Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 1 of 96
`Case 1:16-cv—02690-AT Document 121-14 Filed 08/05/16 Page 1 of 96
`
`E-4
`
`13—4
`
`

`

` wmmimMm.**m.?
`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 2 of 96
`
`Network Analysis Corporation
`
`10 1 L
`i
`100
`
`-<—I—I I I 1111—
`2 3 4567891
`1,000
`
`NUMBER OF TERMINALS
`
`Figure 7.3: Preliminary Estimates of Terminal Connection Costs
`
`TIP
`
`AREA OF HIGH
`TERMINAL DENSITY
`
`Figure 7.4: Network Design for 100 Nodes, 10bps Traffic, TIP in N.Y.C.
`
`7.9
`
` ..:.■,„....-.. ^....■. - ■■ ■ ■■ ■- -■ -- ■- - ■ ' ^-"
`
`

`

`f^Sl^lliPPjpBpwll^pfpi^wswwwjsaww'^nw^^
`
`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 3 of 96
`
`Network Analysis Corporation
`
`n AREA OF HIGH
`TERMINAL DEI* ENSITY
`
`Figure 7.5: Network Design for 200 Nodes, 10bps Traffic, TIP in N.Y.C.
`
`AREA OF HIGH
`TERMINAL DENSITY
`
`Figure 7.6: Network Design for 500 Nodes, 10bps Traffic, TIP in N.Y.C.
`
`7.10
`
`^---■-
`
`--- — —■.
`
` -„.„....,..,,.-,—. —, -^
`
`

`

`■(^W!!iP!«Vf|IPWW,«!*»«MlLJlW i|i üWIPW n J .l.".Pi!^,M4 "M "'-^ ■•..,« li^ iUüi",.^«ffip?pWiHPJ|.lipJ,W.IA»Milll^|.l im -PI«.. JJI .!JI.IJ!H1A.,I.I1^ ^^m^mKf. \ Wm.i'-mL ,U! i.. .vr wwrw.^ JHHHWWMIJLIJ &W'im™*mJ*m^'--* *". i- M'*.'^ .i j ii J 11 ■ iPii pi*! -—"—"-T-TT. .-.«^^^KH,^^^*.. ■w.y.i.^jfl^^^
`
`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 4 of 96
`
`... ■--.-..
`
`Network Analysis Corporation
`
`TIP
`
`AREA OF HIGH
`TERMINAL DENSITY
`
`Figure 7.7: Network Design for 500 Nodes, 100bps Traffic, TIP in N.Y.C., Chicago, L.A.
`
`7.11
`
`»«. - ■•-—■.-^—~—..—^—.^^^^^. ..... -^
`
`

`

`g!l"l!P*w»WW»^»""™»^"!»«»WPWIB!«PIIBPllpiwlPPP(P"!lR1^^
`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 5 of 96
`
`Network Analysis Corporation
`
`Chapter 8
`LOCAL ACCESS-A RING DESIGN EXAMPLE
`
`For local transmission of signals from a nationwide interconnecting network, the user's
`technical problems are complex because many of the techniques are in the experimental
`stage. The problem is not just one of configuring facilities, but actually designing the
`channel. The classical technique of using multidrop lines with polling concentrators as
`described in Chapter 7 is available and in many cases the best strategy. But, new tech-
`niques such as the use of rings or random access multiplexing offer better prospects in
`many cases. However, neither of these are standard techniques and hence protocols and
`hardware are in a developmental stage. Furthermore, new physical links are becoming
`available. One of the most promising of these is the coaxial cable of cable television
`(CATV) systems.
`
`To illustrate some of the complexities and surprises awaiting the designer of local systems,
`we present one example of a ring design. In Chapter 9, we present a detailed considers '
`tion of the use of CATV systems for local data transmission. In the remainder of this re-
`port. Chapters 10 through 15, we discuss the use of broadcast packet radio techniques for
`handling the local access problem.
`
`Let us illustrate just one of the problems with a ring network, inflexibility in routing that
`results because there are no alternate routes. The analysis will show that although the
`ring may accommodate a large throughput when high traffic points are close on the ring,
`there is no flexibility in adapting to redistribution of traffic requirements. The example'
`is carried out for a mixture of tape transfers and interactive traffic.
`
`One of the traffic models developed by Hayes and Sherman [31 ] is used to analy;; the
`ring design. We consider the design of a single slotted ring to which sources of traffic are
`connected through an interface. The source can represent host computers, terminals, or a
`combination of these. The interface is assumed to receive packets from the source, store
`them, and multiplex them onto the ring. A header which addresses the packet to a par-
`ticular interface on the ring is added to the packet; the packet size on the ring is there-
`fore larger than that on the line. In the reverse direction, the interface removes packets
`from the ring addressed to it, removes the "ring header," and transmits these packets to
`the source. It is assumed that an interface can remove a packet from the ring and then
`feed a new packet into that same slot. In this case, the traffic on the ring seen by the
`interface is in the location marked by X in Figure 8.1. That is, it includes only the traffic
`
`8.1
`
` ■■■-'■ ■■ • — ■ -■ ■■ ■' '-■ - - ■ —
`
`UUMaitfiiMliH.twiyila!
`
`.. ^ *. ~.y.
`
`

