View Full Version : Capture Cards: Hardware or Software MPEG2

5th June 2003, 16:56
I have been using ATI AIW capture cards for years now. Started with the original AIW Pro in 98' and got the Radeon 7500 AIW last summer. Now that I have a DVD burner my demand for quality in my caps has jumped significantly from VCD resolution to full DVD resolution. I havent been very impressed with ATIs "dvd quality" captures using MMC8.5. The only option I have for quality caps is with VirtualDub using Huffy codec and then encoding to MPEG2 with TMPGEnc or CCE. This produces acceptable results but it takes forever on my 1.2Ghz machine. One hour of video processes for five hours. Also I cannot use my capture card in a scheduled Tivo-like way. One hour of raw Huffy video will nearly fill my disk up. Ive been doing research and noticed there are a lot better cards out there than the ATI. Some include hardware MPEG2 encoding as opposed to the ATIs software encoding. My question is does hardware MPEG2 encoding make any difference whatsoever. Is the quality of the cap cleaner? Or does this just mean that the processor will be used less since the chip on the card is doing all the work. Also if it is a hardware encoder on the card does this only work for "realtime captures" using their included software or will this also allow me to encode using TMPGEnc or CCE in realtime?

5th June 2003, 18:11
Hardware MPEG-2 encoding will result in much higher bitrates than software real-time encoding unless you have a really fast CPU. You can always capture at a high bitrate (8-10 Mbits per second) and later re-encode with TMPEGEnc, etc. for VBR. The encoding quality also depends on the motion-search algorithms. Possibly, lower bitrates will be more sensitive to this. Really good MPEG-2 hardware encoders are very expensive.

I do re-encoding with TMPEGenc on a 1.5Ghz Athlon XP CPU and it may take from 6 to 10 hours to VBR re-encode a DVD movie - I just leave it running overnight ... or during the day when I am at work ... ;)



6th June 2003, 21:44
I have upgraded my pc for just this reason.
I capture directly into mpeg-2 dvd full res using MMC 8.5 and I am quite satisfied with the results. I have jump the settings up a bit from the dvd high default preset that ATI gives. I use 8mbit cbr and set the motion search setting to highest setting
then demux with tmpgenc. 10 Min
re-encode audio to ac3 using besweet 5 Min
re-encode video using cce (2 hours takes me roughly 1:15)
Then input them into Maestro. 5 Min
build and burn with recordnow Max. 40 Min
Total Time after capture: under 2:30

6th June 2003, 23:37
Why capturing in MPEG2 if you do some reencoding? This does not make sense. Yuo can have better Results capturing with a cheap TV-Card using MJPEG...

9th June 2003, 14:50
I did not know it was possible to re-encode an MPEG2 video in CCE. What bitrate and resolution do you re-encode it at? You must have a very fast machine if you can do two hours in less than realtime.

9th June 2003, 16:14
When it comes to capturing Mpeg2.

You might find some of FredThompsons posts useful and informative!

15th June 2003, 05:13
Normally, you dont want to reencode mpeg2. Kika was referring to capturing directly to avi (mjpeg is an avi codec) and then encoding to mpeg2.

Originally posted by JeromeERome
I did not know it was possible to re-encode an MPEG2 video in CCE. What bitrate and resolution do you re-encode it at? You must have a very fast machine if you can do two hours in less than realtime.

16th June 2003, 03:52
It's probably best to consider just what you mean by "MPEG-2." Think of the phrase "MP3." There are a lot of different encoders and bitrates. Some MP3 files sound like junk, some are very pristine. The same is true with using MPEG-2 for capture.

The cheapest way is to have your CPU try to do the encoding. Cheapest because you've already got a CPU so it, supposedly, doesn't cost you anything to use it. However, this will generally give you quality and speed on the low end.

There are some good MPEG-2 capture cards and there are some cheap junk ones. Forgetting the junk, people like Canopus and Pinnacle make pretty good cards. They're also expensive. Anything internal is going to be subject to a very noisy environment. That's one reason you'll see various external encoders. A good dedicated encoder will have plenty of buffering because MPEG frames relate to each other and more than one must be available for processing at a time.

