1╛ л >═ Э╧╧╧╧╧ ╨
Reviewed and approved by
WILLIAM H. KIRN
PRESIDENT
H & K COMMUNICATIONS, INC.
ALL INFORMATION CONTAINED HEREIN IS PROPRIETARY AND IS NOT TO BE
REPRODUCED WITHOUT THE WRITTEN CONSENT OF THE PRESIDENT OF H & K
COMMUNICATIONS, INC.
COPYRIGHT 1986, -1995 H & K COMMUNICATIONS, INC., ALL RIGHTS
RESERVED KNET THEORY 1
1.2 Design goals & standards, philosophy and methodology
1.2.1 Design goals
1.2.2 National Standards.
1.2.3.1 Manufacturability
1.2.3.2 High reliability design
1.3 Basic external specifications & capabilities
1.4 Comparative analysis of the KNET twisted pair bus architecture
1.4.1 The wiring problem
1.4.2 Wireless alternatives
1.4.3 Cable topologies
1.4.4 Multiplexing techniques
1.5 General KNET system theory of operation
1.5.1 Overview
1.5.2 KNET devices
1.5.2.1 KNET Phone
1.5.2.2 CO Trunk Interface Card
1.5.2.3 VCIF (KNET Controller InterFace)
1.5.2.4 KNET Controller
1.5.3 KNET Bus
1.5.3.1 Power distribution
1.5.3.2 Control link
1.5.3.3 Voice channel multiplexing and switching
1.2 Design goals & standards, philosophy and methodology
1.2.1 Design goals
The KNET Telephone System was created to bring the performance,
convenience and low costs of the modern consumer electronics
revolution to the small business and home office user. Current analog
loop technology was invented in 1876, touch tone and the 1A2
multibutton key set, in the 1950s. Although modern PBXs and Key
systems are more sophisticated, they have debatable performance
improvements in light of comparable or increased costs. The
stubborn persistence of the analog single line set and the 1A2
Key set in the market is proof of this contention. H & K
Communications believed by appropriately applying the full
spectrum of modern electronics technology, a business- home office
telephone system could be designed with an improvement in costs
(direct and life cycle) and performance (improved productivity and convenience).
1.2.2 National Standards.
The system conforms to all applicable U.S. standards and design
guides. Specifically:
EIA Standard RS-464, RS-464-1, RS-470, & RS-478;
FCC part 68 and part 15; NEC Article 725;
ANSI C62.41/IEEE Standard 587- 1980;
ANSI/IEEE Standard 518-1982, 820-1984,
269-1983, 743-983, 753-1983; ANSI C63.4-1981,
1.2.3.1 Manufacturability
System components are designed to allow low cost, high volume
production. All KNET components use multisource commodity parts.
There is a high commonalty of parts among circuit modules to
lower the cost associated with handling and stocking different
components. The basic KNET System has a minimum of different
assemblies to manufacture and inventory. No circuit adjustments
or selected components are required. The designs are based on
worse case component tolerance runout in order to realize high
manufacturing yields.
1.2.3.2 High reliability design
All KNET components are reliable and hard to break in the full
operating, manufacturing, and service environment. This means
power supplies are short proof, or if they do fail, protection
circuits will limit KNET bus loading. Failed devices won't
disable working devices. Circuit node impedances are held within
upper and lower bounds to minimize humidity leakage without
compromising power consumption. All circuitry is designed to be
ESD (ElectroStatic Discharge) proof (Up to 25KV). All software
systems are monitored by watchdog circuitry which will
automatically re-boot a crashed system. Misaligning plugs or
making connections while powered up won't cause damage. Modular
plug connections can be doubled to improve reliability. No
batteries are used in devices for memory retention. Softfail
architecture means that a controller failure will not immediately
terminate a call in progress. The system or phone will recover
automaticly from a power fail or disconnection without user
intervention.
1.3 Basic external specifications & capabilities
The KNET System is a telephone PBX or a Key System, with the
following external characteristics:
All voice, control, and power distribution occurs through a
single common twisted pair bus, the KNET.
The voice path capacity of the KNET bus is currently 28
simultaneous, single voice channels, allocated in adjacent pairs
to make 14 full-duplex voice paths. These pairs have a 200 Hz to
5khz audio bandwith which is 50% more than standard telephone
systems. The bandwith may be further increased with a reduction
in number of channels by changing components in KNET devices and
with suitable software changes.
The maximum range of voice paths is over 3,000 feet on 26 to 22
guage twisted pair. Range is a function of channel frequency and
twisted pair gauge.
Power is supplied at the CO INTERFACE BOX, or at additional
points on the KNET by power boosters. The power source is a 50-
60 volt .5 -1.0 ampere source buffered by a circuit designed to
prevent loading of the voice and communication channels. In
addition, it has circuitry designed to prevent damage to a
standard phone inadvertently plugged into the KNET. Each KNET
phone pulls about 2.5 watts maximum (less during idle) and each
3-CO trunk card pulls about 4 watts. Loading calculations
indicate about 16 phones evenly distributed along 1000 feet of 22
guage can be driven from a single power source at one end. All
KNET devices have low voltage indicators to aid in determining
the need and location of additional power boosters.
