TBC still required for RF capture? (VHS-decode)

[lordsmurf]

Quote:

Originally Posted by Novgorod

I agree that it's plenty sufficient for SD RGB (or YUV) video, but I meant trying to put a raw RF waveform through any type of (lossy) image or video compressor. Technically any waveform signal can be interpreted as a grayscale "image" but the high-frequency RF signal will behave very poorly with lossy compression that is optimized for "human vision". That said, it may be worth exploring lossy RF compression with a modern audio codec like opus or aac with a custom quantization table optimized for the spectral information distribution in the RF signal. This should significantly reduce the file size compared to FLAC with a negligible quality loss. But I think that's not of the highest priority for the RF capture/decode project at the moment ...

I've always wanted to like AAC, but everything can go sideways so quickly with it. I know it's matured some in the past decade (overall), but not enough. The problem comes from certain implementations, as caused by trying to avoid licensing, petty dislikes, or even fanboy usage. Too many egos, too much cheapness, not enough focus on quality.

AAC is like MP3/AC3 version 2. Same BS, different day.

FLAC is larger, but has none of the stupid.

[Novgorod]

Quote:

Originally Posted by lordsmurf

I've always wanted to like AAC, but everything can go sideways so quickly with it. I know it's matured some in the past decade (overall), but not enough. The problem comes from certain implementations, as caused by trying to avoid licensing, petty dislikes, or even fanboy usage. Too many egos, too much cheapness, not enough focus on quality.

AAC is like MP3/AC3 version 2. Same BS, different day.

FLAC is larger, but has none of the stupid.

I'm not pretending to be an audio codec expert, it's just that from all I've read about it, Opus seemed to be the "best" in terms of "quality per bitrate" and latency .. But I really have no idea what's the best lossy codec for high-quality archiving when bitrate and latency both aren't a constraint. But at that point you might as well go lossless (FLAC). What are a few gigs of audio between friends anyway?

[timtape]

Speaking of audio comparisons, does anyone seriously suggest capturing the actual 1000 kHz carrier wave of an AM radio station broadcast for improved audio quality? Or the 100 MHz wave on FM? Im not aware of it. If there were considerable losses in conversion from carrier wave to the much smaller audio wave wouldnt conversion to audio in radios have long ago been implemented by a dedicated digital chip?Sure the analog conversion of a video signal off tape is much more complex but has anyone bothered to do the work of quantifying the actual quality losses incurred at each stage of analog hardware down conversion? At least doing that would describe the extent of the alleged problem at each step and by how much. I'm surprised there seems little interest in doing this preliminary analysis. To me it seems fundamental. It just seems assumed by proponents and even by some detractors that analog decoding is by definition lossy. Where can we find the engineering and photographic evidence?

[lordsmurf]

Quote:

Originally Posted by timtape

Speaking of audio comparisons, does anyone seriously suggest capturing the actual 1000 kHz carrier wave of an AM radio station broadcast for improved audio quality? Or the 100 MHz wave on FM? Im not aware of it. If there were considerable losses in conversion from carrier wave to the much smaller audio wave wouldnt conversion to audio in radios have long ago been implemented by a dedicated digital chip?Sure the analog conversion of a video signal off tape is much more complex but has anyone bothered to do the work of quantifying the actual quality losses incurred at each stage of analog hardware down conversion? At least doing that would describe the extent of the alleged problem at each step and by how much. I'm surprised there seems little interest in doing this preliminary analysis. To me it seems fundamental. It just seems assumed by proponents and even by some detractors that analog decoding is by definition lossy. Where can we find the engineering and photographic evidence?

Excellent post.

The truth is that you start to have diminishing returns, nth % gains. Where and how diminished would make for an interesting objective read. That would involve some advanced imagery quantification, the kind you get with high-end reviews for high-end DSLRs. Or filters or lenses. It's easy to point out junk, but you start to get tradeoff scenarios at the higher end.

Even ignoring imagery gains/differences, you have added signal quantification differences and concerns (demodulation/SNR/etc), and that's where some of the RF/FM/vhs-decode issues are found, as it's unfinalized software that performs it. They're trying to build a better mouse trap, but first have to recreate the existing mouse trap to build on.

But noting it can be difficult enough just to get a normal person to use more than an Easycap and thrift store VCR. Those same folks confuse scientific reviews with BS like "vinyl sounds better, because something something, narf poit zork". So nichest and niche readership, a minority of even the RF fans.

