My first vhs-decode comparisons!
 

A few weeks ago I got my hands on a new AG-1980 for my VHS Decode/traditional capture project. Tonight I finally got everything working, so I thought I would share my (rough and still in progress) results with the community. Obviously there will be changes in the works because I think the Panasonic gives a pretty low rf signal, so I'll probably need to amplify it. I should also note that for the traditional capture I used an ati usb 600. The colors are very screwed up on the decode one, but I think it might look a little sharper? However, there is some obvious haloing in the decode capture. Also not sure why the aspect ratios are slightly different. Thought I'd post it here anyway before I go to bed.

Halo/ringing ("ghosted double vision") is a problem seen on 99%+ of all vhs-decode samples. That alone is a good reason not to use it. Note that it's vastly worse on NTSC than PAL, but the presence on both is obvious.

I don't notice any sharpness gains on the RF. For example, look at the thermostat (?) by the front door. The RF isn't sharper, but rather than image is light. That gives false perception of "sharpness" because you can clearly see "more detail" due to the difference in exposure. That's an old photo trick (darkroom days).

My hypothesis for "better sharpness" from the method has been proven false. There are certain situations where a lack of a specific noise will give a "better sharpness" perception. However, in many cases (literally almost all), in an overall frames, the -decode is adding more noise than not. So any gains are massively offset by destructive losses.

It's ironic how the best VCRs have the "worst" RF. And the cheapest crap VCRs have the "strongest RF". I tend to think there's a reason this is the case. I have books on signal theory, but I don't have the time to read them (again).

This could just be compression, but the more I'm looking at the traditional sample, especially around the guy holding the present's right body. It looks like there is a wee bit of haloing going on. But the decode sample sharpens? it to the point that it way more noticeable. I've had my fair share of haloing on traditional captures, but I feel like VHS-Decode would sharpen those already noticeable halos, making them even worse. Making the halos look like they were from a badly made second gen copy.

The worst offender, and it's probably because VHS decode bypassing this process from the VCR, is all those rainbow sploches. I guess that's consider chroma noise? Basically looks like an ugly mess. Do remember seeing a VHS decode sample in which my response was "ew, look at all that chroma noise, it's crap"

I do like this more fair comparison, while it is rough, it shows that at least for now, for this source and for the VCR you're using, that a traditional capture gives better results.

The ghosting looks like an RF echo. Does VHS-decode need either 75 ohm cables or terminating resistors?

The highlights are blown, fixing that should bring the chroma noise down a little, and VirtualDub's Camera Color Denoise filter should help even further.

I think the VHS-decode recovered a line or two at the top, and I like that the head switching noise at the bottom is significantly reduced.

Look again. The normal capture has the usual blur stripe, but it's encroaching on the picture less. The RF version doesn't have the blur, but it's a wiggly mess, and encroaches more into the image. So from a mask/crop perspective, the RF is worse by probably 5 pixels to about 16 pixels here (which is a lot for masking bottom overscan).

This goes back to two of the things I've said for years now:

1. The RF/-decode projects are currently trying to appease everybody (especially every cheapskate). The focus is not on quality VCRs, but rather every old VCR, especially junk/thrift VCRs. It isn't "honing in" on specs/processing to prevent these artifacts. And these are ancient artifacts, like we often saw on local cable broadcasts from the 70s/80s/90s. RF/vhs-decode a huge step backwards in analog processing quality.

2. This is why the project needs dedicated hardware, not just "Radio Shack" instructions to random DIY. Sure, that gives "a signal", but quality is lacking or completely gone.

Wouldn't this just make the "echo" worse? I know boosting the signal should clarify, but the RF noise just seems pervasive to this project.

Open question:
How many of you have seem 60s-70s color TV broadcasts? This sort of echo/ghosting was quite common on transmissions back then. It was hidden on tiny screens, especially B&W, so nobody really cared at home. But the industry cared. It got better in the 80s, and was almost entirely gone in the 90s. Software probably cannot recreate this, it all lies in hardware.