`

`Network Analysis Corporation
`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 6 of 96
`
`Figure 8.1: Ring-Source Interface
`
`which passes through the interface and not the traffic which is destined for that interface
`or which originates from it. It is assumed that durations of idle periods on the ring are
`exponentially distributed.
`
`In the calculations of the buffer content and the delay, we assume that the traffic flow
`from a source to its interface is at a constant rate, equal to the average rate. The per-
`formance of the system is characterized by the buffer size at the interfaces, the delays at
`these interfaces, and maximum throughput which can be obtained.
`
`8.2
`
`-- - ----- _ . —.
`
`^^^M^^^
`
`■"■■ --' ■-'■--■■--—■-^■-■- ----^.:-.-.
`
` ' - - ■■ -.^
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 7 of 96
`Network Analysis Corporation
`
`For all cases, we used 1.5 Mbps speed on the ring, a packet of 784 bits on the ring, and
`768 bits on the lines to the interfaces. Table 8.1 gives the input data to the interfaces,
`and Table 8.2 gives the distribution matrix PJJ, that is the fraction of traffic from inter-
`face i destined to interface j. An important parameter is the average number of packets
`per message. We assume an average o: 14 packets/message when all traffic is of an inter-
`active type, 65105 packets/message when all of the traffic is tape transfer, and an average
`of (1500 4- 65105) packets/message for other interfaces depending on the fraction of tape
`transfers that it included.
`
`For the given data we analyze two designs referred to as System 1 and System 2. System 1
`connects the interfaces in order 1 through 4, and the direction of flow on the ring is
`counter clockwise. Interface pairs with high traffic requirements are relatively close. For
`System 2, the ring is reconfigured for flow in the following direction: 1 14 8 6 7 9 12 10
`4,13,11,3,2,5,1. ■-<,..
`
`Tables 8.3a and 8.3b show the results for System 1 and 2. Each table shows utilization
`of the ring seen by an interface, the rate of packets/sec. on the ring seen by an interface.
`
`Table 8.1: Data at Interface
`
`Inter-
`face
`No.
`
`Line Speed
`Bits/Sec
`
`Rate In
`Bits/Sec
`
`Rate In
`Rate/Sec
`
`Average
`No. of
`Packets/Msg.
`
`Ratio of
`Source To
`Ring Rate
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`230000.
`
`100000.
`
`100000.
`
`100000.
`
`100000.
`
`100000.
`
`100000.
`
`87754.
`
`59554.
`
`61646.
`
`41554.
`
`30785.
`
`10154.
`
`17754.
`
`114.3
`
`14974.
`
`77.5
`
`65105.
`
`80.3
`
`54.1
`
`40.1
`
`13.2
`
`23.1
`
`14.
`
`65105.
`
`26042.
`
`65105.
`
`65105.
`
`230000.
`
`133754.
`
`174.2
`
`29948.
`
`100Ü00.
`
`230000.
`
`100000.
`
`100000.
`
`100000.
`
`17015.
`
`69754.
`
`30154.
`
`41015.
`
`25354.
`
`100000.
`
`65554.
`
`22.2
`
`90.8
`
`39.3
`
`53.4
`
`33.0
`
`85.4
`
`26042.
`
`44922.
`
`14.
`
`26042.
`
`14.
`
`65105.
`
`.15655
`
`.06807
`
`.07071
`
`.06807
`
`.06807
`
`.06807
`
`.06807
`
`.15655
`
`.06807
`
`.15655
`
`.07071
`
`.06807
`
`.07071
`
`.06807
`
`8.3
`
`i ■ - -- - -- - —• - - - ■ —- — . -.. .- J.,„.. _ ■^■_., ... ^ . . -— . ^ .
`
`

