LIVE NIRVANA GUIDE TO DVDS: TECHNICAL INTRODUCTION - HOW DVDs ARE CREATED
In this document, Jim Hirte has provided a technical description of the art of creating a DVDR. He has provided full details of the transfer process and has appended his own computer configuration at the end of the document. This guide assumes a moderate level of expertise in computers and may be incomprehensible for those with little knowledge of computer hardware. However, we hope that there is sufficient detail here for others to create DVD transfers of low generation video cassettes.
Jim begins with a preamble about how he became interested in transferring videos to digital formats:
"Back in April 2000 I was looking into video capture as a way to archive all the vids in circulation ... I found a guy's site who was making SVCD/VCD's. Using his suggestions I picked up an Avermedia TV98 card. This piece of equipment uses a BT878 chipset like most other basic capture cards. Capture rate is pretty good, specifications are ok and what not, but the problem was I lost frames like no tomorrow, even on the PC I'm currently using.
"So I got frustrated with this card, and after dealing with their technical support who told me the card was "a toy" and that I was lucky it worked. I decided to get rid of it.
"Then I picked up a Pinnacle DC10 Plus. This card was much better, although it only captured a maximum of 640x480 pixels (meaning I couldn't use it for DVD creation unless I stretched the video). I rarely lost frames with this card, and the output was fantastic. I used this card to make a few SVCD's. For some reason I got a hair up my ass to get a Pinnacle DC30 Pro, after reading much about it, read it can be a pain to get to work properly, etc. I decided to get one ... Ebay is your friend.
"So I sold the DC10 Plus and around early December I received the DC30 Pro. This card was difficult to get functioning, but after Pinnacle updated the drivers (v1.42) all problems I had were gone. This card captures at 704x480 (cropped D1) at a datarate of 7.125Mb per sec. This is in MJPEG BTW (same as DC10). This is really high, great for DVD output.
"I used this card to make a few more SVCDs, and several DVD-Rs. Quality is great, and if your running Win98 I'd really recommend this card.
"So once I got my system the way I wanted it, I started playing with the different MPEG encoders out there, trying to get the best quality. For the longest time I used Tsunami's TMPGENC v12 as it was the best. All other versions of this encoder (including 2.0) suck. Hands down no question about it, it doesn't handle interlaced video very well in the newer versions.
"I tried CinemaCraft's encoder, some call it the best, but in my working with it, I think it's about the same as TMPGENC. BBMPEG, and all the others I tried were crap. Finally, I came across another: Ligos LSX MPEG suite version 2.0.
"This encoder is the best one I have found for DVD output. It comes in a few formats, a stand-alone encoder and a plugin for Adobe Premiere (video editing application). You can do more things with the standalone, but the basic settings to adjust your MPEG stream are still there in the plugin. Since I was using Premiere 5.1c by the time I got this encoder, I used nothing but the plugin.
"After playing around with all the other MPEG encoders I eventually came across some settings which game me the best results out of anything I've done before and anything that I had downloaded off the net done with the encoders. No they're not a secret, just adjusting the Group of Pictures and some other nifty little settings in the encoder. Playing makes perfect right?
"When I captured video with all the cards mentioned above I used a program called VirtualDub. Its a free program for video capture and minor editing which can be obtained at http://www.virtualdub.org. This program is very cool in how it captures."
Jim explains further:
"There is some funny rules when you capture video. If your drive that you're capturing the video to is formatting using FAT16, you cannot capture more than 2 Gigabytes of data. FAT32, no more than 4 Gigabytes. NT File System (Windows NT and Windows 2000 are capable of using this, its a proprietary format, NT uses version 4, 2K v5) you have a lot more room, if I recall its like 2TB (terabytes). These are limits of file sizes in the respective formatting of the drive, not a limit in Windows or anything."
At the time of writing, the only MS operating systems to support NTFS are NT, Windows 2000, and Windows XP.
"What's cool about Virtualdub, is it will automatically segment your video files into 2 Gigabyte chunks if your capturing a lot of video (and at a high enough datarate that you need to exceed that file size limitation). This is a good thing. Vdub is awesome. When I first started using it, I used it for editing as well, just snipping ends and stuff. When I started creating DVDR I realized I needed to use something else."
Jim eventually discovered a solution:
"Welcome Adobe Premiere 5.1c.This is the ultimate in video editing applications. It's got so much stuff to it that there is no way I will use it all for what I am doing. I got lucky when I picked up my DC30, as it came with the software AND the manual for Premiere. Without the manual, I'd be screwed. Pretty much any Adobe product has a pretty steep learning curve, Premiere being one of them. I recommend anyone who uses this program, or wants to, that they RTFM (Read The Fucking Manual).