The KNET communication link has a range in excess of 3,000 feet.
The link has a basic rate of 9600 bits/s and an effective data
rate of 150 to 600 characters (8 bit) per second. It was
designed to handle simultaneous, multiple contending devices,
quickly. Specifically, it can handle up to 70 independent
control messages per second. The link can send a message to or
from any of up to 256 devices within 20 ms. The link was
designed not to be a bulk data carrier, but to have sufficient
capacity to cost effectively handle control and display (32
characters or less) packets for up to 32 Key telephones.
The KNET Phone has 49 software assignable buttons and a 32
character alpha-numeric LCD display. A single full-duplex voice
path is switchable to any KNET voice channel. In addition, it
has a speaker for on hook dialing, monitoring, and ringer. The
audio circuitry provides for 16 volume levels, 2 tone generators,
1 white noise generator, numerous cadences, signaling beep, mike,
and speaker switching.
The KNET 3-CO Cards provide audio interfaces up to 3 Central
Office trunks. It provides for trunk and call progress
signaling; specifically, ring detect, loop connect, wink detect,
touch-tone and dial pulse dialing, and call progress tone energy
and cadence detect. It also provides for automatic line loss
compensation and touch-tone decoding for automatic extension
dialing or remote feature activation. It has conferencing
circuitry which allows external and internal conferencing,
although internal conferencing may require extra trunk
interfaces. It has a full duplex speakerphone circuit per
line. And finally, the 3CO Interface has automatic line
balancing circuitry to increase line hybrid return loss so as to
allow volume control without excessive sidetone, amplified
conferencing, and improved speakerphone operation.
The VCIF (KNET Controller Interface) provides an isolated
communications interface between the KNET Controller (typically a
Personal Computer) and the KNET. There is only one VCIF per
system controller (there may be back-up controllers). The VCIF
performs KNET communication protocol translation to the PC bus of
the KNET Controller.
The KNET Controller is an unmodified IBM PC compatible computer.
The control code is written in "C" and operates under MS-DOS.
The Personal Computer can also serve as a powerful interactive
terminal for attendant functions, and additional application
software and capabilities such as voice mail and SMDR. The KNET
Controller can execute other software in the foreground while
controlling the KNET in the background.
1.4 Comparative analysis of the KNET twisted pair bus architecture
1.4.1 The wiring problem
Currently most PBX, Key systems, or central office systems
distribute voice, control, and power with 1-4 pair wiring starred
out of a central switch. Old mechanical (1A2) key systems use a
25 pair cable (thick wire) as a bus (also often starred) out of a
KSU (Key Service Unit). Modern electronic key systems typically
use 2-4 pair (skinny wire) starred out of the KSU. These type
systems are sometimes called hybrid key systems because they have
PBX structures and related features. Very simple electronic key
systems with 3 or less outside lines may use a 4 pair bus (3 CO
lines, 1 intercom) wiring scheme without a KSU.
In all of the above systems voice is distributed as a single full
duplex path per wire pair whether or not the voice signal is
digitized. In systems with extra lines, the lines are used for
signaling, and power. This wiring produces considerable long
term costs in skilled labor to install or to make "moves and
changes".
With long term system cost reduction as the primary design goal,
various digital, frequency, and hybrid multiplexing schemes were
investigated. Also, wireless techniques, such as infrared,
ultrasound, leaky coaxial cable, induction loop, radio, and power
line carrier were considered.
1.4.2 Wireless alternatives
Even though a fully wireless system would be ideal, there is
still a need for power. Without a line cord, a battery system
would need to be recharged or replaced periodically. Assuming a
phone has a 10 mW average draw (a very optimistic and costly
design goal) and a 20 watt-hour battery (4 "D" cells, probably
the maximum practical), the battery would need to be recharged
every 3-6 months. Any longer interval with a nickel-cadmium or
nickel-hydride battery wouldn't be possible because self-
discharge will cause it to lose half its capacity in 3-6 months.
Throwaway alkaline batteries might triple the service interval.
Another problem with batteries are there poor reliability which
can become a significant service problem in any but the smallest
system. H & K has had experience with low power drain
communication systems and determined that although theoretically
possible such a battery system would be too expensive and
compromised in performance for use in most business phone
stations in a system. Any practical wireless system would still
have a power cord either to the unit or a charger. Of course for
mobility needs there is reasonable justification for some phones
in a business system being cordless.
Ignoring power considerations, the wireless techniques of
infrared and ultrasound suffer from line of sight and practical
range limitations of 20 to 100 feet. Infrared is range limited
due to the poor signal to noise ratios of the receiver which is
caused by high background ambient noise multiplied by the
ultrawide detector bandwith and its very low omnidirectional
signal capture area. Ultrasound, on the otherhand, suffers from
bandwidth limitations in air (over these distances) of 50 to 100
kHz due to attenuation with increasing frequency. It also
suffers from severe multipath and directional propagation
characteristics. With enough local relay stations tied into a
central switch (like cellular radio), it is possible to get
around the limitations of infrared and ultrasound, but to do so
is not economically viable for business phone applications.