`

`Network Analysis Corporation
`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 8 of 96
`
`in
`in
`q
`
`o
`o q
`ö
`
`o
`c;
`q o
`o
`ö
`
`o
`o
`o
`o
`
`o
`co
`o
`
`o
`o
`o
`o
`
`o
`o
`o
`o
`
`co
`CN
`CM
`
`O
`in
`
`o o
`
`Cl
`in
`
`O o
`
`o
`o
`o
`o
`o o
`o
`o
`o
`o
`o o o
`
`o
`o
`
`o
`
`G)
`CD
`O
`
`o
`co
`o
`
`CM
`CD
`O
`
`ci
`
`o
`o
`o
`o
`
`o
`o
`o
`o
`
`o
`o
`o
`o
`
`CM
`
`q CM q CM q CO q
`
`CN
`in q
`
`CM
`in
`
`co
`o
`
`CN
`o CN
`
`in o co
`
`o
`o
`o CD o
`o
`
`o
`o
`o
`d
`
`in
`o
`co
`
`o
`CD
`o
`
`co q
`
`o
`o
`o
`d
`
`co
`CN
`o
`
`co
`
`t-
`
`(M
`
`T—
`
`
`
`0- Cl
`2
`o
`o
`w-
`rg
`
`CO
`
`r~
`
`0) u
`CO ^-
`
`c
`o
`1-
`o
`*+-
`
`1-
`O
`c
`o
`
`o
`o
`o
`d
`
`o
`c
`o
`d
`
`ö
`q
`
`in
`
`in q o
`
`o
`o
`
`o
`o
`o
`d
`
`CD
`
`o
`CD q
`
`in
`
`10
`
`CM o
`
`05
`CM
`O
`
`q
`
`co
`o
`
`CM
`in
`o
`
`o
`o
`ü in
`o
`
`CM
`in CD
`
`o
`o
`o
`d
`
`o
`o
`o
`d
`
`o
`o
`o
`d
`
`o q
`
`CN o
`c
`o
`d
`
`CN
`CD q
`
`CN
`
`CN o
`
`co
`CD
`CM
`
`cn
`in
`
`CM
`
`o
`o
`o
`d
`
`c-
`C0
`
`o
`o
`o
`o
`
`CD
`co
`o
`
`in o o CD
`ü
`
`CN
`o
`
`o
`o
`o
`o
`
`O)
`o
`o
`
`o o
`o q a
`o
`d
`
`o
`o
`o
`o
`
`CM o
`o
`
`o o
`o
`O
`o
`O
`o
`o
`O
`o O o
`
`o
`o o
`o
`o o
`o
`o
`o o
`
`o
`o
`o
`o
`
`o
`o
`
`CN
`CO
`
`C35
`to
`
`CD
`co
`
`a>
`o
`o
`
`co
`co
`
`C
`o
`o
`o
`
`o
`o
`o o
`o
`o
`
`o
`o
`o
`o
`o
`o
`o
`o
`o
`o o o
`
`ai o
`o
`
`o
`o
`o
`o
`o o
`o
`o
`o o
`
`o
`o
`o
`o
`o
`o
`d d
`
`o
`o
`o
`o
`o
`o
`d d
`
`co
`o
`
`co
`vr
`o
`
`CD o
`o in o
`
`o
`00 o
`
`in
`o
`
`co
`CN o
`
`o
`co
`o
`
`co
`o
`
`o
`o o in
`o
`o
`
`CN
`
`o
`co
`o
`
`CD q q
`
`o
`o
`
`o
`CD q
`
`O)
`co
`
`r--
`co
`o
`
`C31 q
`
`L1
`
`co q
`
`o
`o
`o
`o
`
`co
`co
`
`CJ o
`
`CD
`co
`
`CN
`
`o
`o
`o
`d
`
`o
`c
`o
`d
`
`co
`r--
`o
`
`(D
`o
`
`q
`
`o
`o
`o
`o
`
`CN
`CN
`
`in
`
`o
`o
`
`o
`co
`o
`
`»♦- 0
`c
`0
`ü
`w
`IL
`
`(0
`L.
`LL
`
`CN
`00
`03
`
`_
`
`o
`o
`o q
`o
`o
`d d
`
`o
`o
`q
`d
`
`co
`q
`
`o
`o
`o
`d
`
`co
`CO
`
`O
`O q
`d
`
`io
`—
`
`in
`
`^r
`
`o
`o
`o
`
`co
`q
`
`CD
`CD q
`
`co
`
`CM
`CM
`
`CM
`
`^
`
`CD q
`
`o
`o
`o
`
`o
`o
`o
`o
`
`o
`o
`o
`o
`
`o
`o
`
`CD
`q
`
`<o
`o
`
`3 o
`o
`
`o
`o
`o
`d
`
`in
`CO
`
`o
`o
`o
`o
`
`CD
`
`LO q
`
`o
`o
`o
`o
`
`(D
`
`io
`co
`co
`
`CM
`00 o
`
`CM
`O o
`
`o
`o
`o
`o
`
`CM
`O o
`
`o
`o
`o
`o
`
`o
`o
`o
`o
`
`o
`co
`
`q
`
`CM
`CM
`
`CN
`o
`o
`
`co
`in
`
`o
`o
`o
`o
`
`o
`o
`o
`o
`
`o
`o
`o
`o
`o
`o
`o o
`
`8.4
`
`iL-t^,..^^.!^;,.^-...^ ^.^^-i-..,-. --.^hi-ii'n ■:-|ill'|||,ii1
`
`,
`
`rv. jiii ..^-.■..- ■■.,.-,- ... ...:..-.w-.^.n...!...^- ^ /.-^^^.^^rti^i^i^-L
`
`• -■ ■ ■
`
`.■^.. ..^^ J^. ^ ^.^-.i^^.
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 9 of 96
`Network Analysis Corporation
`
`Table Ü.3: Comparison of Ring Designs
`
`Interface
`Number
`
`4
`
`13
`
`4
`
`13
`
`Utilization
`Of Ring
`Sean By
`Interface
`
`.221''
`
`.2288
`
`.1811
`
`.1550
`
`Pack/Sec
`On Ring
`Seen By
`Interface
`
`423.04
`
`437.60
`
`346.41
`
`296.59
`
`Average
`Number of
`Packets In
`Interface Queue
`
`.44
`
`.28
`
`40.28
`
`29.95
`
`Average
`Delay Per
`Packet In
`Seconds
`
`.00821
`
`.00854
`
`.74447
`
`.90726
`
`Table 8.3a
`Sy»tem 1
`
`Table 8.3b
`System 2
`
`the average number of packets waiting to be multiplexed onto the ring, and the average
`delay per packet. An important point to notice is that in System 2 at interface 4, the
`average number of packets in the queue is over 40, a severe degradation in performance
`caused by a redistribution in traffic requirements.
`
`8.5
`
`■AU'"^'- '- - ^^ ^__ , i i *_. , . ■ . ^:.^r../-i..u1^^i^;--,^..-.^:v.i.^^.'.-i.^^^.w^- /- ■■■-.-'■'- ■ ..^V.,.. .
`
`^■l*".'"--.. L_ U. ' ■ ' -■■ - ■■■■ - ■ I - "-^^ Ä
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 10 of 96
`Network Analysis Corporation
`
`Chapter 9
`CATV SYSTEMS FOR LOCAL ACCESS
`
`9.1 Introduction
`
`A wide variety of system configurations such as loop structures and various multiplexing
