capture vhs tapes

[sander815]

i am capturing 10 older vhs tapes with a shvs videoplayer
What is better, use the composite video output, or is it better to buy a s-vhs cable, transfer that through a DV cam with firewire to a pc?

[Guest]

S-video is preferable to composite to avoid luma/chroma crosstalk (aka rainbowing and dot crawl).

[jmac698]

Hello,
Sorry to disagree, but I do recommend composite. I will explain my reasons and you decide.

First of all, let's be practical. You can just try it both ways and see what you like, that is the *only* correct answer for *you*. One thing to look for, is what looks like crawling dots in areas with strong colors. You could also pay attention to sharpness and noise.

I do have an S-VHS VCR and have experimented with this (in fact restoring analog recordings is my area of research - however I can't answer for *sure* right off the top of my head now).

Regarding quality of cables, I have found no difference for short lengths of 1m.

As for the theoretical answer, most people will recommend SVideo cable as it's capable of a higher quality signal, as a general rule-of-thumb.

The reason for this is that the svideo cable carries two separate signals; one has pure colour, the other has the B/W portion.
A composite cable carries one signal, colour and B/W at the same time. This is not ideal because the two mix, causing the crawling dots pattern.

VHS does store color and B/W separately inside. An S-VHS VCR provides a connector to this internal signal.
However, imagine if the S-VHS tape was recorded through the composite or antenna input - the signals are mixed to begin with, so there is no advantage to recording from the SVideo output.
But the real reason I recommend composite in this case, is because the SVideo *input* to your TV Card will accept an SVideo signal with no additional processing.
The composite input on your TV Card will apply "anti-mixing" filtering to remove the crawling dots effect. Therefore the composite input is best because it will filter the mixed signal in hardware for you, while the SVideo input will not.

Then you might argue, well the SVideo output of the VCR might do this filtering as well. True, but the one in the TV Card is going to be much better, because it is modern, digital, and hardware.
I also don't think there is any filtering on the SVideo output of the VCR (there is however on the SVideo output of Laserdisc players).

If, and only IF, the original tapes were recorded through the SVideo input of the S-VHS VCR, then use the SVideo input of your TVCard.

This is the only case! Use composite for LaserDiscs and VCR's, but never for anything else.

Does an S-VHS VCR playback normal VHS tapes better? Not necessarily, though you might expect so - with mine, the picture is shaky, and I like one of my other VCRs better. The quality does differ however, one of my VCRs is noticeably worse and blurry. For the record, I have 4 VCRs and I've tested and compared them all :) (Though as I say, I can't say for sure off the top of my head.)

[Guest]

@sander815

To refute that nonsense, just try it both ways.

[jmac698]

ps If the tapes were made in a video8 or svhs-c camcorder, the tapes were originally recorded in s-video, so please digitize them through the s-video outputs only. The camera is connected directly inside with s-video connections.

[sander815]

this is the stuff i get through composite:
http://i50.tinypic.com/2u3wlyo.png
http://i47.tinypic.com/jph54l.png
no idea if svhs is better, because i don't have a svhs cable here.

[Ghitulescu]

The signal as recorded on the tape is always separated, regardless how was it aquired.

If the S-VHS recorder has a native S-Video output use it.

Don't use a DV camcorder in pass-through. See my posts why.

Use a TBC if available.

If S-VHS-C then TBC is mandatory if the tape is used in a standalone and not in the original camcorder.

If you doubt whether RCA is better or not vs. Hosiden/DV, I assume you have a capture card in your PC that has only composite input. The answer in this case is S-VHS/DV-camcorder/FireWire, no doubt at all.

It's hard to analyse a video after 1 snapshot - the tapes look like the 1st generation, but I'll still use a TBC.

There are one and only one exception to the unwritten rule of S-Video cable better than composite cable: when the composite signal comes from a studio gear (BNC) to another studio gear (BNC) compared to a non-native S-Video signal to another consumer gear. The latter will have not only the composite defects (because it was originally composite and the comb filter are not the best in consumer gear) but also colour shifting defects due to timing errors (there are two pathways and one signal can be delayed).

