Note: Descriptions are shown in the official language in which they were submitted.
is:~~..~e~~~.
This invention relates to an apparatus for
transmitting data over a digital transmission system having
discrete data channels with a predetermined base bit-rate
lower than the bit-rate of the data to be transmitted. More
particularly, the invention provides a method and apparatus
for allocating a number of base rate channels to a virtual
channel capable of carrying signals having a bit-rate higher
than the base rate.
One data transmission system to which the invention
is especially applicable is the Integrated Services Digital
Network (ISDN) for which the standards have been defined by
the International Telegraph and Telephone Consultative
Committee (CCITT). ISDN makes use of a 2.048 megabyte per
second primary rate TDM channel onto which are time-division-
multiplexed 32 sub-channels (DSO). Each sub-channel has a
base bit-rate of 64 Kb per second. The primary rate TDM
channel carries 8000 frames per second, each divided into 32
time slots carrying one byte (8 bits) of data from each sub-
channel. The first time slot in each frame is used to
identify the start of the frame, and another time slot, which
constitutes the data channel, carries routing instructions.
The remaining 30 slots are available for carrying data
normally as 30 discrete channels.
The primary rate TDM channel might, for example, be
used to connect a private branch exchange (PBX) to the public
telephone network. A single primary rate channel will
therefore give the subscriber access to thirty 64 Kbps base
rate channels.. Situations often arise, such as in the
transmission of image and video signals, or high volumes of
computer data, where it is desirable for the subscriber to
transmit the data at.a rate higher than the base rate. For
example, it would be desirable to have the capability of
sending a 128 Kbps bit stream over two parallel 64 Kbps base
rate channels. Unfortunately, because of the switching
requirements and propagation characteristics of the public
- 1 -
2~D~.~3~1.
network a channel which occupies a particular time slot in
any given frame of the transmitted signal does not
necessarily occupy the same time slot at the far end. The
base rate channels are subject to different delays through
the network. As a result, if the channels are merely
reassembled sequentially at the far end, the transmitted data
is scrambled and unusable.
International Patent Application No. WO 85/04300
describes a system wherein prior to data transmission
synchronization signals are transmitted along each of the
base channels to determine the delays applicable to each
channel. A reframer unit then takes into account these
delays to reassemble the transmitted data in the correct
order. The problem with this system is that once the virtual
channel has been established it cannot be changed without
being completely reset. Furthermore if the delays for the
various channels change during transmission, the data becomes
unusable. The system cannot therefore be regarded as
reliable.
U.S. Patent No. 4,805,167 describes a system for
providing a variable data rate aggregate channel. In this
system marker signals are sent on the base rate channels to
specify the order of transmission of the sub-signals to as to
ensure correct reassembly at the far end. One problem with
this system is that it can only be used for packet
transmission since it requires there to be idle time slots in
the data channels to carry the marker signals. It cannot
therefore be used with continuous signals, such as video
signals because there are no slots in which to insert the
marker signals. Also, since the marker signals can only be
inserted when idle time slots are available, it does not
permit.the delay characteristics of the network to be
continually monitored.
- 2 -
~~12361.
An object of the invention is to provide an
improved transmission system which does not depend on holes
being available in the data stream and which permits
continuous monitoring of the channel delays.
According to a first aspect of the invention there
is provided a digital data transmission system normally
providing discrete data channels having a predetermined bit-
rate, said channels being subject to different propagation
delay characteristics through the system, a method of
transmitting. signals having a bit-rate higher than said
predetermined bit-rate rate, comprising allocating a group of
n+1 said predetermined bit-rate data channels as a virtual
channel for the transmission of said high bit-rate signals,
one of said n+1 channels being designated as an overhead
channel and the remaining n channels being designated as data
channels, dividing said high bit-rate signals into n sub-
signals having a bit-rate equal to or less than said
predetermined bit-rate, transmitting said n sub-signals over
said n data channels, transmitting delay calibration signals
at intervals over said overhead and data channels, the delay
calibration signals being transmitted in said data channels
in slots normally allocated for data, said data slots being
transmitted over said overhead channel while said calibration
signals are transmitted in their place, and using said delay
calibration signals at the far end to reassemble said n
subsignals into said original high bit-rate data signal.