`schemes have been proposed for communicating data on future CATV Systems |39l.
`
`A circuit switched video system has been developed by Rediffusion International Ltd. in
`Great Britain [24]. Multipair cables are used with each pair being dedicated to a separate
`subscriber. He may then select the program of his choice by means of a telephone type
`dial. The Rediffusion System presents interesting tradeoffs between initial investment,
`flexibility and reliability. However, since this type of system has not made significant in-
`roads into the U.S. market at present we will not consider it further here.
`
`We first present a very brief introduction to the structure common to most of the 3000
`current U.S. CATV Systems [57],
`
`Signals are received at an antenna located for ideal reception and are then relayed from
`this "head end" to individual subscribers via a distribution system of coaxial cables, broad-
`band repeater amplifiers, and subscriber taps.
`
`A cable television distribution system generally consists of a trunk section and a feeder
`section. The trunk section contains trunk cable connecting the head end to distribution
`points from which the feeder cable emanates. Located along the trunk cable are high-
`quality repeater amplifiers, which provide gain along the trunk and to the feeders. At
`the termination of the trunks there are distribution amplifiers. Along the feeder cable
`there are lower quality amplifiers called extender amplifiers and subscriber taps that pro-
`vide signals to drop cables leading to home receivers.
`
`With recent broadband amplifiers, the full Sub-UHF spectrum from 5 to 300 MHZ has
`been used. Partitioned into 6 MHZ channels for television, only a small amount of this
`spectrum is currently used for TV signals.
`
`FCC regulations now require that new CATV Systems must have two-way capability.
`Practically speaking, this does not mean that all new systems are two-way systems, but
`rather that amplifier units are installed with forward amplifier modules in place and with
`
`9.1
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 11 of 96
`Network Analysis Corporation
`
`distances between amplifiers constrained so that at some future date reverse amplifier
`modules can be installed for two-way operation. However, a number of actual fully two-
`way systems are presently being built and the number is increasing rapidly. Most present
`two-way systems use the configuration in Figure 9.1a. Filters at each end of the station
`separate low (L) and high (H) frequencies and direct them to amplifiers. Two possible
`"two-way" configurations |33| are shown in Figures 9.1b and c.
`
`Of course, two-way CATV Systems are themselves in an experimental stage so that there
`are still implementation problems in achieving written specifications. Some of the classi-
`cal electrical and communication bugs are being removed at present-ringing around loops
`through band separation filters, tuning return AGC's, alignment procedures and construc-
`tion problems.
`
`Figure 9.1a: Two-Way CATV Repeater
`
`rO»«««D TBUhK 4MPi i Mi
`
`OIHEC'IONAL COUPLER
`
`FOKWAMD
`TWUNK '"I [HITWUNK
`(«TUHN | |L|F,LTE"
`muNMoum
`
`BRiOGt»
`
`I I FORKrAHD
`H TRUNBIHI . TRUNK OUT
`-•\ 1 FILTER ;Tl
`
`I TRUNK IH )
`
`ACTUNN TRUNK AMPLIFIER' ,
`DIRECTIONAL COUPLER^
`
`Figure 9.lb: Two-Way CATV Repeater (With Feeders)
`
`'CFOPTWARD AMPLIFIER
`
`OI»E(mON«L COUPLE»
`
`V TRUNK
`
`54-260 '
`
`BTFORwAPO-pk
`174-260 iS!'"^
`
`^'B'TWUNK Ip] TWIJN«
`
`5-108 V\ ' '"
`
`RETURN TnuH'
`
`»'TRUN«
`haZ* 54-260
`
`—-JfEEDERM»KER -5 — J U
`
`FILTER h
`
`-^ I
`
`^OIRECTIQNAL COUPLER
`
`5-108
`
`Figure 9.1 c: Dual Trunk/Single Feeder Station
`(Suburban Boston Configuration)
`
`9.2
`
`..■.. ,..!...-.—..^. .. ,..-...■..■...■■■..■-... L - ■■■ ^. ^^J..:_jU.. J^^^„.,.^_:_^__.J^^
`
