate variation of home DVD recorders.
Specially for test was choosen 3 hard frames and 3 simple frames to compare.

You can look at these frames. Here are some links...
Examples of test frames comparison (record mode: 4 hours per disk):
1. simple frame after hard: "upside down" fish; (fixed link)
2. simple frame after simple: tied fish; (fixed link)
3. hard frame after hard: race-meeting - stadium; (fixed link)
4. hard frame after simple: scared jellyfishes... (fixed link)

Good tests are live action shows

I agree with you. I will take it into account. Perhaps, method of testing will be extended by recording test of live action shows (HD trailer or smth else).

What are your suggestions?

P.S. Sorry me, but it is not my fault. The comparison page did not worked today. There was unexplaining trouble of SpyLOG counter. Counter is removed now. The page is works at this moment.
P.P.S. If problems with slow site loading repeated, please let me know. Here (in Russia) loading of big comparison page with test frames (about 600 KB) last about two seconds.

[Andrey /MAG/]

Sorry for inconvenience...
Now new location of Testing Project available.
The mirror of Site should work faster.
The New adress of Testing: http://mag.irk.ru/index_en.html.
Links above are fixed.

[spectroelectro]

Once again have to agree with Smurf. Using HD is also a weird choice. How many folks can view HD on a hi-def TV?

Much better test would be full D1 720 x 480 (or possibly 720 x 576 if you're in Europe) with live action. Scene dissolves, light on water, these all burn up a lot of bitrate and reveal problems clearly.

FWIW my own rule-of-thumb experience with software encoders vs hardware encoders is that TMPGenc 2-pass VBR with reasonable numbers (2K min, 5k av, 8K max) yields roughly 1.85 gigs per 45 minutes of encoded live video. Mainconcept VBR with the exact same numbers yields roughly 1.65 gigs per 45 minutes encoded live video. What's really amazing is the result from the JVC DRM10 LSI hardware encoder: 1.45 gigs per 45 minutes of encoded live video, and the video looks _better_. There are subtle motion artifacts during slow pans from TMPGenc or Mainconcept or CCE encoding (not Procoder "studio" encoding AFAICT) but none from the JVC LSI chip hardware encoding. That's pretty amazing. A real-time hardware encoder can beat not only the quality of software encoders that run at multiples of real time, it can even produce a smaller size encoded mpeg file.

Panasonic DMR E80 encodes seem to fall in the midrange with about 1.6 gigs per 45 minutes of encoded live video, but of course the Pannies introduce those encoding artifacts everyone knows about so it's not a fair comparison.

[trhouse]

Any idea what the p/n is for the LSI chip?

[trhouse]

The reason I ask is the LSI web site has testimonials from four recorder companies that use the LSI chips. One would think those recorders would have similar encoding performance.

"Consumers are looking for the latest in DVD technology, and the LSI Logic DMN-8600 has enabled Samsung to deliver this in our DVD-R5000 recorder. The DVD-R5000 gives consumers superior picture quality, multi-format playback capability, and unique DVD recording features, such as time-shift, to maximize their television watching experience."

-Mr. Y.S. Park
Mr. Y.S. Park, Vice President
Digital Video System Division, Samsung Electronics

"LSI Logic has not only worked with us to design a feature-rich, cost-effective recorder, but they've also enabled us get to market quickly."

-Ancle Hsu
COO
Apex Digital, Inc.

"LSI Logic's DoMiNo network media processor is the most powerful video processing engine in the market, delivering more than four times the performance of the existing codec solutions today while consuming less than 2.7 watts of power."

-Hiroyuki Takekura
General Manager
JVC

"LSI Logic's DiMeNsion 8602 DVD recorder system processor allows us to deliver powerful video recording and editing capabilities to our customers, like the world's first DVD Recorder+VCR with YesDVD technology."

-Ed Brachocki
CEO
GoVideo

[lordsmurf]

You are correct, the VBR uses of LSI by Apex and JVC are almost identical in quality.

The LSI chips perform pretty equal, software permitted.

I say this because the DVD+VR implemtentations by LiteOn (and clones like ILO/Daytek/Gateway) use the true DVD+VR spec of CVBR instead of VBR, which is somewhat lesser quality, though only noticed towards the 4-hour range. It adds mosquito noise at scene changes sometimes, and it you have rapid-fire scene changes, it'll turn to blocks. The 3-hour mode or less is best setting on these units with CVBR. It may take a hack to get it, though all LiteOn's had at least one model that was 3-hour factory set (special firmwares).

It should be noted that the Apex had a rare gamma error that showed it's ugly head on some S-VHS tapes. Faulty Apex firmware, but was uncommon. Never really saw this from VHS or off-air.

Some LG units also use LSI, looks excellent.

As do some of the Zenith units. The Zenith does an excellent job like JVC and Apex, in terms of visual quality. I think it's also CVBR, but is closer to true VBR.

The GoVideo equipment is crap for other reasons (other parts failed frequently, company is infamous for junk all the way back to VHS days), but the few that use LSI looked quite nice. The LSI line was shortlived, they are now using some cheap Cirrus or ESS crap. And CVBR instead of VBR.

Same for Samsung, few models used LSI, then they went to cheapo chips for latter models.

It's always easy to spot an LSI chip. Look for a low-noise, no-noise signal, and if the source was VHS, look for it to be relatively chroma-noise free. And unlike some of the other companies out there, LSI chip users do not hide their chipset specs, proud to have a good chip. Easy to find info on who is using them.

[Andrey /MAG/]

...Using HD is also a weird choice. How many folks can view HD on a hi-def TV?
Much better test would be full D1 720 x 480 (or possibly 720 x 576 if you're in Europe) with live action...

Yes, you are rigth. I didn't going to record it in HD format.
HD source is necessery to get maximum quality of source video data. First step of prepearing video data is converting it to DVD resolution (720x576) and DV format. After that DV data are sending to DVD recorder though FireWire cable to minimize losses.

...What's really amazing is the result from the JVC DRM10 LSI hardware encoder...

As I understood from your's and lordsmurf's posts, there are several chips which embedded in different hardware recorders (LSI, ESS, Cirrus...). Where can I get the detailed info about it?
Did anyone already test and compare their's quality?

[Znet_[Sneg]]

The MAG uses Cartoons for the comparative analysis of quality of coding, and it, in my opinion, is quite admissible. Also test images contain fine details and various color transitions. The good project.

[lordsmurf]

Bumping a year old post?

[Andrey /MAG/]

Thnx Znet_[Sneg].
In a web this topic about recording quality apears occasionally.
The project is still alive. Number of tested DVD-recorders is 16 now.
see http://mag.irk.ru/results_en.html and http://mag.irk.ru/compare_en.html

[shakabrah]

Thnx Znet_[Sneg].
In a web this topic about recording quality apears occasionally.
The project is still alive. Number of tested DVD-recorders is 16 now.
see http://mag.irk.ru/results_en.html and http://mag.irk.ru/compare_en.html

some of the contents on the site is still in russian.