I tend to think the quality of the RF from this project is like something out of the 70s, and not at all like the 90s. Many of the VCRs are from the 90s, and the internal VCR processing (which is not used) is 90s. But the net result of amateur tinkering degrades the signal, even if the raw RF is, in theory, "better" than the VCR. Theory vs. practice strikes again.

How viable this becomes truly depends on dedicated hardware, and focusing on certain models.

I've long been afraid that AG-1980/1970 wouldn't do well for RF/-decide methods, but it's largely because the project is focusing on using junk VCRs. This confirms it. But perhaps Slushly can tinker more, get better results next time?

From day one when I first saw VHS-decode samples I said the frame is stable in terms of line timing and frame timing, Never been impressed by the picture quality itself luma or chroma, Keep in mind Laser-decode and VHS-decode are very different due to the difference in the media medium, The Doomsday project and the laser-decode hardware had much better success because LD optical format is more robust than magnetic tape, magnetic tape is inherently noisy even during playback not just recording, But because LD is a composite format in nature tapping to the robust signal from the disc makes the video looks much better than the processed composite one. Try this yourself, capture a DVD from composite and digitize it, then compare it to backup copy that was ripped from the DVD to computer directly.

On the other hand, there is not much you can do to the tape signal to improve it better than what a S-VHS VCR is already doing in terms of processing Y and C and keeping them separate all the way to the digital phase, So the VHS-decode can only match or be close enough to a S-VHS VCR in terms of signal timing as it stands, Even when VHS-decode picture quality improves it will still be recommended only for low end VCRs.

I will only believe it when I see it that a hobbyists' software can over perform a video processing system from JVC or Panasonic that was designed by full time engineers on a multi-billion budget department. I know we are better equipped today technologically but there are limits to what technology can do when the budget is almost zero.

Slushly, can you provide some video samples? Understanding that what we've seen so far is probably not a best case scenario, I'm still academically interested in having a peek at the levels and such just as a point of reference.

I only did very limited testing with VHS_Decode about a year ago and I always kind of wondered if I was doing it right. I don't think I had that much chroma noise (though the tape I was using was a commercially produced SP tape in basically unplayed condition could have something to do with it), but the video produced did seem rather "light" to me at least with everything at the defaults. I basically had to set the dip switches for the lowest signal for the AG1980 and that's the only VCR I tried it on at the time. AG1980 is kind of odd in that the RF tap point is pretty far away from the head itself - though I have seen alternative tap points mentioned that might be physically closer to the heads and you might get a different result by using those.

I suppose the next test would be to try it on a run-of-the-mill basic VCR that might have stronger RF to begin with and see if the picture is much different playing that same tape.

The argument has always been that VHS_decode should give superior results to high end equipment now and possibly do even better as the decode software improves later. The standard for comparison should really be whether you can get as good or better picture output with a cheap VCR and VHS_D compared to this AG1980. This AG1980 being brand new also has not had any TBC recapping which always made me wonder if those SMT caps just go bad due to prolonged heat exposure, or due to age and electrolyte drying out, so the answer seems to be that unused SMT caps can fare ok 25+ years, at least in some cases.

Also keep in mind that the AG1980 has a sharpness slider which does more than you think it might. If you go too far with it, you'll get some diagonal full image noise usually. I'd probably not press that farther than half way between normal and sharp.

One area where VHS_decode should really shine is with dropouts since it can look a frame or two forward and back to figure out what the missing line "should" look like. Though with my short sample I did, it seemed like the AG1980 had already done dropout compensation on some lines as I saw some vertical line repeats in the VHS-decode-created-video. When I posed that question to the community, they thought it could be that the original production master may have had those dropouts compensated and that all the resulting tapes made from it would also have that effect, so who knows. I really should have tested it with more variety of tapes and VCRs when I had the setup working to learn more.

Very much looking forward to your full results as you do some more testing.

