Capturing VHS to PC with a high definition resolution...?

[edDV]

@vhelp

For CCIR-601 4:2:2 based 720x480 8bit uncompressed video, there is an average of 2 Bytes per pixel. Each pixel gets a Y sample, UV samples average to one byte per pixel.

So you need to multiply your 37,325MB x2 for 74,650MB.
To that you need to add ~590MB per PCM stereo audio pair.

So figure ~75GB/hr. for raw data.

For 10bit add 20% for ~90GB/hr.
SDI contains some additional bandwidth for more audio and data channels.

SMPTE 259M-C (4:3 SDI) uncompressed serial 4:2:2 = 270Mb/s = ~97 GB /hr data flow
SMPTE 259M-D (16:9 SDI) uncompressed serial 4:2:2 = 360Mb/s = ~130 GB /hr data flow

[edDV]

1280x720 30 fps has 921,600/345,600 = 2.7x pixels so ~data would be

8bit =~ 75GB x 2.7 = 203GB/hr. plus audio and data
10bit =~ 90GB x 2.7 = 243GB/hr plus audio and data

Double that for 60fps 1280x720p59.94 as broadcast by ABC, ESPN and FOX in USA.

These links show ATSC and production HDTV standards
http://www.hdtvprimer.com/ISSUES/what_is_ATSC.html
http://www.miranda.com/library.en/Digital%20Standards/Std_Poster_INV.pdf

[BobK]

That, and a recent consumer survey found more than 50% of people would support Blu-Ray over HD-DVD.

I'd be very suspicious of that survey and how the questions were asked. Over 90% of people don't know what the difference is Blu-ray or HD-DVD.

And the important question would be- what percentage would support Blu-Ray over whatever they're already using?

[SatStorm]

Back to the original question

With the current hardware, any analogue framesize capture beyond full CCIR, is a product of scaling, a fake thing.

In case you don't like mpeg 2, use mpeg 4 at the same bitrates. An xvid file from a good VHS source at 720 x 576 framesize with ~6000 bitrate looks amazing good

A filtered VHS @ 3100kb/s with 352 x 576 framesize looks as good too.

Macroblocks are always present on any mpeg based solution. It is the nature of the "family". Viewable especially when you zoom at any frame at 8X (something some wierdos actually do...)

[Wilbert]

With the current hardware, any analogue framesize capture beyond full CCIR, is a product of scaling, a fake thing.

No, that's not true (well it's true if you mean downscaling, but i guess you mean upscaling). For bt8x8/cx2388x cards for example the sample rate is: NTSC: 28.64 MHz, and PAL: 35.48 MHz, which is well above 13.5 MHz.

[junkmalle]

With the current hardware, any analogue framesize capture beyond full CCIR, is a product of scaling, a fake thing.

No, that's not true (well it's true if you mean downscaling, but i guess you mean upscaling). For bt8x8/cx2388x cards for example the sample rate is: NTSC: 28.64 MHz, and PAL: 35.48 MHz, which is well above 13.5 MHz.

No matter how fast you sample, there are still only 480 or 576 discreet scanlines.

[Nelson37]

SFAIK, there is no capture card currently available which would capture an analog source at any resolution higher than 720x480.

Debating Blue-Ray vs HD-DVD is a totally pointless, complete waste of time and effort at this stage of their development.

Best quality capture method - Huffy or uncompressed AVI @ 720x480, depending on the card. TBC and /or hardware filtering may be needed for VHS.

Best encode method - multipass, and VBR because multipass is not offered for CBR.

Best encode settings - maximum bitrate the format will allow, this is where the Blu-Ray would have a slight advantage of the full 15Mbps bitrate. I do not know if I-frame-only is within spec for either blue-ray or current DVD.

[junkmalle]

TMPGEnc can be set for MPEG2 bitrates as high as 80,000 kbps. Whether those high bitrates get you a better result I don't know. But it does produce the expected file sizes.

I just ran a quick test: 720x480 interlaced DV to 720x480 MPEG2, Constant Quality 100, min bitrate 1000, max bitrate 80,000, 4:2:2 encoding. The average birate of the resulting MPEG2 file was about 27,000 kbps.

[Wilbert]

No matter how fast you sample, there are still only 480 or 576 discreet scanlines.

Did i say there aren't? I was talking about the horizontal resolution.

[junkmalle]

No matter how fast you sample, there are still only 480 or 576 discreet scanlines.