Performance of a good capture device will be better than your AIW card because there is more money put into the capture components. The All-in-Wonder (I've had the original and the Radeon versions) are fine for general video vieweing but the quality of what they can cpture just doesn't match up with more modern circuits.

That just makes sense. The whole consumer video market is moving very rapidly to pure digital. The need for inexpensive, high-quality chips is a lot greater now than it was a few years ago. Oh, I assume you meant capturing from an external source and not the tuner module of the All-in-Wonder...

There are also external DV converters, some of which get very high ratings in the reviews. The Canopus box is universally priased with the caveat that it is supposed to boost the signals a little bit with analog out. I've never used one and don't know if this is a true problem. The difference must be quite visible or it wouldn't have been mentioned.

Some camcorders will also do DV passthrough. I've got a Canon ZR65MC which does that. It does a great job with color but the sharp edges of CG screws it up. Same thing with my ZR25MC when using tape as atransfer medium. The picture can be quite sharp but if the edges of logos tend to wiggle it's annoying. The ATI and my MPEG-2 encoder don't do this. It might not be a big deal if you do a lot of homemade stuff or theatrical source but consider how many broadcast channels use logos and such.

There is a new JVC HD camcorder that stores the stream as MPEG-2 on a standard miniDV tape. Don't know much about it except what I read in Popular Mechanics (I think.) It's supposed to have 480i resolution. It's also $4000. Probably has some darn good encoding circuits.

Are you wanting DVD resolution because you have source at that resolution? 1/2 D1 and SVCD can yield incredibly good results and have a benefit of greater bandwidth per frame given the same final file size.

Regarding your CPU, uh, news you probably know but don't want to be reality. It's not that powerful when you want to do video work. If you start using filters and multi-pass encoding, it'll seem like a dog. Sorry, that's just the way it is.

Regarding re-compressing MPEG2 to MPEG2, yes, it's doable but I wouldn't do it directly. Any analog source is going to be noisy, no matter how good a capture system you have. (Well, maybe not if you have access to a broadcast studio...) MPEG2 is easily used by the latest AviSynth and VirtualDubMod releases. You'll still want to filter out noise, be it chroma from tapes or whatever. I wouldn't capture at a low bitrate, though. For a quick and dirty grab, sure, it'll still probably look better than your VCD but burning to DVD means you'll want to do compression for a target file size most fo the time.

Oh, even if you're working with straight MPEG-2 source (like a hacked DVR or satellite feed) you'll probably want to consider re-encoding in the case where you get lots of blockiness. The BlockBuster filter adds noise on purpose in an attempt to prevent the encoder from deciding a large area is all one color. (think of those swirling almost-blacks you see in credits of a satellite transmission.)

If you're happy with the image quality you have from the AiW with VirtualDub captures, maybe you should play with other codecs besides HuffYUV. If the image quality isn't good enough, you'll need different hardware. Also, consider the resolution of your source material.

Yes, CCE can do fairly fast encodes. 3-pass VBR on my Athlon 2100 is about 3x realtime. CBR is faster than realtime, something around 1.3 or 1.4 when going from D1 to 2/3 D1.

16th June 2003, 18:31
Hi Fred

The JVC camcorder sounds interesting. If it stores the data stream in Mpeg2 instead of avi, you should get way over 60/90 mins per miniDV tape.

That is, if you are able to fiddle with the capture settings!

Have you got to grips with your ADS 2.0 external Mpeg2 capture device?

It sounds like the ADS would be just the thing for these guys!

16th June 2003, 19:11
The article was a little skimpy on details. There's maybe 5 paragraphs and some pictures. It talk about being able to read the tapes on a standard DV camcorder. Maybe it's dual-format, maybe they just mean read the stream through firewire, I don't remember.