[sander815]

this is the only way i can capture it:

i have a svhs video, that i have connect through composite to a dvcam, then through the dvcam's firewire into the pc.
The other choice would be svhs->svhs from the dvcam and then into the pc via firewire

the svhs recorder has a TBC button en 3d noise reduiction or whatever? should that be on?

[sander815]

i have a sample online here:
right click, save as (http://www.sanderkerkhoff.com/test_composite_capture.avi)

[Ghitulescu]

So it's a JVC if I understand your "has a TBC button en 3d noise reduiction or whatever" as "has a TBC and 3D noise reduction button". For JVC this issue is a hot one, as it's model dependent, maybe lordsmurf knows better this issue: since 3D and TBC are coupled, if one of them goes wrong then you have to switch both off. Some JVC models degrade the image quality with 3D ON...

So do your tests with ON/OFF and keep the best shot. Yes, if it's a DVCAM (and not a DV CAMcorder) use it with the S-video cable. A DV camcorder is sensitive to synch errors...

[sander815]

no, its a panasonic
ill buy a svideo cable, see if it makes a difference

[Ghitulescu]

Then there are two buttons, one for TBC and one for 3D NR.

Set the TBC to ON (it works well on Pannies) and test it with 3D ON/OFF (I would let it however OFF, probably this is how the clip was captured).

See if using S-Video you'll still have the colour shift at the beginning and end clips of your video (http://www.sanderkerkhoff.com/test_composite_capture.avi). If you don't see it then it's Ok, go ahead, quality is personal.

[sander815]

this was captured with 3dnrt off and tbc on yes

what do you mean colourshift?
these are recordings from different days, so the colourshift could also be from different settings on the camera, if you mean the light image at the start and end

[Ghitulescu]

Just buy the S-Video cable, it's not a fortune ...

[2Bdecided]

However, imagine if the S-VHS tape was recorded through the composite or antenna input - the signals are mixed to begin with, so there is no advantage to recording from the SVideo output.
But the real reason I recommend composite in this case, is because the SVideo *input* to your TV Card will accept an SVideo signal with no additional processing.
The composite input on your TV Card will apply "anti-mixing" filtering to remove the crawling dots effect. Therefore the composite input is best because it will filter the mixed signal in hardware for you, while the SVideo input will not.Have you actually tried this?

I ask because it's a wonderful theory, but
1. You're talking about comb filtering or similar. Many devices with comb filters disable them when they receive a VHS signal - because the signal is too temporally unstable for the comb filter to work
2. The signal probably really is too temporally unstable for the comb filter to work! Whether a good TBC gets it close enough, I'm not sure.
3. Most importantly, the cross-colour artefacts aren't recorded accurately enough on VHS tape (or even S-VHS tape) for them to still be the same problem that the comb filter is designed to work with after they're played back from tape. It may look similar, but fundamentally, mathematically, frequency-domain wise, the signals (and their problems) are now completely different.

So, if you've made this work, I'd love to see it - because I can't make it work. Fine luma detail causes rainbows on composite. These composite cross-colour "rainbows" are recorded to tape in low resolution, and can't be restored to fine luma detail later.

Cheers,
David.

[zilog jones]

The composite input on your TV Card will apply "anti-mixing" filtering to remove the crawling dots effect. Therefore the composite input is best because it will filter the mixed signal in hardware for you, while the SVideo input will not.
Comb filters in capture cards are not created equal - generally they are 2D comb filters of average quality and will still leave some noise (there's a comparison of comb filters in another thread here). Few have decent 3D comb filters, and even then this is still a compromise compared to having the chroma and luma separated in the first place.

I also don't think there is any filtering on the SVideo output of the VCR (there is however on the SVideo output of Laserdisc players).
VHS VCRs don't need comb filters on the s-video output because as you already know the luma/chroma signals are already separated on the tape recording. S-VHS VCRs do however have comb filters on the composite inputs. LaserDisc players have comb filters on their S-Video outputs because the video is stored as a composite signal on the discs. Some late high-end LaserDisc players have pretty decent 3D comb filters which may be better than your capture card's (if it's only 2D or worse).

Does an S-VHS VCR playback normal VHS tapes better? Not necessarily, though you might expect so - with mine, the picture is shaky, and I like one of my other VCRs better. The quality does differ however, one of my VCRs is noticeably worse and blurry.