Did i say there aren't? I was talking about the horizontal resolution.

I was afraid that someone who doesn't understand would read your message and think they would get more vertical resolution.

By the way, can those cards actually transfer higher resolution to the computer or is it scaled down to 720 pixels in the hardware/firmware? I know some people capture PAL at 768x576 to get square pixels -- I don't know if 768 is simply scaled from 720 in software though.

[SatStorm]

From what I know, the non bt8xx based cards capture at 704 x 576/480 and scale from there....

768 x 576 is a stupid thing usefull only when you cap with a bt8xx card and don't know how to calculate the "native" framesize of your card

[Red Hat]

Ever thought of just capping in a high resolution then resizing it down to like 480x360 or 512x384 on compression?

[edDV]

dark_myuutwo's goal seemed to be capture at the highest practical sample rate that will be ready for future Blu-Ray AVC codecs.

720x480/576 4:2:2 YUV is the logical target resolution if the capture card can handle it. This is the SDTV reference size for all current and next generation AVC codecs. Uncompressed YUV or lossless compressed huffyuv make the ideal intermediate storage format. dark_myuutwo said he has the disk space to store at that size (either 75 GB/hr or ~30 GB/hr.).

High quality 8 bit A/D assumes the input video has been analog processed for levels and timebase stability. 10 or 12 bit capture would provide more dynamic range for digital image processing later but would require specialized capture equipment and more intermediate storage space.

He should wait for further encoding until the Blu-Ray AVC codecs (VC1 and h.264) are further standardized and refined. Uncompressed input will give best future encoding results.

I think the above is a practical no-compromize approach for SDTV source capture.

[vhelp]

And, to add to edDV 's note:

720x480/576 4:2:2 YUV is the logical target resolution if the capture card can handle it. ...

Even if he encoded to MPEG-2 he might benefit more if he were to encode
his source (using the above) to 4:2:2 by setting (only TMPG seems to
have this feature) TMPG:

** Profile & Level to: [4:2:2 Profile & Main Level (422P@ML) ]
** YUV format: [4:2:2 ]

My guess would be that if blue-ray does win the new standard, then
I am thinking that MPEG-2 will continue to rain as the prefered standard
for video content.

But, (assuming that it does win) The question I would ask is, what
Profile will they use. Will they keep the 4:2:0 or go with 4:2:2 ??

If they go with 4:2:2 then the above is a good preporation route.
Not everything has to be so perfectly GOP 'itized you know. I mean.
You don't have to always follow the old golden rules, use vbr. But
then, there is nothing wrong with it. Its just that the majority of
sources out there are interlaced. And, when we capture it, through
a (lets assume hardware device) an hardware MPEG board, it will be
interlaced.. not IVTC 'ed to progressive.

It would seem that everyone is slowly starting to wake up to the
reality of quality, and the concirns are only now coming to light.
Just because the bitrate will be higher, does not mean that its ok
to skimp lower on a few things. I say, go higher.. as high as you
can go for the given content and media you plan to archive to.
So, if you are able to encode to 15MB, by all means, do so. In the
name of Quality. And, if you can do 20MB, that is even better.
Well, 720 x 480 will look darn good with these bitrates
And, if future re-encodes is in the possibility, then that is the
bonus for using CBR over vbr.. always I-Frame 'like quality

-vhelp 3538

[lordsmurf]

MPEG-2 will very likely continue to be a preferred compression scheme, as it is currently not even close to being used to it's max potential.

The only problem is that fast enough decompressors and adequate space has always been lacking except for the most expensive and sophisticated pro/broadcast equipment.

MPEG-4 (current implementations of XVID, DIVX, WMV) is not as advanced on the high side, not as far as I am aware of. It gets better compression on Full D1 images, but at what visible cost? Some say nothing, but many see errors in how color and pixels are handled. Not to mention it's too "open" of a standard (almost no standard at all). It can also take much higher processing needs than MPEG-2. Not good for a low-dollar mass-use device like a DVD player.

H.264 is outside the scope of what I pay attention to at the moment.

[vhelp]

Yes, I agree w/ LS

Personally though, I don't like the quality from those XviD and
divX and mp4 codecs. I've seen nothing but chisel-chodd nonsense
in many areas. I have several of what is considered the best
XviD encode (cartoon) and I can tell you, though it looks fairly
good at a distance on a pc monitor, one good examination of the
actual video frames, and you can see the chisel-chodd all over
the place. Then, play this on a TV set, and you will no doubt
see them floating all over the place on your TV set. I've been
there with divX and so on. I've tried it and did not like it in
the end.

But, with MPEG, there is a difference. And the quality is next
to AVI if done properly. some time ago, I did a VHS sample clip
and made an MPEG-2 clip. Link is no longer available. But, it
was a good clip, even from VHS. Actually, I have another one, if
you want to see it. Just click my signiture link for latest clip.

Anyways.

There is more to it, then just taking a given source and encoding it.
There is a lot of: technique/knowledge/skill/basic insticts; that
go into the final quality. And all that (and more) comes in time,
through what I consider, scaring.

-vhelp 3540

[edDV]

As far as Blu-Ray goes, current DVD specs will carry forward for SDTV resolutions. There has been no announcement of higher bitrate, higher chroma sampling, 10bit or any other advanced support for SDTV resolutions but that may come. All this needs to be supported in hardware in all players.

Sony already markets a SDTV pro broadcast version of Blu-Ray camcorders that approach Digital Betacam in quality but I forget the precise format spec. I'll look it up.

1x Blu-Ray transfer rate is 36.5 Mb/s for data. Normal HDTV rates being discussed range from ATSC rates (~14-19Mb/s) up to 25 Mb/s and more.

It is expected that the Blu-Ray commercial HDTV DVD spec (MPeg2) will be based on 1080p24 (1920x1080 progressive 24 fps) for film material. Additional interlaced standards will be supported.

In addition to all of the above, VC1 and H.264 have been selected as advanced standards. By advanced I don't think they intend higher quality, but much higher compression that will require very sophisticated MPeg4 hardware decoding in every player. The AVC standards will probably evolve over time with different classes of players. Current HDTV rates for experimental AVC codecs are in the 6-9Mb/s range and falling.

[davideck]

Does anybody know the encoding specs for Mini-DV?
At 25MBit/s, it looks virtually lossless.

The encoder can't be that expensive, because there's one in every Mini-DV Camcorder.
It would be great if the next generation would provide an equivalent quality.

[edDV]

* This is nothing to do with the Blu-Ray consumer spec but shows some of what the re-write Blu-Ray technology can produce. Sony XDCAM camcorders record with these specs:

72 Mb/s max data transfer (per optical head)
23.3 GB per side

Formats supported:

DVCAM 4:1:1, 8bit, 25Mb/s, 85 minutes per side.

MPeg2-IMX 4:2:2, 8 bit, 30/40/50 Mb/s speeds for 68/55/45 minutes per disc side.

http://bssc.sel.sony.com/BroadcastandBusiness/markets/10014/xdcam_info.shtml

[edDV]

Does anybody know the encoding specs for Mini-DV?
At 25MBit/s, it looks virtually lossless.

The encoder can't be that expensive, because there's one in every Mini-DV Camcorder.
It would be great if the next generation would provide an equivalent quality.

DV, DVCAM, DVCPro 4:1:1 @ 25 Mb/s look very good but there are higher forms of DV

DVCPro50 @ 50Mb/s 8bit 4:2:2 (twin codecs)
DVCPro100 @ 100Mb/sec HDTV 1080i and 720p (quad codecs)

The hardware codecs are designed for use in parallel to get these higher rates. See these links.

http://www.adamwilt.com/DV.html
http://en.wikipedia.org/wiki/DVCAM#DVCPRO

[lordsmurf]

Does anybody know the encoding specs for Mini-DV?
At 25MBit/s, it looks virtually lossless..

No more than a high bitrate MPEG-2 file. Consumer DV (at least NTSC) is lossy because of the colorspace compression of 4:1:1. Alright for shooting, sucks for conversion.

[edDV]

Does anybody know the encoding specs for Mini-DV?
At 25MBit/s, it looks virtually lossless..

No more than a high bitrate MPEG-2 file. Consumer DV (at least NTSC) is lossy because of the colorspace compression of 4:1:1. Alright for shooting, sucks for conversion.

Some thoughts,

All formats are a tradoff. MPeg2 in particular has many tradeoffs to adapt its use to varying applications. MPeg2 and DV are both YUV CCIR-601 derived. CCIR-601 is uncompressed 8 bit YUV component video sampled 4:2:2. In serial form, this data stream is equivalent to ~180 Mb/s (22 MB/s) which would fill a DVD blank in about 200 seconds. It has been shown that 4:1:1 (and its cousin 4:2:0) sampling provides adequate chroma bandwidth for most editing and display uses. 4:1:1 (4:2:0) cuts the uncompressed data rate from 180Mb/sec 25% to ~135Mb/s.

For commercial DVD, highly asymmetric MPeg2 encoding is used (i.e long encode cycle, easy decode on a DVD player) to compress about 22x (4:2:0)*. Very high quality sources are used. Specialized processors allow high intraframe compression combined with interframe compression for high quality video display at DVD bitrates (4-9Mb/s VBR).

DV is optimized differently as a realtime acquisition and editing format. The DV format relies on a hardware codec that provides mild intraframe compression (about 5x) and almost no interframe compression** during encoding. This is done in realtime. The same hardware codec is used for accelerated playback decompression in realtime. DV is very good quality and is highly editable.

The 4:1:1 (4:2:0) compromise allows very high luminance quality at 25Mb/s. If 4:2:2 was used instead at the same bitrate, compression would be >7x instead of 5x. Luminance quality is more important than chroma resolution for perceived image quality.

When MPeg2 is used for realtime acquisition in a consumer personal computer or standalone DVD recorder environment, quality is usually compromized severely vs. DV format.

First, bitrates are chosen to match DVD SP, LP and EP (~8/6/4 Mb/s) expectations in realtime. To get there hardware codecs are required but compromises are implicit in 18x-22x-33x realtime compression with thin memory and processing budget.

Realtime software MPeg2 encoding is even more compromised (except memory) but is possible with a fast computer.

The lastest round of standalone DVD recorders with HDD use a three step MPeg2 process that looks promising. The first step captures video realtime at 15 Mb/s to HDD. My guess is this step is mostly intraframe compression. Second, light audio/video editing can be done on the HDD. Third, a second MPeg2 encoding pass is made as the edited video is recorded to the DVD. There is no need for this second pass to be realtime. More time could be used to increase quality.

Whether DV or MPeg2 encoding is used in the first step, the main problem with consumer recording remains the quality of the source. The only quality sources available at home are camcorders and DTV or HD tuners.


* If 4:2:2 was used for DVD, compression would be 30x rather than 22x.
** DV has limited interframe bitrate balancing used to smooth motion.

[Fos]

Mpeg 2 is the best nowdays for backup a vhs tape. As long as it it has become with pro equip.
For me :

-Best mpeg2 hardware :
1º Dc1000 -> can capture mpeg2 at 25 mb ( yea very near to hd quality)
remember that :

Mpeg 15mb = DV qualily
Mpeg 25mb -> better than dv.

-Best mpeg2 software:

1º Ligos lsx
2º Pro coder 1.5 based on ligos

[Wilbert]

I know some people capture PAL at 768x576 to get square pixels -- I don't know if 768 is simply scaled from 720 in software though.

No, it's not. The capture chip samples at 27 MHz (or higher, depends on the chip[*]), which equates 2*720=1440 pixels. If you request a capture size of 768 pixels from VDub, the capture chip will resample from 1440 to 768. (Well not exactly, because it throws away some info before the downsampling.)

There's a utility which can get you the raw VBI data (bypassing the chip's resamping routines). I will look it up for you.
[*] I don't know at which rate the ATI samples. For the others you can check http://www.doom9.org/index.html?/capture/sizes_advanced.html (plus references).

From what I know, the non bt8xx based cards capture at 704 x 576/480 and scale from there....

So, this is false (at least for non-ATI cards). See doom9 link above for the sample rates.

768 x 576 is a stupid thing usefull only when you cap with a bt8xx card and don't know how to calculate the "native" framesize of your card

If by native you mean the capture window such that you get 4:3 footage, then it depends. If you encode to DVD, it will be better to cap at native resolution, because you don't need to resize after that. For MPEG-4, SVCD you probably want/need to resize anyway.

If by native you mean, the raw unsampled data (unsampled by your capture chip), then no. Your cards always DOWNsamples, regardless what capture size you reguest.

A bit nitpicking

@vhelp,

I agree with your 4:2:2 comments. 4:2:0-interlaced is fundamentally broken. But current dvd players already support 4:2:2. I doubt that they will use it in the future.

Personally though, I don't like the quality from those XviD and
divX and mp4 codecs.

mp4 codecs?

But, with MPEG, there is a difference. And the quality is next
to AVI if done properly.

You mean MPEG-2 vs MPEG-4 asp. You are aware that you can put MPEG-2 inside AVI, don't you?

[Cornucopia]

DVD players are working towards more functionality and versatility, not less. I think more players will support 4:2:2. Has mainly to do with the smarts of the MPEG decoding chip, which are constantly improving. May not always be advertised, but it'll be there.

Personally, if you're capping to futureproof, you should cap and save at the best currently available quality. So I would go with 720x480, YUV, 4:4:4 or 4:2:2, uncompress (or lossless) etc. like edDV said.
Obviously, all HD stuff is already digital (not counting Japanese HiVision, etc), so ALL analog cap cards start with SD quality. So make it the best damn SD you can, then upconvert once you have all the HD tools in house. Who knows, in 5 years, there may be an Analog HD-quality cap card--then you can redigitize to get better Horiz resolution (no way to get more vertical lines except upsampling).

Scott

>>>>>>>>>>>
edit: 'Course, who's to say that 4:2:2 would be part of the HD-DVD or Blu-Ray specs. If it's not, it's a moot point.

[Wilbert]

DVD players are working towards more functionality and versatility, not less. I think more players will support 4:2:2.

Well they should, because it's a part of the dvd specs.

[edDV]

Mpeg 2 is the best nowdays for backup a vhs tape. As long as it it has become with pro equip.
For me :

-Best mpeg2 hardware :
1º Dc1000 -> can capture mpeg2 at 25 mb ( yea very near to hd quality)
remember that :

Mpeg 15mb = DV qualily
Mpeg 25mb -> better than dv.

I'm not so sure of that. It depends on the encoding algorithm and intended use. The only way MPeg2 can beat DV is with interframe compression which compromises editing performance to some degree.

DV intraframe compression is standardized and is the same from a $250 Digital8 to a $25K DVCPro broadcast camera. Cheap camcorders and transcoders get to use the same parts because the cost is spread over many standardized units.

MPeg2 capture with all I frames becomes very similar to DV in theory but realtime intraframe compression may or may not be as good as DV. At the consumer level intraframe compression quality is often compromised in favor of speed and cheap parts. Broadcast MPeg2 realtime encoders optimize for performance with tradeoff of size and cost. Open one up and you will see expensive parts including memory and custom DSP. Open a consumer hardware MPeg2 encoder and you see a consumer Conexant encoder chip.

Over time the consumer chips become more sophisticated. Before the home DVD recorder, there wasn't a high demand for consumer hardware MPeg2 encoder chips. DVD recorders are providing some volume but the customers are more cost than quality sensitive, so I think we will continue to see compromise in future generation Conexant and other consumer focused MPeg2 hardware encoders.

[edDV]

DVD players are working towards more functionality and versatility, not less. I think more players will support 4:2:2.

Well they should, because it's a part of the dvd specs.

I think you are confusing 4:2:0 with 4:2:2.

4:2:0 is the DVD MPeg2 standard (NTSC and PAL).

4:2:2 is not supported. If it were, you would need to increase compression 25% or reduce play time by the same amount.

4:2:2 is a production format. Research shows no advantage for 4:2:2 vs 4:2:0 for playback display quality (for human viewers not machines).

[lordsmurf]

I would agree the loss from 4:2:2 to 4:2:0 is minimal, to human eyes. The first two numbers are the important ones. This is why I (and many others) take issue with 4:1:1. It was a stupid mistake. If PAL can be 4:2:0 then there is no really compelling reason for NTSC to not also be 4:2:0, aside from greed of compression space. That lone "1" in the middle of the equation does some havok on color and contrast quality. You simply cannot mash the hell out of chroma without suffering the side effects. But again, you notice it mostly with conversion of sources previously in another colorspace. When shooting DV, not so much (although bright color scene, neon ... you can still see oddities at times, especially in the interlace fields during motion).

[Wilbert]

I think you are confusing 4:2:0 with 4:2:2.

No, i wasn't.

4:2:2 is not supported.

Yes, i guess you are right: http://www.answers.com/topic/ycbcr-sampling

4:2:2 is a production format. Research shows no advantage for 4:2:2 vs 4:2:0 for playback display quality (for human viewers not machines).

Well, i agree you won't see any difference at realtime playback, but if you scroll the frames i'm sure you will see a difference in case the source is interlaced.