THX 1138 "preservations" + the 'THX 1138 Italian Cut' project (Released) — Page 55

Does the 704 exhibit "CLV smear"? Most laserdisc players do.

Very few don't. A few examples of laserdisc players that don't exhibit "CLV smear" are: Panasonic LX-900, Runco LJR-I/II, Pioneer HLD-X0, Pioneer HLD-X9, and a few others.

"CLV smearing" looks like bad DNR artifacts sometimes.

The following post claims that the 704 smears like crazy. http://www.tomshardware.com/forum/38342-25-d704-sony-ghosting#1817894

AntcuFaalb said:

Does the 704 exhibit "CLV smear"? Most laserdisc players do.

Very few don't. A few examples of laserdisc players that don't exhibit "CLV smear" are: Panasonic LX-900, Runco LJR-I/II, Pioneer HLD-X0, Pioneer HLD-X9, and a few others.

"CLV smearing" looks like bad DNR artifacts sometimes.

Yeah, I've read about the infamous "CLV smear", it's a good thing you own a player who is said to not exhibit smear then.

SilverWook said:

I can only speak to the titles I own, but I have music videos that smear pretty badly on the 704. I archived them with my PR-8210, as most were analog sound anyway. In a movie like THX, I could see some smear around Pedro Colley's head in the white void sequence. I have read elsewhere even with all the DNR turned off, the 704 still applies some processing to the video.

msycamore, I'm pretty sure I transferred the pan and scan disc with the 8210. If you can hear audible buzzing on the soundtrack, that's what it sounds like with CX turned off. The disc didn't have CX encoding.

It wouldn't surprise me if someone slathered VNR all over the video master while censoring it for the Japanese release. Keep in mind it's a much older film transfer, even the telecine could be the culprit.

I can unfortunately confirm that there is indeed NR present in the captures, it may be present in the actual LD-master as it's there in all caps including the two UK LD caps. For the most part of the film it's fine but there are a few scenes which have clearly lost fine detail due to noise reduction. The thing is, when it occurs there's not a lot of movement in the picture and it's in these instances where it's also is as worst in the early pan & scan LD which makes me suspect the combfilters in these DVD recorders could be the culprit. It appears that sometimes the 3D adaptive filters fail to switch to 2D during motion when it suddenly appears in otherwise quite static shots, it very much reminds me of what I noticed when I first tried to capture my Technidisc SW-LD through the comb. Anyway, it's definitely not the end of the world, the WS LD-caps looks great, it's just a few scenes which exhibit this.

However, I began going through these captures again systematic looking at where they go out of sync with each other and I've encountered dropped frames in several places, actually there's fields missing in both the 2's and in the 3's in the PR-8210 US which are present in PR-8210 JAP and vice versa, including US from the 704. So you have 2:3:2... and suddenly 1:1:1.... in short bursts where it happens, still need to check the others so we can determine if it's a LD player hiccup or the fault of your DVD recorder... fucking weird.

msycamore: Speaking of 3D comb filters, do you notice any checkerboarding on the first frame of a sharp scene/shot-change?

The user poita and I have been looking at different 3D comb filters and we've seen this artifact a lot.

My Panasonic DMR-ES10 has an amazing 3D comb filter, but it exhibits an ugly checkerboard artifact on the first frame of any sharp scene/shot-change.

See more about the DMR-ES10 here: http://forum.lddb.com/viewtopic.php?p=33347#p33347

SilverWook said:

Maybe I should try a capture into my DV camcorder one of these days? (Just on the problem scenes as a test.) The DV input of my DVD recorder might bypass the filters used for analog video.

On the other hand, without the DVD recorders combfilter we would have captures with bad dotcrawl and rainbowing instead. So you did the right thing of course, the combfilter in the recorder I tried out was horrible, combined with the lousy composite signal of my player it was basically unwatchable.

That is definitely not the case here, there's only a couple of scenes I've noticed smearing and that it may be the combfilters fault is still only a guess of mine so it would maybe be nice to get it proven right or wrong instead of speculation, if you'd like to go through the trouble  I can post examples of the problematic scenes for you to check. What really baffles me though, is the dropped frames/fields in the video, so weird.

AntcuFaalb said:

msycamore: Speaking of 3D comb filters, do you notice any checkerboarding on the first frame of a sharp scene/shot-change?

Unfortunately yes.

I can capture my THX 1138 LD when my Panasonic LX-900 gets back from the shop.