S-VHS VCRs are generally of higher build quality and use higher quality heads, so technically they should be "better". But VHS is such an imperfect art. For starters we're mostly dealing with equipment that is 10+ years old now, so it's hard to know how badly these machines have aged - especially with second hand purchases where one doesn't know usage history, etc. (only professional decks will record operation time). And due to poor tolerances in typical consumer VCRs, recordings may play better on one VCR much better than another, especially if the VCR is the same or similar to what the recording was made on. This is true even moreso for LP recordings (don't have much experience with EP myself), which are more likely to play terribly on anything but the VCR that made the recording.

[jmac698]

@Sander815: thanks for letting us know your progress. Once you've captured to your satisfaction, we're here to help you make more improvements in software, if you're interested and willing to learn. Good luck! Don't worry about our technical discussions; some of us are engineers and love this stuff :)

@all:

Let's trace out each step in this process

Step 1a: A composite source from TV tuner or composite in
Step 1b: A true SVideo source is input.

Step 2a: The chroma carrier frequency is down converted to 629KHz in all VHS formats (Beta uses 688KHz, 8mm uses 743Khz). This is done by an analog version of multiplying two sinewaves. In electronics it's called heterodyning. The formula is sin(f1)*sin(f2)=1/2cos(f1-f2)-1/2cos(f1+f2).
In other words, multiplying two frequencies gives you the sum and the difference of those frequencies. We multiply the raw composite signal by an oscillator around 3.58MHz. Then we take the difference, which will be 629KHz, and record that to tape (the higher sum frequency is filtered out).
There may be an analog comb filter before recording.

Meanwhile the luma is FM modulated as a frequency of 3.9MHz +-.5MHz (5.9MHz +-.8MHz for SVHS). This is controlling the frequency of a sinewave with the input. It's a way to store the exact waveform and read it out again. There is nothing preventing the full composite signal being recorded here.

Step 2b: A true Svideo source is downconverted/FM modulated, but basically untouched and recorded to tape.

Step 3: Signal is played back. The SVideo outputs contain raw signals from Step 2. If this was originally an SVideo source, it's untouched, there is no question here. If the TBC switch is on, the signal is stabilized. I believe this is a good thing and should always be turned on. There is some stabilization in digitizer chips (BT878 tv decoder), but it doesn't appear to be as effective as the one in the VCR (I've actually seen this and have samples to post).
Also there may be a 3d noise reduction filter. This is probably digital, but could be affected by the stability of the signal. I wouldn't turn this on, because the variety of avisynth *software* filters will probably work much better, in any case you have full control over software.

Step 4a: Digitizer of device has composite input, which is digitally comb filtered in the tv encoder chip (such as BT878). This digital, hardware filtering is applied on top of any analog comb filtering already done in the VCR when recording. It is almost certainly better than the analog circuit of a 10+ year old device.
Step 4b: Digitizer of some device (camcoder, TV Card) has an SVideo input, which does no comb filtering. In this case, the only comb filtering in the entire chain was that which was done in Step 2, which is an analog circuit from a 10+ year old device.

We must compare the analog comb filtering of Step 2a to the digital comb filtering of Step 4a. Even if the digital comb filter is relatively "bad" (compared to other digital comb filters) it is still "good" compared to a 10 year old analog design. This was my original point.

Step 5: If digitized by a camcorder, the digital result is compressed as a DV format, which adds compression artefacts. These artefacts will probably be multiplied by additional compression in to mpeg2/dvd or some kind of AVI.
If digitized directly by a TV card, you are able to save uncompressed 4:2:2 or 4:2:0 signal (YUYV or YUV).

