3 Warp Points On Siemssen!!!!

[telemachus]

Sorry for shouting, but i could not believe my eyes.

I have just upgraded the firmware (available at siemssen.de go to the german section, choose support download model 5000)

I tried on Bond's CD and the 3 warp point GMC works smoothly and with no jerks at all.

They have also fixed the audio-video sync problem.

Had to let you know
Mark. :D

[BoNz1]

Do you mind posting a direct link to the firmware? Their download page is not accessible to me :(.

[telemachus]

http://www.siemssen.de/admin/files/399622291420cdb6e2637d.zip

If it does not work, i can stick it on my site.

mark.

[SeeMoreDigital]

This is great news!

The Siemssen must now be the most flexible Mpeg4 player available today :)


Cheers

[bond]

cool, i already saw someone talking about that on the german dvdboard, but the person wasnt really specific on this

the readme of the upgrade says:
"improved Xvid GMC Playback (not 100% now)"

so there might be still problems?

anyways if it really supports 3 warppoints than the Zoran Vaddis 776 (which is used in the siemssen player) is the first chip to support the full MPEG-4 Advanced Simple Profile :)
and next to it also AAC and .mp4!

afaik the siemssen is the only player which uses this chip till now, hopefully we will see more of those soon

[SeeMoreDigital]

Just did a quick search on Google and found the following link about the Rowa RDVD104 Mpeg4/DVD player: -

http://forums.eyo.com.au/showthread.php?t=87831

http://ausdivx.ms11.net/rdvd104/rdvd104_mainboard.jpg


Cheers

[Zhnujm]

The Cyberhome 462 wich i had a while ago also has this chipset.
But no mp4 support.

[bond]

Originally posted by Zhnujm
The Cyberhome 462 wich i had a while ago also has this chipset.
But no mp4 support. hm are you sure it doesnt use the version 7 (and not 776)?

[SeeMoreDigital]

Originally posted by bond
cool, i already saw someone talking about that on the german dvdboard, but the person wasnt really specific on this

the readme of the upgrade says:
"improved Xvid GMC Playback (not 100% now)"

so there might be still problems? Yes, you've just reminded me.

I am able to play NeroDigital 3WP GMC encodes using my Xcard but not XviD 3WP GMC encodes!

Can somebody here test the Siemssen player again please using NeroDigital and XviD 3WP GMC, and both the AVI and MP4 containers?


Cheers

[telemachus]

Seymourpixie

i have tried bond's, but if you create a test file or point me to one, i will gladly try it out.

mark

[Zhnujm]

Originally posted by bond
hm are you sure it doesnt use the version 7 (and not 776)?

Afaik there is only one Zoran 7 version, and thats the 776.
Heres a pic from the 462:
http://img12.exs.cx/img12/6382/Vaddis776.jpg

[Sagittaire]

SCO 5000 ND History

Feb, 04 2005
new Release which fixes or adds following points:

- improved Support of add. DVD+RW
- Jpeg Display now optional in Original Size or "Fit to screen"
- New Skin Design of Setup Menu
- New Data Menu Design
- AUTOPLAY for ISO File-CD or -DVD
- Improved Nero-Digital Playback
- Detection of not supported Nero Digital Profiles. Supported are: Mobile-, Portable-
Standard- and Cinema*-Profiles
- Detection of Nero-Digital (DRM) protected Files
- Display of File extension Icon as AVI, JPEG etc.
- improved Xvid GMC Playback (not 100% now)
- improved external Subtitle and AV-Synchronisation on DivX and Xvid Files

Update Procedure:

Please burn the image file (NERO) on a clean CD. Then load the CD and choose START to update the software automatically. The tray will open after approx 20 sec. Take out the Update CD. Do NOT turn off the unit until the tray will close again.
To make sure that all settings will updated please use the Factory Reset function in the Menu.


This player is the best MPEG4 ASP player in the world now ... !!!

Detection of not supported Nero Digital Profiles. Supported are: Mobile-, Portable- Standard- and Cinema*- Profiles ... ???

[bond]

Originally posted by Zhnujm
Afaik there is only one Zoran 7 version, and thats the 776.
Heres a pic from the 462:
http://img12.exs.cx/img12/6382/Vaddis776.jpg hm i am not sure about that, but well dunno :)

[SeeMoreDigital]

Originally posted by Sagittaire
SCO 5000 ND History

- AUTOPLAY for ISO File-CD or -DVD This is interesting!


Cheers

[BoNz1]

Originally posted by telemachus
http://www.siemssen.de/admin/files/399622291420cdb6e2637d.zip

If it does not work, i can stick it on my site.

mark.

