Those solutions are not always cooperative with consumer sources like VHS. That's the main difference. The TBCs aren't "the same" as options generally suggested for this task. While I'm glad you found a solution that works, and it's definitely decent, it's not going to be what everybody needs. I refer to as a "closed loop" for a reason, aka single workflow within a device. But if/when there are issues, it becomes an expensive capture card, and external TBCs will be needed. Given the non-consumer source intended, errors are likely with anything other than out-of-camera originals. Sometimes not even out-of-VCR, as even 1st gen VCR recordings could have awful issues.

I wish I could say it's an upgrade, but it's not. It's just different.

And so does your suggested gear, not always cooperative with consumer tapes. As always being said tapes are case by case scenario and every tape has its own problems, but the majority of consumer tapes can be transferred with proper gear whether consumer components or a full fledged pro device, I'm looking forward for someone to send me a problematic tape so I can tinker with its issues, The learning never stops.

As to cost, actually both methods are expensive but the consumer one can be even more expensive, a fully working TBC-1000 is over $2000 and if it doesn't fix the tape you have to get another TBC and hope it will fix it.

I don't disagree with that -- aside from the word "majority". What % is retail, vs. VCR, vs. camera-shot? The first two don't work well with certain workflows, and the latter is indeed case-by-case due to recording gear used. Mostly the era: camcorder quality, tape grade. It is messy. Camera-shot perfection (as "perfect" as you can expect for VHS, at least) is generally okay-ish with lesser TBCs, though you'll still want to scrub it for errors. Actual frame and line TBCs (not "also has TBC" devices) specifically made for consumer sources generally has less hand holding, less monitoring, and "just works". Generally, few narrow exception always exist.

I wish I could do this for you. I'd have to recreate some bad 90s methods to crank those out. Every project has bad tapes like this, and I have a library of bad tapes that took decades to build.

People overly focus on that exact TBC model. It's good, excellent even, but not perfect, not the best. But the best are rare units.

"But the best are rare units."

Such as......?

I guarantee you the reason the Datavideo TBC-1000's and the AVT-8710's are so incredibly expensive is because those were the two touted around video forums as the best money can buy. Some insight is appreciated.

Several are Cypress variants (and possible reverse engineer + upgraded Taiwan-origin knockoffs) that you've never heard of, from companies that long ago disappeared. And in each case, certain sub-versions/variants therein.

The most "common" (not really) TBC is the 4th gen DataVideo TBC-3000, and 2nd gen 4000/6000 (using modified 3000 boards). The production run was more limited, especially on the 4/6 models. It had almost perfect transparency, strength, somewhat decent/accurate proc amp, and luma/chroma values were accurate. I have 3 of these, and it took years to hunt and acquire. The monkey wrench here is condition, not just acquiring the right variant.

These TBCs are generally unobtainium.

The main reasons the TBC-1000 and AVT-8710 were so suggested is
(1) sub-$1k price
(2) easily obtained new
(3) my posts in forums, along with various guides

- First has been gone for years, add inflation, etc. Now about double that max.
- Second has been gone for well over a decade, nothing good new since 2000s.
- Advice still applies, but there are many adds for latter TBCs, more TBCs, budget/frankenstein setups, etc.

The 1T-TBC, for example, can be considered a AVT-8710 "clone", but it has both "green" and "black" (mostly black, post-2010, bad!!!) versions. That TBC is less fiddly than AVT-8710, due to construction, but cost more at the time, almost double that I recall. Not that it was 2x the TBC, but rather sold to a market that afforded 2x the cost. The last time I saw a good 1T unit was back in early 2020. Several actually, dripped into my workflows into 21. Lots of junk units in 2021-2022.

I could go on ... but time doesn't allow. :2cents:

It would be good to capture those using a simple DAQ, similar to what this guy has in mind, so that you can "play" the S-Video signal back through a TBC and capture card you're testing, long after the physical tape has crumbled to dust.

Has anyone done this kind of thing? There's the Domesday Duplicator but it captures the RF from the video head(s) instead of the S-Video signal coming out of the VCR.

I don't see anything special about that link, Using FPGA is not new, it's over a decade old and real devices do exist not just math formula's, His approach is to sample at 16MHz, As I mentioned in post #34 SingMai samples at 27MHz, that's almost twice or 4 times a consumer card. The wheels has been invented and re-invented several times throughout the years. The only exception that VHSdecode has a completely different approach since it grabs the head RF signal which it was never done before, only if they have a working model.

Can the SingMai save the raw voltage samples to disk?