Agreements
----------
Ghitulescu and I agree on Step 2, and Step 5 with DV compression, and Step 3 with turning on the TBC. I also share the concerns of Ghitulescu in regards to 3d noise filtering, I believe it's much better to do this yourself in software.
As for 2Bdecided, I have no knowledge of a TV encoder chip (aka BT878 etc.) turning off comb filtering. The spec sheet doesn't mention this. As far as it not working with an "unstable signal", that's not true, depending on how the comb filter is implemented. A notch filter as described in BT878 is immune to unstability (that is, jittering of the line). As far as a 3d comb filter, it may have an effect but I'd have to do more research.
Zilog Jones: we agree that comb filters differ by tv encoder chip. BT878 uses a simple notch filter. I agree that *TRUE* svideo is better, but at the same time I agree with Ghitulescu that there are often phase errors in SVideo, which leads to a shift in tint (which can easily be corrected and should be consistent).
"VHS VCRs don't need comb filters on the s-video output" I agree.
"S-VHS VCRs do however have comb filters on the composite inputs", now I know, thanks. But can we agree they are not as precise as the digital version of any tv encoder chip?
"LaserDisc players have comb filters on their S-Video outputs because the video is stored as a composite signal on the discs." I totally agree. I have a laserdisc player too :)
"S-VHS VCRs are generally of higher build quality and use higher quality heads" I agree as well. I think the problem is how old my equipment is. Also playing on the same VCR you recorded on could be better (due to azimuth issues.)

Disagreements
-------------
@2Bdecided: Your point 3, you seem to have a good intuition here! The downconverting in Step 2a has additional effects; obviously there is less color resolution from this step. Here is the calculation, 15.75KHz/629KHz=25, so 25 "cycles" of color per line on VHS. If you adjust for the active video portion, 83%*25=21. There may be also a small correction for "lines per picture height" (the analog version of PAR) depending on how color resolution is reported. This gives very close to the 20 lines of color often reported.
That was the technical explanation for your intuitive guess that "artefacts aren't recorded accurately enough".
I've wondered about this myself, and to test it, I recorded a frequency sweep and played it back, expecting to see some rainbows at a certain point. I didn't find it however, and haven't figured this out yet.


References
----------
www.conexant.com, 878a datasheet:

built-in TBC:
"The Fusion 878A employs a proprietary technique known as UltraLock to lock to
the incoming analog video signal. It always generates the required number of
pixels per line from an analog source in which line length can vary by as much as
a few microseconds. UltraLock’ s digital locking circuitry enables the
VideoStream decoders to lock on to video signals quickly and accurately,
regardless of their source. Since the technique is completely digital, UltraLock
can recognize unstable signals caused by VCR head switches or any other
deviation and adapt the locking mechanism to accommodate the source.
UltraLock uses nonlinear techniques that are difficult, if not impossible, to
implement in genlock systems. And unlike linear techniques, it adapts the locking
mechanism automatically."
However when I actually tried it, the TBC of an SVHS worked better.

Comb filter:
Section 2.3:
"Figure 2-2 illustrates Y/C separation and chroma decoding. Band-pass and notch
filters are implemented to separate the composite video stream. The filter
responses are illustrated in Figure 2-3. The optional chroma comb filter is
implemented in the vertical scaling block. See Section 2.4.
Figure 2-4 schematically describes the filtering and scaling operations.
In addition to the Y/C separation and chroma demodulation illustrated in
Figure 2-2, the Fusion 878A also supports chrominance comb filtering as an
optional filtering stage after chroma demodulation. The chroma demodulation
generates baseband I and Q (NTSC) or U and V (PAL/SECAM) color difference
signals.
For S-Video operation, the digitized luma data bypasses the Y/C separation
block completely and the digitized chrominance is passed directly to the chroma
demodulator."

This *proves* that no comb filtering is done on the SVideo inputs.

[Ghitulescu]

no, its a panasonic
ill buy a svideo cable, see if it makes a difference

And put another sample of the same part of the cassette, so a comparison can be duly made.

[lordsmurf]

@sander815
To refute that nonsense, just try it both ways.
Indeed. You'll find s-video to be better at least 99% of the time -- less chroma crosstalk (easiest to see). Only once in a great while, on a really screwed up tape -- usually mult-gen copy of something -- composite will look better. But that's still largely dependent on the VCR in use.

...

I'll also remind those of you who want to do "capturing by the numbers" that the theory/math behind capturing and the actual process itself have little overlap. The math is great to know when you're building the device, but in a studio it doesn't mean much. Nobody cares about theories that should work. This has always been a "doom9 thing" from my observations through the years. It's one reason I didn't post here very often in years past -- people wanted to theory me to death instead of get useful advice on capturing. Don't get sucked into the theory trap. If you visit a site like Creative Cow, you'll find many skilled pros that don't know the engineering/math (and frankly don't care -- it doesn't help them anyway).

