This is meant to be a simplified overview of the various pieces that together compose MythTV. It is hoped that after studying this, users new to MythTV will better understand how it works and be able to make more intelligent planning decisions about their own installations. It would actually be a good idea to read this document carefully, and then read the installation documentation in it's entirety before you actually begin any work at all.
MythTV is basically a big multimedia jukebox primarily meant to record television programming to be watched later ("time-shifted") when it's more convenient to the viewer. There are plugins to handle video files that have been manually added to the system (MythVideo) and plugins to play music that has been manually added to the system (MythMusic).
- 1 Operational Breakdown
- 2 Various Technical Explanations
...or "My God, it's full of pieces!"
...also known as "the free parts".
There are three main pieces to the software end of MythTV, but only two of them are actually the MythTV distribution. These can all run on a single machine (provided you have the 1.2Ghz or so of CPU power) or on separate machines, depending on what you'd like to do. It's important to understand what role each of these pieces plays in order to be able to make correct decisions about what parts to run on which pieces of your equipment.
MythTV keeps track of programming schedules, configuration information, upcoming recording events (i.e., what you wanted it to record) and the inventory of (both recorded and imported) media MythTV has in a relational database. The database is probably the most important component of MythTV because without it, absolutely none of MythTV can function.
Work is progressing on porting things to work with Postgres, but at the moment MySQL is the only supported (and reasonably safe) relational database that works. MySQL 4.1.x is probably "best", but with MythTV 0.20 it became reasonably safe to use MySQL 5.0.x. Don't get crazy and try using beta or development versions of MySQL (like 5.1.x) or Postgres unless you're a very experienced programmer and fully intend to take on debugging and patching things to work without any outside help. (This is directed at you, Gentoo users.)
The database server is going to need some disk space to store the actual database information. The amount of disk space it needs varies depending on a number of factors (whether or not you have a massive music collection, or whether you only have four or five broadcast channels to worry about) but ranges between 256Mb of disk space (comfortable minimum) and 1Gb of disk space (not many people will ever get beyond this). Also, if there's any part of your setup that you'll want to arrange regular backups for, it's the database files. Without them, MythTV will no longer have any idea of what things it has recorded, and there's no easy way to put things back into the database once those records have been lost. (Good thing it will usually all fit on a single rewriteable CD!)
Unlike the other components of MythTV, there may be only one database server involved. If you want to (for whatever reason, probably masochism) spread the database out to multiple machines using some kind of clustering or replication scheme, go right ahead, it's your equipment--just understand that MythTV only needs one database server and doesn't really have a use for any others.
The backend server is the drab little workhorse part of MythTV that takes care of managing tuner cards, recording actual programs, and starting up jobs like transcoding, commercial flagging, deleting expired programs, and downloading updated TV schedules. Since the backend needs to talk to the database on a somewhat regular basis, it cannot be started until the database is up.
The backend is typically where you're going to want to have the bulk of your drive space allocated for storage. The average compressed video that MythTV initially records is going to take up about 1.1Gb per hour, so it's not considered excessive to go out and buy a nice 250Gb drive for this purpose, but even the first TiVos only had about 40 hours of storage capacity on them, so you might be quite comfortable with as little as 40Gb of space.
Since the backend also does the bulk of the CPU intensive tasks like commerical flagging and transcoding, it's also the place where you probably want the fastest CPU you have involved. It can run multiple jobs at the same time, so it can take advantage of a multiprocessor or dual-core machine, unlike the other parts (frontend, database) which are mostly low-CPU or single-threaded operations.
...also known as "where has all my money gone?"
CPUs and RAM
Various Technical Explanations
...so you don't let the nice salesmen talk you into buying something gold-plated and obsolete.
Issues for recording
Framebuffers and firewire and hardware compression, oh my!
There are three basic ways for MythTV to record broadcasted content, and which ones are usable is tied to how the user gets their TV signal in the first place. The first (and probably least desireable) is a technique called "framegrabbing", the second is over a firewire connection (ideal but not widely applicable), and the third (and least troublesome) is from a tuner card that uses hardware-accellerated video compression.
Framegrabbing is what you're stuck with if your tuner card is one of the older (and/or simpler) tuner cards made that uses a Brooktree chip called the Bt848, Bt878, or Bt878a (although there are a few other variants). These cards are the cheapest type of tuner cards one can get, but this inexpensiveness has it's own hidden price. Framegrabbing (in almost layman's terms) is when the tuner card is given a (nearly literal) window of memory used by your video card to paint the recieved video image into in realtime. The CPU has to "grab" each frame as it appears and then copy it to the software compression codec in order to compress the data so that it doesn't take up absolutely ludicrous amounts of disk space. The audio usually has to be fed through the line-level input on the sound card to be recorded, although there are some cards that support sending audio directly to the CPU through DMA transfers. These cards are excellent for just watching live television broadcasts, but the task of grabbing the images and audio and compressing it in realtime makes framegrabbing a very CPU-intensive task. Generally an SD broadcast requires at least 1Ghz of processing power and about 256Mb of RAM in the machine to record just one stream.