Radio is the most viable wireless technique and includes
induction loop and leaky coaxial cable which are near-field
antenna techniques. Suitable radio frequency spectrum for this
application is presently not available except for some bands on
which spread spectrum techniques may be used. There are near
term possibilities (5 to 10 years) for spectrum for PCN (Personal
Communications Networks) which can be though of as mini-cellular
phone/modem systems with ranges of about 1000 feet to a local
node. PCN systems may be under the control of a public
communications companies which will tariff the use at a rate of
about 1/5 to 1/10 the cellular telephone rate. The FCC may
decide that there is not enough spectrum space to allow usage of
this spectrum without the disincentive of charging per minute
especially when effective and lower cost wired techniques are
available.
Gaining spectrum from the FCC is a longterm bureaucratic process
requiring demonstration of technical viability and compelling
public priority over other radio needs. Priority usually means
there is no other viable way to perform the service without
radio; such as in mobile dispatch, aircraft and marine
communications, widearea public broadcast, etc. The presence of
viable wired telephone systems would disincline the FCC to
approve channels for the same purpose. For example, cellular
telephone, only recently, got spectrum space a decade after
demonstrating technical viability and strong public demand; and
then only because, it was more spectrum efficient, used less
desirable short range UHF frequencies, and was a premium service
capable of serving only 1-2 % of the telephone population. On
top of that, it took another 5 years of "legal turf wars" to
become commercially available.
Radio systems suffer, relative to wired systems, from the major
problem of highly variable propagation characteristics (each
frequency band has its own peculiar problems). The radio signal
may not be able to penetrate through the walls, ceilings, and
floors of a building to a nearby extension but is picked up by a
similar system 10 times as far away in an adjacent building.
Part of the propagation problem is the wide 100 dB+ dynamic range
of two-way radio signals. These signal ranges coupled to
spectrum crowding make interfering spurious responses expensive
to overcome. Spread spectrum techniques can make a significant
improvement in these propagation problems but at significant
premium in cost and power consumption. Although for cordless and
movable (in building) data applications, spread spectrum systems
are available but at significant costs.
There is some interest in using radio to provide customer access
to a telephone carrier by replacing the wire link between
building and pole (or pedestal) with a multichannel digital radio
link. Such a system may be cheaper and more reliable by
eliminating the need for the installation labor by the telephone
company, by allowing multiplexing all the way down the local
loop, and by improving the resolution of who's at fault when the
phone doesn't work. In addition such a system makes it
economically possible for another telephone carrier to compete
against the local telephone company because it needs only to
signup a small fraction (a few per cent) of the total number of
customers within the pole coverage range (~1000 to 2000 ft
radius) in order to cost justify the pole or pedestal transciever
whose capacity would probably increment in about 20 channel steps
(a single multiplex digital transciever). In order for the
customer to gain radio access to such a service, he may need to
mount an antenna with a unblocked radio path; such as, in the
attic, through a window, or outside the building.
Such radio access to the telephone network systems will not
obsolete the need for PBXs or Key systems. In fact, they may
spur a need for a multichannel PBX or Key system on the
customer's side. Since radio replacement of the pole to customer
connection may use digital multiplexed radio, it may be possible
to give the customer several channel capacity for very little
incremental equipment cost. The radio access telephone carrier
would have a very strong incentive to ensure that customer line
access equipment have a maximum cost effective capacity (probably
at least 2 channels). For a radio access telephone provider
there is no cost of individual fixed wire loop per telephone.
Since virtually all of the costs will be associated with usage,
the radio telephone provider might only charge for usage and
eliminate or reduce the fixed monthly fee. This means that the
radio telephone provider would like to encourage the use of
"virtual" lines in the home, allowing as many users as possible
access to the network through a single radio link. Because
customers might have access to 3 or more "virtual" lines
simultaneously without paying the monthly fee for these lines,
they would have an incentive to install a key system phone system
to give them access and control of these "virtual" lines.
Induction loops (used on low frequencies) and leaky coaxial
cables (used on high frequencies) are sort of half-way solutions
between radio and wire systems with characteristics of both.
They can usually be designed to get around FCC problems. They
have better one-way propagation characteristics, but are not
secure from nearby radio eavesdropping, or immune from
interfering with a similar system in an adjacent office. The
dynamic range is similar to regular radio if signals go directly
between phones. If all signals go through a relay hooked to the
coax or loops, the dynamic range is less but the expense is
higher. Also, the coax or loop is not a labor improvement over a
simple wired system because the loops or coax needs to be
correctly strung, and fed signals at the right points by a
skilled technician. Most commercial applications of these
techniques are used to augment radio services (such as bringing
radio into a mine, tunnel, or basement building or providing
local paging) rather than replacing wire based signals.
