Star Wars Laserdisc Preservations. See 1st Post for Updates. — Page 15

I can't wait to see a sample of your software decoded video!  Keep up the good work. :)

A few screenshots. Ld-decode vs the R7G(5x average). Ld-decode has the 3d comb and noise reduction turned off.  The R7G has comb filter set at 50%.

Obviously there’s going to be a lot of artifacting with the ld-decode picture because the 3d comb filter is turned off but notice that the ringing is always better than the R7G. The JSC has terrible ringing in some scenes on the death star. This can be tweaked to further reduce the ringing but this was the default settings with ld-decode.

LD-D

R7G

LD-D

R7G

LD-D

R7G

LD-D

R7G

Impressive result. The software decode seems very promising. A clear improvement! There is obviously less video noise and dot crawl in the R7G screenshots due to the steps it went through (3d comb & 5x average) but it also looks like the noise reduction did more harm than good.

So does your R7G work again? Sorry to hear the SWE LD was just another ISR LD. They are not completely useless though, they contain a lot of frames which the JSC LD lack so they can be used for patching holes in the JSC print to some extent if that is desirable.

g-force said:

althor,

So LD decode is the software decode? I don't see any artifacts, but I do see much more detail, and reduced ringing as you said! There are looking great! Can't wait to see how far you take this.

-G

I guess the screenshots don't show it but when watching the video there's a lot of dot crawl and rainbowing happening but that's to be expected when you basically disable the comb filter :). Yes, Ld-decode is the software version. You basically capture the full RF stream with an sdr. This means the raw file contains all of the video,audio,CC,LD-G and whatever else information is on the laserdisc. The software then decodes this raw file entirely in the digital domain. I think right now video and analog audio is implemented with digital audio decoding soon to follow.  I'm not sure about ld-g and cc support but I guess anything is possible. PAL might be implemented soon as well. The guy doing the coding seems to really know his stuff and he's been very helpful with getting me setup to use the software.

msycamore said:

Impressive result. The software decode seems very promising. A clear improvement! There is obviously less video noise and dot crawl in the R7G screenshots due to the steps it went through (3d comb & 5x average) but it also looks like the noise reduction did more harm than good.

So does your R7G work again? Sorry to hear the SWE LD was just another ISR LD. They are not completely useless though, they contain a lot of frames which the JSC LD lack so they can be used for patching holes in the JSC print to some extent if that is desirable.

 The R7G also had 30% YNR strength from the player I believe. The 5xaverage is quite beneficial normally I believe. The R7G denoiser does tend to bring forth a digitized look and as always a loss of detail. I will probably not do a 5xaverage with the LD-Decode versions because the filesizes are huge and decoding takes a seriously long time.  For example 1 side of a CAV disc is about 50GB in RF data. You decode that to rawvideo with RGB48 colorspace and the filesize is about the same or bigger and the decoding takes something like 5 hours. Anyway, the LD-Decode method introduces far less noise since digitization takes place way earlier in the chain and it never strays back into the analog domain. 

The R7G is still not working.  I'm maybe going to let it collect dust for awhile since I just got a cld-s315 that works great and since I'm tapping the RF it doesn't matter how crappy the player is since you are nabbing the signal right after it leaves the optical pickup. I will probably buy a cheap oscilloscope and try to do a full service adjustment on it but that's going to have to wait until I have some money. The LD-D screenshots were captured with the cld-s315 btw :).

EDIT: Speaking of pal is there anybody that can confirm these are the same as the coffret trilogie that Antcufaalb recommended to me here?

That is the right set.  That one should have the French dub.  There is also a version with English audio and French subs that looks very similar.

althor1138 said:

The R7G also had 30% YNR strength from the player I believe. The 5xaverage is quite beneficial normally I believe. The R7G denoiser does tend to bring forth a digitized look and as always a loss of detail. I will probably not do a 5xaverage with the LD-Decode versions because the filesizes are huge and decoding takes a seriously long time.  For example 1 side of a CAV disc is about 50GB in RF data. You decode that to rawvideo with RGB48 colorspace and the filesize is about the same or bigger and the decoding takes something like 5 hours. Anyway, the LD-Decode method introduces far less noise since digitization takes place way earlier in the chain and it never strays back into the analog domain. 

Well I think your R7G capture is superb but switching between the samples you posted I almost see the kind of difference I saw between your HF9G and R7G. :) Was checking out that thread over at lddb and the samples he posted are really impressive. Amazing what can be done really.

althor1138 said:

The R7G is still not working. I'm maybe going to let it collect dust for awhile since I just got a cld-s315 that works great and since I'm tapping the RF it doesn't matter how crappy the player is since you are nabbing the signal right after it leaves the optical pickup. I will probably buy a cheap oscilloscope and try to do a full service adjustment on it but that's going to have to wait until I have some money. The LD-D screenshots were captured with the cld-s315 btw :).

EDIT: Speaking of pal is there anybody that can confirm these are the same as the coffret trilogie that Antcufaalb recommended to me here?

Oh, I somehow missed that you had bought a PAL player, that's great.

VideoCollector's site is nice for visual info on various SW releases: http://swonvideo.com/ Looks like the ones with french subs displays a V.O on the bottom left corner of the case. The coffret trilogie did go through DVNR like the Faces set but I guess it's the best set the PAL-system can offer.

msycamore said:

The coffret trilogie did go through DVNR like the Faces set but I guess it's the best set the PAL-system can offer.

I'm pretty sure it's:

1994 No DVNR: http://www.lddb.com/laserdisc/31358/856235/Star-Wars:-Coffret-Trilogie:-Edition-Collector

1995 DVNR: http://www.lddb.com/laserdisc/35598/168035/Star-Wars:-Coffret-Trilogie

Ah, so I was wrong.  That set is not the one you're looking for...