Tn a preferred embodiment the delay calibration
signals are transmitted successively over the respective
channels in a rotational pattern. For example, in the case
of a virtual channel consisting of four data channels arid one
overhead channel, it would take successive frames to complete
one rotation. In a first frame the delay calibration signals
are transmitted in the time slot corresponding to the
overhead channel. In a second frame the delay calibration
byte is transmitted in the time slot corresponding to the
- 3 -
2os~3s1
first data channel and the data that would normally be sent
in that time slot is instead sent in the time slot in the
overhead channel. Likewise in the third frame, data from the
second channel is swapped into the overhead channel time slot
and the delay calibration signal sent in its place, and so on
until after the delay calibration signals have been sent in
the fourth data channel the next delay calibration signals
are sent in the overhead channel, whereafter the cycle is
repeated.
Each data channel (DSO) may be delayed differently
by the network. The delay calibration signals, hereafter
referred to as delay calibration byte (DCB); form a framing
pattern in each channel which can be extracted at the
receiver. The contents of the DCBs permit the relative delay
to be determined. In the scheme just described every end
byte on the far end channels constitutes the framing pattern.
The rotating calibration may be performed continuously, or
used initially and then discontinued.
To determine the relative delay between channels
the overhead byte sends out a rotation count (LSB). A
rotation starts with a DCB sent out on the overhead channel
and ends when the last data channel in the virtual channel
has sent its DCB. At the receiver this creates the
appearance of every (N + 1)th byte on the channel forming a
framing pattern.
To accommodate 48 Kbps data channels, only the six
most significant data bits of each byte are used, the
remaining bits being set to one to ensure correct ones
density on Switch 56 networks. Six bits provide 64 unique
bytes for the DCBs, but since the count wraps around only
half the rotation count can be resolved. One byte rotation
count allows differential delays of 32 frames to be detected.
Since the DCB would occur every 2N overhead slots the maximum
- 4 -
201361
delay resolution would be 2N*32 frames, which in practice is
insufficient.
To overcome this problem, a second slot is used to
send the most significant bits (MSB) of the rotation count,
which results in a 12 bit delay calibration count. This
permits 2048*2N frames to be resolved. In the worst case,
where the channel size is three (N=3) this results in 12228
frames, which is equal to 1.528 seconds. This is sufficient
for mixed terrestrial satellite networks.
The overhead slots are used for the LSB count, the
MSB count and general purpose overhead information. The LSB
appear every second overhead slot. The slots not used for
LSB alternate sending the MSB and an overhead byte. The MSB
follows even LSB counts and the OHB (overhead byte) slot
follows odd LSB slots.
Another aspect of the invention provides in a digital data
transmission system normally providing discrete data channels
having a predetermined base bit-rate, said channels being
subject to different propagation delay characteristics
2o through the system, an apparatus for transmitting signals
having a bit-rate higher than said predetermined bit-rate
rate, comprising means for allocating a group of n+1 said
predetermined bit-rate data channels as a virtual channel for
the transmission of said high bit-rate signals, one of said
n+1 channels being designated as an overhead channel and the
remaining n channels being designated as data channels, means
for dividing said high bit-rate signals into n sub-signals
having a bit-rate equal to or less than said predetermined
bit-rate, means transmitting said n sub-signals over said n
data channels, means for transmitting delay calibration
signals at intervals' over said overhead and data channels,
the delay calibration signals being transmitted in said data
channels in slots normally allocated for data, said data
slots being transmitted over said overhead channel while said
- 5 -
CA 02012361 2000-11-17
calibration signals are i~ransmitted in their place, and means
at the far end for using said delay calibration signals to
reassemble said n subsignals into said original high bit-rate
data signal.
The invention will now be described in more detail,
by way of example only, with reference to the accompanying
drawings, in which:-
Figure 1 is a :schematic diagram of a virtual
channel operating in accordance with the invention;
Figure lb show=~ the scheme. for sending DCBs in the
successive time slots of a single DSO data channel;
Figure 2 is a block diagram of a data transmission
system incorporation a virtual channel in accordance with the
invention;
Figure 3 is a flow chart illustrating the operation
of a frame state unit;
Figure 4 is a diagram illustrating the consequences
of a base channel slipping within a virtual channel;
Figure 5 is a simplified block diagram of a four-
node transmission network:; and
Figure 6 is a block diagram of two-node network.