Power line carrier systems which use the AC power line as a
transmission medium are alright for low speed local data
transfer. However, for voice signals the medium is too noisy due
to load switching and poor line balance, and it has insufficient
reliable bandwith due to transient and RFI (Radio Frequency
Interference) filtering in both power distribution strips and AC
loads. Also, the signal range is usually constrained to the
outlets fed from a single tap on the power line distribution
transformer. The user may not be able to communicate between the
two power plugs on either side of a room without adding high
voltage bypass capacitors across power lines at a breaker box.
1.4.3 Cable topologies
The simplest and least costly telephone interconnect topology is
a single twisted pair bus. However, due to multiplexing
difficulty, virtually all modern telephone systems use star
topologies with twisted pair wiring. Such systems place the
switching circuitry at the hub of the star rather than in each
phone as in KNET. Mechanical key systems, the only significant
exception, place most switching in each key set using the common
KSU for auxiliary functions and using an expensive 25 pair cable
as a common interconnect bus. If compared against the KNET
system's distributed switching, a maximally loaded central switch
with standard phones or a plain key system may offer a slightly
lower switching cost per phone. If not fully loaded, the fixed
cost of the central switch or KSU may make the KNET less
expensive. Most compelling though, any disadvantage KNET has in
capital cost against star topologies is overwhelmed by the life
cycle costs. Industry figures indicate the cost to install the
average business phone is $100-$150 per phone and it is moved
once a year at an average cost of $75+. The KNET bus structure
should at least cut these figures in half by significantly
lowering labor and material costs through reductions in cable
lengths, wiring complexity, and number of connectors. It should
also be possible for a large percentage of users to install or
make "moves and changes" themselves with considerably less
disruption or delay in service. Future downward cost spirals of
electronic equipment and upward spiraling labor costs will make
this analysis even more viable through the next decade. More
complex wiring schemes were considered. Some of these provided
somewhat lower equipment costs but suffered increases in typical
life cycle costs.
Despite the apparent advantages of the single twisted pair bus
preliminary analysis indicated it could be potentially less
reliable than a star structure. In a star structure an
unreliable cable or connector will cause only a single phone or
trunk to fail, but in a bus structured system such a problem may
cause the whole system to malfunction. Device malfunctions can
produce bus failures. (Ring topologies do even worse since node
failures are usually fatal to system operation.) Consequently,
the KNET bus structure was designed to be as reliable as a star
structure. If a KNET device fails there are a number of
protective mechanisms designed into it to prevent disabling the
KNET.
Cabling problems are the most likely source of failure for the
KNET. Since connector opens are the overwhelming source of cable
problems, these typically limit failure to individual phones.
(Most of these are due to modular cable crimp failure or
conductor breakage due to flexing cycles.) The only time shorts
are likely to occur are during installation or changes in wall
wiring. Shorts will not harm the KNET which has automatic
resettable fusing. Consequently, it is not expected that the bus
topology of KNET will result in lower reliability than star wired
phone systems.
1.4.4 Multiplexing techniques
Digital techniques were thoroughly investigated and found not to
be optimal for local voice distribution. Digital voice coding
techniques although expensive may be appropriate in large
systems. In such systems they provide immunity from incremental
degradations that analog signals experience when going through
numerous long distance links and switches. Small systems
typically have less need for noise immunity because noise pickup
tends be a function of system size and complexity. Large systems
have larger wiring "antennas" and more sources for cross talk.
1) Digital coding in current telephone systems
Digital voice encoding tends to be expensive due to the added
cost of analog to digital conversion, anti-aliasing filtering and
time slot assignment circuitry. This fixed cost per phone may be
offset by savings on switching and intersystem trunking costs if
the system is large enough. Digital switching and trunk
multiplexing techniques have a lower rate of cost increase with
size although they have a higher minimum system cost than analog
techniques. Because telephone system circuitry per phone
increases with system size, at some break even point (~100
phones) the savings on digital coding offset the conversion and
minimum system costs. (This analysis doesn't apply to long
distance trunking where bandwith and repeater spacing
considerations may give analog systems cost and capacity
advantages.)
In fact all low to moderate cost phone systems of less than 60
phones use analog switching techniques. These systems have no
worse and often better (greater bandwith and lower distortion)
audio compared to digital systems. On such systems the main
source of audio degradations are the analog loop trunk lines from
the central office. Only systems large enough to bring digital
trunk carriers (more than 24 trunks) directly into the switch can
avoid this problem.
There is an industry trend on high end digital systems to do the
analog to digital conversion at the phone and send the digital
data via proprietary high speed modems to the central switch.
This is also the goal of ISDN (Integrated Services Data Network)
which will be provided by the local telephone companies. This
saves nothing on costs, but adds the expense of the modems. It
does provide digital communication connect capability at the
phone but usually with a several fold increase in cost over a
data only connect system. Perhaps the only viable advantage
these systems have over standard systems is that the integrated
data path allows implementing sophisticated feature phone or
voice/data functions. Some of these systems potentially can save
some money by avoiding separate cabling for the data system.
