[mythtv-users] front/back-end vs 'MegaMythTV'

[James L. Paul]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1



Bob Swithers wrote:
| I'm interested in getting into Linux, and want a house-wide PVR solution
| so have been following this forum and doing some homework for about a
week.
| I was first quite attracted to MythTV's backend/frontend possibility.  I
| have 5 TVs scattered around the house (I admit that's nuts), and want my
| wife and I each able to record programs as we wish (so possibly multiple
| concurrent recordings), and would want to watch live TV or any recording
| from any of the 5 locations.  As I don't want fans or big heat sources
| in our main living areas, the possibility of a lightweight frontend
| seemed attractive.  The more I think about it, though, it seems like it
| may not be the best solution.

Always a good idea to examine the wisdom of the status quo, I think.
Your goals probably aren't very different from those of many of us, just
a minor difference of scale.

| First, each front-end looks like it might cost $180 or more.  (Adrian's
| recent post of his $130 front-end excluded power supply, keyboard,
| remote control, etc., and many might opt for a slightly beefier
| processor.)   At 4-5 front-ends, that adds up.
| Second, there are places where putting a front-end is just not
| convenient -- like for the small TV on the kitchen counter.  (And the
| nearby cupboards are full and don't have electricity.)  Last, there is
| the significant challenge of running new Cat-5 wiring to each of the TV
| locations.

I used to have a similar number of TVs, now I'm down to a maximum of 3.
Size/heat/noise are a big deal for me, we have less than 1000 square
feet and ambient inside temperature is in the mid-80's F. Several of my
computers are water-cooled by necessity.

| An alternative would be a single 'MegaMythTV' which does all the work on
| a single box and distributes the video over the existing cable wiring.
| I'm writing to ask you whether this is feasible.

Ambitious, but unlikely to be technically practical. Distributing audio
with video by using RF Modulators and tuning all your TVs to the
ubiquitous channel 3 is an obvious goal. Perhaps an IR remote extender
that also uses the TV coax for transmission is a goal. The only hardware
you would need to add at each TV would be a presumably small gadget that
connects inline to the coax and has a wire with a IR receiver on the
end. I'm speculating. I don't even know if such a product exists. It's
certainly not a hard thing to build, but I don't know if anybody has
productized it.

| Are there restrictios in either Linux or MythTV which would prohibit
| running multiple (e.g., 5) display cards with TV-Out?   Similarly, are
| there any restrictions on the number of tuner cards?  (I'm ignoring the
| question of the number of free slots in a PC chassis.)   Certainly the
| processor would have to be sized accordingly, but the incremental cost
| of any faster CPU is smaller than 5 times the cost of the motherboards
| for the frontends.  If this is all possible, then each TV locations
| could be provided with an RF remote control with USB connection, and I
| think I've found one for $35 each retail.  Also, are there any
| restrictions in lirc support of multiple remotes?

There are several dependencies as I understand it. There aren't inherent
limitations in Linux or MythTV aren't possible to overcome, but there
are practical considerations that might make things non-trivial to get
working. Many video drivers might assume there is only one adapter
installed and make it difficult to address more than one. Same story
with many sound drivers, I imagine. You would need to research the
abilities of the drivers carefully. I'm curious if the USB-based RF
remote you found has a driver that works with multiple instances of it's
hardware installed. Just because USB makes it easy to connect multiple
devices doesn't mean that the drivers were written well.

It would be nice to suspend the problem of PCI slots for discussion, but
we just can't. We can't really include the unusual video adapters that
natively support multiple monitors, even if we ignore the TV-out issue.
So, we're faced with 5 video adapters. That's a hard thing to do,
considering you won't be able to use AGP or bus-mastering PCI cards. I
don't know of a solution to that offhand, not to mention you will also
need 5 video capture sources that presumably also need PCI. There are
multi-channel sound cards that might reduce your audio requirements to 1
or 2 slots, but I don't know how well supported they are under Linux.

Based on the limited number of slots, the only practical mass-market
solutions that come to my mind are researching firewire and high-speed
USB products with Linux support. If there are any V4L-supported devices
for these busses, then the problem at least moves back onto the table.
You would still possibly have serious CPU resource issues even if you
did manage a solution with 5 TV-out boards.

| I do know there are some things that many people might want that this
| doesn't provide.  E.g., you need the front-end, for example, if you care
| about generating local 5.1-channel surround sound -- if you care.  Also,
| if you want a local DVD-player you'd want a distributed system.
| However, if the above concept works, one could still mix and match.
| E.g., one large backend master, plus one backend slave with a DVD player
| and good audio in the main entertainment rooms, but have the remaining
| rooms have their video served directly from the large multi-video-out
| backend.

This may seem counter-intuitive at first, but my experience has been
that using multiple small machines often is cheaper and more practical
than trying to meet the demands of specialty hardware. This seems to
happen at all diffferent levels, I call it the Google factor. (Why use a
few expensive supercomputers when thousands of cheap small computers
will do?)

I fear that even if you do find a way to fit 5 TV-out, 5 TV-in, 5
Audio-out, 5 IR-in into one box with enough CPU to handle it all, you
will find the solution to be more expensive that 5 discrete boxes.

My own approach to such a project would be to build a custom enclosure
for 5 motherboards with about 1000 watts of power supplies. I'd set up
one MB as a "master" with a couple 120GB drives and LVM, and all 5 with
the video in/out/IR/network hardware. Four of the boards would boot to
the network and each operate from their own partitions on the "master."
No, it wouldn't be cheap, but all things are relative. The balance point
between special hardware and another motherboard is about $120. If it
costs me more than $100 to use a special video card and IR and sound, I
might as well instead buy another MB for $40, another CPU for $40, and
more RAM for $40. Ballpark, subject to a lot of things like using a PVR
card so I can buy older cheaper components from over a year ago that
will work just fine. (My current myth box is build entirely from
components I've had for over 2 years, except for the PVR cards for $89
each.)
|
| In summary, the questions are:
| 1.  Is there any restriction on multiple display cards with TV-Out?   In
| Linux?   In MythTV?   For any other reason?   Is there a better way?

Sorry I don't have specifics for you. You will need PCI cards, not AGP,
that have TV-out and Linux drivers that work fine with more than one
card installed. I don't know of any, but that doesn't mean they don't
exist. (I've been surprised before! :) My suspicion is that since this
isn't a mainstream thing, it's not likely to be supported very well or
at all.

| 2.  Same questions regarding TV tuners?

There are several of us using multiple TV Capture cards, and while some
may be tricky, this is not uncommon. The PVR-250 flavor cards I have
claim to require bus-mastering slots, which limits me to 3 in my
machine. I think. I have 2, they work _fabulously_ well so far for the cost.

| 3.  Same questions regarding remote controls?

I believe LIRC wants to see only one receiver, although you can simply
hack the driver source to change the driver name and run as many as you
want that way. I personally use a homebrew receiver using the circuit
described on the LIRC site and elsewhere, 5 could be built for about $40
~ total easily. You need a real serial port for each though, no USB
serial adapters, so connecting 5 to one machine short on slots would be
tricky.

| Your answers, thoughts, opinions will be appreciated.

I love to think. Especially about problems I don't really have to solve!

|
| ------------------------------------------------------------------------
|
| _______________________________________________
| mythtv-users mailing list
| mythtv-users at mythtv.org
| http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQE/iiN1T8BYaKRUpkQRAqQBAKCMynOS75j105t5ednfyebyllQocwCgmJDh
7ZUk1vHEt8RDn/X8JUsXzTI=
=dqnf
-----END PGP SIGNATURE-----