The invention will be described with reference to
an ISDN network having a primary rate TDM carrier of 2.048
Mb per second carrying thirty data channels and two system
channels. As discussed above the primary rate channel
carries 8000 frames per second, each base channel (DSO)
having a base rate of 64 :Kb per second.
- 6 -
2012~6~
Figure 1 shows a transmit section T and a receiver
section R. It should be noted that while for convenience the
two sections are designated as the transmitter T and receiver
R respectively, the system operates in the full duplex mode.
The system is symmetrical and either section can serve as
transmitter or receiver.
A 256 super-rate Kbps bit stream 1 is divided into
four parallel bit streams 2, 3, 4, 5 each running at 64 Kbps
(256 = 4 = 64). The four bit streams 2, 3, 4, 5 are applied
to a Virtual Channel Resource (VCR) unit 6 in the transmit
section T which outputs four base rate bit streams 7, 8, g,
10 serving as data channels and an additional channel 11,
which normally serves a data channel, but which in the
virtual channel serves as an overhead channel.
The five base rate channels are transmitted over
five base rate (DSO) channels on the public telephone network
to a Virtual Channel Resource (VCR) unit 12 in the receiver
section R, which includes delay buffers 13 respectively
receiving each of the incoming channels and a a reframer unit
14 for reassembling the incoming bytes in their proper
sequence so as to reconstitute the four original bit streams
as 2', 3~, 4~, 5~. Figure 3 is a flow diagram showing the
operation of the receiver framing unit 14.
The reconstituted bit streams are then recombined
into the original 256 Kbps super-rate bit stream.
The VCR transmit unit 6 inserts delay calibration
bytes (DCB) 15 among the data bytes 16 according to the
scheme shown. In frame 1, the first delay calibration byte
(DCB) is inserted into the time slot corresponding to the
overhead channel. Ln the next frame, the DCB is inserted in
the time slot corresponding to the first data channel 7 and
the contents of this time slot are transmitted in the time
slot corresponding to the overhead channel. In frame 3 the
202361
contents of the time slot corresponding to the second data
channel 8 are transmitted in the overhead channel, and the
DCB is transmitted in this time slot. The scheme is repeated
in the third and fourth data channels 9 and 10, whereafter in
the sixth frame the DCB is transmitted in the overhead
channel time slot, and the pattern is repeated on a
rotational basis.
Figure 1b shows how the bytes constituting the DCBs
(LSBs' MSBs, and OHBs) are sent successively in the DCB slots
of a single DSO channel. The first LSB, representing the
rotation count, is sent in the first DCB slot. There then
follows three regular data slots followed by the MSB in the
next DCB slot. The next DCB slots contain successively the
LSB = 1 byte, the OHB, the LSB = 2 byte and the next MSB.
Figure 2 shows the hardware implementation of the
virtual channel. Data Transmit Unit (DTU) 20 of the transmit
section T receives the incoming super-rate bit stream at 256
Kbps and outputs four parallel 64 Kbps bit streams on ports
0, 1, 2, and 3. These bit streams are connected by switch 21
to the Virtual Channel Resource (VCR) transmit unit 6 in such
a way that the lower circuit number on the data transmission
unit 20 is connected to the lower circuit number at the input
of the VCR transmit unit 6. The VCR transmit unit 6 outputs
five channels (four data channels + one overhead channel) to
switch 22, which is in turn connected to the line termination
transmit unit 23. This is connected via a time division
multiplex line 24 forming part of the public digital
transmission network to the line termination receiver unit 25
of the receiver section R. The receiver unit 25 outputs five
channels (including the overhead channel) to the switch unit
26, which is connected to the VCR receiver unit 12, which
extracts the overhead information and outputs four data
channels to the switch unit 27. The switch 27 connects the
lower incoming circuit number from the virtual channel
resource receiver unit 12 to the outgoing lowest circuit
_ 8 _
21~1236~.
number on the DTU 28, which in turn outputs the original
super-rate 250 Kbps bits stream.