However, analysis of typical users suggest this saving is small
due to the following factors: 1) Phones outnumber terminals by
more than 2 to 1 in most environments. 2) Although adding data
paths to a location is a major cabling problem, another major
problem is adding additional voice paths to a location. 3)
Because of wiring uncertainty, multiline cable of sufficient
capacity is usually run anyway to any phone or terminal
termination to allow a reasonable increase in users. 4)
Voice/data PBXs can usually only service a subset of the common
data terminals. They do not provide universal connectivity.
2) Digital coding on a bus based system
Digital transmission on a bus based system is the basis for a
number of commercial LAN (Local Area Network) data communication
products. Telephone PBXs have been experimentally demonstrated
on such products. (With 50 milliseconds of voice buffering a 10
Mbits/s Ethernet can support 60 simultaneous conversations.) Such
high speed bus systems suffer from high equipment costs ($1000+
per tap), costly cabling and connections, cable sensitivity to
kinks and termination quality, and without repeaters the
inability to branch and go beyond a limited range (1500 feet). To
send voice, the added cost of analog to digital conversion would
also be incurred. For such systems to share data communication,
they would need to limit data access to avoid causing voice delay
disruptions.
A number of digital systems optimized for voice were considered
in attempt to get around some of the above problems. However,
two fundamental problems appear to make a digital bus have poor
cost to performance.
The first major problem of a long twisted pair digital bus is its
range/bandwith limits. Twisted pair attenuation increases with
bandwith, for 22 gauge being 19 dB per mile at 400 kHz and
increasing to 31 dB per mile at 2 MHz. Since digitizing the
voice signal typically increases its bandwith by tenfold, doing
so reduces the bus capacity manyfold over FDM techniques. Of
course higher bandwith cabling such as coaxial cable or glass
fiber can solve this problem but only with significant (X 10)
increase in material and labor costs. Techniques such as
multilevel encoding, pulse amplitude modulation (PAM), or pulse
position modulation can reduce the bandwith but make the signals
significantly more vulnerable to pulse reflections.
The second major problem of a long digital bus (longer than a 1/4
wavelength of the highest frequency) is noise and pulse
distortion errors due to reflections from impedance mismatches.
This problem gets worse as the bus length increases due to the
rise in signal dynamic range caused by attenuation. This means
digital bus terminations or taps must be made with care and
branches are prohibited unless expensive adaptive echo canceling
circuitry is used. Such circuitry is increasingly expensive and
power hungry with rising frequency. ISDN interfaces are
expensive because of the need for 140 kbit/sec echo cancelers.
Echo canceler designs for a 20 channel system at 1+ Mbits/s would
require nanosecond signal processing. This would result in
prohibitive costs per device and raise the power consumption per
device beyond practical line powered levels.
The difficulty of controlling digital reflections due to
impedance discontinuities is the major reliability problem of
high speed bus LANs such as Ethernet. In fact, LAN designers
have backed away from both of the above major problems by various
performance compromises such as lowering the bit rate (Appletalk,
HP-IL, Cheapernet), going back to star wiring and doing bus
contention on a short bus (Cheapernet, Star Net), going to daisy
chained ring topologies (IBM token ring, HP-IL), or encoding data
on a radio frequency subcarrier (Sytek, Wangnet).
Long (more than 1/4 wave) digital buses of indeterminate lengths
also suffer from other problems. One is uncertainty in
propagation time which makes digital time slot assignment more
complex than on a short bus. To get around this requires storing
data, and sending minimum length data blocks so as minimize
propagation dead time. Also, each device needs to generate its
own synchronization from data on the bus. Probably, the more
proven technique than time slot is packet switching. (Most LANs
are packet switched.) Theoretically, this can double the
throughput over a time slot system because most conversation is
half-duplex, and the non-talking side of the conversation is
wasted in a dedicated time slot system. However, packet switch
systems incur overhead from packet addressing, control,
propagation time, and collisions. This can be mitigated by
making the data blocks long enough. The major problem (in
addition to the two above) with packet switch systems is the high
cost of the complex control circuitry at each connection to the
bus.
Practical digital multiplexed twisted pair bus systems are
possible if the length and channel capacities are limited. Such
systems typically have much less capacity and range than the KNET
system. Phillips has demonstrated such a digital system of
medium cost (?) which has a capacity of 8 half-duplex (4 full-
duplex) channels over a range of 300 meters of twisted pair.
The KNET's use of FDM (Frequency Domain Multiplexing) solves all
of the above problems efficiently, economically, and without loss
of relevant performance. Because all 28 KNET voice channels are
below 427 kHz a total bus length of 5,000 feet or more is
possible. Most important though, any topology is allowed with
special terminations unnecessary. Impedance reflections still
occur but typically only cause fixed amplitude variations of less
than 6 dB in the voice carrier. This is because reflections
which have significant phase shift must travel a distance which
causes an attenuation of 2 to 6 dB. Worse case reflective
interference occurs when a signal at the maximum KNET frequency
of 435 kHz reflects from a 1/4 wave open causing phase inversion.