Just a reminder. :)

[zilog jones]

I agree that *TRUE* svideo is better, but at the same time I agree with Ghitulescu that there are often phase errors in SVideo, which leads to a shift in tint (which can easily be corrected and should be consistent).
I haven't experienced phase errors myself, is this an NTSC only thing?

"S-VHS VCRs do however have comb filters on the composite inputs", now I know, thanks. But can we agree they are not as precise as the digital version of any tv encoder chip?
My Panasonic NV-HS960 has a 3D digital comb filter, probably one of the reasons why it cost so much back in 2000. My Philips SAA7134 capture card has only a 2D comb filter. However I don't record much VHS tapes these days so it's not much use to me (the comb filter only works when recording to tape AFAIK).

[jmac698]

Hello,
thanks for comments. I understand what you mean. I agree, the practical result is always the winning answer (in fact, this is the first comment I made in my reply).
As for the theory, don't mind me, I am just enjoying learning how things work. If I'm wrong, that only makes it more interesting. However, I think if you come to a good theory it can predict the reality, after all that is the scientific method and it lets us build space shuttles :)
I am learning from your comments. I'm just finished some other projects and next I am developing a comb filter test cd - one of the first free ones. Watch in other threads!
And your comments challenge me to find the true answers.
Anyhow, it will still be interesting how his project works out, and I will help further if I can.

[juhok]

Can someone post a sample to these so called s-video phaseshifts? I do a lot of VHS and this kind of error never popped up.

Then you might argue, well the SVideo output of the VCR might do this filtering as well. True, but the one in the TV Card is going to be much better, because it is modern, digital, and hardware.It's not better BECAUSE of those reasons. It might be better, might not. I have many modern digital composite inputs and they all suck. One being 2009 model Blackmagic <$1000 professional card. Or did you explicitly mean *TV* cards?

[jmac698]

I was thinking digital vs analog. I never thought of the very last models of S-VHS; these were modern enough to have digital and even 3d comb filters.
I say digital is better because the tolerance of analog circuits is at most, 1%. Digital, even in 8 bits, is far more precise and repeatable, so in general digital circuits are better. There are also things you can do in DSP which are just not possible in analog.
So I was thinking of a simple analog comb filter compared to even a simple digital comb filter.
In fact my S-VHS has no digital comb filter, so I was unfamiliar with this feature.
And amongst digital filters, yes some are 3d and better than other digital ones.
I will be able to prove this soon anyhow.... after I watch my country demolish yours at the olympics :)

[juhok]

Alright. That's reasonable. I have VCR's that are newer than BT8x8 etc. Latest models have DV decoders, MPEG2 encoders, HDMI connectivity and 1080i playback etc :p Not every deck is made in the early 90'ies.

[jmac698]

Really? I had no idea... can you send me some product links? Where I live, VCR technology is obsolete. Maybe a year ago I saw some that were combo with DVD.
I have heard of the D-Theatre, but that is also quite old.
Pointless however. My interest is in analog - I dunno, for some reason I always liked pushing the limits of systems. In fact, one of my projects was a new VHS encoding. I actually tried this. I placed component video into a VHS tape, and it can also be watched normally. I found a way to record live component video into 3xVCR's with simple equipment. Cost me a total of $20 :)

[juhok]

MPEG2/DV; JVC D-VHS HM-DR10000EU (PAL)
D-Theater MPEG2/1080i/etc JVC HM-DH5U D-VHS D-THEATER (NTSC)

There are other interesting models but with these I have experience with.

[zilog jones]

In fact, one of my projects was a new VHS encoding. I actually tried this. I placed component video into a VHS tape, and it can also be watched normally. I found a way to record live component video into 3xVCR's with simple equipment. Cost me a total of $20 :)

W-VHS (http://en.wikipedia.org/wiki/W-VHS) recorded analogue SD or HD (1035i) component video with just one VCR :). Of course analogue "Hi-Vision" was a Japan-only thing so W-VHS VCRs are ridiculously rare.