Although the invention has been described in
connection with a 64 Kbs base rate network, in order to
accommodate 48 Kbps data channels, which are used in some
countries such as the United States, only the six most
significant data bits of each byte are used according to the
scheme described above.
A rotation starts with a DCB sent out on the
overhead channel 11 and ends when the last data channel 10 in
the virtual channel has sent its DCB, i.e. at frame 5 in
Figure 1. At the receiver R this creates the appearance of
every Nth channel forming a framing pattern.
The Overhead Bytes (OHB) are described below. As
discussed above, these are sent in the DCBs alternating with
the LSB rotation counts. The slot following rotation count
LSB = 0 is used for rotation count MSB and may not be used
for overhead information.
The Overhead Bytes are six bits to allow for 48
Kbps transfer. Overhead information is sent via a one or two
byte pattern. The OHBs are described below. The unused bits
are set to l to comply with Switch 56 ones density
requirements. The overhead bytes consist of:
~ Status Words
These are used to advise the sender of the status of an
incoming virtual channel.
_ g _
201~36'~.
OOabcdll Channel Status Word
a = virtual channel synch (1=in synch)
b = virtual channel state bit 1
c = virtual channel state bit 1
d = data synch (1=in synch)
State Virtual Channel State
00 In Service
01 Calibrate and leave
l0 Out of Frame
~ Channel ID Bytes
These are used to indicate the channel/virtual channel
numbers of the transmitted virtual channel (VC) to the
receiver.
0lnnnnll nnnnn = 0 following byte is channel #LSB
1 following byte is channel #MSB
2 following byte is VC #LSB
3 following byte is VC ~MSB
4 channel mode = start up mode
5 channel mode = continuous mode
6 channel mode = transparent mode
The following overhead byte codes may only be sent in the
overhead channel (channel number 0):
lOnnnnll nnnn =0 overhead channel message start
1 overhead channel message end
llabcdll ~ abcd => ABCD signalling bits
In a given 100 ms interval each channel must send the OH
bytes listed below one or more times.
- l0 -
~
channel status word
~ channel number
~ virtual channel number
~ channel mode
overhead channels in message mode are exempt from this
requirement.
In order to correct delay at the receiver the
sequence in which the rotation count is sent and the
assignment of channel number to circuit numbers must be known
by the receiver. The channel rotation count is sent first on
the overhead channel and then on the data channels in order
of increasing channel number. The channel numbers are
assigned to the sender data circuits in the following
fashion, which is given by way of example with reference to a
Newbridge 3645 Mainstreet system. The channels must be
numbered in some fixed manner which is the same at each end:
~ in an ST-BUS the earlier timeslots get the lowest numbers
~ in MX streams simultaneous ST-BUS timeslots are ordered by
PE slot number
~ in MX links the timeslots are ordered D1 < D2 < CBI < CB2
This protocol requires that data circuits from the data
source (DTU) 20 to VCR transmit unit 6 be mapped in the same
fashion as the circuits from the VCR receive unit 12 to data
receiver unit 28. Switches 21, 22, 26, 27 bring this about,
and this arrangement ensures that data interface ports are
properly connected.
The circuits between the VCRs need not be
consistently mapped. . The VCR receive unit 12 will determine
the channel number assignments and send the data out in
increasing timeslot order on the data circuits to the data
receiver. This allows independence of the network (56 Kbs
networks are known to jumble group circuits).
-- 11
2012361.
In the continuous calibration mode, discussed
below, the overhead channel (channel number 0) may be used to
send messages from end to end. The messages are sent in the
overhead DCB slots following an overhead channel message
start byte pattern. The overhead channel on byte pattern is
repeated 10 times before the message channel is active. Once
the channel is active a message of not more than 128 bytes
may be sent. The message is terminated by sending overhead
channel message end byte (repeated 10 times). While the
overhead channel is active no other messages may be sent on
channel 0.
The overhead message packets are not defined. Any
suitable packet based protocol can be used to send messages.
They are used for messages for:
~ Performance summary (slips, LSB count errors etc.)
