This lecture is based on Wu's Chapter 8 of the same title.
CATV Components:
Query 21.1: Define 'coaxial cable'. Why is coaxial the medium
of choice for high bandwidth signals? What is the principal alternative,
and why is coax more expensive?
The guerilla marketplace's response has been to sell pirating devices in flea markets and want-ads all over the world. Robin Hood is alive and well (rob the rich and give to the poor, and charge the poor for the service....)
Now that digital technology is beginning to arrive, the number of ways to cheaply scramble signals is infinite. Encryption is the high tech general solution, because you can require each subscriber to have a unique paid-for ID number. But the hackers will sell boxes for this, too.
Telecommunications Act of 1996 let the regional telephone companies go into video, and the cable companies go into phone service.
Key problem for CATV: One way distribution systems. Upgrades usually involve taking fibre optics further down the hierarchy, and the replacement of the distribution amplifiers and optical/electrical converters.
The Satellite Alternative to CATV. There are a limited number of locations for geosynchronous satellites that can serve North America. They must be 22,600 miles up, which imposes a delay of 45,200/186,000 seconds (plus other signal processing delays, typically adding up to 0.5 second in each direction), irrelevant for one way applications but very important for synchronous communications such as telephone, or acknowledged systems like TCP/IP. (more below.)
Satellite systems can offer enormous economies of scale, providing hundreds of broadcast channels to continental-sized audiences. However, recently, local broadcasters have successfully blocked satellite operators' free use of the major networks (and also to some extent the CATV operators' right to use these signals) because the majors' deals with their local affiliates include the affiliates' exclusive advertising rights in the local market. That is, an Orlando station makes part of its money by running local ads with national programs. This contract term's value to the local broadcaster is violated if NBC's signal is also available elsewhere.
I'm not sure how this one will sort out.
There's another service that is called "wireless cable TV". This uses local broadcasting equipment at extremely high frequencies, which are capable of multiplexing as many signals into one signal. Wireless cable doesn't have the local/national network political problems that satellite TV has, but it has technological problems that to date have kept it from displacing most in-place CATV systems.
Bidirectional Communication by Satellite. Therefore most new telephonic services are using multiple low-orbit satellites rather than synchronous ones. Lower cost links to third world countries still use geosynchronous systems (e. g. the one I used to Belize last week.)
Considter the problem of TCP/IP via satellite. Remember the "sliding window" of TCP packets which require acknowledgement. If the delay is 1 second the window is essentially however big the bandwidth of the signal might be, which could be enormous. Furthermore, home satellite systems don't have uplink transmitters, so you have to use a telephone circuit for the uplink anyway.
Hughes' DirectPC system uses a "fake TCP/IP" system to overcome this problem. At the up-station, the encoding equipment sends back acknowledgements for incoming TCP packets, so the sender is happy. The system then transmits the packets to the user, and generates its own error checking protocol. The downlink unit adds on its own 'pseudo' TCP control information and sends it to the user's PC. The downlink unit receives the user's acknowledgement traffic but doesn't send it directly to the sender. If DirectPC's own error check protocol, or the user/downlink communication process determines that a problem occurred, then the telephone system is used to request a retransmission.
The above situation is also in effect in hybrid Cable-Modem situations (discussed below) where the uplink is via telephone, downlink via radio frequency (RF) signals in the coaxial or fibre cable.
********************************
Query 21.2: Draw the analogy between this "Fake TCP/IP" system and the extrapolation techniques used in SIMNET. Identify the key similarities and differences in the two approaches.
Yes, I know you hated the last question like this which was on the midterm exam. That's why you need the practice.
Here's how to draw an analogy between System A and System B.
0) First, express the key idea in a sentence or two. "DirectPC and SIMNET are two telecommunciation systems which use a local model of the remote sender's state, to emulate deterministic (acknowledged) transmission by cheaper means. These techniques introduce some degradation of the service but include mechanisms for limiting the amount of degradation.
1) Identify the components of System A. In this case, A is DirectPC and B is SIMNET. A's components are the sending computer S, the uplink system U, the downlink system D and the receiving computer R. B's components are tank simulator X, its network control computer Xc, another simulator Y and its network computer Yc.
2) Identify the correspondences between systems' components:
DirectPC SIMNET
Sending Computer S
Simulator X (initiating actions)
Uplink U
Simulation computer Xc
Downlink D
Simulation computer Yc
Receiving computer R
Simulator Y (displaying actions of many other tanks including X)
3) Identify key relationships between the components of System B, if it's the one that is best understood (i. e. it's the one which is being used to explain the operation of System A.)