I use a 2D adaptive comb filter made by Faroudja: the VP-100. It has a chroma delay adjustment that I use to correct an ~1px chroma delay inherent to all LX-900s.

Two LX-900 => VP-100 captures averaged together exhibit no dot crawl and rainbowing whatsoever! It looks like I passed the LX-900 through the most advanced 3D comb filter in the world except it's artifact-free!

I know this is BigTalk™, but I can post some samples tonight to backup my statements, if you'd like.

msycamore said:

AntcuFaalb said:

msycamore: Speaking of 3D comb filters, do you notice any checkerboarding on the first frame of a sharp scene/shot-change?

Unfortunately yes.

Which DVD recorder was used?

I'm testing the 3D comb filters in ones I come across (at thrift stores, etc.), so I like to keep a mental blacklist of what to not test.

I don't remember which model he used but it's one of the better ones going by various tests done over at Videohelp forums, Silver will tell you. That odd checkerboard pattern is visible sometimes but it's so insignificant on his captures that it's not a real problem, on the one I tried to cap the Technidisc however, holy shit was it bad, a newer Panasonic model.

It would be great if you could post a sample or two from your LD. :) I'll try to point out which scenes I'd like to see samples from a little later. Thanks!

EDIT: misread your post, I see you haven't access to your player at the moment. Well that testpattern posted over at lddb shows that you have really fine equipment, interesting thread.

msycamore said:

EDIT: misread your post, I see you haven't access to your player at the moment. Well that testpattern posted over at lddb shows that you have really fine equipment, interesting thread.

I was going to post two separate captures (of the same scene) to demonstrate the Faroudja VP-100 2D Adaptive Comb Filter and how averaging the two separate captures together looks indistinguishable from a 3D comb filter.

Spaced Ranger said:

AntcuFaalb said:

I was going to post two separate captures (of the same scene) to demonstrate the Faroudja VP-100 2D Adaptive Comb Filter and how averaging the two separate captures together looks indistinguishable from a 3D comb filter.

Very interesting. Please do! And I wonder if you'd post a sequence, from the worst possible example(s), of raw cap to individual 2D caps to combined-2D cap (in full-size PNG, if posting stills).

Do you want a test pattern or real footage?

JEDIT: test pattern posted.

I've tested a lot of comb filters, including 2D adaptive and 3D ones.

If you're accustomed to seeing the S&W Zone Plate with a 3D comb filter, then what you see may shock you (in a bad way). I ask you to keep your eyes on the averaged sample up top as much as possible. :-)

The averaged sample was generated with a simple c1.Merge(c2).

You'll notice a loss in diagonal resolution, but it's impossible to notice with real footage. Most 2D adaptive comb filters (like the Faroudja VP-100) lose diagonal resolution. You need a 3D comb filter to get it back.

The two captures were taken with the exact same settings, hardware, everything. The conditions were identical.

I powered everything off and then on again before doing the second capture, however.

Small analog differences are the reason that the dot crawl and rainbowing artifacts cancel out.

You can see these differences more clearly by doing c1.Subtract(c2).

Also, the Faroudja VP-100 is nothing compared to the 3D comb filter in my Panasonic DMR-ES10. You can see my review of that here: http://tx0.org/5pq

DisgruntledFan said:

Antcu,

Since your source was a DVD player, it's output luma phase must have jumped 180 degrees between captures (a bit like pal).

Won't help you with LD captures, since the phase is already baked in.

I am surprised it canceled out so neatly. If you can get the same trick to work with a laserdisc capture - i'd be extremely surprised.

Some background before I ask my question: The Panasonic LX-900 runs the video signal through its own internal comb filter before recombining it for the composite out (i.e., the composite output is not "pure"). The simple 2D comb filter it uses is non-destructive, so it doesn't harm diagonal resolution like 2D adaptive ones do.

Since the LX-900's internal comb filter is non-destructive, I can recombine the signal outside of the player before passing it to a better comb filter. (This is what's currently being done, except I'd be doing it outside of the LX-900 rather than inside.)

What if I use my Panasonic DMR-ES10 to recombine the s-video signal? It non-destructively digitizes the input signal internally for its frame-sync/line-TBC to work, so wouldn't the DAC used to convert it back to analog change phase like my DVD player does?

OK, I'm going to stop hijacking this thread. I'm sorry msycamore and SilverWook!

Let's continue the discussion here: http://originaltrilogy.com/forum/topic.cfm/Comb-Filter-Testing/topic/15421/