At this frequency the signal is attenuated about 2.5 dB
traversing the 1/2 wave (1/4 down & a 1/4 back) 22 guage distance
of 650 feet. Because of phase inversion the reflected signal
subtracts from the forward signal and reduces its amplitude 12
dB. For lower frequency signals such as a 100 kHz the
attenuation for the 1/2 wave distance which is 2800 feet
(reflects cable phase velocity) increases to 5 dB, which means
the resulting interference is less, causing only about 7 dB of
attenuation. Because the system has a dynamic range in excess of
20 dB, these levels of signal fluctuation shouldn't be a problem
except when operating at the upper KNET frequencies and the
maximum range of 5,000 feet where the attenuation is about 15 dB.
Should the system diagnostics sense such a condition, it is
possible to assign lower frequency channels to avoid the problem.
Smaller guage wire will have less reflection problems but will
have greater attenuation which may limit the range on the upper
frequencies by a few thousand feet. Practically, very few
installations should push these limits or encounter these
problems.
The flexible KNET topology is not a problem for the control link
even though it is a bus based system because the bit wavelength
exceeds 35,000 feet. Due to the nature of the modified NRZ
coding worse case reflections off of line end opens are in phase
with the originating transition and potentially can only
interfere with a following bit transition after 1.3 bit intervals
of delay. Such a reflection would travel over 40,000 feet of
twisted pair with a loss in excess of 30 dB. (This is composed
of 24 dB of 22 gauge cable loss and at least two reflections or
traverses past the VCIF load with at least 3 dB per pass.) This
reflection is 10 dB less than the minimum receive signal
threshold which is 20 dB below the maximum signal.
The KNET, although ideal for local phone distribution in a
building, is not recommended as a central office local loop
replacement. This is due to insufficient range and crosstalk
problems. Although the KNET voice and control channels have a
maximum range equal to most local loop distances, the power range
is significantly less. In a local loop application the KNET
would typically require several remote power boosters. However,
the most serious drawback to the local loop application would be
the high cross talk between KNET cables in the same multipaired
distribution cable. Crosstalk isolation in such cables is
adequate over several miles for audio frequencies but becomes a
serious problem for the higher frequency KNET voice channels and
is further compounded by the twisted pair loss at those
frequencies. If need be, these problems might be cost
effectively overcome by using heavier guage, lower loss, shielded
twisted pair cabling. Of course, none of these things should be
problematic in the business-home office environment where the KNET technology
provides a more efficient telephone distribution system than the
standard local loop technology. It does demonstrate that
technology appropriate for one environment is not necessarily
optimum for a related environment.
1.5 General KNET system theory of operation
1.5.1 Overview
The KNET multiplexes power, control, and many voice channels on a
single twisted pair bus by frequency domain multiplexing (FDM).
The frequency bands are 0-200 Hz for power, 1-14 kHz for baseband
communication control link, and 49-427 kHz for voice. KNET
devices receive power and control over the KNET. They have
frequency agile receiver-transmitters which are used to perform
audio switching and communication between devices over the voice
channels. The KNET controller assigns frequencies and sets audio
characteristics for the device. In addition, it supports
terminal functions.
1.5.2 KNET devices
1.5.2.1 KNET Phone
The KNET Phone basically contains 1) a special power buffering
circuit in conjunction with 2) a power supply, 3) a baseband
control (or communications) link modem, 4) a controlling
microprocessor, 5) keyboard and display components, 6) a
receiver-transmitter and 7) a frequency determining programable
synthesizer, 8) programable audio circuitry, and 9) programable
tone generators and attenuators.
The microprocessor (8049) is the local intelligence for the
phone. The processor performs two principle functions, 1)
communication link protocol handling and 2) input/output command
processing.
The KNET Phone's microprocessor transmits and receives commands
to and from the controller via the baseband control link modem.
In order to reduce communication and system complexity (and
consequential cost), the KNET devices only talk to the controller
and never to each other. (They do sense when another device is
transmitting but ignore the content.) KNET devices also use a
complete primitive command set (much like a microprocessor) which
allows building higher level phone/terminal functions by
stringing commands together.
In response to commands from the KNET Controller, the
microprocessor sets the receiver-transmitter to an assigned
channel. (The controller will send commands to another KNET
device to set it to mating frequencies so as to form a voice path
connection between devices on the KNET.) It sets the audio
circuitry to the appropriate state; such as, turning the speaker
on, setting volume or ring cadence, etc. It scans and debounces
the keyboard, sending the struck key number to the controller.
It drives the display by passing on the received ASCII data
directly to the smart alpha-numeric display. It also handles
various housekeeping and diagnostic status messages.
The KNET Phone's power supply converts the more efficiently
transmitted 50 volt line voltage to the 5 and 12 volts required
by the digital and analog circuitry. Special power buffer
circuitry provide noise, load, and protective isolation from the
KNET.
1.5.2.2 CO Trunk Interface Card
The CO Trunk Interface Card provides audio connection path and
line interfaces to 3 CO (Central Office) trunks. It contains
much of the circuitry (some duplicated for each trunk) and
software of the KNET Phone. It is principly different in that it
doesn't have a keyboard and alpha-numeric display but it does
have additional audio circuitry for DTMF generation and decoding,
line isolation, line supervision, and conferencing circuitry.