`- ' "-— --- '
`
` - - -—^
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 12 of 96
`Network Analysis Corporation
`
`A number of companies have developed system concepts and subscriber hardware to im-
`plement digital home response modes for existing CATV Systems |7, 8|. Among these
`arc: Fheta-Com. Jerrold, Rediffusion Electronics, CAS Manufacturing Co., Hughes Aircraft
`AMECO, Scientific Atlanta, and Cable Information Systems. Several of these companies '
`are running prototype systems in cities throughout the U.S.-EI Segundo, California-
`Dennisport, Mass.; and Orlando, Florida, among them. In addition MITRE of McLean
`Va., has installed an experimental system in Reston, Va., which incorporates a "frame'
`grabbing facility" to enable the viewer to store a frame of video data produced by a char-
`acter generator. Data frames are sent every l/60th of a second interlaced with standard
`video frames [56|.
`
`In most of these systems an FSK or PSK signal occupying a 4 MHZ bandwidth is used at
`about a 1 megabit per second rate with different carrier frequencies to and from the cen-
`tral antenna site. In each case, customers are polled at regular intervals to determine ac-
`cess to the channel. Typical proposed uses of these systems are opinion polling, meter
`reading, shopping, systems diagnostics and aLrms. Acceptable response times are in the
`order of several seconds or in some cases, even minutes [30|. We will investigate data
`transmission on existing CATV systems with required response times of tenths of seconds
`and with up to 100,000 interactive users.
`
`To illustrate our points in detail, we will consider a specific design for the Suburban
`Boston complex. The design will use the "feeder backer" configuration shown in Fig-
`ure 9.1.c with the frequencies assigned as specificly indicated. The design techniques for
`the CATV Systems themselves are well known applications of classical communications
`techniques [16, 21 |.
`
`9.2 Data Error Rates on CATV Systems
`
`Two-way CATV systems permit input from virtually any location in the network. The re-
`sult is a large number of noise sources being fed upstream toward a common source CATV
`amplifiers have a noise figure of about lOdb for a 5 MHz channel. Cascading amplifiers
`can increase effective system noise figure by 30db or morn. Nevertheless, we shall see that
`system specifications on signal-to-noise ratio for CATV systems are strin6ent enough so that
`data can be sent with existing analog repeaters, and no digital repeaters, such that bit rate
`error probabilities are negligible.
`
`For example, if the worst signal-to-thermal noise ratio is limited to 43db and the worst
`cross-modulation to signal ratio is limited to -47db, system operators may want to limit
`data channel carriers to a level of 10 to 20db below TV operating levels in order to mini-
`mize additional loading due to the data channel carriers [51). Accepting these restrictions
`in the worst case, we would be limited to 23-Jb signal to thermal noise ratio and -27db
`cross-modulation to signal ratio. Let us con- Jer both of these sets of restrictions to deter-
`mine the resulting CATV system performance for random access packet transmission
`
`9.3
`
` _■.-...■ _- ... .. . . - -^——«-—--^.-.
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 13 of 96
`Network Analysis Coiporation
`
`We calculate error rates for a FSK system with incoherent detection to determine a lower
`bound for system performance. The error rates for coherent detection or phase shift key-
`ing, of course, would be even lower.
`
`Let S = Signal power
`N = Noise power
`Nc = Cross-modulation noise power
`Nr = Thermal noise power
`t = Average synchronization error time
`T = Bit width time
`
`T'-en the sign to noise ratio is:
`
`S - Sq
`N Nc + Nr (Nc/S) + (Nr/S)
`
