Path: utzoo!utgpu!water!watmath!clyde!bellcore!decvax!decwrl!reid
From: reid@decwrl.dec.com (Brian Reid)
Newsgroups: sci.electronics
Subject: Re: House Automation (other approaches)
Message-ID: <276@bacchus.DEC.COM>
Date: 10 Feb 88 19:45:07 GMT
References: <745@PT.CS.CMU.EDU> <20522@felix.UUCP>
Reply-To: reid@decwrl.UUCP (Brian Reid)
Organization: DEC Western Research
Lines: 55

In article <20522@felix.UUCP> dennisg@felix.UUCP (Dennis Griesser) writes:
>One thing that made me sad was that we seem to be working harder but not
>smarter on this kind of project.  Yes, I want to rip up the walls too.  But
>isn't there a less dramatic way to do it?

There are different kinds of smarter. I am perfectly capable of designing
custom integrated circuits to operate my doorbell, and such. I just decided
it was better engineering practice to do otherwise. Let me explain why I did
what I did. My going-in principles were as follows:

1. Every component must be off-the-shelf commercial material. No experimental
   or home-made stuff. I need to be able to sell the house someday.

2. Do not use active components where passive components will do. A piece
   of wire is much less likely to fail than a LAN emulating a piece of wire.
   Corollary: use low-tech active components rather than high-tech active
   components wherever possible (I built my furnace controller with relays
   and not with TTL, because relays are far less likely to fail).

3. Given several different ways of doing the same thing, always choose the
   one that is the most ordinary and/or the least expensive. This is why,
   for example, I went with 25 pair cables. 25-pair wire is the standard
   of the commercial telecommunications industry, and therefore is cheaper
   than 8-pair wire. Also, end connectors and jacks for 8-pair wire are
   many many times more expensive than 25-pair connectors. As an example,
   I bought a 1000-foot roll of 25-pair cable for $130 at a local
   telephone supply store. A 1000-foot roll of 16-conductor wire at
   the local electronic supply house is $400. The 25-pair connectors from
   Amphenol are $4 each, and there is a crimping tool that will install
   them. The 16-conductor connectors are $14 each, and must be soldered.

4. It has been my experience that most automation projects live and die
   by interface or I/O problems. The issue is not conceptually how to
   multiplex a doorbell circuit on top of a packet-switched network, but
   rather how to connect the wires from the doorbell button to the
   network. My 25-pair cables have ordinary telephone company modular jacks
   on the end of them, and it is easy to plug a telephone, a computer
   terminal, an intercom, a thermostat, a doorbell, or anything else I
   need into the modular jacks. The conceptual problem of how to implement
   the (physical or virtual) circuits pales in the face of the physical
   problem of how to get a nice-looking wall plate that can be used to make
   and unmake simple connections to it. 

5. I tried not to solve problems that were just dreams. I do not at the
   moment own a toaster with an LAN connection, nor does any company
   sell one, as far as I know. Therefore I didn't worry about LAN toaster
   connections. However, I do own several telephones and several computer
   terminals, and I have frequentlyu wanted to be able to move them from
   one place to another, so I tried to solve the problem of comm-device
   mobility within the house. I didn't build a rocket landing pad in the
   back yard, but I did build a bicycle rack. Same principle. Be realistic
   about what the future is likely to bring. Wire will continue to be more
   useful than packet switches well into the next century.

Brian