Laserdisc players - screenshot comparison

I think this topic will be interesting for many...

I decided to do some capture tests (see my previous thread) but before them, I compared different players and different laserdisc edition.

Hardware used:

Pioneer CLD-D925 PAL/NTSC LD player

Pioneer CLD-2950 PAL/NTSC LD player

Pioneer LD-V4300D PAL/NTSC LD player

Pioneer DVL-909 PAL/NTSC LD player / PAL only DVD player

Pioneer CLD-600 PAL LD player

Pioneer DVR-320S DVD recorder (with 3D Y/C comb filter)

SAA733x based capture card in a Sony VGX-XL201 PC Core 2 Duo E6300

Notes: unless stated, the captures were not color corrected.

Let's start:

* choose which laserdisc platter to compare *

All are PAL OUT THX (no DVNR master) EXCEPT the "pyramid" edition that is not THX.

As you can see, the Pyramid french edition is the worst - older than the other, and no THX benefits. Has completely different colors, less details and cropped image. Discarded immediately.

The german edition seems to have a little more definition than the french one. The spanish shows a little less definition. So I choose to test the spanish.

Why I didn't choose the german one? Because it's too easy to test players with the best image, and I wanted to see which player shows the best with the non-top discs...

* test comparison one - using the spanish LD *

The first picture is taken from the Pioneer CLD-600 composite out passed through the Pioneer DVR-320S - note the difference VS the second picture, direct composite output?

Part of the merit is certainly of the 3D Y/C comb filter (but I presume it works only with NTSC video, and in PAL it *may* be a 4 or 5 lines 2D comb filter), but quite surely the second picture suffer from a DShow capture, instead of a higher quality direct capture, because DScaler don't "see" the composite in - note also the different image size. From now on, I'll use only the S-Video input of my capture card.

The third picture is from the DVL-909. To me, face colors are better and also the definition seems a little more better than the fourth picture, taken from the CLD-2950; by the way, the last picture shows a better white.

* test comparison two - using the spanish LD *

The first picture is from the Pioneer LD-V4300D routed through the DVR-320S, just because the first has not an S-Video out.

The second picture is from the same capture, but after some very basic Avisynth scripts, to correct color and cancel some noise. It seems it has better colors now.

The third picture is from the composite output of the CLD-D925 routed through the DVR-320S. It seems the blacks are "blacker" but the details are quite the same of the first picture.

The last picture is from S-Video output of the CLD-D925. Colors are less vibrant, but it looks like there are fewer artifacts than the previous picture.

Here you are the results.

What is the best picture?

Waiting for your comments,

Thai Panther

Very interesting test.  It would be even more interesting to see how the 1050 stacks up against this lot as I think it may well prove to give the best and most detailed PAL picture.

The only other thing I would add, is that I have 2x 4300D players. One of which shows slightly unstable colours which looks a bit like 'blotching' (can't describe it any other way).  The other one is a lot more stable and looks much better.  I only mention this as it looks like the 4300D you used has the similar colour blotching issues that my first player had.

It would also be very interesting to see the test repeated on a frame with saturated red or blue to see what difference the comb filter will make to the picture, the 925 shows really nasty combing effects on high saturation colours.  Any idea if the 909 is any better than the 925 for comb-filtering of colours?

What were the settings for the comb filter in the DVD recorder?  Did you have this set to 3D or 2D?

Keep up the excellent work mate!

Do you have the PAL Jason and the Argonauts disc? The test pattern at the end of side 4 is an excellent way to evaluate both frequency response (i.e. resolution) and comb filter performance.

I tested the D925's composite and s-video outputs with both a Panasonic DVD Recorder and a BT878-based capture card.  

1. D925-composite-Panasonic (right-click/view for full size):

Complete loss of detail at higher frequencies (right hand side of the waveform pattern), bad dot-crawl between colour blocks.

2. D925-s-video-Panasonic:

Dot crawl improved, but still present. Much better definition, although there is some residual reflection of the waveform pattern into the solid grey on the RHS - this should be flat. There's also a small amount of vertical colour bleed (tops of the colour blocks into the grey bars above).

3. D925-composite-BT878:

Awful dot crawl, and terrible rainbows (cross colour artefacts) on the high-frequency part of the waveform pattern.