`where q = (1 - y )2
`
`Let Pe be the bit rate erroi probability.
`Let m be the number of keying frequencies in a multiple FSK system
`Then (49|:
`
`-jS/N)
`p - m - 1 2(m-l)2
`e " m e
`
`We assume that each packet carries its own synchronizing bits and hence there is no need
`to synchronize every terminal to a master clock. Therefore, temperature, pressure and
`humidity variations which have approximately the same effects at all frequencies do not
`enter into the calculation of t. The group delay variation over a six Megahertz bandwidth
`is less than .2 seconds [48|. For a 1 Megabit pulse rate T = 1 second and t = .2 ^seconds
`Hence, q - .36. We, therefore, have the error probabilities in Table 9.1 for the suburban
`Boston complex.
`
`For effective sig,,al-to-noise ratios above 20db there is a threshold effect for error proba-
`bilities. This is borne out by the negligible error rates. Even for the degraded specifica-
`tions the error rate is low enough for the most stringent practical data requirements Fur-
`thermore, at a rate of 10^ pulses/second the FSK signal will occupy the 6MHz bandwidth
`with negligible intermodulation into TV channels.
`
`Consideration of reflections, intersymbol interference and 60 cycle hum also lead to the
`conclusion that CATV systems are excellent media for packet data transmission.
`
`The signal levels ir. a CATV system are controlled via AGC and dual pilot carriers. Ripples
`are kept to less than Idb over the whole frequency band. In any case, frequency shift
`
`9.4
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 14 of 96
`Network Analysis Corporation
`
`Table 9.1: Error Rates for FSK
`
`System
`Label
`
`Type of
`Specification
`
`<NC/S)
`
`(S/Nr)
`
`m
`
`P.
`
`A Undegraded -47db 43db 2 1/2e"5'148:s1/2X10"2'239
`Specs.
`
`B Undegraded -47db 43db 4 3/4e"57i:s:3/4X10'248
`Specs.
`
`C Undegraded -47db 43db 8 7/8e"'O4'5:7/8X10"45
`Specs.
`
`D Boston Specs. -27db 23db 2 1/2e'51as.5X10*22
`- degraded by
`20db
`
`keying is insensitive to small amplitude variations. The effect of group delay error has
`already been taken into account in the use of q in the formula for error probability. The
`remaining source of intersymbol interference is the reflection of pulses and the effect of
`the reflected pulses on the transmitted data. There are three types of disturbances due to
`reflections. In each case, we shall -ee that video restrictions are certainly stringent enough
`to ^void any difficulties for data t.ansmission.
`
`Periodic changes of minute magnitude uniformly distributed along the cable length, the
`magnitude of changes being essentially equal from period to period, due to the nature of
`the manufacturing process, cause reflections which add in phase at certain frequencies.
`The signal strength relationship of the reflected wave to the incident wave is referred to as
`structural return loss (SRL). Typical values for the magnitude of SRL are better than
`-26db [43].
`
`Assuming that the reflected signal is always of an opposite sign to the original signal, the
`signal level is degraded by at most S-a, where a is the amplitude of the reflected signal.
`The signal-to-noise ratio becomes [53]:
`
`c
`In other words ^ is degraded by (1 - |)
`
`N N U S'
`
`For a reflected signal of -26db, (1 - |) is .9975-quite acceptable.
`
`A localized change or changes on the cable cause echo phenomena. Low reflection coef-
`ficients of active and passive devices and the use of directional couplers at all subscriber
`taps ensure that the magnitudes of reflected pulses are in the "no ghost range" of Fig-
`ure 9.2 [45, 40). These are translated into critical distances for different types of cable
`in Figure 9.3. Thus, for example, considering the reflection on .412 inch cable at Chan-
`nel 13 the critical distance is about 250 feet and the ratio of the magnitude of the reflect-