Thanks dude. I looks like people outside of europe have problems accessing parts of their website. It got the file now using an anonymous proxy somewhere in Germany :P.

[Sagittaire]

After testing GMC 3WP (XviD) isn't completely supported (with this new firmware sample (http://multimediacom.free.fr/Video/10-ASP-GMC-3wp.mp4) play but with artefacts for SFrame)

Here my CDR test for Nero Digital hardware Player
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part1.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part2.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part3.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part4.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part5.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part6.rar


ISO MPEG4 CDR test (for Nero compatible hardware player)


1) ISO 14496-1 (Container)
test 01: MPEG4 ASP + LC-AAC + chapter*
test 02: MPEG4 ASP + LC-AAC + subtitles* Eng & Fre
test 03: MPEG4 ASP + LC-AAC + anamorphic
test 04: MPEG4 ASP + LC-AAC 2.0 Eng + LC-AAC 2.0 Fre
test 05: MPEG4 ASP + LC-AAC 5.1 Eng + LC-AAC 5.1 Fre
test 05: MPEG4 ASP + LC-AAC 5.1 Eng + LC-AAC 2.0 Fre
*No MPEG4 compliant

2) ISO 14496-2 (Video)
test 01: MPEG4 ASP + LC-AAC
test 02: MPEG4 ASP 1 Bframe + LC-AAC
test 03: MPEG4 ASP 3 Bframe + LC-AAC
test 04: MPEG4 ASP 1 Bframe packet bitstream (DivX) + LC-AAC
test 05: MPEG4 ASP 3 Bframe packet bitstream (DivX) + LC-AAC
test 06: MPEG4 ASP Adaptative Quantisation + LC-AAC
test 07: MPEG4 ASP Custom Matrix + LC-AAC
test 08: MPEG4 ASP Qpel + LC-AAC
test 09: MPEG4 ASP GMC 1WP (DivX) + LC-AAC
test 10: MPEG4 ASP GMC 3WP (XviD or Nero) + LC-AAC

3) ISO 14496-3 (Audio)
test 01: LC-AAC 2.0 32KHz 64Kbps CBR
test 02: LC-AAC 2.0 44KHz 80Kbps CBR
test 03: LC-AAC 2.0 48KHz 96Kbps CBR
test 04: LC-AAC 2.0 48KHz 90Kbps VBR
test 05: HE-AAC 2.0 48KHz 96Kbps CBR
test 06: HE-AAC 2.0 48KHz 50Kbps VBR
test 07: LC-AAC 5.1 48KHz 192Kbps CBR
test 08: LC-AAC 5.1 48KHz 192Kbps VBR
test 09: HE-AAC 5.1 48KHz 192Kbps CBR
test 10: HE-AAC 5.1 48KHz 147Kbps VBR
all test with PNS actived

[SeeMoreDigital]

Originally posted by Sagittaire
After testing GMC 3WP (XviD) isn't completely supported (with this new firmware sample (http://multimediacom.free.fr/Video/10-ASP-GMC-3wp.mp4) play but with artefacts for SFrame) It would seem that you generated this sample using XviD's 3WP GMC and not NeroDigital... can you confirm please?

My own tests have revealed that XviD and NeroDigital's versions of 3WP GMC perform quite differently in hardware... but why this is, is a mystery to me!


Cheers

[Sagittaire]

It would seem that you generated this sample using XviD's 3WP GMC and not NeroDigital... can you confirm please?GMC 3WP with latest XviD beta ... :)


- improved Xvid GMC Playback (not 100% now)Readme firmware say exactly that: improved but not totaly supported. Perhabs ND GMC isn't really 3WP ... !!?

[BoNz1]

Originally posted by Sagittaire
GMC 3WP with latest XviD beta ... :)


Readme firmware say exactly that: improved but not totaly supported. Perhabs ND GMC isn't really 3WP ... !!?

Yes that is correct, I think I remember one of the ATEME people on IRC saying that it uses some sort of optimization where it can chose 2 warp points over 3. Video encoded with XviD probably will be quite problematic when you get lots of s-frames in a row.

[SeeMoreDigital]

Originally posted by BoNz1
Yes that is correct, I think I remember one of the ATEME people on IRC saying that it uses some sort of optimization where it can chose 2 warp points over 3. Video encoded with XviD probably will be quite problematic when you get lots of s-frames in a row. Hmmm!

If you de-mux a NeroDigital Mpeg4 stream containing 3WP GMC to AVI and run it through MPEG Modifier, it is clearly reported as containing "3 warp-points".