4. D925-s-video-BT878:

Similar to the Panasonic image, but without the vertical colour bleed. Best results at the moment, but dot crawl could be improved and again got those reflections of the waveform leaking into the solid grey.  

I still have my other LD player to test, plus comb filters in an LG DVD recorder and a JVC SVHS deck; I also have a more modern capture card I've yet to install. Hopefully I can update this test sometime this year...

Holy crap, this really is excellent work guys!

It looks like Mothers idea to use the test pattern from Jason and the Argonauts LD is the best way to empirically test which LD player is best and what connection to use.

It will be interesting to see how the 4300D stacks up against the D925 with that test pattern.

(Slightly off topic, but interesting - at least to me, anyway.)

The pattern above, from the Jason LD, is actually the BBC's Test Card G. A slightly modified version of a pattern from the Philips PM5544, it was first broadcast in 1971, and was the first electronically generated pattern to be used by the BBC.

The frequency sweep (the waveform pattern below centre) ranges, I think, from 1.5 MHz to 5.2 MHz. The upper limit of LD bandwidth is listed in various places as around 5.5 MHz (or an equivalent resolution of 440 TVL) so any player that shows some definition in the 5.2 MHz region is doing very well. The D925 shows a little definition here when using the s-video output, but it's still almost flat.

(I've also since found the PM5544 pattern used on a clock and a pillow!)

_,,,^..^,,,_ said:

... the "pyramid" edition that is not THX.

As you can see, the Pyramid french edition is the worst - older than the other, and no THX benefits. Has completely different colors, less details and cropped image. Discarded immediately.

Which disc are you referring to as the "pyramid" edition?

This one from 1989, which is reported to have very bad quality, or this from 1994 which is apparently the same master but a better pressing?

Moth3r said:

_,,,^..^,,,_ said:

... the "pyramid" edition that is not THX.

As you can see, the Pyramid french edition is the worst - older than the other, and no THX benefits. Has completely different colors, less details and cropped image. Discarded immediately.

Which disc are you referring to as the "pyramid" edition?

This one from 1989, which is reported to have very bad quality, or this from 1994 which is apparently the same master but a better pressing?

is it related to the 89/92 tececine or to the pal master used for the coffret set

Moth3r,

I refer to this one.

As you can see in the pictures, the video quality is the worst of the PAL editions I have, but maybe with a better player, cable, comb filter, it could be watchable... and it maybe "smear-free" too!

_,,,^..^,,,_ said:

Moth3r,

I refer to this one.

As you can see in the pictures, the video quality is the worst of the PAL editions I have, but maybe with a better player, cable, comb filter, it could be watchable... and it maybe "smear-free" too!

I believe the contents of that boxed set are the same discs as the 1994 individual releases - so, according to LDDB, a bad master but better pressing than the 1989 release.

I suspect that it's not worth the effort, and you were right to discard it.

Moth3r said:

_,,,^..^,,,_ said:

Moth3r,

I refer to this one.

As you can see in the pictures, the video quality is the worst of the PAL editions I have, but maybe with a better player, cable, comb filter, it could be watchable... and it maybe "smear-free" too!

I believe the contents of that boxed set are the same discs as the 1994 individual releases - so, according to LDDB, a bad master but better pressing than the 1989 release.

I suspect that it's not worth the effort, and you were right to discard it.

But it has from star wars to jedi in the set, only pal release on laserdisc though i doubt it comes from a pal master.

I have no way of checking my copy as i have no pal laserdisc player but i believe that either it is dubbed in french or has burned in french subtitles, kind of like how the Japan release has burned in subtitles.

Moth3r said:

(Slightly off topic, but interesting - at least to me, anyway.)

The pattern above, from the Jason LD, is actually the BBC's Test Card G. A slightly modified version of a pattern from the Philips PM5544, it was first broadcast in 1971, and was the first electronically generated pattern to be used by the BBC.

The frequency sweep (the waveform pattern below centre) ranges, I think, from 1.5 MHz to 5.2 MHz. The upper limit of LD bandwidth is listed in various places as around 5.5 MHz (or an equivalent resolution of 440 TVL) so any player that shows some definition in the 5.2 MHz region is doing very well. The D925 shows a little definition here when using the s-video output, but it's still almost flat.