No, still not done tweaking the settings. I don't have a standalone DVD player so I think the only way to do this is to make a DV tape with various test images like the color bars then, with the camcorder on pause, do a series of captures then sample to test the retrieved colors. I've been kind of dragging my feet on it because it will be tedious. Should happen in the next couple of day, though, I have a PAL tape that has to be read and this is the only way I can do it accurately. Also, finally have a good torture test sample. The title sequence from WWE Smackdown! 2 weeks ago was a fantastic mix of saturation, motion and scene changes. A lot of their title sequences are routine but every couple of months there's an incredibly good artistic one.

I'm also interested to see if the ADS can be forced to I-frame only...

Yeah, the DV devices are nice but also cost more. I'll try to get a sample to show the blockiness characteristic I hate about doing conversion with my camcorders (Canon 25 and 65.) The History Channel and Fox News have very square logos. I can probably grab from there tonight to show the bloom. The ADS doesn't do this which is a good thing and one of the reasons I chose NOT to get the Canopus DV unit (the other being file size.) The only stuff I have right now is already SXVCD and it went through a temporal cleaner so it's not so obvious.

edit - here's a quick sample showing the DV passthru problem. It's not as extreme as I've seen on some other stuff. Bright subjects on a dark background, especially if they're zooming, really show this better: http://www.geocities.com/fredthompson6/Screenshots/CanonDVPassthruEdgeProblem.png

That logo is static and opaque. See how the edges are screwed up? Each frame is different. This sample just shimmered, other stuff will bloom out for a pixel then jump back.

16th June 2003, 19:42

The site's gone down!

Also, I found out today that the disc I promised to sent you was only posted today!

Sorry for the delay.

16th June 2003, 20:10
OK, site's working now. They want an HTML file, not just file storage.

18th June 2003, 20:16
Hi Fred,

It's still not working!

18th June 2003, 20:26
Try the homepage: http://www.geocities.com/fredthompson6/

There's something goofy with the advertising scripts GeoCities forces. Ignore the errors. There's nothing I can do about it.

20th June 2003, 22:02
Thanks Fred

There's some interesting stuff there.

But come on, you've had your ADS 2.0 for a few weeks now and still no word on how good (or bad) it is!

I think you should lay off the beer and women for a few nights and spend a few productive hours on something worthwhile for your fellow man!

Seriously though. Keep up the posts, they are most informative.

Thanks again.

20th June 2003, 22:54
Believe me, I'm trying to get back to it. I've finally figured out how to make a decent color bar sample. I'll use VCDEasy's MPEG Stills utility then duplicate the frame with AviSynth then export to DV and use the pause button on the camcorder.

I've also promised some All-in-Wonder samples for the general thread.

Beer and women? Whutsat? I know work and exhaustion ;)

20th June 2003, 23:50
Hey Fred

I managed to lift some decent (well they look ok to me) colour bar samples from a THX DVD.

There are few useful audio and video test segments on these discs.

21st June 2003, 00:41
You're talking about the THX Optimizer, right? The blue glasses thing really does work for a TV. Think I've also found a good VHS tape to use for comparison grabs, it's got lots of facial closeups and the people look normal. I'm also going to test with some fast motion from the WWE titles. Would love to know how they do their editing. Some of their opening montages are incredible. Last Thursday was a real good one for dramatic use of color. I've got a few quotes and paperwork to put together to ail out tomrrow then I'm planning to tune the ADS and All-in-Wonder and do the grabs tonight.

21st June 2003, 13:35
Yes I know what you mean!

A few months ago there was a TV commercial running here in England for Audi cars (I think it was Audi). I've never seen so many different colours on a screen at the same time in my life. It even had a Jimi Hendrix sound track!

It would make a great test file.

I'll ask a friend of mine who did it. If it was produced by an British 'ad-house' I might be able to obtain a copy.