It must have been hard to sync up 3 VCRs for simultaneous playback/recording. I know high-end decks have edit control interfaces but are these even frame-accurate?

[Lyris]

I dunno, for some reason I always liked pushing the limits of systems. In fact, one of my projects was a new VHS encoding. I actually tried this. I placed component video into a VHS tape, and it can also be watched normally. I found a way to record live component video into 3xVCR's with simple equipment. Cost me a total of $20 :)
That is really cool purely from a "messing with your mind" point of view :D

[jmac698]

zilog: I didn't explain details. There was no experiement which played back true component live, but I could record component live.
I'm afraid we're going off topic. There's probably a forum rule to this. Mods don't ban me! I'm the one who stopped it :)
Maybe I'll write it up somewhere...

[lordsmurf]

I understand what you mean. I agree, the practical result is always the winning answer (in fact, this is the first comment I made in my reply).
As for the theory, don't mind me, I am just enjoying learning how things work. If I'm wrong, that only makes it more interesting. However, I think if you come to a good theory it can predict the reality,

Signs of video wisdom. :goodpost:

Always refreshing to communicate with people that understand things (or are trying to understand).

[jmac698]

@Zilog: I thing betacam SP used a type of interleaving like: Pb,Pr,Y... samples along the video line, then separated them into pixels again. Something like analog YUYV. The article I read said it was some kinda secret interleaving scheme. I could simulate this also, but at a big loss of resolution. You need at least 6 pixels to read one steady pixel. EIA608 knows this, I wonder how big their pixels are?

[zilog jones]

W-VHS uses a similar sounding interleaving of colour difference signals, but requires metal particle tape like Betacam SP (apparently D-9 cassettes are compatible). I don't know any more detail about how the multiplexing occurs. (Sorry for going OT!)

[jmac698]

I did some tests...
the crawling dots was slightly better in composite. Some sharpness was better in svideo. I haven't tested rainbow yet. Tests are inconclusive so far. I think the theory that vhs is too blurry to even make dot crawl is possible. The vcr's inputs does good 2d crawling dot removal.
I am also doing some other cool experiments, like recording rotated anamorphic video. The results are excellent! I get a real resolution of about 480x240, SVCD quality.

[Ghitulescu]

If the piece of video you put it on the net was a direct capture, then the tapes are edited, but quite well. It might be that the artefacts you're seen now are already on the tape and only a vdub/avisynth/premiere/avid filter (or similar) can correct them. I thought to be the originals judging by the stills only, before you put the video. The chroma shift indicates that either two camcorders or two VCRs have been using to shot/edit the tapes.

BTW, 480x240 is not SCVD resolution, it should be 480x480i for NTSC or 480x576i for PAL.

[jmac698]

Yes, but I'm comparing to a letterbox SVCD.

[Ghitulescu]

I haven't experienced phase errors myself, is this an NTSC only thing?

Tint changes yes, it's an NTSC only issue, due to how the standard was designed (I mean, it was not designed to have tint issues, it's just a side effect).

Phase changes and color shifting are kept under control if the S-Video cables are short enough and well manufactured. A TBC also sorts this issues out, if correctly configured.

[lordsmurf]

Some people confuse tint with "lousy camera work", too. The green-looking man may not have been green, but the video signal may not be the reason why. Just something to keep in mind.

[sander815]

If the piece of video you put it on the net was a direct capture, then the tapes are edited, but quite well. It might be that the artefacts you're seen now are already on the tape and only a vdub/avisynth/premiere/avid filter (or similar) can correct them. I thought to be the originals judging by the stills only, before you put the video. The chroma shift indicates that either two camcorders or two VCRs have been using to shot/edit the tapes.

BTW, 480x240 is not SCVD resolution, it should be 480x480i for NTSC or 480x576i for PAL.

this was a direct capture from the vhs tapes. However this has been recorded on large 180Me dv tapes and later on copied to vhs tapes. Too bad our dv player stopped working, so we are forced to use these captures

[Ghitulescu]

If you wait for an used DV deck on eBay, use it, it would save you a lot of trouble (2nd generation, instead of 4th). Or rent one for one day.