Have an idea where those would be? I'd be interested in tapping them.

Out of curiosity I tried it on my JVC HR-3500u and it came out significantly worse and emitted weaker RF (and in my opinion, it plays almost as good of quality as the 1980 in SP, just more squiggles), so I don't think standard VCRs would make a huge difference. I have heard there are some decks from other brands that emit stronger RF, but I don't know if those give a high quality signal.

All of the JVCs are comparable at the heads, so non-TBC and with-TBC will not give overly different results.

Most of the project was developed in PAL lands, and NTSC decks overall perform quite poorly. Always have, and maybe always will. It's not because NTSC is inferior in any way (old myth), but rather that the format doesn't get any attention. Even many of the better PAL decks perform poorly, as the project focuses more on thrift/junk decks. This is the problem with giving cheapness the top concern, and not quality.

Certain Sony decks are often cited, but even then it's lackluster.

And it's such a newbie/childish argument, too! Anybody that has been around IT for decades knows this is complete BS. "Later" never comes. What exists now is all you can rely on. The "later feature" and/or "better version" is always disappointing, assuming it even comes. This is exactly what I was referring to with my vaporware comment from several years ago, as vhs-decode promises could be complete vaporware longer term.

I'm dealing with this "better soon/later" scenario right now, for the forum upgrade. XenForo 3.0 was pushed, 2.3 was released instead, and 2.3.x seems to have broken as much as it fixed/added. 3.0 is easily several years away now. And understand that XF is a paid platform, with decades of forum coding/usage experience, not some weekend warrior hobbyist project. I can't wait around for the version that's "finally working as promised", but instead have to run with what's best right now.

... and in our video capture context, it's quite clear that vhs-decode is inferior. The best option is to use what we currently have: budget TBC(ish) workflows, or quality/ideal with-TBC workflows. So JVC/Panasonic type S-VHS VCRs, DataVideo/Cypress type fame TBC (or budget end ES10/15, DVK, etc), and quality ATI/Pinnacle type capture cards. Some variations allowed, but in general it's going to be s-video VCRs, some sorts of TBC(s), non-lossy capture cards.

I agree with all this. It's just good science. Add constants (known knowns), verify components.

Why do you think this? The JVC DOC was quite excellent, and Avisynth can repair in software.

This is an aspect where I consider the "all in one" approach (single piece of software) to be foolish. A dedicated DOC software should exist, not some "feature" tucked away into the armpit of some bigger generalist "capture" software. This is a complex problem, and next complex solutions. Not a toggle switch that half-ass "fixes" the problem.

That's such a rare situation that it almost doesn't exist. To me, that sounds more like a person trying to cover up weaknesses of vhs-decode, shifting blame to some obscure unlikely cause.

Yes, same here. I'm glad to see that serious testing is finally being done, as opposed to the misleading newbie tests on Youtube. And based on outcomes/findings here, I'll be able to decide what deck I want to commit to this, if any. Now I'm glad I didn't bother repairing my AG-1970, and may just sell it off cheap since non-working.

I think dropout compensation should be better with VHS_Decode because it can identify what is a dropout and what isn't based on how low the level of RF is on the area in question and it will know for sure that anything displayed there is noise. With traditional captures, static is captured as regular video and AVISynth to my knowledge isn't that great at knowing what is a dropout and what isn't - best filter I've seen is "remove dirt" but I don't think it's quite the same. that, and most VCRs will try to mask it with their own dropout compensator which create what look to me like slightly discolored purple lines of video and that won't appear to be static to the different AVIsynth filters because it isn't, it was modified by the internal dropout compensator before leaving the machine. I've posed the question a while back about dropout identification and masking and some of the more experienced AVIsynth users didn't seem to think it was really possible to do ideal dropout masking without frame by frame editing which VHS_Decode definitely avoids.

As far as alternative tap points, the AG1980 is a K-Mech, so the Panasonic NV-HS1000 should have a similar tap point.