The raw samples are the actual digital signal, Samples are just pixels in a scan line, each scan line has a number of 720 pixels in a 480 scan lines raster. The chroma resolution is different it is sampled as 4:2:2 not 4:4:4 to save data, there isn't enough chroma details to begin with anyway. The VBI and HBI signals are also digitized and used to assign each pixel a H and V positions in the raster.

The frequency oversampling of pro devices has something to do with the way ADC chips work, dithering and bit depth and that kind of stuff, Supposedly oversampling produces clean signal at 14bit and 12bit, The final resolution is always 720 samples @ 10bit or 8bit.

I don't want the pixels, I want the raw analog voltages from the luma and chroma wires sampled at a sufficiently high frequency and bit depth to capture all of the information in the signals, and then be able to replay them back through the TBC and capture card as I explained above.

You can't store voltages, they have to be digitized first. But they are not actually voltages, it's frequency. if you want to keep the frequency as analog just store the tape and watch it as is whenever you like but it will degrade and wear out, that's the whole idea behind digitizing tapes.

That's what capture cards exactly do, they take the Y and C signals and digitize them.

There are online books on how video is stored on tape and converted to Y-C signal for display, recording or capture, it is a good idea to read up a little bit, it's a lot of fun.

Another way of saying it, is that the tape recorded directly the OTA broadcast (NTSC/PAL). So what's contained on the tape is directly what controls the cathode ray tube at the back of your old TV. In pure analog glory. It doesn't get more direct than this. The tape is just a replay of the same broadcast, yes with some details lost due to lower bandwidth and compromises of VHS.

That's why I like the comparison with SDR (software defined radio). They do the same thing, digitize the whole RF spectrum (in practice a small-ish band of it) at a chosen bit depth and sampling frequency. Then later it can be post-processed at your leisure to demodulate or whatever, but you depend on the earlier choices made, for how much details of the original analog signal you've retained. If you didn't use a bit depth or sample rate high enough you may not be able to demodulate the more complex modulations.

Digitizing an analog signal will always be a question of where you put the limit on how granularly you've discreticized a continuous signal.
.
Will you have enough details to differentiate 0.000000001 volt from 0.000000002 volts? Or 0.00000000000001 Hz? Or 0.00000000001 dBm? Or whatever continuous/infinite list of possible values you are trying to categorize/bucketize into x finite possible digital values.

It depends on what you meant by "raw voltage", when here we are concerned ultimately by the magnetic flux of each grain deposited on a plastic tape.
I guess one could argue that the particule size acts like a "digital scan résolution" at a physics level. And if you go into quantum mechanics you could say the same about the magnetic flux of each grain is "digital". But then the Heisenberg principle throws it all into wack because the mère act of trying to measure something so small will influence the reading you get from it.

So when is it good enough? Why try to digitize at the magnetic level? You expect we have better gear today to capture this information, then letting the magic of analog be preserved all the way to the capture card and let the digitization happen at that point in the form of pixels?

At this point, it becomes mostly a philosophical debate to me. I understand the argument that implies we can get a more pure signal closer to the source before any circuit influences it. But it assumes at the practical level that we get a cleaner and more precisr way to measure magnetic quantas in 2022 than the noise imparted up to the capture card. And this hasn't been demonstrated afaik.

My own 2 cents. Hopefully I made sense 🤣

By "raw voltage" I mean the waveforms from the luma and chroma wires in the S-Video output of the VCR. I'm not concerned with how the information is stored on tape.

@jeremfg

Yes, some of what you posted makes sense, But there are facts that need to be addressed:
Video on tape is divided into scan lines, You can't capture more than say 525 lines which 480 is the actual visual frame, the rest is signaling, data such caption, teletext, and VBI. Now, yes each scan line is represented by a frequency signal lasting from the first HBI pulse to the end of the scan line, plus the chroma signal and chroma burst for timing it with the luma, You could argue that you can sample this at as high as you can but Shannon sampling theorem proved that you can't gain any more data than what's on the original frequency if you sample at more than double the maximum of the original frequency value, Meaning that if you have 12Mhz max, you sample at 24MHz, anything above 24MHz is a waste of bandwidth and the original frequency is perfectly reproduced.

Again that's what capture cards do.

Capture cards turn the signal into video frames. I don't want that, I just want the raw waveform.

@latreche34
I completely agree. I was only concerning myself of the information within a line, because yes vertical resolution is already a discrete digital notion.

And yes, the Shannon theorem is what I implied with my oversimplification of "good enough". At some point more details doesn't help. A concept hard to grasp/accept for some of the analog purists and audiophiles out there.