(I've also since found the PM5544 pattern used on a clock and a pillow!)

That's a very good and helpful test pattern, some TV-stations in Sweden is still using it sometimes and have done so for as long as I can remember.

5. D925-composite-LG (again, you can right-click/view for full size):

and

6. D925-s-video-LG:

With this recorder, there is virtually no difference between the results from the composite and s-video outputs. This recorder probably has a decent comb filter in it. It's also the image with the least noise and sharpest transitions. The frequency bursts still show the reflections onto the grey background, so this is probably due to the D925 itself - as is the very slight rainbowing at the top and bottom of the waveform that also appears on the other s-video images, if you look closely.

The main disadvantage is the weird interlaced chroma problem at the top and bottom of the colour bars. Might be due to a sub-sampling error in the MPEG-2 encoding/decoding.

7. D925 -> composite -> SVHS deck "passthru" -> s-video -> BT878:

This JVC prosumer SVHS deck apparently features "a Y/C filtering technique incorporated under license from Faroudja Laboratories Inc.". Sounds good, and although it's much better than the results from using the cap card's built-in comb filter, the y/c separation is no better than the comb filter built into the LD player. Dot crawl is about the same or worse, and the rainbows are much more noticeable at the top and bottom of the frequency bar.

However, luma transitions seem to be improved (less ringing/overshoot visible, but the diagonal black edge of the circle has picked up a halo.

Anyone still following this? ;-)

4(a). D925 - s-video - BT878 (using btwincap drivers with sharpness maxed out)

I've included this shot because the setup is similar to what I used back in 2004 when I did my first LD capture. Increasing the sharpness setting produces an image that initially looks more detailed. However, if you look closely (and compare with picture No. 4 above) you will see that the added detail is an illusion caused by the addition of false halos that are not part of the real image. It also has the effect of increasing the overall noise level; I really dislike this oversharpened look now.

4(b). D925 - s-video - BT878 (after AviSynth processing)

This shows the result of a simple levels tweak in AviSynth with a 5-capture average (compare to the original No. 4).

The black level was originally too high, the levels tweak has brought it down to where it should be.

Multiple capture averaging is a technique for reducing noise that can, in theory, be totally non-destructive. All other denoising methods can lead to a loss of detail, either spatial (smoothed textures) or temporal (blurred motion). Even though the latest motion compensated algorithms can help, such denoisers can never be completely non-destructive.

Yes noise levels are reduced (notably in the chroma) and there is no loss of detail, but what impressed me about using this method is the virtual elimination of dot crawl artefacts at colour boundaries. You could argue that you don't actually need a 3D comb filter as this method produces results that are just as good.

grisan said:

The BT878 is a pretty dated chip, do you have another capture card? Years ago I replaced my BT878 card with an SAA7134 based card and the difference in picture quality was pretty big.

Yes I have a newer card, based on the SAA7231 chip. I'm going to do some tests with my other LD player first (an analogue-only V4100) then swap the cards over and test again.

To be honest, I cheated slightly with this. Because it was a still frame on the laserdisc, I just used TemporalSoften(2,255,255) on the capture.

The principle is the same though, for a real multiple capture average you would use Interleave, TemporalSoften, then SelectEvery.

Moth3r said:

[...]

With this [LG] recorder [...]

The main disadvantage is the weird interlaced chroma problem at the top and bottom of the colour bars. Might be due to a sub-sampling error in the MPEG-2 encoding/decoding.

Looking at this again, the issue is obvious. DVD recorders encode to MPEG-2, which uses 4:2:0 YUV (YV12) colour subsampling. The recorder has no way of knowing whether the material you are recording is interlaced or progressive; this particular model assumes it is interlaced and flags each frame as such.

When upconverting from YV12 to RGB24 for a screenshot, DGIndex looks at this flag to decide how it should upsample the chroma, the error seen on those screenshots above are due to incorrect assumptions used for the upconversion. If you force progressive type upsampling, the problem is fixed.

Does anyone have screen grabs from the German letterbox silver screen collection pre THX?

The first two shots don't look very good at all, which disc was it that you captured? Also, why did you resize them to 1024 x 600?