I just tested to which pin on the larger ribbon connector coming off of the head amp has continuity with the test point pin - it's the 3rd one in from the right when looking at the larger 14 pin ribbon connector from the front of the VCR. This point is electrically equivalent to the original test point, however, it does not go through a foot of what appears to be unshielded ribbon cable first. Furthermore, I believe that you can completely disconnect that ribbon cable and the VCR will continue to output the RF we want via that pin plus ground and you eliminate any noise that could be picked up from the cable. It should not be expected for any harm to come from a disconnected ribbon cable there as those ribbons are often forgotten to be re-plugged in after servicing and that specific 14 pin ribbon connector doesn't carry any power signals, just the freshly amplified RF signals from the head amp which is still within its radio-shielded metal box up until it goes into the ribbon cable. My guess is that the VCR has a way to noise cancel similar to how balanced audio works so the unshielded cable may not matter to the VCR's processing, but it could matter if the goal is to directly record that RF. It's also possible with that ribbon cable disconnected that the amplitude is much higher directly at that point which is what we'd want since the RF at the test point seems weaker than some other VCRs described.

If there's a concern about running the head amp without a load (ribbon cable completely disconnected), you could always just place a thin piece of Kapton tape over ONLY that ribbon lead to block just that signal from conducting into the ribbon cable, then attach the VHS Decode box to the output at that same ribbon connector instead again using the 3.3uf capacitor suggested in the VHS Decode documention and then nothing is running "unloaded". In the "tap list" they just use the leg of that capacitor sandwiched into the ribbon contact which can still be done with the caption tape method or by just having the whole ribbon cable disconnected.

In the tap list, they just mention using a 3.3uf capacitor inline with the signal for I believe for DC filtering purposes as capacitors pass AC signals, but not DC signals. Here's a picture I took showing an AG1980 that continues to play with the 14 pin head amp cable disconnected as none of those signals are apparently required for the servo system to stay locked into playing the tape at the correct speed. I can't say for sure if there'd be any risk of damage to the head amp if all of those signals stay unloaded for a long time, but guessing that is not the case. Also of note, they don't mention detaching the ribbon cable, but my guess is they didn't try it as one would assume that the VCR may stop playing if all output from the video head suddenly stopped going to the mainboard, but this apparently not the case at least for the AG1980. It does periodically flash an error code of 00311 and then goes back to showing the tape counter still ticking up as usual, so the VCR knows something is wrong with the ribbon disconnected, but it doesn't stop it from playing - probably for ease of servicing if you don't want to have everything always connected would be my guess.

Attachment 18562

I do think there's something we are missing with VHS_Decode still. I've seen some pretty good transfers that weren't light at all and also lacked that chroma noise, so could be some setting or impedance matching is off in the cables maybe. I know if the fancier kit you buy, they provide 50 ohm BNC connectors and cable and not the typical 75 ohm that is used for most video cable, so that could be part of the reflection if the wrong impedance cable or connectors is being used here to do the tapping with. I would share whatever results with the VHS_Decode discord as they'll point you in the direction of why the capture currently looks less than ideal.

The JVC DOC works the same. Only the image-based (Avisynth) would not. I highly doubt the vhs-decode DOC is better (at least vs. the JVC), but may beat out inferior decks, especially (obviously) non-DOC decks. A lot of PAL decks lack DOC, including Panasonic S-VHS with TBCs.

Perhaps. But that does remove the 1:1 concurrent comparison being run, requiring subsequent plays for the comps. That adds variables that must be accounted for in the comp, and leaves room for doubt.

I'm going to order an ADA4857 FM RF Amplifyer board from Harry and tap that to the regular RF point to see if there is an improvement because that's what the discord guys told recommended I do. What I might do while I'm waiting for that to arrive is buy another ribbon cable and just rip the pad off that one point so I don't have to screw around with the connector.