24th June 2003, 14:53
The original question by JeromeERome could have been written by me. I have a 1.2 Gig machine and an ATI AIW Radeon. I also had exactly the same issues with trying to capture at high DVD resolutions - 720 x 480.
I opted for the Canopus ADVC-100 and have not regretted it. An option would be to upgrade to a faster machine. The problem I had going that route was justiying another computer; having built and invested a lot of time in the ones I already have. Also, the unanswerable question - how fast is enough to gaurantee good results with the AIW, your OS and your software overhead? Read the FAQ's on video capture.
Some have claimed the AIW works great on an 800Mhz machine. But, often if you read carefully, thy are capturing in VCD or SVCD resolutions.
Oh - by the way. If you decide to go the ADVC-100 route, it sounds like you are going to need a much much bigger hard drive. (With Win98SE, my capture s/w has an 18 minute - 4 Gig file size limit per capture.) Then there is the Fire Wire card you might need to add if you don't already have one.
Good luck.

24th June 2003, 15:33
So what does this $300 Canopus card have to offer over the ATI AIW. Im trying to find a reason to justify the cost of the Canopus. What do you use yours for? I see that you also only have a 1.2Ghz machine. Since the Canopus is a hardware based encoder it must not matter much how slow your PC is right? Oh...hard drive space isnt an issue for me. Total disk space in my pc at the moment is 260GB with about 50-60GB free most of the time on my 7200RPM Western Digital Caviar SE.

25th June 2003, 02:41
So what does this $300 Canopus card have to offer over the ATI AIW.
I think I paid more for the AIW card than I paid for the CANOPUS ADVC-100! It was (probably still is) available at www.macsolutions.com for $250 with free S&H. That said, the Canopus works - the ATI AIW doesn't on my machine. Many above have answered your question. I will put it in my own words below in less technical terms.

Im trying to find a reason to justify the cost of the Canopus.
I can't do that for you. There are so many options. Sevral friends at work suggested I'd be better off buying an iMAC and dedicating it to do Video. It's really a personal question. I like the challenge and the experimenting. Buying a faster Mobo & processor was an option. Whose kidding who - its only a matter of time.

What do you use yours for?
Primarilly to archive old VHS home movies. Some are over 20 years old.

I see that you also only have a 1.2Ghz machine. Since the Canopus is a hardware based encoder it must not matter much how slow your PC is right?
You missed a subtle point made in several of the responses to your post. Although both devices are technically Analog to Digital Video Convertors (I'm guessing that is why Canopus calls it the ADVC-100), there is a vast difference between how they send the digitized video to your machine and what happens to it afterwards. The ATI AIW is essentially a 'pass through' device and it relies on your Computer's processor to convert the digitized frames into an MPEG2 file on the fly. That MPEG2 conversion process is very processor intensive. More pixels per frame - the harder your processor has to work. But the MPEG2 compression is fairly efficient. Consider that you can fit about an hour of full 720 x 480 video onto a single 4.7 Gig DVD-R.

Now contrast this to the Canopus. The Canopus ADVC-100 works similar to the electronics in a Digital Video camera. That is why many people suggest using a DV Camcorder with Analog 'pass-through' to do a somewhat equivalent job to the Canopus box. The Canopus box contains circuits that convert the analog video coming into the box into Digital Video (DV) format that is virtually identical to the output of a DV Camcorder. Just like a DV Camercorder, it connects to your computer using a firewire port. 4 Gigs of video in this format is only about 18 minutes! The big difference is that your Processor doesn't need to crunch many numbers like when its trying to compress video down to MPEG2 size. The downside is that the AVI Video files are huge by comparison - but your processor was able to take it easy. But that said, I wouldn't try to do any on-line gaming while capturing - even with the Canopus.

Oh...hard drive space isnt an issue for me. Total disk space in my pc at the moment is 260GB with about 50-60GB free most of the time on my 7200RPM Western Digital Caviar SE.[U]

Look at the file size that 2 hours of DV video would require. You are also admitting that you have over 200 GBytes worth of 'stuff' on your drives. You might want to consider deffraging before any serious caturing.
Once again - Good Luck
Ray S

25th June 2003, 02:56
I'm editing DV around 6 hours per day right now and will recommend if you're going to do this a lot, dedicate a drive to it, don't try to share with other uses. Actually, use two drives if you're going to do a lot of processing. It goes so much faster when you read from one drive and write to another, regardless of interface type.