`ed signal to the magnitude of the original signal is -23db.
`
`9.5
`
`__, -
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 15 of 96
`Network Analysis Corporation
`
`-
`
`
`
`1 1 I
`GHOST VISIBLE
`i i i
`
`V
`s
`\
`
`s s <
`N N W
`s»
`
`"~"
`
`1
`
`—1—
`
`\
`
`-
`
`> s ss
`s s s »«
`
`NC G H 3S T S
`
`—
`
`s
`
`m "
`Tl
`
`CD
`m
`
`5 "
`s
`52 JO
`
`GNAl
`
`z
`a -"
`
`CD
`
`1 • 1 ) X ■ 1 0
`
`i M 10 0 10 o
`
`too
`
`1000 >000
`
`MM
`
`tODOO
`
`—Li-Li
`
`K>0 OO
`
`TIME DELAY IN NANOSECONDS
`
`Figure 9.2: Curve Showing Perceptibility of Ghosts
`
`z
`Z <
`I u
`
`LU
`-i
`z
`<
`X
`o
`
`_l
`z
`2
`<
`I
`Ü
`
`i—
`CO
`
`_l
`tu
`Z
`Z
`<
`I u
`
`n CN
`»—
`
`_j
`
`Z
`z
`<
`I u
`
`LU z
`z <
`I o
`
`(N
`
`_)
`LU z
`z
`<r
`i
`
`(N
`
`_J
`LU
`
`z
`<r
`i o
`
`50 70 100 200 iOO 700 1000
`
`?OOO
`
`i ' i
`5C00 I0C00
`
`DISTANCE FROM POINT OF MISMATCH IN FCET
`
`Figure 9.3: Graph for Determination of Critical Cable Lengths
`
`9.6
`
`■■ f.jt -.-■-... ■ - -
`
` _
`
`-- - ■ - ^
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 16 of 96
`Network Analysis Corporation
`
`Randomly distributed changes of random magnitude wnich persist throughout the cable
`length cause reflections which do not add in phase. These can be taken into account in '
`noise calculations and are usually negligible.
`
`Cable system amplifiers are powered by low voltage 60 Hz power through the co-axial
`cable This power may be as high as 60 volts (RMS) and currents may run to 10 ampheres
`(RMS) with peak currents even higher. There are significant harmonics of the power line
`frequencies present. Some amplifiers use switching mode power supplies with switching
`frequencies in the 10-20 KHz range. Hash from these switching regulators also finds its
`way mto the cable. However, both the 60 cycle harmonics and hash limit only the area
`of very lew frequencies which are generally avoided for data transmission.
`
`9.3 Other Performance Criteria for CATV Systems
`
`Data users may find cable system reliability quite poor when compared with the common
`earner facilities with which they are familiar. One of the major problems with data trans-
`mission on CATV Systems is that there is no redundancy of path cables or amplifiers
`There are no government or industry minimal standards for acceptable performance- hence
`performance will vary from system to system. Many old systems were built to extremely '
`loose specifications on noise and cross-modulation and have serious reflection problems be-
`cause of the use of unmatched subscriber taps. Fortunately, systems in large cities and
`new buildings are much newer and are required to meet more exacting standards.
`
`Even with these systems, the construction norms are still those which satisfy casual TV
`viewers, not data users. Thus, loose connections cause intermittant transmission condi-
`tions, and momentary "disconnects." These would cause only minor •■flashes" on a TV
`picture but constitute major data dropouts in a high speed data circuit. Finally systems
`may be inadequately tested, and hence, in some parts of a CATV System noise and cross-
`modulation levels may not meet written system specifications. The limiting factor in de-
`termining the performance of the system will not be Gaussian noise interference but a
`number of practical factors which provide interference, generally categorized as ''impluse
`noise." These factors are difficult to characterize and include phenomena such as loose
`connections, cracked cable sheaths, and R-F leaks.
`
`Two factors dominate the specification of any data transmission mode on a CATV System.
`
`a. That data is sharing a transmission medium with video signals.
`
`b. Thai there will be a large number of users.
`
`9.7
`
`- ----- —. . .—.^^—^J^.—^J.».-^». .w - ^ „^—.^ .. ^ — .__.... ^_ - - -
`
`