I built my duplicator, but I know what you're talking about. I'm currently using 50ohm alligator clips which likely didn't help, but I ordered the 50ohm cable that the domesday wiki recommends. When that comes in I'll have to test it. Additionally, I haven't put the duplicator in its case yet because I haven't ordered the copper tape to line it, so that may be a contributor to chroma issues.

Thanks for showing me that other point though because that's likely where I will tap for my final setup.

Also, 12 days ago was the official launch for the MISRC v1.5 development board. I haven't read much about it so does anyone know what its benefits are aside from recording 2 channels?

-- merged --

After further analysis, I think my alligator clips weren't properly attached and also I think the chroma issue is from bad ground because when I was recording a different tape tonight, I screwed around with the ground wire on duplicator and the chroma got worse after looking at that part in the video. I'm probably going to solder to the points for the next capture. The video I have now is better, but still needs work. I did notice some crosshatching in the video as well so I'm not sure how to fix that.

What do you guys think a good test for both methods would be?

Is playing a normal tape a good test or does it need to be a tape with a lot of timebase errors?

How many tapes are going to be tested?

Was the crosshatching with only VHS decode specifically and not visible with regular playback? I still wonder if tapping directly at the head without the ribbon cable installed would get rid of some additional noise that might be picked up by the cable itself. I still think the VCR must use some sort of noise cancellation to use long unshielded cables like that and recording raw RF from a point that is a full foot away from the source seems less than ideal?

I might have to dig out my VHS Decode rig again and play around with it again. When I compare a bunch of VCRs against each other eventually, it'll be included, but wasn't really planning on messing with it until I had to so that software would be most up to date with hopefully some better Mac support by then.

First, I'm just going to test on my family's home videos. Eventually I'll use an s-vhs tape. Right now I'm just testing normal playing tapes. I don't really have any that are consistently bad. VHS decode works really well with the time based errors now from what I'm seeing. Even if there is barely anything on a piece of the tape, the video will still be slightly visible in all the static, which is super impressive.

Yes

After some more reading, I think you're right and something else I read was that removing the tv modulator pack puts less draw on the power supply unit which gives slightly better noise on the power feed. I'll have to try that. What I think I'm going to do is directly solder the amp I'm getting into that one pin you were talking about.

Not sure how the Mac support is doing, but the windows version is officially stable now and I haven't had issues so far.

I've thought about desoldering the RF modulator, but didn't know if it would affect the operation somehow. Certain power supplies don't like it when certain voltages have "no load", though that'd only apply if the power supply makes a certain voltage that is *only* used for the RF modulator which is probably not the case.

I'd be curious how much power use drops without the RF modulator - would be easy enough to check with something like a Kill-A-Watt or similar, would just need to check before/after on it and subtract.

I suppose you could probably just look at the schematics and assuming there's only DC power going in on a pin or two, you could probably just disconnect those pins as opposed to actually removing the part. Since it's mounted upside down on the main board, you'd have pretty immediate access to the pins without any disassembly really required beyond taking off the top cover - I'd probably just use a multimeter to identify any DC power going into it and unsolder those pins, though insulating the pin from the hole it is in might be a little tricky. Would be helpful if someone else can verify that it doesn't cause any issues by doing that since I can't say for sure it won't cause problems.

This is about as far as I got with my AG1980 with vhs-decode. I'm not about to start building an amp board just to get a capture i'm not impressed with in the end.

At this point it seems my Domesday Duplicator is already obsolete and I need to go down the CX card route and pick a more compatible VCR. Really need a "Buy this VCR, buy this card, do exactly this" guide over there.

One thing to keep in mind with the AG-1980. Its head output signal is very susceptible to noise from the power supply. When I was working on a unit here, I had forgotten to tuck all the ribbon cables coming out of the head amp behind the chassis. they were laying on top of the power supply instead and I was getting pretty bad chroma noise on the output.

All of those ribbon cables need to be tucked in/routed between the head amp casing and plastic partition. Any tap wiring for RF decoding should be well enough away from the power supply as well.

Did you use a domesday duplicator ?