I've done some pre-demodulation cleanup to more closely match what LD players actually did, and it improved color saturation a lot.

Megapixie - on the color bar test you should analyze RGB levels/color saturation as well as signal/noise ratio. Both my previous version and the 4300D have similar undersaturation which leads to off colors. The 2950 is better, so maybe it's something they did to get such high frequency response on the 4300D...?

I also changed the NR to be standard noise coring, and reduced it's strength to 1 IRE luma only. (If 1 IRE is enough for VE, it's probably enough for anything but the noisiest disks...) This is the same type of analog noise coring many players use (their usage of it depends by model, though... some only use it on CLV and the early bits of CAV disks - others like the S1 have it enabled all the time, even in test mode)

(For another aside, per the service manual the RCA SJT-400 CED player has an undefeatable 5 IRE coring.)

New screenshots, with the least flattering (zone plate) shots first, and the best (cleaned-up ballerina shots) last:

http://imgur.com/a/JQDql - 1IRE NR

http://imgur.com/a/wVQm3 - no NR

edit: Looked again at the original de-emphasis timings which boosts frequency response - at the expense of overshoot. Made it an option for ld-decoder.py (-h) - samples here: http://imgur.com/a/pxNOS

Awesome stuff HC.

You are quite right: I've never looked at chroma phase/saturation. Once I've got up and running with capture - I'll do another article.

I'll take a look in a bit and re-run the numbers.

Mega.

Starting to figure out how to use OpenCV and took a first pass at optical flow-based 3D comb filtering - samples from VE here:

http://imgur.com/a/3t5nD#QWfTXx1

Also improved the RF capture driver by turning off the s-video chroma ADC, and slowly working on PAL LD decoding thanks to samples from robwhar and megapixie. Much more discussion etc over on the google doc.

I like these samples a lot, in particular the zone plate is really good!

Here's the 1986 Laservision demonstration. I improved my post-decode encoding settings (h264, flac audio, yadif+mcdeint) in addition to the other tweaks I've been doing on the way. Audio's slightly regressed because I turned down the filtering and tried to speed it up, and it's got some pops/distortion (or it might just be the disk is a little scratched)

Digital sound isn't nearly finished yet, but being worked on. And from a purely preservation point of view, digital tracks can already be saved without RF decoding.

https://www.youtube.com/watch?v=JXQcPwLpqwo - 720P is recommended for 60P deinterlacing.

G-R-E-A-T-!!!

Been playing with PAL a bit - I don't have a working comb filter yet alas, but I'm making some headway.

On a lark I got a cheap PAL disk off eBay (Eric Clapton's Unplugged - which I have an undocumented WEA US pressing of that I need to look at ) and the CLD-V2800 can be persuaded to 'play' it in test mode for at least a bit without puking (outside of test mode for ~30 seconds or so until it goes ) , and I can process it's RF (the odd look of it's from the opening screen and not a software glitch )

Great news. It's pretty amazing to have this software available. Thanks for persevering with it!

Thanks (and thanks for all your help/motivation, too!)

Got the PAL comb filter somewhat working today:

http://i.imgur.com/ewuEDTx.png

edit: new samples

http://imgur.com/a/YS1VA

Not bad, not bad at all!!! (read: good, very good!)

Here are some RGB line comparisons between ld-decode PAL, the V4300D, and CLD-2950. RF used for LDD (via 4300D) and the test pictures are all from megapixie. ld-decode is noisier but has much higher frequency response, and it has better color.

http://imgur.com/a/Qo6Qc

Is it possible to see actual images from all three - 4300, 2950 and lddecode?

ld-decode: http://i.imgur.com/eZ7GChe.png

from http://notonbluray.com/blog/ -
2950: http://notonbluray.com/blog/wp-content/ ... n-bars.png
4300d: http://notonbluray.com/blog/wp-content/ ... n-Bars.png

I don't think the potential of PAL LD was ever properly tapped (a CLD-97 level player would have been great, much less an X0-level one!), hopefully ld-decode will show a bit more of what it can do. But I will need to decode digital audio and implement an adaptive 2D comb filter.