I actually have an AG-1970 that I'd be willing to repurpose for vhs-decode testing, and a decent PAL test library (as NTSC is still craptastic). I doubt many have the gear collection I do, for comparative testing. My issue is time. Let''s see where I am next year.

[Novgorod]

Quote:

Originally Posted by timtape

Speaking of audio comparisons, does anyone seriously suggest capturing the actual 1000 kHz carrier wave of an AM radio station broadcast for improved audio quality? Or the 100 MHz wave on FM? Im not aware of it. If there were considerable losses in conversion from carrier wave to the much smaller audio wave wouldnt conversion to audio in radios have long ago been implemented by a dedicated digital chip?

Funny you mention it - SDRs have been doing that since their inception ..

You're also comparing very different signal types and applications. AM/FM radio is extremely narrow-band and the 100MHz FM carrier is very impractical for direct capture so it needs at least analog down-conversion, which is what SDRs do (or any other tuner). And on the application side there's just no demand outside of hobbyist tinkering and government-level surveillance (it's convenient to be able to listen to all radio stations at once, I suppose). It's not the 70s anymore, nobody is recording analog radio nowadays when digital streams are available, and wherever analog radio is still used, the FM demodulation is probably done with some "dedicated digital chip" after baseband downconversion (like in your smartphone) and that's far from being the quality bottleneck.

In contrast, RF digitizing of video tapes is purely a conservation effort. Nobody is suggesting doing that to the few remaining analog TV broadcast channels. More importantly, the RF signal on the tape has very little in common with an FM radio channel (other than some tape standards including FM audio either as a mixed-in band or on a separate magnetic layer). The signal consists of at least two differently modulated components (luma FM and chroma QAM) and the modulation bandwidth is close to the carrier frequency for both components, i.e. the signal is as broad-band as it can physically be. The proper band separation, demodulation and de-emphasis filtering in order to recover the original Y/C video is very complex to say the least, and then you still have TBC and image processing left.

The purpose of RF capture is to have, in principle, access and control over all these steps and to abstract the decoding process away from the physical hardware and tape (after it has been digitized). So the key advantage is the flexibility you get from preserving a raw unprocessed tape image that you can tinker with on the signal level instead of having to commit to a particular implementation from some manufacturer with little adjustability. The down side is that you have to tinker with the raw signal and can't just dump your tape library into a magic software blackbox that spits out the best possible video (yet )...

[latreche34]

In my opinion the team should have started from a later model JVC VCR with built in line TBC and tap the timed digital scan lines from the TBC's memory, That's easy to achieve and doesn't require a lot of signal manipulation since the signal is already digital. Also doing other formats should follow the same process, just knowing the signal pins of the line TBC, V8/Hi8, SuperBeta/ED Beta, uMatic with TBC card, Betacam/Betacam SP .... etc. Once that's perfected they should move to RF acquisition for very old formats that don't use digital timing first, such as reel to reel, V2000, CVC and the likes, then expand to more formats.

[Shakedown St.]

Quote:

Originally Posted by Novgorod

So maybe some SDI capture hardware might be bastardized for RF capture, but I think at the moment nothing is economically competitive

I would honestly much prefer work, effort, and time being spent by the community going in this direction rather than trying to make a software. This sounds like a much better idea being honest lol.

On a side note... not even RF capture. The JVC DVHS decks (3000/4000 etc) that were made in the late 90s have SDTV mpeg encoding in 720x480, that accept analog video signal from s-vid. Back in that time, it was being used as a popular method to convert old hi8 tapes then displaying on early high defintion sets. Best of all the JVCs allow you to select multiple compression/recording time choices up to 30 hours.

Nope, these decks don't have built in TBC... but take the Pan 1980... with playback heads good enough that on most tapes you have TBC turned off. Record the Pan deck with DVHS tape running from another unit... you now have a digital recording, which can then be transferred by DV firewire @ 9.2mpbs (avoid) or HDMI on back of unit. Totally negating the need for both line and frame TBC, which is the argument that software TBC is more transparent or "sharper" because less steps involved.

P.S This is something I have been experimenting with and am anxious to see the results in comparison to my TBC setup. If you're a pro transfer house and you have the best of the best that VHS has to offer, DNR is good enough in 9/10 cases. This is another method of making digital recordings and possible implementation of rf capture.