The CO Trunk Interface housed in the CO BOX, along with the KNET
power supply, and common equipment card.
1.5.2.3 VCIF (KNET Controller InterFace)
There is only one VCIF (KNET Controller InterFace) per KNET
system. (There can be a total of 256 KNET Phones, CO Trunks, or
other devices.) The VCIF has three functions. These are 1)
communication protocol conversion and 2) buffering KNET messages
to the KNET Controller, and 3) and watchdogging the KNET
Controller.
The VCIF has a microprocessor which in conjunction with a
baseband modem and support circuitry acts as the bridge or
gateway between the KNET and the KNET Controller. This frees the
controller from immediately servicing the VCIF when it receives a
message. The VCIF handles much of the low level KNET
communication protocol which requires rapid time critical
response such as acknowledge response, retransmissions, error
checking etc. In addition, the VCIF provides various KNET status
functions to the Controller. To improve controller reliability
from line power fluctuations or software bugs, a VCIF watchdog
program monitors the KNET Controller and forces a reset on a
system crash. (The VCIF and all KNET devices have their own
hardware watchdog monitor which resets it if its own software
crashes.) To aid in system troubleshooting the VCIF provides
diagnostic indicators to help determine which box or cable has a
problem.
1.5.2.4 KNET Controller
The KNET Controller is the central intelligence for the KNET
system. Typically it is a commodity IBM compatible personal
computer. Unlike the KNET devices whose simple control programs
are small assembly language programs optimized for speed and
limited memory, the KNET Controller's much larger program is
written in "C", a high level language. The use of such a high
level language facilitates program development on complex
programs required for telephone controllers and allows program
portability to other types of computers. By concentrating system
intelligence and all adjustable parameters in a single powerful
computer and by using phones with a complete primitive command
set, software changes are easier, and high virtual phone
intelligence is possible at a much lower system cost.
1.5.3 KNET Bus
1.5.3.1 Power distribution
Standard phone systems typically send power for the phone set
down the same twisted pair that the audio path uses. The power
is usually 50 volts current limited to 60 mA. The phone only
draws power when it goes off hook. Then, it only uses a fraction
of this power because it clamps the line to 2-8 volts. The bulk
of the power is wasted in the line or the current limited source.
KNET power is 50-60 volts, bus distributed from an electronically
protected voltage source which drives efficient switching power
supplies in all KNET devices. KNET Phones although consuming
more power (KNET Phones consume 2.5 watts) than standard phones,
typically load the system power supply less. KNET power is
sourced and picked off the KNET through electronic choke buffer
circuits which act to prevent the power circuits from loading the
control or voice channels. Because of the slow response of these
circuits (less than 200 Hz), KNET devices with sudden load
changes need to energy buffer with large electrolytic capacitors.
These same buffer circuits also act to prevent noise from the
KNET device switching supply from getting into the control or
voice channels. All KNET devices draw power from the net through
full wave rectifier bridges. This makes KNET wiring polarity
insensitive and ensures correct control link data polarity.
1.5.3.2 Control link
Unlike standard phone systems which communicate to and from the
phone with voice band tones (touch tone, dial tone, busy, ring
back, re-order, etc.), the KNET System uses a dedicated digital
channel (control link) to KNET Phones, CO Trunk Interfaces, VCIF,
and all other devices connected to the KNET. This control link
has a much higher signaling rate and is continuously usable, not
just during connection setup. This control link passes control
information between the KNET controller via the VCIF and the KNET
devices (phones and trunk interfaces). All control messages are
strictly to and from the controller, the system intelligence.
KNET devices are like "dumb" terminals and take all commands from
the controller. The only independent actions KNET devices take
are for maintaining control link communications integrity and
allowing for graceful call disposition following link or
controller failure.
The communications control link is a 9.6 (actually 9.579) kbits/s
CSMA (Carrier Sense Multiple Access) bus topology LAN (Local Area
Network).
The communications control link uses an alternate phased impulse
coding technique which results a significantly simpler modem
circuit than other modulation/coding techniques without
compromising system performance. More robust coding techniques
were found to be unwarranted in the KNET environment which
suffers only limited transmission impairments of phase shift,
group delay, frequency response, attenuation, noise, and
distortion. The low speed of 9.6 kbits/s makes for termination
and bridged tap insensitivity. Any reflections caused by such
transmission line impedance changes die out well under a bit time
and thus do not cause bit errors. Analysis suggests than any
line impairment that would significantly disrupt control
communication would also disrupt a standard telephone signal
using the same line.
The KNET communications protocol is simple, efficient, and
robust. It uses a standard bit rate and asynchronous byte format
(one start bit and two stop) to allow the use of standard
communications components and testers. Even though synchronous
coding can provide improved noise margins, its extra complexity
was not warranted because of the high average signal to noise
ratio of the channel. Also, because of the short length of KNET
data blocks, the normal coding efficiency improvement provided by
synchronous protocols would not have been available. Data blocks
need to be longer than 5 bytes to gain improvement from
synchronous protocols by eliminating start and stop bits in
exchange for adding one or more synch bytes.