"Before I started using LSX to encode, I was using TMPGENC. It doesn't have a plugin for Premiere, which is unfortunate. I do not know if Hori Sans (the creator) will make one at some point, but he should. In order to get the video from Premiere's timeline to a standalone encoder you have to use another program called AVISynth to "frameserve" it to your encoder. What this does, is it acts like a middleman between your programs. Handing frames that premiere is exporting to whatever encoder you use. Very cool.
"When LSX came into the picture, I could cut out the AVISynth from the picture. Now the one thing that really sucks about video encoding, is the time it takes to do it. Even on my PC which is pretty powerful it takes a long time. For SVCD creation it took about 6-8 hours to encode an hour or so worth of video. For DVD creation it took more, roughly 12 hours. That's 12 hours of my time where I cannot use the machine that's doing the encoding, something some people can't stand (myself included)."
Jim used his hardware expertise to solve this problem:
"Well, having two extra motherboards, I built another machine to do the encoding. Granted I'm editing on the machine with the DC30, all I had to do once that was done was copy the files to the other PC via my LAN. This can take time to, as you can image copying 30 Gigabytes of AVI files is a lot."
Jim's setup for capturing video cassettes to his computer was therefore:
VCR>DC30(via SVHS cable)>editing in Premiere 5.1c (add fades, remove junk before and after, etc)>LSX plugin for Premiere. The result of this process is a high quality MPEG2 video file.
Capturing the audio part of a video required a different process, as Jim explains:
"When I capture the video, I capture the audio through my soundcard. For all intents and purposes the soundcard I have works great. The only beef with DVD as my goal, is I need it to be in 48Khz sampling. For SVCD and VCD 44.1Khz is fine. One common thing all these formats have is audio. Audio can be in MPEG 1 layer 2 for all formats. Default bitrate for audio for VCD/SVCD is 224K.
"While this is fine (mp2 audio is better than mp3 IMO) I always used a higher bitrate with SVCD's. 256 or 320K in most cases. The downfall, is audio at a higher bitrate will create a larger file size, which will in turn mean you might not get all your video on one CD if making VCD/SVCD.
"For DVDR I have not used MPEG1 layer 2 audio. I use a program from SonicFoundry called Softcode. What this thing does, is create DolbyDigital audio files in AC3 format. This is VERY high quality audio compression. I emphasise the word COMPRESSED. It is true that you can use PCM audio (wav files) for DVD, but it eats up space that can be used for video, the bitrate I use is 448Khz. This is around four times as high as what you see most MP3s at. There is no artifacting like in MP3 either. So needless to say, I love this program!
"For mastering audio (or remastering as most people like to refer to it as) I use CoolEdit Pro 1.2a. It seems this is the standard audio editing application these days. I've been using it since Windows 3.1... I wanna say version 1.5. I have 1.53 on floppy disc still! Back then it was just called CoolEdit. The Pro version came out later, and has multitrack support.
"Anyway for audio editing, what I do is export the audio from the timeline in Premiere, open it in CEP. I much like most other people who master their own audio EQ it to liking, adjust volume, and also use noise reduction. Most people use too much noise reduction IMO, its one of those things you have to play around with for a long time to get good at. There are a few people who use it well, Kris Sproul, maintainer of the Nirvana Live Guide, being one of them. I'm one of the people who would rather leave a little noise in there rather than have artifacting in the audio, but that's just me and my preferred way of doing things. Everyone is different in what they do, and as long as it sounds good in the end. That's all that really matters.
"Once my audio is finished, I save it and encode in Softcode. I also downsample to 44.1 and create a audio CD if the video is one I do not have an audio disc of. Whatever you do, don't burn a disc with 48Khz audio. It sounds really weird upon playback."
CREATING THE DVD:
Jim now describes the final steps in creating the DVD.
"At this point, I author the files for DVD output. What this means, is basically I create the menu (click here for an example), the credit scene using Premiere and then burn a test disc on DVD-RW. For authoring there are a few programs out there. I'm not one for fancy menus and stuff (takes up disc space, remember that!) so I just use a image from the movie for a background and add text to it using Paint Shop. Then add buttons, set chapter points, etc and burn it. If everything looks good, then everything is done. More often then not I find something I want to change and will make the appropriate changes.
"Now if someone reads this and wants to start archiving their concert videos, etc., they need to do a lot of reading. I've spent over a year and a half with this stuff and still don't understand all of it. Here is some useful info I've found that I needed to know."
MPEG2 video maximum bitrate of 9.8mbs (megaBITS per second, remember 8 bits equals one byte, you do the mathematics!). resolution: 720x480 (reffered to as D1 spec) OR 704x480 (cropped D1). 29.97FPS (NTSC) or 720x576 or 704x576 25FPS (PAL). [See http://www.nirvanaclub.com/nfc_faq.htm#42 for basic information on video formats.]