And before you ask... MPEG Modifier is able to detect Mpeg4 streams containing "2 warp-points" ;)


Cheers

[bond]

Originally posted by BoNz1
Yes that is correct, I think I remember one of the ATEME people on IRC saying that it uses some sort of optimization where it can chose 2 warp points over 3. Video encoded with XviD probably will be quite problematic when you get lots of s-frames in a row. afaik xvid also uses 2 warppoint gmc

[SeeMoreDigital]

Originally posted by bond
afaik xvid also uses 2 warppoint gmc
All I know from experience is, both NeroDigital and XviD 3WP GMC work very differently in hardware. Despite MPEG4 Modifier reporting them both as containing 3 warp-points!

So I guess, it would be useful to obtain so kind of official confirmation from both parties in this regard :confused:


Cheers

[Mr_Schizo]

Originally posted by Sagittaire
[B]After testing GMC 3WP (XviD) isn't completely supported (with this new firmware sample (http://multimediacom.free.fr/Video/10-ASP-GMC-3wp.mp4) play but with artefacts for SFrame)

Haven't tried it yet but the artefacts could be also caused by the XviD GMC bug.

[telemachus]

Originally posted by Sagittaire
After testing GMC 3WP (XviD) isn't completely supported (with this new firmware sample (http://multimediacom.free.fr/Video/10-ASP-GMC-3wp.mp4) play but with artefacts for SFrame)

Here my CDR test for Nero Digital hardware Player
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part1.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part2.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part3.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part4.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part5.rar
http://multimediacom.free.fr/Download/Test%20ISO%20MPEG4.part6.rar

hello sagittaire,
i downloaded all your files, and i'm not sure i understand why they are called Part1-6 when most of the archives are the same!
I will test it tomorrow, thanks for the test files.

by the way, have you noticed that the sync problem is gone, but as soon as you >>FF or <<RW a movie and then play agian, the sync is out again.

till later
mark.

[Sagittaire]

Originally posted by telemachus
hello sagittaire,
i downloaded all your files, and i'm not sure i understand why they are called Part1-6 when most of the archives are the same!
I will test it tomorrow, thanks for the test files.


Multifiles part cutting with winrar (restauration 10%) because 80 Mo files are too big for my serveur ... :D

[bond]

can someone simply try a nero gmc stream on the siemssen before we speculate around about xvid bugs...

[SeeMoreDigital]

Originally posted by bond
can someone simply try a nero gmc stream on the siemssen before we speculate around about xvid bugs... Agreed!

OK, I fed Sagittaire's XviD 3WP clip into Recode2 and generated this NeroDigital 3wp clip (http://homepage.ntlworld.com/seemoredigital/SMD_ND_3wp.zip), for you Siemssen player guys to try!


Cheers

[telemachus]

Originally posted by SeeMoreDigital
Agreed!

OK, I fed Sagittaire's XviD 3WP clip into Recode2 and generated this NeroDigital 3wp clip (http://homepage.ntlworld.com/seemoredigital/SMD_ND_3wp.zip), for you Siemssen player guys to try!


Cheers

hello seemoredigital,
your file plays perfectly, no artifacts at all, smooth from start to finish.

all sagittaire's video files play perfectly, his 3wp file "MPEG4 ASP GMC 3WP (XviD or Nero) + LC-AAC" plays almost perfectly, there is very, very mild artifacting in the top 1/10th of the image towards the latter part of the clip, but it is most certainly viewable without disruption.

mark

[Zhnujm]

Well, SMDs file even plays smooth on my mediatek based players....
Muxed to avi, of course.

Seems to be no real 3Warp Point GMC then.

[bond]

Originally posted by Zhnujm
Well, SMDs file even plays smooth on my mediatek based players....
Muxed to avi, of course.
Seems to be no real 3Warp Point GMC then. strange, but theoretically a codec doesnt have to use 3warppoints if he thinks its not necessary. so it could be the smd clip uses only 2 or less wps

edit: ups, thats not correct! a 3wp encoder will either use always 3 warppoints or no gmc on a frame

[SeeMoreDigital]

Originally posted by telemachus
...your file plays perfectly, no artifacts at all, smooth from start to finish.Thanks for the confirmation Mark :)

Originally posted by Zhnujm
Well, SMDs file even plays smooth on my mediatek based players....
Muxed to avi, of course.

Seems to be no real 3Warp Point GMC then. I thought it might run okay... seeing as though I can play it fine using my Sigma Xcard (with their 13 month old driver/software set).

Originally posted by telemachus
all sagittaire's video files play perfectly, his 3wp file "MPEG4 ASP GMC 3WP (XviD or Nero) + LC-AAC" plays almost perfectly, there is very, very mild artifacting in the top 1/10th of the image towards the latter part of the clip, I wonder if the effects you are seeing is due to Sagittaire's clip containing mattes!