True. I think he only wants the ADC part of the capture card.

To which you must follow up with "what more does it do"?
In essence, nothing. A video frame is just that, an array of "voltages" numerically represented by a color and luminosity.

@traal
If you want the original raw waveforms, just take the picture frame, extract the chroma and luminescence of each pixel, and there you have it, the "raw waveform" as captured by the ADC. There's no further magic. All the information is there, lossless as offered by the digital domain.

Ok so if I have a glitchy HuffYUV capture because I didn't use an external TBC, I can play that file through the S-Video connector on my video card, connected to the input on an external TBC, and take the output of that TBC to a capture card capturing to a second HuffYUV file, and it will fix the video same as if I used the external TBC in the first place, correct?

LordSmurf, do you concur?

Don't make me say something I didn't say.

In theory, yes. In practice, no.

Your typical capture card has "smarts" in it that will result in dropped frames if it doesn't understand the signal it receives. It's trying to recognize a specific number of lines to generate a video frame for example.

That's why we use a TBC in the first place, to modify the signal, to "patch" it. Fix what was broken on the tape or the VCR. A TBC knows what the signal is supposed to look like and surimposes itself in a "I know better than you what the signal should look like". So the capture card is happy and receives a always happy perfect signal. The TBC only messes with the timing/sync parts encoded in the signal, like the "returning the electron gun to the top of the screen", aka as the vertical blanking interval.

A dumb capture card, a simple ADC, would work like you suggest. But then you can just fix the signal in software in post instead of converting back to analog again to feed it to an external TBC, and recapture again. That's the theory behind what drives the VHS-decode project. The software will not do better job than the hardware. The real advantage is that you don't need all the hardware/gear we need today.

If your capture chain flawed you can't recover a corrupted signal, My assumption on all what I've said above is the capture device either digitizes a good clean Y-C signal or it is already built in TBC so it replaces the HBI and VBI by new signals from a weak Y-C signal and add it to the video signal, This has nothing to do with waveform or pixels. The reason VHS-decode can be fixed later is because the hardware is already built in digital TBC, So it's just a matter of combining the timing data and the video data into a useable YUV signal for displaying video. You are treating the Y-C signal as if it is an original complete signal like RF, it is not.

No. Capture cards and video decoder ICs in general will normally only give you the 480/486/576 visible lines of a SD video frame, with the color decoded into two color channels, following the ITU-R BT.656 standard. The data for the non-visible lines, and the horizontal synchronization area of each line are replaced by blank bytes or status codes and won't be in the final file. Additionally, there is often some facilities for capturing parts of the vertical blanking area (the lines above/below the visible image) as a separate stream for decoding teletext, closed captions and the like, either raw output of it or letting the video ic decode that.

To be able to do what you say one would need a way to be able to grab the whole signal, not just the visible bits. The non-visible parts are what actuallly signals where a field and where a line starts, and the reference color data which is what's needed for a tbc or similar to adjust to compensate for instability and corrupted data. Playing back a captured video file on some composite output will just give you a clean video signal with the non-visible parts freshly generated by the video output IC, any errors will have been baked into the video when it was captured.

Digitizing a composite/s-video signal directly is something we've looked into a bit via vhs-decode, but we've had some issue finding a device that will do that properly. The conexant CX2388x-based pci capture cards are capable raw signal capture, but if used with composite input they tend to do some weird re-sampling causing distorted output and we haven't managed to find out how to get around that so it's a bit hit and miss if it works. It's not an issue when using them to capture e.g head amd rf for vhs-decode, so it seems to be something they do when detecting composite input. Most/all software-defined radio devices won't work either as they're designed for radio signals and don't capture the lowest frequencies. Same with the domesday duplicator hardware that was designed to capture the laserdisc rf output (at least not without hardware modification.)

There may be other video decoder ics that have a similar raw output mode as the cx2388x ones, though datasheets are not full always available for either the IC, or even if it is, it may still lack for the USB/PCI bridge chip that is used (some use a bridge chip, some have a combined video decoder and PCI/PCIe/USB bridge chip).

It may or may not also be possible with some devices to set the registers on the video IC to output a raw Rec. 601 component (Y/Cb/Cr) stream without the non-visible parts being replaced with blanks and status codes so one could process the full component signal rather than just the visible parts though it's not something I've looked deeply into.