SIMNET's key concept: Simulation computer Yc contains a model of what it thinks Simulator X is doing. Simulation computer Xc knows what Yc thinks, and only transmits (broadcasts) update information when its behaviors change such that Xc knows Yc's information is now wrong. (Periodic unconditional transmissions provide error recovery if signals are missed.) So, Yc supplies simulator Y with a 'pretty good' picture of what X is actually doing.
4) Seek similar relationships between components of System A, the one we're explaining.
DirectPC's key concept: Uplink U contains a model of what it thinks downlink D is doing (namely, successfully receiving each TCP packet.) Downlink D only transmits a request for update information when it gets an actual error. Therefore, D can tell R exactly what sending computer S originally sent it.
5) Isolate and describe key differences between the systems if any exist.
In this case, it's the SIMNET sender who is responsible for detecting potential errors and transmitting corrections. The receiver has no means of asking for corrections. Therefore the SIMNET system is inherently less than perfect.
In DirectPC the downlink on the receiving end of the relationship is responsible for detecting errors and requesting corrections. Therefore DirectPC has the potential for error-free communication.
There. I hope you can make analogies about other subjects now.
**********************************
Video On Demand was tried in seveal markets including Longwood/Altamonte Springs, by Time Warner. The central servers contained a rack with hundreds of hard drives with MPEG-2 encoded movies, and software which fed the movies into an ATM switch system. Each set-top box consisted of a stripped-down and specialized Silicon Graphics Indigo computer to decompress the MPEG-2 signal.
Pure VOD means you can ask for any movie, anytime. The implication of this is that you need one channel per subscriber (though this bandwidth requirement is much less than one full channel's worth, because CATV channels aren't usually MPEG-2 compressed.) Most CATV users would presumably continue to watch shared conventional channels.
Why did VOD experiments fail? In Moshell's opinion, the basic sales model was too centralized. Time Warner expected to be able to sell combinations of packages like CATV traditionally does, but also to sell many services such as community information, and video games. In essence VOD was to be an exclusive information superstore with a price on everything. But the combination of Blockbuster video, CATV and the satellite alternatives, and most importantly the Internet/World Wide Web provided an "open market for information" analogous to Highway 436 with its multiple kinds of shops. Time-Warner's model was the equivalent of the Sears Department Store that tried to sell you everything you want under one roof-an impossible task.
Cable Modems will provide WAN service, using small fractions of the cable's bandwidth (especially on fiber links.) ATM will be the basic protocol system, with IP and Ethernet components.
Quadrature Amplitude Modulation uses phase AND amplitude to encode information. Consider slicing the available voltages in a wire into 16 levels. You can then, in effect, send 4 bits of information in each "time slice". If you also allow the width of the time slice to vary in any of 4 options (e. g. 1, 2, 3 or 4 nanoseconds) you can transmit another 2 bits of information in that signal parameter. This "width modulation" is called quadrature because it uses four potential values. I don't know what the 64-bit in "64 bit QAM" means.
ATM over Cable. Shall we assign every user a unique frequency? Too expensive. So we have to use a routing and channel allocation strategy. In essence if there are m down-channels each with capacity c bits/sec, and n users, at most m out of n users can have downlink going on at any time. However "any time" can be very short, so you can provide all n users a continuous service but each has use of at most mc/n bits/sec on the average.
How to handle the uplink is always an issue, given that uplink for homes will usually be orders of magnitude smaller than downlink. (The main currently anticipated exception is video telephony.) Page 274 discusses how constant bit rate and available bit rate (CBR, ABR) services will be supported by exclusive use of portions of the upstream resource (time slices or frequency slices.) The ethernet approach would use a "contention region" of the upstream, for a cheaper but less QoS oriented approach.
The @home Example on text pages 282-84 shows how a substantial
investment in network operation centers, with local caching and appropriate
crossovers to Internet, is essential.
xDSL where x varies across A, R-A, H, S V and I. (IDSL=ISDN.) but ISDN requires external power for its modem and is a switched service; the others are point to point services. (In essence, IP does the switching per "micro-call".)
ADSL: Assymmetric: more downlink (1.5 to 8 mbps; higher bandwidth, shorter distances) then uplink (16 to 640 kbps).
R-A: Rate-adaptive. Seeks out highest reliable rate for any given twisted-pair.
HDSL: High speed symmetric. 1.544 mbps on two pairs, 2.048on three pairs.
VDSL: very hi-rate, only 1000 feet or so for max rate of 51-55 mbps.
xDSL supports plain old telephone service (audio POTS) at the low frequency end of the spectrum while doing this data magic on the high end.
Query 21.3: What are the principal advantages for CATV in the coming data wars? What are the principal advantages for ADSL?
Back to the course index
Back to the course syllabus
Back to the previous lecture
Onward to the next lecture