Aspect ratio is specified in the stream as 4:3 or 16:9 (widescreen).
MPEG1 layer 2 up to 384Kbs, PCM audio (WAV), AC3 audio(dolby digital, can be multichannel up to 760Kbs). N.B. It HAS TO BE 48Khz sampling rate!!! ALWAYS!
"Video can be either interlaced or progressive scan. What is the difference, you may ask? Well, back in the early days of making TV, the manufacturers ran into a problem with half the image disappearing as the lower half was being drawn on the screen. The phosphors in the tube would fade the image on the upper half while the lower half was being drawn. They therefore came up with the idea of splitting the image into two separate fields, a upper field and a lower field. What happens is the upper field is drawn out first in 1/60th of a second, then the lower field is drawn out.
"Imagine it like this: lines 2, 4, 6, 8, etc. are drawn out on the screen, then 1, 3, 5, 7, etc. is drawn. All this happens so fast that your eye doesn't catch it, and the whole image was drawn on the screen without any of the image disappearing like they would before. Pretty neat, eh? Well, neat as it is, this is no longer a problem, but we still continue to use this technology in TV today. Everything you watch that's being broadcast is interlaced video. I imagine this will be the case for a few more years as well.
"Progressive scan on the other hand is like a movie you see in a theatre: one image displayed on the screen 29.97 frames per second (25 for PAL and SECAM). Image looks pretty good right? right! Remember that your monitor is non-interlaced. One image, not 2 fields (upper and lower). Some TVs these days also use the progressive scan methods. What I believe most do is to lose one of the fields and interpolate a new image using the remaining field. Most deinterlacing filters in video editing do this.
"The only thing about this when using captured video for DVD (or SVCD output) is that interlaced video encoded into progressive video does not have as good of motion as interlaced video (which has twice the rate on the screen as progressive; remember 60 fields per second vs. 30).
"A lot of encoders support interlaced video but don't handle it very well, TMPGENC being one. LSX handles it much better in my experiments between the two. Also keep in mind that your MPEG stream has a flag in it saying that its either interlaced or progressive."
"For everything you see on TV (the news for instance) is 4:3. Movies (widescreen) are 16:9. This gets complicated, and I won't explain it here as there are webpages out there (for example: http://www.doom9.org/aspectratios.htm) which explain it much better than I could.
"Say you capture video at the resolution of 720x480. This is going to look stretched on your monitor on your PC. Why? because your monitor uses square pixels, where your TV uses pixels that are not. If you played back your captured video on your TV, it will have the correct proportions. When you encode your AVI file to MPEG2, it has a flag that tells it what aspect ratio. If you played the MPEG2 on your PC using some software DVD player, it will be played back in the proper 4:3 aspect ratio because of this flag. Confused yet?"
"As you can probably figure out, higher is better. However with DVDR as your output, your limited to 4.7 Gigabytes of data on a single sided, single layered disc. Most DVD's you purchase are dual layered discs, meaning they can hold TWICE that amount. Dual layered DVDR's are 2 years in the future (so they say). So we are currently limited to that magic 4.7 Gigabytes. If you encode at 9mbs you'll fill this disc in under an hour if you used CBR (constant bitrate). What I generally do is use 8mbs CBR for hour long shows. This gives outstanding quality. For longer shows, I generally use 6.5-7mbs VBR (variable bitrate). What this does is uses varying bitrate according to the source, say dark scenes with no motion, they don't need a lot of data to encode that so it might go down to say 2 mbs, but once there is a lot of motion going on or really bright or something it goes back up, but not exceeding your limit that you set. This provides excellent quality as well, and you can get more video on your disc. DVD's that you buy are VBR at around 6.5-8mbs. I've not come across a movie that was done with CBR.
"There is a whole ton of pages out there that explain how MPEG2 video works (for example: http://www.pinnaclesys.com/.../WhitePapers/DC1000-DVD1000MPEG2whitepaper.pdf), GOP's, intra frames non-intra frames, etc. I won't go into how that works, as I'd probably just end up confusing you as well as myself in the write up. If your really interested in how it works, do some searching on MPEG2 video, and I'm sure you'll see what I mean. I'll include links to some good ones in the future.
"There is also a lot of information on how video is created on, different coloring methods, subsampling, etc. Again I'll include links at a later date.