I got my PAL Cape Fear today and I've had some success improving the TBC to handle it better. There are definitely still glitches (one single-line error every few frames), but given that all the servo systems are probably degraded while playing a disk the player wasn't meant to play at all, it's quite good. (That said, I'm glad I didn't spring for that set of PAL disks on eBay US)

---

I found an existing project from 2005 to decode CD audio last week and got in touch with it's creator (the source is now on my github as cd-decode), and hopefully soon CD and digital audio will be decoded. Looking at a raw capture from a CD now and still working out the filtering, which is kinda worrysome for the LD digital audio case since much more filtering is required. But it's definitely possible, and having code to handle the subcodes etc is a big help.

Wow, very cool !

Can you decode PalPlus with a software ?

Thanks! Given the specs and a good RF sample, it's possible. Not sure when I'd get around to it though.

Sidney Cadot's setup taps the CD audio EFM after analog filtering, which means it was capturing a nice clean signal - so for my application filtering code will need to be properly worked out.

As an SDR enthusiast...I'm sad I missed all these developments.

I have..maybe one question that was probably answered but I don't have time to read 10 pages....ok..it's maybe two...or three...anyway...

What's the width of the actual RF signal on the LD's? I know the RTL-SDR dongle doesn't have enough resolution to do it; so I'm wondering just how much I need. 8mhz? 10mhz? more?

How much dynamic range are you needing to get a decode from the RF signal? I assume it's a pretty strong signal...so 8 bits might be enough range.

I'm assuming there's no carrier to speak of...so we're basically dealing with RF signals contained in a "baseband"...the same way MPX, RDS, and SCA carriers are part of an FM baseband signal.

Has anyone tried any "IF DSP" effects or have they not been necessary since we're dealing with a direct feed?

I'm kind of wanting to get in to this...I feel I could get better captures than from that Dazzle USB thing I've got...which I don't think could be modded to work in the function to capture this stuff.

I'd say about 13.5mhz to capture all audio and video stuff in one go. For just video you can capture 3.5mhz-13.5mhz.

I have an 8-bit signal now (with some noise, around 8-10 at full amplitude) - I'm looking at ways to improve it. I may actually finally roll my own RF amplifier and do something with that LPCLink2 once I have time.

If someone can find a magic USB3 ~50mhz/12bit ADC, noise levels should improve quite a bit and the output quality could have a decent chance of taking on the best players. Right now the noise is a bit much for that and the 2D comb needs work especially on PAL.

(Pioneer's PAL players are so bad that even my early version beats them easily... but it won't be a complete package until I have CD/LDD audio)

If you want to try this yourself, you should have a decent quad-core Linux system (you can get away with running it in a VM if you have something with VT-d like a non-K i7) and the right capture card (Connexant CX23880 to 23882).

13.5mhz makes sense; I've seen points marked 14mhz on the board for my 505/504. I'm guess I should assume it's referenced from DC, meaning I would need a tuner capable of DC - 14mhz...I could probably heterodyne that up to an IF the SDR can tune (in fact, my existing upconverter with 125mhz LO would probably do it since information below 100khz probably isn't important and filtered out anyway.

The BladeRF runs 40msps at 12-bits...so that's close to your 50mhz 12-bit ADC setup. I don't see why you would need an entire 50mhz of bandwidth; 20mhz @ 12-bit would suffice for a bit more dynamic range.

My current setup is ~28.636mhz@8bit, the 50-56mhz@12bit is just what I'd like to have I don't think you can go much below 27, you need 2x the frequency you're trying to capture. (might be different capturing IQ)

The 14mhz is actually 4 times the NTSC color carrier frequency (~14.318). It's much easier to run a digital comb filter at that frequency for phase/calculation reasons, so the TBC and a lot of other stuff is timed off it.

BladeRF with the upverter board would be nice, but it's about $500-600 more than I can afford right now

Much different capturing in I/Q, your sample-rate basically equals your actual bandwidth. So 1msps actually captures 1mhz rather than 512khz as shannon nyqist tells you. However, you're dealing with two signals vs just one; one channel is in phase with the LO and the other is 90 degrees out of phase with the LO and the system is able to determine positive or negative deviation from center. So if you need 28.636mhz; you can do it with a 30msps ADC that uses I/Q (quadrature) sampling.

I came from the digital audio world, where Shannon-Nyquist fully applies; *granted* you're still effectively using same data rate (since you have two channels of information)....if your ADC can only handle 30msps, at least your stuff will handle 30mhz.