TServo2049 said:

The 1994 one is the "pyramid" set, right? That uses the same pink-shifted transfer/print as the earlier French letterbox LDs, right?

That one is still interesting and deserves preservation IMO. Nobody has ever truly analyzed that version, what frames are there, and so on.

I don't recall it being call the "pyramid" set, but I wouldn't be surprised.

I'm pretty sure it uses the same transfer as the earlier French LBX LDs, but I'm not sure.

Well it's on it's way to me now.  If anybody sees a copy of the 94 version feel free to let me know.  

TServo2049 said:

One of the box sets is known as the "Pyramide" set because of the distinctive shape of the box. I think it's the 1994 one, the LDDB pic seems to show the sloped sides.

Ha, weird! o_O

The pre-THX one might be free from DVNR but as far as I know that one used an atrociously bad master. I assume the one seen in this post...  http://originaltrilogy.com/forum/topic.cfm/Preserving-French-Original-Trilogy-ANH-V10-released-ESB-in-progress/post/577965/#TopicPost577965

...is the one you're referring to?

AntcuFaalb said:

msycamore said:

The coffret trilogie did go through DVNR like the Faces set but I guess it's the best set the PAL-system can offer.

I'm pretty sure it's:

1994 No DVNR: http://www.lddb.com/laserdisc/31358/856235/Star-Wars:-Coffret-Trilogie:-Edition-Collector

1995 DVNR: http://www.lddb.com/laserdisc/35598/168035/Star-Wars:-Coffret-Trilogie

 I "studied" both PAL and NTSC editions, and I discovered that the "pyramid" (can't remember if I found the name, or named it myself...) is DVNR-free, while all THX PAL has light DVNR, not even nearly to NTSC versions...

I think that if we set DVNR=0 to the "pyramid" set, and DVNR=100 for the DC (and Faces), the THX PAL versions should be set around 30 or 40, no more... there are only "traces" of it, and, with no direct comparison, one could argue there is no DVNR at all - as I did the first time I watched them...

msycamore said:

The pre-THX one might be free from DVNR but as far as I know that one used an atrociously bad master. I assume the one seen in this post...  http://originaltrilogy.com/forum/topic.cfm/Preserving-French-Original-Trilogy-ANH-V10-released-ESB-in-progress/post/577965/#TopicPost577965

...is the one you're referring to?

Yes, quality is quite low - apart the pink tint that, strangely, affects only parts of the movie, not all; maybe the only way to squeeze out more detail is the software decoder, because there are no high-end PAL laserdisc player, like HLD-X0 or X9, or even CLD-G7... the best we have is the Pioneer CLD-D925, that it's the PAL equivalent of CLD-606...

Thanks for the information Andrea. I've found the THX PAL to be pretty "acceptable" compared to the ntsc versions too.  There is only 1 or 2 scenes where I have noticed it's effects.  The stormtroopers breaking through the blown up door on the tantive is one.  The laser bolts here show dvnr effects but not as bad as the ntsc versions.  They are rather good pressings as well with not very much rot or dropouts. I have the 97SE pal now too.  Someday I'd like to compare it's dvnr vs the other sources to see if it's any better.

If I capture using my s315 and the ati theater750pcie it gives a really decent picture actually.  It looks about as good as capturing ntsc except at scene changes where it will blend a bit if the luma is similar. The extra resolution really helps alleviate rainbowing and twinkling star fields. I'll probably end up using software decoding for these though once it's fully developed since I can fully tweak every setting instead of letting the capture card decide.

althor1138 said:

If I capture using my s315 and the ati theater750pcie it gives a really decent picture actually.  It looks about as good as capturing ntsc except at scene changes where it will blend a bit if the luma is similar. The extra resolution really helps alleviate rainbowing and twinkling star fields. I'll probably end up using software decoding for these though once it's fully developed since I can fully tweak every setting instead of letting the capture card decide.

...think if you could capture them with a PAL high-end laserdisc player, as good as HLD-X0! That's why is really probable that software decoding will be even better!

I'm sure we are talking about Happycube's work; is there available a version that a human being could use with Windows? (^^,)

And, what's about the hardware?

Yeah I think once he gets a working pal TBC and comb filter implemented it'll be way better than any market player available. 

I don't think it'll every be available on windows or even have a gui but it didn't take me too long to install ubuntu and figure it out. There is a WIP document that is sort of a help document for using the software.

The hardware requirements are a capture card with the cx23880 chipset.  You will also need to tap the raw rf signal on your player and make a custom cable to run to the capture card.

No big problems with Ubuntu - I used it in the past; nor with command line - even if I prefer a simple GUI... about hardware modifications... ehr... do you think it's possible to use the AC3 RF output, with no mods? (I like to dream, sometimes...)

Can you list some cx23880 capture cards? I know I'm asking too much...

The ac3-rf output was thought to be a possibility but nobody has had any success on any machines yet that I know of so it seems that maybe a filter is attenuating the rf signal before leaving the machine. Drilling a hole and mounting a bnc chassis connector and hooking up a coaxial cable to the test point actually turned out to be pretty easy.  I've now tapped all 3 of my machines this way.

You would still have to make a custom cable anyway because the luma pin of the s-video input of the capture card is the pin you need to feed the raw rf signal into.

Here is the card that Happycube and I are currently using. I think just about any card with the cx23880 will probably work with the custom cxadc driver according to happycube though.

Hauppauge WinTV 34132 NTSC/NTSC-J PCI