"After all this information, there is one thing that still comes back at you no matter what. Encoding takes a long time using software! I was getting tired of spending hours and hours waiting for encodes to finish. So I decided to do something about it. I purchased a new capture card that captures directly to MPEG2 video, not AVI. This allows me to save a lot of time, and I can get work done more quickly. While the quality of output my DC30 provided me was outstanding, I just wanted to get this stuff done faster. The card I purchased was a Pinnacle DC1000. Its hardware MPEG2 is outstanding. (for info on it check out http://www.pinnaclesys.com). With it I can capture long videos in one file, which means I can use Premiere as my program to capture now (in fact I have to with this card). Not only that, but I can use Windows 2000 as my OS (DC 30 has no non-beta drivers for Win2K as of this writing). Win2K means I can use NTFS file system, which means files bigger than 4 Gigabytes.
"Also rendering of fades and any effects are real time, not over several hours. I highly recommend this card to anyone who is seriously considering video editing for DVD production."
Other equipment needed to capture other than a fast PC and a high quality capture card:
VCR: "I recommend most Panasonic and JVC models. I use a JVC S5900 SVHS, which is very nice. If you want to use PAL tapes (assuming your in the US or Canada) you'll need a multisystem VCR. I've read good things about the Samsung models, and have heard mixed feelings about the Aiwa model. However I think both will be fine."
TBC (timebase corrector): "Remember the interlaced video I described above? Well sometimes on older tapes the sync order of fields can get messed up, and when you encode video whose field order gets outta whack, it looks awful. So, how can you fix this other than manually going thru your video frame by frame reversing the field order? By using a TBC of course! What this does is recreates the field order, meaning your captured video will have proper field sync (upper lower upper lower, etc.). This can improve the quality of some older tapes as well, since it corrects a lot of jitter problems. You can use these when dubbing tapes to, which will help preserve the quality. Also if you seem to lose frames a lot when capturing (not due to your PC being slow, etc.) a TBC can alleviate this."
"I use a Datavideo TBC1000 which runs about US$300. Once I got it, I could capture video while fast forwarding the tape, and not lose a single frame. Something which is impossible otherwise.
"Also with most hardware MPEG2 cards you can have issues with audio being out of sync with your video on older tapes... this is because of the field order being out of sync... TBCs will correct this.
High quality cables: "Monster Cable makes good ones albeit expensive ones. RadioShack actually makes some really decent cables for pretty cheap."
Good quality source (video) material: "Remember, garbage in, garbage out!" LiveNIRVANA.com recommends that you use the lowest generation video cassette you can obtain. If you do not, your efforts may be wasted if someone later transfers a lower generation cassette."
"The last thing you need is patience! If you want to start editing video, be prepared to do a lot of reading, more than you though you would want to, and more technical stuff then you probably wanted to learn! Keep in mind practice makes perfect (which is particularly apt in this case), and you'll come out alright!"
For your reference, Jim has also provided a complete description of his computer set up, so you can compare your equipment with his. This also serves to give an idea of the technology needed to produce DVD transfers. His current PC configuration is as follows:
Motherboard: ASUS K7V-T (same as a K7V, except modded to take thunderbird CPUs).
CPU: AMD Thunderbird (SLOT A) 850Mhz overclocked to 1 Gigahertz ( = 1000Mhz)
RAM: 512Megs PC133
Hard Drives: Western Digital 45 Gigabyte (Operating System/apps/storage) Maxtor 60 gig (video). Jim adds: "This is important because HD space is needed in large quantities. Video takes up a considerable amount of disc space. I also have another machine which I use for video storage it has a 60 Gigabyte hard drive drive in it as well.. This one is mainly for data backup and also so I can play games while rendering video :).
Sound Card: Creative Labs live! platinum. Jim explains: "I got this mainly because at the time I wanted the 4 channel output with digital in/outs, etc. I don't use it to capture audio anymore though. My new capture card handles audio capture when I capture video. It can capture at 48K without resampling.
Video card: ASUS V7700 GeForce 2 GTS 32meg DDR. Jim elaborates: "Got this for gaming mainly. For video editing you can use pretty much any video card. I'd recommend something with AT LEAST 8 megs RAM.
NIC: 3com 3c905b-TX. Jim adds: "NICs are not really important for video capture/editing, but nice if you want to use the internet (if you have DSL/Cable/whatever) and if you have a LAN so that you can move files back and forth).
DVDROM: Creative 12x Encore kit, comes with DXR-3 decoder card, not a necessity, but useful.
DVDRW: Pioneer A03. "Let me stress that this is the best solution for DVDR/RW. not only because it burns DVDR and RW, but also it doubles as a CDR/RW drive (as well as a DVDROM). This is the one drive that does all, replacing whatever burner you have. IMO it kicks ass and is totally worth whatever the going rate is. I paid 598$ US for mine in 9/01
LiveNIRVANA.com would like to thank Jim for both this interview and his groundbreaking foray into the world of NIRVANA DVDRs.