All correctly received (no parity, framing, time-out, or check
sum errors) transmissions are acknowledged. Failure to receive
acknowledgements causes retransmission. Packets have low
overhead and recover rapidly from collisions and burst noise so
that system response is fast with many users accessing the
channel.
1.5.3.3 Voice channel multiplexing and switching
The voice channels are multiplexed using radio type Frequency
Domain Multiplexing (FDM). Audio is encoded on the 28 voice
simplex channels using companded amplitude modulation. Amplitude
modulation (AM) is used over frequency modulation (FM) because it
is more efficient in the twisted pair radio environment.
Wideband FM only achieves quieting improvement over AM at the
expense of additional bandwith. The twisted pair does not have
sufficient bandwith without paying a range and termination
handicap to justify wideband FM. Narrow band FM is used in the
VHF and UHF mobile environments where the frequency of multipath
fading is high. At these fading frequencies AM's automatic gain
control (AGC) can't keep up while FM limiters are more effective
at suppressing the amplitude modulation of rapid multipath
fading. Because the KNET twisted pair is not a rapidly
changing multipath environment, expensive (requiring linear phase
delay filters), narrow band FM is unnecessary. And, by using
audio companding techniques, the desirable signal to noise ratios
of wideband FM is possible even at maximum range.
The voice channels are 14 kHz apart. Adjacent channels (one for
each direction) are paired to form a full duplex channel.
Channel bandwidth is 7 kHz and passes both sidebands of the audio
signal from 200 to 5000 Hz. A doubling of channels is possible
by using single sideband or suppressed sideband techniques but
with a significant penalty in cost (due to more expensive
filtering).
The voice channels were kept below 500 kHz to avoid potential
interference from AM Broadcast stations. Above 500 kHz it
becomes harder to suppress pickup due to two factors: 1) The
twisted pair balance deteriorates significantly. This determines
the coupling between common (longitudinal) mode (the dominant
antenna pickup mode) and differential signals (KNET is
differential). 2) Common mode pickup of these potentially
interfering signals also increases significantly because wiring
runs approach 1/2 wave dimensions. There are radio sources
(usually navigation signals) below 500 kHz which might interfere
on a channel. This is a low probability event requiring long
(more than a 1000 feet) KNET runs and close proximity to the
station (less than 2-3 miles) to be noticeably audible (although
not necessarily disruptive). If such an interferer is disruptive,
the associated channel can be blocked from the system with only a
small loss of capacity, or it can be given a "last to be used"
priority.
The main limitation to range is the loss associated with
increasing frequency (about 35 dB at 435 kHz down 10,000 ft. of
22 gauge). Beyond about 40 dB of dynamic range spurious signals
become difficult and expensive to overcome. Typically most
spurious responses are 70 dB below the level of the transmitter
loaded into a long line. This means that with worst case loading
on the top-most channel, interfering spurs if present will be 35
dB below the signal. Because of the channel frequency generation
scheme virtually all spurs (including those from the switching
supply) fall between channels where filtering and companding make
a 35 dB down spur minimally audible.
The KNET safely complies with FCC part 15 radiation limits. All
signals are below 500 kHz which gets around the tightest
specified band from 500 kHz to 30 MHz. All conducted common mode
signals are at least 30 dB below maximum FCC specifications. All
digital circuitry is moderate speed CMOS, which along with short
bus runs accessed at low duty cycle ensures low RFI.
on. Amplitude
modulation (AM) is used over frequency modulation А ╪ w s 9 o Г k j g Ь c Ё _ 0 [ ╥) W * S ╝У O юУ K юУ >═ w s 9 o Г k j g Ь c Ё _ 0 [ ╥) W * S ╝У O юУ K А И К Т Ф ╔ ╦ # , \ b d Ф ╟ · № M ▒ ║ ╦ ╘ ╘ ▐ ч & { Н Ц Я и ▒ ║ ├ ╠ ╒ є " + O X t t К │ ь = A Ъ ф Q s u ~ И ┴ ├ ╬ ╨ H H ╢ є 2 Г ╣ Є & a Ь ╔ · +
i
░
ц
L ╬ ╬ ╨ ╪ ┌ ▄ ▐ р ч 0 2 4 V X и ы 5
}
└
F Й Й ╠ V Щ ы э я ` А ╕ х
7 h j У Х ╪ ╪ ^ б ф ' j н ─ ╞ ╚ ў ∙ < ┬ H Л ╬ T Ч ┌ ` г ц ) 8 : < > ~ А ├ ш ъ - X Z Z Э р # f й ь
R Х к м о ё 4 w ║ ¤ @ Г Г ╞ L П ─ ╞ L П ╥ X Ы ▐ !! [! ]! а! у! &" &" i" м" я" # # L# П# ╥# $ X$ Ы$ ▐$ !% d% б% ф% '&