~ Dynamic channel sizing
In order to accommodate a variety of network types
several modes exist. All of these modes assume that the
virtual channel is full duplex. The channel mode is sent in
overhead bytes in all channels. Examination of any one
channel is sufficient for determining the mode. The mode
should be debounced for 4 super-rotations.
1. Continuous Calibration Mode
Data and DCB patterns are inserted at start-up and
continued for the life of the channel. For transport of N
data channels N+1 circuits between VCRs must be allocated.
The overhead channel should be used to continuously send the
following information:
- 12 -
201~30?~.
~ Channel Number (LSB & MSB)
~ Virtual Channel Number (LSB & MSB)
~ Data Channel Status
~ Continuous Mode Channel
2. Start-Ub Calibration Mode Only
This mode exists to allow overhead-free
transmission of super-rate data following initial set-up. In
this case for N data circuits only N inter-VCR circuits are
required. These channels are numbered from 1 to N. This
mode starts with the transmission of the standard rotating
DCB pattern. The overhead DCBs send (continuously):
~ Start-Up Mode Channel
~ Channel Number LSB
~ Channel Number MSB
'15 ~ Far End Channel Data Status
~ Far End Virtual Channel Status
To simplify framing at the far-end all ones are
sent in place of data. Once the receiving VCR has framed and
extracted the necessary information from the overhead bytes,
it sends a data synch = 1 in its Channel Status Word, and
virtual channel synch = 1 and state = In Service in the
Virtual Channel Status Word.
The sending VGR upon receipt of virtual channel
synch sends a transfer to transparent signal in the data
status word of channel 1. This signal consists of 8
repetitions of~the Data Status Word in successive OH bytes
with mode set to Transparent. Once the signal is complete
the sender then shifts to transparent mode.
- 13 -
20123f ~.
3. Transparent Mode
This mode allows for direct super-rate virtual
channel connection without continuous delay calibration.
This mode can be started immediately, or entered after a
delay calibrating start-up sequence. While in the
transparent mode the receiver continues to monitor for a
framing pattern. If framing is detected and maintained for
500 ms on all of the channels forming a virtual channel then
the receiver change modes to correspond to the mode type in
the data status word of the incoming data.
Ideally the maximum reframe time for a single DSO
within a virtual channel should be 500ms, so the maximum time
to re-calibrate an entire virtual channel would be 2.5 secs.
The maximum time to correct a frame slip on a DSO would be
100 ms. For large values of N, these targets may not be
attainable.
Channel Operation
In normal operation each channel sends data status
and virtual channel overhead bytes. During start-up the
channel watches for far end Virtual Channel synch. If synch
is not received within 10 secs, a recovery signal is
broadcast. If VC synch is received within 10 secs the
channel declares In-service and continues monitoring or moves
to the transparent mode.
In the event of a data channel out-of-frame the
receiver sends an out-of-synch signal in the channel status
byte and all other channels within the virtual channel send
Virtual Channel Out-of-Synch (VCOS) in their virtual channel
status.words. If VCOS is detected in the incoming streams
and persists for 5 secs, the sender declares Far-End-Out-Of-
Frame and enters the recovery mode. If the receiver cannot
- 14
2~01:3~~.
frame on the incoming stream within 5 sees, it declares Near-
End-Out-Of-Frame and continues to maintain far end alignment.
Frame slips occur as a result of asynchronism
between network elements. A frame slip will result in either
the duplication of deletion of a frame of information within
the transmission facility. If a single channel within a
virtual channel slips then it will become permanently
misaligned with the other members of the virtual channel as
shown in Fig. 4. In the virtual channel information the data
all starts within one synchronization domain. This leads to
two slip scenarios:
~ phase error slips due to an asynchronous
~ network span
~ continuous slips due to different network synch
~ sources
The phase error slip case results in a maximum
delay adjustment of +/- 1 frame. In Fig 5 node 2 runs
slightly slower than the network. Data from 1 -> 2 arrives
too quickly and the receiver must periodically throw away a
frame of data. At the 2 -> 3 interface data is arriving too
slowly and a frame must be duplicated occasionally to
correct. The net effect on the channel is that it will
periodically slip in one direction and then eventually slip
back. Eaah time the slip occurs the delay of the affected
channel must be changed.