I seem to remember reading somewhere (on the forum) that GMC works more efficiently when the mattes are cropped away...

...Anyway, I think it's safe to say that there are differences between Nero's and XviD's 3WP GMC implementations. As you can see from MPEG4 Modifiers analysis. The NeroDigital clip contains a lot more S-VOPs than XviDs...

http://img154.exs.cx/img154/1585/3wpgmc7rb.gif

I wonder if we "end-users" will ever get to know what the differences are?


Cheers

[bond]

SMD, bobololo meant your clip doesnt use 3wp even if it signals it! because of the blackborders it doesnt use 3wp, but instead uses 0wp gmc, which is the same as if no gmc is used, but still gmc gets signalled or so :D

so to make it short: your sample didnt use 3wp gmc even if it signals it

try another sample with cropped borders!

[SeeMoreDigital]

Originally posted by bond
SMD, bobololo meant your clip doesnt use 3wp even if it signals it! because of the blackborders it doesnt use 3wp, but instead uses 0wp gmc, which is the same as if no gmc is used, but still gmc gets signalled or so :D

so to make it short: your sample didnt use 3wp gmc even if it signals it

try another sample with cropped borders! Thanks for confirming this again bond ;)

I've generated the NeroDigital 3WP clip again, without mattes (cropped to 720x304) (http://homepage.ntlworld.com/seemoredigital/SMD_ND_3wp_(720x304).zip) and.... yes, when played using my Xcard, it now performs in much the same way as Sagittaire's XviD 3WP clip does - with mattes!

MPEG Modifier also reports that this new clip now contains a lot less S-VOPs than the previous NeroDigital clip: -

http://img222.exs.cx/img222/6761/smdnd3wp720x304mp42nl.gif


Cheers

[bond]

Originally posted by SeeMoreDigital
I've generated the NeroDigital 3WP clip again, without mattes (cropped to 720x304) (http://homepage.ntlworld.com/seemoredigital/SMD_ND_3wp_(720x304).zip) and.... yes, when played using my Xcard, it now performs in much the same way as Sagittaire's XviD 3WP clip does - with mattes!so it always was neros "fault" and not xvid's :D

hopefully someone can test your new clip on the siemssen soon

[SeeMoreDigital]

Originally posted by bond
so it always was neros "fault" and not xvid's :D Hmmm!

I wonder whether Nero's GMC encoding approach can really be considered as being "at fault"?

GMC is a weird implementation at best. DivX did not appear to do a very good job with their original 1WP implementation and look like they are still struggling to get it right (if the latest beta's are anything to go by).

If GMC really does work to its fullest potential when there are no big black mattes in the encoded frame to confuse it, then maybe Nero's approach is correct and everybody else's approach is flawed :eek:

The same might also apply to Qpel :eek: :eek:


Just a thought!

[bobololo]

Just a few notes about GMC and warping work and why some GMC clips play flawlessly on hardware that aren't supposed to do so.

When enabling GMC, an encoder depending on its internal implementation can exploit a number of warping points (3 for XviD & ND, less for others). This number is written in the stream configuration (ie the syntax element 'no_of_sprite_warping_points' in the so called VOL header).

Then during the encoding process, the encoder chooses the type to be used to code a frame : I, P, B or S-VOP (GMC predicted) depending on how efficient is the type accordingly to the content. So here you can understand that even with GMC enabled, you can find some clips without any S-VOP because they don't help.

In a S-VOP frame, the encoder tries to estimate the sprite trajectory (ie the global motion) and computes the warping points describing it. These warping points will then be use to compute a transformed reference frame for the prediction. The sprite trajectory which means the WPs are written in each S-VOP header.

For an encoder supporting only 1 WP, this leads to a global translation motion vector. With 2 and 3 WPs, the transformation is more complex (rotation, etc).

However, even when using 3 WPs, the global motion estimator can find that the sprite trajectory is a simple translation rather than a complex transformation. In such case, the 2nd and 3rd WP are (0, 0). And the process involved to compute the predictor is the same as a 1 WP S-VOP.

Hence, for each S-VOP, the sprite trajectory actually involves 0 to N WPs (N being the number of warping points in the VOL header). That explains why SMD's 3wp clip is played correctly on the standalone. In this clip, due to the black mattes, the global motion estimator finds that 1 WP decribes sufficently well the trajectory.

Moreover, within a S-VOP, each MB can be selected as gmc predicted or standard temporal predicted depending on how good is the global motion compensated predictor compared to the un-transformed temporal reference. That also means that a S-VOP could have 0 GMC predicted MB which is such case could be decoded without the support of GMC N warping points.