The datavideo and avt/cypress TBCs use run of the mill video decoder (SAA7111 and SAA7114 respectively) and encoder ICs. I don't know how they are set up exactly, we don't have any ROM dumps of these TBCs. Either they output standard ITU-R_BT.656 and the CPLD/FPGA chip decides what do do with the output of that if the sync parts aren't detected, or they output decoded component and some more stuff is handled by the CPLD/FPGA chip. The video encoder chip will then output a signal with freshly generated sync parts based on the video output from that. Some other TBC units like some of the leitch/DPS ones featured a custom video decoder implemented on a FPGA as far as I know, very old ones had the decoding tasks split up between many chips so those are more complicated.

Not entirely sure what part in the capture cards is responsible for "dropping" frames either, or if that is done by the drivers when they don't detect start of field/line codes or how it works exactly, as it's not done int he video decoder IC at least.

VCR tbcs on decks like the NV-HS1000 are a bit simpler, judging by the datasheets, they digitize a signal delayed by 1 line into a buffer. The Y/Luma part is digitized directly while the chroma is demodulated into the two chroma components. Then there is some chip that looks at the sync parts compared to a reference and modified the output timing of the ics that turn the digitized video back to analog. I think the TBC might also be re-creating the sync pulse but not sure. There is also some logic to compensate for head switch based on the head switch signal and trick play modes etc. Later VCRs like the oft used JVCs have all this combined in a single IC combined with noise reduction and other features. Afaik these won't re-create the vertical sync part that tells where a field starts or ensure a stable 25/29.97 fps stable output so you can still get frame drops and such downstream.

The TBC in many camcorders like the Sony and hitachi Hi8/D8 ones, and the late model JVC VHS-C one I have seem to also have some function to re-generate vertical sync based on head switch if it's missing/corrupted. Some JVC vcrs have a similar thing called video stabilizer but it does not work if the TBC is active on those.

Some of the "pro"/broadcast decks there seems to be able to buffer full frames and integrate drop-out compensation with the TBC also.

No.

Don't confuse line timing with frame or field timing. Not the same. Dropped frames are the consequence of frame issues, which may or may not have anything to do with bad line timing. That's why both TBCs are needed, essentially timing correcting (TBC'ing) both axis, both intra and inter, both temporal and non-temporal.

Well ... usually. Bad signals, TBCs can get confused. The capture card gets confused with or without TBC on really bad signals.

Noting that nothing exists, and may never exist.

Correct.

That remains to be seen. I still call BS on it. That project will be stillborn vaporware without a dedicated hardware kit for capture, TBC, various assist, etc, and requiring specific VCRs. It's simply too much of a variable mess right now, no consistency among results.

Correct.
Or if the source (tape) is bad.

Decode continues apace, it can be continually developed unlike hardware.

The criticisms are fair to an extent, but I've had best results with some very lowly late model junk machines: sample of 1.

Cards are now commodity, I'm still working on impedance conversion with good results. The final impedance matching amplifier card won't cost more than $50US or so and I'm happy to open source it pro bono if people want to improve or roll their own.

I need to nail the design down, and work on reducing a few parts, but to do a basic buffer to match isn't rocket science if you know a bit about amplifier design .

I can not say it's vaporware, I can not say it will wholesale change the video capture landscape.

But it's coming, and people far smarter than me are working on it.

@lordsmurf
Thanks for adding the missing détails and clarifications I was too lazy to add myself. A true expert and professional.

@RobustReviews
My favorite comparison is Greaseweazle, which does a similar thing for floppy disks. It and Kyroflux have revolutionalized the landscape for early computer History preservationists and archivists. Museums and institutions have jumped started that segment.

But they had it much easier. You can just "plug" an old school floppy drive to their hardware. A working one or course, like VCRs there's a whole community of people restoring and repairing them.

Thankfully for them, floppy drives are a lot more common and ubiquitous than VCRs it seems. And way simpler in operation and construction. By comparison, video tapes are the pinnacle of the magnetic-as-a-storage-medium technology complexity, at a hardware level.
With VHS-decode there just isn't an easy turnkey solution when it comes to hardware to read the magnetic medium. Not only you need a working tape transport mechanism and head drum, but it remains very hands on, you need to disassemble a VCR and all. That's why I tend to share @lordsmurf pessimism. A whole side-industry or market of people disassembling old VCRs and creating ready made kits for playing tapes and plugging in to the hardware FPGA/DSP platform will need to exists if there's ever a chance for it to take off. A very tall order. Until then it will remain very niche and obscure, without any meaningful adoption or support hampering it's success. And yes that's a shame because I'd like to see it take off.