`

`Network Analysis Corporation
`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 17 of 96
`
`9.3.1 Interface with CATV System
`
`Two Way Options
`
`The data transmission system must be readily adaptable to a wide variety of existing
`CATV System designs and two-way options.
`
`Data Rates
`
`The data signals must not cause visible interference with video signals.
`
`Installation
`
`if auxiliary data equipment is to be added to the CATV System, it must satisfy the fol-
`lowing requirements:
`
`■ It can be installed with only minor changes in the CATV System.
`
`■ It need be installed in only a small number of locations.
`
`• It can be installed rapidly in early hours of the morning to prevent interference
`with TV service.
`
`Low Cost
`
`To maximize the marginal utility of data distribution over the CATV System any equip-
`ment introduced must be inexpensive. > ? M K
`
`9.3.2 Interface With Population
`
`Population Density Variations
`
`Standard transmission configuration options must be available for systems of various sizes
`population densities and percent of active users. Because of the huge number of potential
`users, all terminal equipment must be simple and inexpensive.
`
`Unsophisticated Users
`
`To minimize user interaction with the system operating mode, all terminal equipment must
`be the same for each location; it must use the same frequencies and data rates; and it must
`have no options for equipment modification by the user.
`
`The MITRE Corporation has patented a system called MITRIX which meets all the above
`specifications [58] and has many other excellent features. Some of the disadvantages of
`
`9.8
`
` , . . . , . ^_ ,
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 18 of 96
`Network Analysis Corporation
`
`polling for terminal-oriented networks |50l are the synchronization delays (32] and the
`T^Z0^1 0f Channel bandwidth occupied by simply polling 100,000 subscribers
`MITRIX overcomes both these problems by using a time division multiple access scheme
`Furthermore, since the number of time slots per second is dynamically assigned to a sub-
`scnber the system also avoids the wasted bandwidth in allocating fixed frequency bands
`(FDM) or f.xed time-slots (TDM) to subscribers who are active for only small amounts of
`time. The users mterface unit requests a certain number of time slots within frame per-
`lods and these are then allocated by a Computer Digital Interface Unit; a DEC PDP-15
`I he system is highly flexible in structure, efficient and inexpensive.
`
`9.3.3 Other Considerations
`
`However there arc still some tradeoffs involved and for large systems improvements are
`still m the offing. In particular, we are still faced with the problem that if a subscriber
`logs in at a terminal and makes a request for a certain data rate then he holds those time
`slots until he logs out whether he is actively typing or thinking and not typing. For small
`systems or systems with a small number of users, this may be an acceptable inefficiency
`in bandwidth use But for systems like the suburban Boston system it may not be accept-
`able. As we shall see the factors involved are the available bandwidth, the average ratio of
`active user time to inactive user time in a logged-in period, and the average number of ac-
`tive users. The alternative which makes more efficient use of bandwidth for high peak to
`average data rates is the random access packed multiplexing method previously derived and
`applied to the satellite channel. With packets, terminals seize the channel only when they
`arc active. Hence, more users can be accommodated. Furthermore, reservations of chan-
`nels are not required. As we have seen the random access feature results in a channel
`ava.labiluy of l/2e of the band or 1/e for a slotted system. This is effective if the aver-
`age peak to average data rate is greater than the number 2e or e, respectively.
`
`9.4 Number of Active Terminals for Random Access Pa.iket Sample System
`
`We have seen in Chapter 5 that for an unslotted random access packet channel the maxi-
`mum number of active terminals kmax is given by (2eA r)^. If we let d be the pulse dura-
`tion and let 7 be the number of pulses per packet, then kmax is (2eX7d)-1 where X7 has the
`imension of pulses/second per terminal. For a two level FSK system, this is the same as
`bits/second per terminal. In Figure 9.4 wc plot the maximum number of active terminals
`versus Xy for the systems in Table 9.2 using the above equations. The curves labeled A
`B.C and D correspond to the slotted system in Table 9.2 labeled A,B,C, and D. The lines
`labeled A , B', C, and D", are for the corresponding unslotted systems,
`
`We can now examine Figure 9.4 to determine system performance under some typical data
`transmission requirements. For a data rate of 40 bits/second per terminal, a single trunk
`can handle 4000 terminals with a slotted system (1 Megabit/sec) with an error rate of
`10 at a signal to noise ratio degraded by 20db. The average number of TV sets per
`trunk in the suburban Boston system is approximately 27,000. Hence, the simplest
`
`9.9
`
` a—i
`
`- - ■ ■
`
`

`

`Case 1:16-cv-02690-AT Document 121-14 Filed 08/05/16 Page 19 of 96
`
`Network Analysis Corporation
`
`4
`
`10 100
`
`PULSES PER SECOND PER TERMINAL
`
`1000
`
`Figure 9.4; Performance of Random Access Packet Cable System
`
`modulation scheme will handle one th