[Novgorod]

Quote:

Originally Posted by latreche34

In my opinion the team should have started from a later model JVC VCR with built in line TBC and tap the timed digital scan lines from the TBC's memory, That's easy to achieve and doesn't require a lot of signal manipulation since the signal is already digital.

I don't think the TBC memory is usually accessible unless the processing chip clocks out the data for debug reasons. So it may work on some models with a fast enough logic analyzer and wouldn't be too hard to implement (except for potential micro-soldering). But this approach would be essentially the same as digitizing the final processed Y/C video output, only omitting one D/A-A/D step and the associated minimal noise figure (depending on the DAC/ADC quality). And S-video capture is pretty much a long solved problem, even with software decoding of the raw digitized Y/C data in case you don't trust your TV capture card (I did that myself). But you're still stuck with the particular hardware implementation of the demodulation, filtering and TBC. RF capture provides a couple more layers of adjustability - if you can implement it correctly.

Quote:

Originally Posted by Shakedown St.

I would honestly much prefer work, effort, and time being spent by the community going in this direction rather than trying to make a software. This sounds like a much better idea being honest lol.

I was still talking about digital RF tape imaging, not the processing and decoding software. The capturing part is really not that much harder than digitizing video and you don't have to bastardize SDI converters or DVHS recorders for that. Any appropriately spec'd SDR capture device will do (like DVB tuner cards with custom drivers on the cheap end) or any other fast enough ADC. To be clear: what you end up with is a digitized RF waveform as an image of the magnetic signal on the tape (as seen through the read head) - more akin to an audio signal (but at much higher frequencies) than video. You still need the right software tools and filter parameters to convert this into a viewable video.

Quote:

which is the argument that software TBC is more transparent or "sharper" because less steps involved transferring.

A TBC is not responsible for "sharpness", it "just" compensates for tape flutter by aligning the horizontal lines in each field and properly dealing with drop-outs. A software TBC (like anything "software") allows for more control, e.g. how the start of a line is detected and what method to use for quantization (converting the sampled analog signal into a fixed number of discrete pixels per line) and optional other corrections. For example, for a Video 8 source I can properly recover the few lines at the bottom of the image after the head switch with a (self-made) software TBC, which the built-in hardware TBC of a Sony camcorder just ignores.

In terms of actual "sharpness", RF capture can at least conceptually retain as much as possible of it by letting you choose the filters to apply at each stage starting from the raw RF signal. Many players (especially camcorders) tend to over-soften the image (already at the de-emphasis stage) in order to get rid of the noise and then add a lot of artificial edge sharpening to compensate for it, which makes it look a bit unnatural. It may very well be a matter of taste, but wouldn't it be nice to have user control over it? In the end RF capture/decoding comes down to meticulous filter tuning with a ton of parameters, which in the long run will be probably more of a hurdle than the physical capturing or even the proper software implementation, so I can definitely understand why it would throw people off - and that's totally fine ...

[latreche34]

Quote:

Originally Posted by Shakedown St.

The JVC DVHS decks (3000/4000 etc) that were made in the late 90s have SDTV mpeg encoding in 720x480, that accept analog video signal from s-vid. Back in that time, it was being used as a popular method to convert old hi8 tapes then displaying on early high defintion sets. Best of all the JVCs allow you to select multiple compression/recording time choices up to 30 hours.

Not really, as I posted above a link to a VH thread I made, D-VHS is not good for digitizing VHS, I've tried it, Its S-Video out with conventional lossless capture method yields much better results, more chroma and luma dynamic, more sharpness.

Quote:

Originally Posted by Shakedown St.

Nope, these decks don't have built in TBC... but take the Pan 1980... with playback heads good enough that on most tapes you have TBC turned off.

JVC made several S-VHS decks with built in line TBC up to the very late 90's using mostly chips SoC, 1980 VCRs are not good candidates for these hacks, too many analog boards, too many wires.

Quote:

Originally Posted by Shakedown St.

Record the Pan deck with DVHS tape running from another unit... you now have a digital recording, which can then be transferred by DV firewire @ 9.2mpbs (avoid) or HDMI on back of unit. Totally negating the need for both line and frame TBC, which is the argument that software TBC is more transparent or "sharper" because less steps involved.

Again check post #77, I've already tried that, With my analog to SDI devices I achieved better results than any USB capturing method, D-VHS or DV, Even through youtube they come out clean, the actual captures are way better.