In the continuous slip case, slips occur in one
direction only. The slips on links A & B will be out of
phase. If an attempt is made to compensate by changing delay
on the channel which.slips, then as slips accumulate so will
the delay. Eventually the buffer space for delaying will be
exhausted and it will be necessary to disrupt service to re-
calibrate. One solution, provided a knowledge of network
topology is available, is to add a delay to channel 1 when
- 15 -
2012361.
link 1 slips and remove the delay from channel 1 when link 2
slips, in which case the need to re-calibrate can be avoided.
A frame slip will misalign data on the virtual
channel. The slips should be detected and corrected quickly.
At the receiver the framing pattern will slip, and the framer
must cope with slips without declaring out-of-frame.
The receiver framing operation will now be
described in more detail. At the base-rate channel (DSO)
level the rotating DCB pattern appears as a DCB every N
frames. This allows each DSO to be considered as an
independent channel which must frame on the incoming DCB
pattern. Delay equalization is then performed by reading the
current DCB count and the byte offset from it from each DSO
in a known time-sequence. This allows the relative delay
between the DSO to be calculated and the variable delay
buffers changed. It is permissible to allow the DSOs to
reframe as a consequence of delay adjustment provided that
the performance objectives are met.
The start-up time and the recovery time from
2 0 protection switches depends on the time it takes the receiver
to frame on each DSO in the virtual channel. The frequency
of delay calibration bytes is a function of the virtual
channel size. As the frequency of DCBs decreases reframe
time increases.
The VCR state framer, for which the flow chart is
shown in Fig. 3, handles slips without re-framing to allow
very fast slip response time. In the start state the framer
selects a byte from the data stream and moves to the Load
First DCB state. N data bytes later the second DCB is loaded
on the transition Load Second DCB state. If the two DCBs
form a valid pattern (the second is one more than the first)
then the Reverse Guard state is entered (Note: if the first
DCB was 0 then the following DCB is a MSB and the one after
- 16 -
2~1236~.
that is an overhead byte, in this case a valid pattern cannot
be declared until the next LSB rotation count is examined).
From Reverse Guard the pattern is checked until GUARD-COUNT
correct DCBs have been detected at which point an in-frame is
declared and the Forward Guard state entered. A 2/4 DCB
error causes loss of frame and transition to Load Second DCB.
In addition to looking in the expected timeslots, the forward
guard state examines the timeslot on either side. If a valid
pattern is detected between the previous timeslot DCB and one
on either side of the present DCB then a Watch for Slip state
is entered. If GUARD-COUNT valid DCB bytes are observed in
the +/- 1 timeslot then slip is declared, the timeslot for
examining framing is moved and SLIP is declared. If 2/4 DCB
error in timeslot +/- 1 are encountered in one of the Wait
For Slip states then the Load Second DCB state is entered.
The parameter GUARD-COUNT is a feature of the
framer. It should be adjustable on a per Virtual Channel
Resource level. The ability to adjust GUARD-COUNT on a per
virtual channel basis is desirable, but not required. The
frequency of DCB bytes rotation time) is N x 125 ~CSec. This
time is calculated for various channel sizes in the table
below:
N Rotation Time Super-Rotation Time
8 1 ms 256 ms
16 2 ms 512 ms
32 4 ms 1024 ms
64 8 ms 2048 ms
128 16 ms 4096 ms
256 32 ms 8192 ms
The overhead frequency is 2 x Rotation Time.
The framer slip detection depends on the size of channel and
the framer parameter GUARD-COUNT. Some times for various
values of parameters are indicated below:
- 17 -
2~12361.
G C Slin Response
8 4 10 ms
8 8 18 ms
32 4 20 ms
32 8 36 ms
256 2 192 ms
256 4 320 ms
256 8 576 ms
These numbers assume GUARD-COUNT + 1 rotation count
l0 LSBs are required for declaration of slip. Time to implement
the delay adjustment is not included.
A main advantage of the system described is that
the method is adaptive and will adjust for changes in delay
during traffic transmission. The system also allows a high
bit rate synchronous channel to detect and recover from
changes in network (i.e. frame slips, protection switches)
during the course of the connection and the signalling can be
passed via ABCD bits in the overhead bytes.
- 28 -