Firewire is one of the primary transport mediums used for digital cameras and other forms of digital video transfer. If you are a subscriber of HD digital cable in the US, there's a good chance the receiver you got from the cable company has a firewire port on it. If it doesn't ask them for one and you'll likely get it (FCC mandate). Depending on who your cable provider is, you may be able to get the video you're watching on the receiver to stream from your cable box to your MythTV box over the firewire cable in it's raw, compressed form (which would be great). At the very least, you should be able to use the firewire connection to make the receiver box change channels (which is less problematic than using an IR blaster and LIRC). Recording video over firewire doesn't require much CPU at all, but with some channels it may be a bit like "drinking from the fire hose" if your machine is underpowered.
Video tuner cards with hardware-accellerated compression that are affordable are somewhat new to the personal computer market, but they're absolutely wonderful now that they're here. What this means is that the tuner card itself handles the "heavy lifting" of compressing the video signal down to an MPEG or XviD stream so your computer doesn't have to. Recording video with these isn't really much more CPU-intensive than just copying a file. If your computer is capable of playing back an SD video in fullscreen without skipping or dropping frames, then it's probably got enough power to handle one of these cards. The most common are the Hauppauge PVR-xxx series cards, which are detailed elsewhere on this site.
Issues for playback
Widescreen or SD (standard definition)
Thankfully, dealing with whether or not you have a widescreen (16:9 aspect) display or a normal (4:3 aspect) display is something that only X (Xorg or XFree86, either will do) only has to deal with. For the most part it's a simple matter of configuring X to use the display at it's native resolution as if it were a normal monitor. If you have a highly expensive display which does 1920x1080 resolution (1080i/p with no downsampling) then that's the resolution you tell X you want to use. If you're using a widescreen LCD monitor that does 720i/p (without downsampling) and has a native resolution of 1280x768, then you configure X to use the display at that resolution. With some widescreens (and some video cards) that X can't quite figure out on it's own, the user may have to manually generate a modeline (or search the wiki for one) that tells X how to use the display, but this is usually not necessary. In almost every case, the video card (possibly with the help of accellerated drivers) and MythTV will be capable of upsampling (expanding) or downsampling (shrinking) the images to fit the display exactly. You'll still get "letterboxing" if you watch a widescreen video on a standard (4:3) display, of course but the fewer resolution changes your video has to be put through before appearing on your screen, the better.
HDMI, DVI, VGA, S-Video, Component, and Composite Output
The type of output your video card can do, and the type of inputs your display device can handle are primarily what dictates what you should use to connect them. From highest- to lowest-quality, the order of consideration is: HDMI, DVI, VGA, S-Video, Composite.
HDMI is a digital connection primarily seen on high-end home theater displays. It also incorporates some technology to thwart naughty video-pirates, but this mostly stays out of our way (for now). If your television and your video card both support this, then it's probably what you'll want to use. It's likely that any display which has an HDMI input also supports DVI input, so you don't need to rush out in a mad buying frenzy to get a new video card if your video card already supports DVI output--the quality difference between them isn't likely to bother you.
DVI is another type of digital connection which is somewhat more common than HDMI. HDMI and DVI connectors look rather similar (in that they're obviously not the 15-pin DIN connector of VGA) but don't get them confused. Most LCD/TFT panel displays support DVI input, and most newer video cards will have a DVI output (some will have two).
VGA is the "normal" monitor cable everyone's used to seeing on computers and monitors. A great many of the LCD TVs currently being manufactured in the US are intended to be dual-purposed as computer monitors and will have a VGA input (and possibly DVI as well), and nearly all (some are coming DVI-only now) computer monitors being sold as computer monitors will accept VGA input. Your video card will almost certainly support VGA output, and even if it only has one (or two) DVI output on it, it should have come with a little adapter that will allow you to use it with the more common VGA display.
S-Video is a multiple pin, round connector type that carries the chroma and luma (color and brightness) parts of the video signal over separate wires, which usually means a very good signal. It looks a little bit like a PS/2 mouse connector, but with more pins. Many, many televisions support S-video input. The cables are reasonably inexpensive, and can be up to fifteen feet long without experiencing notable signal loss. Word of warning: Some video cards have what looks very much like an S-Video output, but has a few more pins involved and is not an S-Video output. These cards will have come with a "breakout box" which will have a number of other connectors on it which often include an actual S-video output, as well as component outputs, and sometimes a composite output as well. Don't get confused and try to force an S-Video cable directly into these, because bent pins really suck.
Component video splits the video signal up into it's three "component" parts, the red signal, the green signal, and the blue signal and sends them over the familiar RCA connections. You can spot these pretty easily because they'll usually be three RCA connectors colored respectively red, green, and blue. This is used by some "studio" monitors, some very expensive theater displays (quite often projection screens), as well as some SUN-type displays and so forth. It's only listed after S-Video because it's not that often you'll come across displays that use this.
Composite video is probably the least desirable connector type of the lot. It's the type of output most video games use, and practically every display supports at least this if you aren't going to have to use an RF modulator with it. It will be a single RCA connector which should be yellow. The other two RCA connections commonly found with it will be white and red, which are the left and right audio channels respectively. (Remember, right and red start with the same letter.) Some of the older nVidia cards have only a composite video connector for TV output. To use the TV output on these cards you will need the nvtv driver as well as nVidia's binary driver. For no other nVidia cards do you need nvtv so don't put yourself through that hassle by mistaking it for something you need if your nVidia card has an S-video output.
...or "How I learned to like stereo all over again".