[latreche34]

Quote:

Originally Posted by Novgorod

In the end RF capture/decoding comes down to meticulous filter tuning with a ton of parameters, which in the long run will be probably more of a hurdle than the physical capturing or even the proper software implementation, so I can definitely understand why it would throw people off - and that's totally fine ...

That's it, Not for everyone. Reminds me of professional camcorders with tons of manual controls vs point and shoot consumer camcorders with everything automatic. Without the need and knowledge of what you are doing, it's better to stick with an automatic system.

Some people can't even handle a multi step lossless capture, They just want a device that converts, de-interlaces and encodes, all in one step.

[timtape]

Quote:

Originally Posted by Novgorod

... In the end RF capture/decoding comes down to meticulous filter tuning with a ton of parameters, which in the long run will be probably more of a hurdle than the physical capturing or even the proper software implementation, so I can definitely understand why it would throw people off - and that's totally fine ...

I'm interested in a typical meticulous fine tuning regime for say an RF capture of a VHS tape. Specifically which parameters are we talking about and again, how much of an improvement will tweaking make to the picture and why, compared to a good analog processed capture.

I'm talking about a range of cases from where it makes little or no improvement to where it makes a large improvement, and examples in between those. A/B still and motion examples are welcomed.

[Shakedown St.]

Quote:

Originally Posted by latreche34

Again check post #77, I've already tried that, With my analog to SDI devices I achieved better results than any USB capturing method, D-VHS or DV, Even through youtube they come out clean, the actual captures are way better.

I see your post now interesting stuff. Giving it a read.

[lordsmurf]

Quote:

Originally Posted by Shakedown St.

I see your post now interesting stuff. Giving it a read.

I see your post, and raise you another post!

We have to have some levity in here, to keep it friendly, we're all too serious at times.

[Novgorod]

Quote:

Originally Posted by timtape

I'm interested in a typical meticulous fine tuning regime for say an RF capture of a VHS tape. Specifically which parameters are we talking about and again, how much of an improvement will tweaking make to the picture and why, compared to a good analog processed capture.

Typically it's tuning a bunch of filters and (optionally) software TBC parameters. VHS-decode does a lot (if not most) of it automatically following the spec which works more or less well depending on the tape and RF pickup. But you can override the defaults and tweak everything yourself - of course it's not intended for beginners, it's still experimental software at this point. I'm working on my own implementation (from scratch) for V8/Hi8 tapes which requires more complicated de-emphasis filters compared to VHS and thus more parameters to tune.

A "typical meticulous fine tuning" could consist of these steps:

• Band-pass filters for the luma (FM) and chroma (color-under QAM) bands (parameters: center, width, steepness)
• Possibly some EQ of the FM-luma before demodulation to flatten the frequency response of the tape (parameters: pass-band shape)
• Static de-emphasis for (demodulated) luma and chroma with shelf filters (parameters: gain, corner frequency, order/steepness)
• For V8 and Hi8: HF dynamic range expansion (a.k.a. nonlinear/dynamic de-emphasis) for luma and chroma; may also apply to VHS (at least the luma part) but I'm not sure about that (parameters: filter settings for the level detection and all kinds of compressor/expander settings similar to audio processing depending on the implementation)
• TBC: can be essentially automatic, but you might want to choose which edge and level of the sync pulse you want to align at; also fine-tuning of the chroma burst timing for Y/C alignment and to select the exact phase reference, since the color-under burst is not as neat and well-defined as the normal PAL/NTSC burst.
• Optional filters: if the signal suffers from relatively narrow-band interference (such as Y/C cross-talk or some other carrier bleed-through), you can apply notch filters anywhere in the processing chain; also low-pass filtering or de-noising can be added here.
• Chroma filter: line-averaging to cancel out phase drift as per the PAL spec, or any other vertical smoothing filter - this and other image corrections should be also available in the traditional workflow, so from here on it becomes the same.

If it sounds like a lot of work, that's because it is. For some parameters you have direct feedback from your signal (luma sync pulse shape, RF waveforms and spectra, correlation plots etc.), others you have to calibrate with a test pattern, but the parameters may change for each recorder (or even for each recording), so it can be hard to verify whether you got "correct" results. Therefore it's always advisable to compare with an S-video capture to see if you can recover more detail with your custom filter settings. This will depend on the particular player and source material, so you're definitely not guaranteed to achieve a significant improvement over some commercial hardware solution and you have to make your own decision whether you really need to improve it and whether it's worth to invest the time to tinker with it (i.e. how important is that tape for you)...