But hey, I still hope you guys succeed. Surely someone somewhere out there will have a need for such development beyond the hobbyist market? Sure VHS was always a consumer-only format unlike floppies, but it's been so widely used for over 30 years that I'm sure there's some archivists out there with tapes and deep pockets, and by that I mean a few institutions, that will kick things up at some point.

At least that's my hope.

Greaseweezle is a very apposite comparison, we do occasionally do data recovery and you're exactly right.

VHS decode is directly analogous, it's a powerful tool but considerations must be made.

For me, that's part of the problem. They definitely know some things I don't, but they're not smarter overall. I know consumer analog formats quite well. I'm not a dummy in awe of video magic. "Wow, they're smart, narf, poit!"

What is being proposed is interesting, and there is a chance it can happen. Just not as it exists now. I can see benefits, but also flaws. I'm not a cheerleader, I'm a realist. A lot of supporters of the project have overly wishful thinking, doe-eyed, even naively so. I'm sure the project devs can eventually rig up something that works passably for themselves, but the issue is what are non-dev users supposed to do? What about all the errors that WILL happen, that current devs haven't accounted for? Therein lies the issue.

I monitor the project. Sometimes I see stuff that gets a "hmm" and nod. Others elicit a "WTF?" with furled eyebrows. It's a mixed bag.

And yet, I'm setting aside some important tapes, to redo better (not the the current captures are bad), should it ever happen. But it clearly won't be anytime soon, not for years.

I think the issue is more politized than it should be. It's pretty simple, hardcore hobbyists and archivists will always find/have the appropriate gear to make good captures, They are not waiting for the holly grail to show up, yes expensive and exotic but it works almost as good as the VHS-decode, Such gear is circulating online, I've sold it and it sells. The average Joe looking for a $5 Chinese device just want a convenient way of getting the video signal into storable state regardless of its quality, those are the masses and those are not going to invest in the VHS-decode if it ever happens, they simply don't have the skills to buy one and rig up a VCR or buy a complete system for God who knows how much but it won't be cheap. I think we should stop fantasizing about it, because such devices do exist already, not cheap but they work pretty well.

Bring in a sub $200 price point, which VHS Decode demonstrably provides and opinion will shift.

The old kit seems to be failing with alarming regularity (look at recent threads here) how long do we push the 'buy it, sell it' paradigm?

I'm not bashing a new device design, I'm just saying there is no market for it even among the well established hobbyists. $200 for what? The hardware? How about rigging it to the VCR, troubleshooting it, calibrating it? Do you think the average Joe would fork $200 for such an adventure?

Buy it sell it may sound like snake oil but it is actually the truth, Analog video tapes are not being made or massively recorded anymore so every person has a limited collection of tapes, If done with his project, what is he going to do with the hardware? Sell it of course.

Yes, but development or the application continues apace. I could sell basic machines, my mod and a card for $200 all day.

I'm quite 'comfortable' to be dreadfully vulgar, I don't have a profit motive.

We're a way from that granted, but as even LS is starting to get a reputation for selling equipment that fails in short order, how long do we cling to the 'buy it, sell it' paradigm?

It's a genuine question, and no sleight intended. It's the cold reality of old equipment, it's entropy in action.

I'm not staying VHS Decode is the defacto answer, but it needs to be appreciated a lot of equipment is moving from 'asset' to 'hot potato' status. It's all well and good until it fails on you.

Then what?

I can understand selling the duplicator for $200, But the whole system for $200? I don't know for how long you can sustain losses.

You're a long way out of the loop if you're talking about the DD :laugh:

That's actually the point, You want a perfectly working VCR. So before you commit to selling it you want to make sure you service it and everything works as it should and will last for at least the duration of the project, quite frankly VCR's in good working condition out of the shipping box are hard to come by, Unless you are willing to service them for free because not only you will not be making profit but you will be loosing money.

I tell what, When you start making those systems, I will buy every unit from you for $200 in one condition, you only sell to me.:congrats:

PM me with an order for 100 and we will have a deal.

I'll get my procurement bods on to you, we will need some details etc.

Where are you in the world? Might be worth a face to face.

Not so fast, without a working porotype there is no deal, Demonstrate a fully working and enclosed system and we can discuss the business details.

That's cool, that's why we arrange meetings.

I'm always happy to do business.

PM me some details and I'll see what I can arrange. If you're genuine we might be able to work out a deal.

I will PM you when I see a working prototype.

What, of a working $200 VHS decode with a machine, CX card and impedance matching?

Let me get back from my current business trip?

You're not a business person are you?:laugh:

There is no business without merchandise.

I wasn't asking you to?

If I genuinely thought you were legitimate you wouldn't ask on a public forum :hmm: