Why are my VCR/LD files so much bigger than sat dish?

[KerryDavis]

When I capture video directly from my dish network receiver, an 'hour' TV show (about 40-45 minutes without commercials) runs 900 meg to 1.5 gig or so. The same running-time length of programming coming in from VCR or Laserdisc player, to the same capture card (ATI AIW 8550dv), same software (ATI MMC 9.02, highest version 'supported' for my card), same bitrates/settings (what they call DVD quality) is always more than double that size in the output file.

I figure that the dish-direct signal is very 'clean' even though it's still analog, while the VCR recording, at least (and it doesn't seem like LD should have the same problem), has a degree of 'artifacts' from recording and playback (such as shade changes in 'black' that should really just be 'black') which are being encoded as 'changes' even though they don't really matter.

Can someone recommend some options in, for example, VideoSoap, to overcome this? I don't want to resort to just lower encoding bitrates for a variety of reasons. I looked into the VideoSoap options myself and it seems to be a very complicated thing that nobody - not even the people who MADE it - really understands. The 'split screen' fiddling method (where you make changes to the parameters and immediately see what it looks like on the other half of the display window) doesn't really help me, because I'm mostly concerned with the sizes of the output files and that can't be measured on-the-fly.

[Evander]

Well I can't help you, but thought you'd be interested to know you're not alone. In my case, I capture in huffyUV, from something called a Broadcast Satellite (not in America) and those captures are much larger than caps from my cable box or vcr.

[spectroelectro]

Digital Broadcast Satellite streams use non-standard MPEG-2 and play some games with the GOP and I and P frames to get a lot more compression than you would normally see. The price you pay with DBS is that you'll see some visible macroblocking.

If you capture the DBS stream and you use to burn DVDs it should work fine for you. But if you try to edit it or transcode it, you'll need some real expertise. It's not standard MPEG-2 at all.

[KerryDavis]

I'm not grabbing the digital stream directly, I'm taking the S-video output from the receiver and running that to my capture card. Then the card turns it (back?) into MPEG-2, but whatever the reason, the result is much more compact - and still much cleaner - than what I can get from other 'analog' inputs.

So what you're saying is that the 5 gig for a 90-minute movie is 'normal?' Then how does 4.7 gig on a blank dvd account for '120 minutes at standard play?'

There must be some way to 'fix' this, and it shouldn't be by reducing the overall quality. That's certainly not how it's done for the 'real' dvds.

[bugster]

The size of an mpeg, or any type of video or audio file for that matter, depends on just two things. Bitrate and playing time. Nothing else matters.

How do you set up the capture card. Can you set absolute bitrate limits like min, max, average, or even a constant bitrate, or is it some form of quality based encode you are using. If the latter, this will be because VHS (and perhaps LD, I don't know about this) has a lot more noise in the signal which uses up bitrate. Thus the overall bitrate needed for a given quality setting is higher and so the file size increases.

[KerryDavis]

That's why I was asking about possible VideoSoap options to reduce the VCR (especially) 'artifacts.' Maybe if I could come up with VideoSoap options that are similar to what the DBS operations use to get their high levels of compression? Reducing overall quality/bitrates/etc is both undesirable and apparently unnecessary since the DBS people seem to have figured something out that works pretty well.

[lordsmurf]

I take it you're using an ATI card.

www.digitalFAQ.com and read the ATI guides

17% despeckle VideoSoap may help some.

[bugster]

Reducing overall quality/bitrates/etc is both undesirable and apparently unnecessary since the DBS people seem to have figured something out that works pretty well.

But they have amuch cleaner source than VHS to start with, so they don't need digital filters to clean up the picture.

[KerryDavis]

I seem to have hit upon a combination that's working well. At first I tried using the 'record cropped video' option that one of the pages (might have been digitalfaq) suggested as a means to not encode all the extraneous stuff, but what gets cropped is actually a good chunk of the picture. So that was a time-wasting disaster.

I found that a 'copy of dvd' from the ATI MMC program with the bitrates adjusted similar to the 'good' settings (variable from 6 to 8mbps), plus 20% despeckle, 20% sharpen, and 20% combo filter 1 has been producing some very good results, and the files are actually smaller than the 'cropped video' option was making! so far, so good.

[ofbarea]

... to the same capture card (ATI AIW 8550dv), same software (ATI MMC 9.02, highest version 'supported' for my card), ...

I been using ATI MMC 9.03 since the day it was released and it works fine with my 8500dv. I think ATI will not support us if I have a problem with this version of MMC, but it works.

[lordsmurf]

IAt first I tried using the 'record cropped video' option that one of the pages (might have been digitalfaq) suggested as a means to not encode all the extraneous stuff, but what gets cropped is actually a good chunk of the picture. So that was a time-wasting disaster.

Uh... no. It masks the outer 7% of the image that disappears into overscan. You lose nothing. That part of the image is never on a tv screen anyway. It is certainly not "a good chunk of the picture". Not even close.

II found that a 'copy of dvd' from the ATI MMC program with the bitrates adjusted similar to the 'good' settings (variable from 6 to 8mbps), plus 20% despeckle, 20% sharpen, and 20% combo filter 1 has been producing some very good results, and the files are actually smaller than the 'cropped video' option was making! so far, so good.

If you over-process the image, you will get temporal trails and a "plastic" look to the video. The 17% despeckle will usually remove "extraneous" analog noise that would otherwise have confused the encoder.

[Nelson37]

KerryDavis, how did you verify cropping "removed a good chunk of the picture"? Smurf is correct on this, it merely removes the overscan area. Are you capturing to "DVD-Mpeg"?

Reboot hit on the essential problem, sat companies start with a MUCH higher quality source than VHS. Very much less noisy.

The file size differences you describe do seem unusual. I have never seen a variance that large on same time recordings, using same bitrate settings.

Just realized something. You mention using S-Video for Sat connection, not many VCRs and SFAIK no LD players have this connection. Are you using composite, or by any chance coax, for these devices? Coax is not only inferior signal but runs thru the tuner on the card, a source of significant signal degradation. Composite not that much different than S-Video.

On the filters. Be very careful not to over-filter, and double-check your results. The removal of a little snow may seem so good that you don't notice the corresponding loss of detail, until later.

[KerryDavis]

Maybe it's just an 'artifact' of how TV sets work, but the results of the 'record cropped' wound up with a noticable 'enlargement' of the visible frame, to the point that the 'corner bug' shown on many networks these days partly disappeared into the corner of the screen. When I looked at the detailed settings, 'record cropped' was actually cutting down the overall size from 720 x 480 to something like 525 x 360, which is a pretty major drop.

If record-cropped just 'masked' the extraneous stuff and replaced it with 'black' to fill the previously-used space with something that doesn't take much space to encode, that would be fine. But if it cuts out the extraneous stuff and then 'stretches' what's left to fill the original space - which is what seems to happen - then part of what WAS the 'good' picture will NOW "disappear into the overscan."

I don't know where other people get their LD players, but every one I've had, has had the S-video connection. I also use the S-video output from my VCR to feed the video capture card just because the S-video is a cleaner signal even if the tape was not recorded as S-VHS.

[lordsmurf]

Maybe it's just an 'artifact' of how TV sets work, but the results of the 'record cropped' wound up with a noticable 'enlargement' of the visible frame, to the point that the 'corner bug' shown on many networks these days partly disappeared into the corner of the screen. When I looked at the detailed settings, 'record cropped' was actually cutting down the overall size from 720 x 480 to something like 525 x 360, which is a pretty major drop.
.

You did not follow the instructions. Read it again. Do not skim.
You must enable the CROP function in two places.
In fact, it gives warning that NOT enabling both gives problems.
One is in the overall MMC display settings, one in recording preset.

If you only enable ONE of them, you get incorrect sized video, which is obviously being re-encoded when it's authored.

If you enable BOTH of them, you get a MASK done on the outside 7%.

Try again.

[KerryDavis]

I've had the Crop option for 'display' enabled since day one. I don't know why mine is behaving differently, maybe it's because I'm using the s-video input which it sounds like most people don't, based on the earlier comments expressing disbelief about my LD player and VCR both having s-video outputs.

[thrasimacus]

The source isn't an issue. Bitrate and time is the only thing that matters for file size. This is happening because you are using VBR. Also make motion estimation 100%.

Sat is clean

VHS is not. Dirt=information. Moving dirt=a lot of information. Hence, more to encode, bitrate jumps up because of VBR and you get a larger file size.

[lordsmurf]

I use s-video pretty religiously for most of my work. You either do not have both errors checked properly, or your version/install of MMC is bad. You should stick to ATI MMC 8.7 if at all possible. It is most stable. However, I will say that I have ATI MMC 9.02 and have zero errors here. And this would be the VERY FIRST TIME out of thousands of users, and several years, that any such error on this nature has been reported on any version of MMC that has both options available.