For me (as I said it many times), it's first and foremost academical to explore what kind of control you can have over the decoding process, just for the heck of it. I also see a tangible advantage for V8/Hi8 tapes because camcorders tend to over-sharpen the image (losing low-dynamic detail) and the built-in TBC cuts off some otherwise useful parts. Maybe a studio-grade V8/Hi8 deck would do a much better job and also allow granular (analog) control over the decoding, but that's unobtainable for me and if it can be achieved with some ugly home-brewn code, it feels like the way to go ...

[lordsmurf]

Quote:

Originally Posted by Novgorod

Typically it's tuning a bunch of filters and (optionally) software TBC parameters. VHS-decode does a lot (if not most) of it automatically following the spec which works more or less well depending on the tape and RF pickup.

I've not seen any evidence of this to date.

PAL samples do always suck less than NTSC, and PAL tends to have less of certain problems (instead having others), so that can skew observations. The head-to-head tends to be faulty comparisons, or lackluster when 1:1'ish.

I recently remembered I have an AG-1970 that can be used, and will look into RF again next year.

[Novgorod]

Quote:

Originally Posted by lordsmurf

I've not seen any evidence of this to date.

Do you mean parameter tunability or "good results"? I can't speak for the results because they're not mine (and my own V8/Hi8 stuff is not finished yet) and they need to be compared with a good-quality S-video capture from the same player, so I agree that the results are still "mixed" at this point. As for the tunability, it's simply the nature of software processing that "everything can be tinkered with". How far this goes in practice, including user-friendliness, depends on the implementation.

[latreche34]

Quote:

Originally Posted by Novgorod

Do you mean parameter tunability or "good results"? I can't speak for the results because they're not mine (and my own V8/Hi8 stuff is not finished yet) and they need to be compared with a good-quality S-video capture from the same player, so I agree that the results are still "mixed" at this point. As for the tunability, it's simply the nature of software processing that "everything can be tinkered with". How far this goes in practice, including user-friendliness, depends on the implementation.

Do you have any samples to share? How's the DOC implementation on those formats?

[Novgorod]

Quote:

Originally Posted by latreche34

Do you have any samples to share? How's the DOC implementation on those formats?

What do you mean by DOC? The 8mm tape spec (IEC 843-X)? It's more of a guideline and the manufacturers don't seem to use the exact filter specifications, at least I have to tune the "spec" values quite a bit to get it right (or it could just be the difference between an analog and digital implementation)...

Anyway, here's a sample which is as apples-to-apples as I could make it. It shows a single field (full-frame) comparing S-video capture with RF capture/decoding (luma only and full color). It's very much still a work in progress and some processing parts are not implemented yet, such as proper correction of chroma phase fluctuations and nonlinear chroma de-emphasis:
https://i.imgur.com/9ustMij.png

Source: Video 8 LP recording, PAL
Player: Sony DCR-TRV238E (used for both S-video and RF capture)

No post-processing was done to both images other than vertical up-scaling (i.e. interpolation deinterlacing using just one field). Keep in mind you're comparing the latest 2000s-era Sony SoC with integrated decoding, TBC and post-processing from more or less the final generation of analog capable camcorders to some caveman code cobbled together in a few weeks, so there's no expectation of a superior image quality. As I mentioned before, I could recover the image parts cut off by the built-in TBC and also have fine control over the filter parameters (which were tuned to roughly match the spectrum and waveform of the S-video reference). Also, the hardware processor in the camcorder seems to swallow the color signal at the end of each line, but only for LP recordings. When the color processing part of the software is finished, I expect a better match, and I also plan to try some different approaches for the dynamic range expander (nonlinear de-emphasis). So it's still pure tinkering at this point.

[timtape]

Thanks for the A/B comparisons. Well done. To my eye, very little difference. Possibly a little more detail in the S-Video luma but maybe just some artificial sharpening at work? A motion picture comparison would be even better.

[Novgorod]

Yeah, the camcorder is doing a pretty decent job at decoding, so the goal (for me) wasn't to make the picture substantially better (that's limited by what's on the tape anyway), but rather to have granular control (e.g. sharpness vs. noise/grain) and fixing "glitches" of the hardware (such as the TBC encroaching on the image, the missing color at the right edge and the occasional long drop-outs after a tape hick-up or recording cut). I'll export video for comparison when I'm done implementing the remaining features because (especially) the chroma processing is not in a usable state yet.