This note was a little bit technical but I hope it will clarify the things though.

-- bobololo

[SeeMoreDigital]

Thanks bobololo,

With regard to generating GMC encodes "with mattes" I'm still interested to know why XviD's implementation works so differently to Nero's?

Cheers



bond,

Me thinks this kind of information might make interesting reading for people over on the "New A/V Formats - Codecs" section of the forum...

[telemachus]

Originally posted by SeeMoreDigital
Thanks for confirming this again bond ;)

I've generated the NeroDigital 3WP clip again, without mattes (cropped to 720x304) (http://homepage.ntlworld.com/seemoredigital/SMD_ND_3wp_(720x304).zip) and.... yes, when played using my Xcard, it now performs in much the same way as Sagittaire's XviD 3WP clip does - with mattes!


hey there all,
i just tested this file, it plays ok, but there is some artifacting on some of the shots, not necessarily the fast moving ones. The artifacts are different from the ones on Sagittarius file, his were only on the top 5 or 10% of the image, seemore's were all over sometimes, and a few blocks here and there. About 60% of the time the film clip was completely free of artifacts.

btw, what do you mean by mattes? Do you mean it is cropped? A "matte" in video speak is like a stencil where you create an alpha channel to determine the transparency of your media.

Mark.

[SeeMoreDigital]

Originally posted by telemachus
btw, what do you mean by mattes? The mattes are the black bars, above and below the video images...

Here's a still from Sagittaire's 720x400 encode - note how he's managed to include information about his encode in the top matte... very nicely done: -

http://img98.exs.cx/img98/8624/sag720x4008lx.jpg


And here's a still from my 720x304 re-encode - note that the mattes have gone: -

http://img109.exs.cx/img109/7589/smd720x3043kn.jpg


I guess most of the macro-blocking, will be due to losses generated during the "re-encoding" process!


Cheers

[bond]

However, even when using 3 WPs, the global motion estimator can find that the sprite trajectory is a simple translation rather than a complex transformation. In such case, the 2nd and 3rd WP are (0, 0). And the process involved to compute the predictor is the same as a 1 WP S-VOP.hm bobololo, but why isnt it possible that an encoder can use different warppoints on different frames? eg one frame uses 1wp, another one 3wp, than 2wp aso?

Originally posted by telemachus
i just tested this file, it plays ok, but there is some artifacting on some of the shots, not necessarily the fast moving ones. The artifacts are different from the ones on Sagittarius file, his were only on the top 5 or 10% of the image, seemore's were all over sometimes, and a few blocks here and there. About 60% of the time the film clip was completely free of artifacts.i assume this means that the 3wp implementation of the firmware altogether isnt finished (not only for xvid)

[Sagittaire]

i assume this means that the 3wp implementation of the firmware altogether isnt finished (not only for xvid)

yes, exactly : I will make a new sample in XviD and ND-ASP (720*400 16/9 for movie) with very various global motion (zoom, rotation, translation with combinaison) ...

GMC 3WP isn't 100% implemented but sample play fine with light artefact (certainely on SFrame with 3WP and not on SFrame with 1WP) ...

[yaz]

Originally posted by Mr_Schizo
... the artefacts could be also caused by the XviD GMC bug. pls, remember this. it seems as if there were sg messy w/gmc in xvid.1.1b1. at least, our tests (http://forum.doom9.org/showthread.php?s=&threadid=88404&perpage=40&pagenumber=3) revealed sg like that, so, it's just some experimental. no devel has commented it yet :-((( if u want workin gmc u should roll back to an early december build.

the bests
y

[Manao]

but why isnt it possible that an encoder can use different warppoints on different frames? eg one frame uses 1wp, another one 3wp, than 2wp aso?The number of warp points is written in the VOL header, so it can't change at each VOP. And anyway, if you've encoded at least one frame with 3wp, you'd better use 3wp always, since efficiency is better.

[bond]

Originally posted by Manao
The number of warp points is written in the VOL header, so it can't change at each VOP. And anyway, if you've encoded at least one frame with 3wp, you'd better use 3wp always, since efficiency is better. but as bobo wrote and the sample showed the vol header doesnt have to be correct always

[Manao]

The VOL header is correct : three warppoints are systematically written, though two are always null in the case of SMD's sample.

[plonk420]

http://www.siemssen.de/working/layout_4/images/images_service/service_04.gif

wow, that's a hot callcenter rep ^_^

[Soulhunter]

Originally posted by plonk420

wow, that's a hot callcenter rep ^_^

Though she should clean this dirty keyboard... :D


Bye