sync

Published on January 2017 | Categories: Documents | Downloads: 60 | Comments: 0 | Views: 365
of 21
Download PDF   Embed   Report

Comments

Content

Synchronization in networks
Content:
• Background / Starting point • Definitions • Levels of synchronization • Digital transmission line • Phase-locked loops for bitrate recovery • Elastic store • Interconnection of plesiochronous network elements • Multiplexing of plesiochronous signals • Switch synchronization • Synchronization of constant length envelopes / ATM • Synchronization of radio systems / GSM / WLAN / Satellite • Scrambling and interleaving
1

Background / Starting point
• Starting point for network synchronization is a circuit switched system, e.g. the «plain old telephone system» (POTS). • The whole network is assumed to be digital, including the subscriber connections to the network. • Starting point of discussion: The digital telephone network, i.e. Streams of 64 kbit/second from a coder via the network to a decoder. • All observations are true for all circuit switched digitale telecommunication networks.
2

1

The simplest possible scenario:

User/network Interface (UNI)

User/network Interface (UNI)

Codec

Codec

USER

Customer network part

Customer network part

USER/SERVICE PROVIDER

• Coder og decoder must be synchronized to not loose speech samples or get pauses in playout of voice at receiver.

- A challenge since the signal passes through many links (multiplexing systems) and switches between coder and decoder.

3

Definitions:
• (All of these are defined in a telecommunication context – the same terms may be used with different meaning in other contexts). Synchronous:

• ITU-T: ”The essential characteristic of time-scales or signals
such that their corresponding significant instants occur at precisely the same average rate (Note: The timing relationship
between corresponding significant instants usually varies between specified limits)”.

• A signal is only synchronous relative to another signal, i.e. this is not a characteristic of one signal. • There is room for very small differences for corresponding pairs of significant instants (e.g. arrival of bits), but not for the average over a long period of time.
4

2

Definitions (2)
Asynchronous (or non-synchronous): such that their corresponding significant instants do not necessarily occur at precisely the same average rate”. • A signal is only asynchronous relative to another signal, i.e. this is not a characteristic of one signal. • There may be large deviations between both corresponding significant instants (e.g. arrival of bits) and between averages over a long time.

• ITU-T: ”The essential characteristic of time-scales or signals

5

Definitions (3)
Plesiochronous («Almost synchronous»):

• ITU-T: ”The essential characteristic of time-scales or signals

such that their corresponding significant instants occur at nominally the same rate, any variation in rate being constrained within specified limits. (Note 1: Two signals having the same nominal
digital rate, but not stemming from the same clock or homochronous clocks, are usually plesiochronous. Note 2: There is no limit to the time relationship between significant instants)”.

• A signal is only plesiochronous relative to another signal, i.e. this is not a characteristic of one signal. • The signals are drifting relative to each other. • Often the best we can achieve in a telecommunication network!

6

3

Definitions (4)
Isochronous: such that the time intervals between consecutive significant instants either have the same duration or durations that are integral multiples of the shortest duration (Note: In practice, variations in the time are constrained within specified limits)”. • A signal is isochronous by itself, i.e. An inherent characteristic of one signal. Isochonous signals are often synchronous or plesiochonous relative to other signals in a network. • The static multiplexing structures are examples of isochronous signals. • Any digital signal with a constant rate is isochronous at the bit level. (NB! But not necessarily at higher abstraction levels).

• ITU-T: ”The essential characteristic of a time-scale or a signal

7

Definitions (5)
Anisochronous (rarely used): such that the time intervals between consecutive significant instants do not necessarily have the same duration or durations that are integral multiples of the shortest duration”. • A signal is anisochronous by itself, i.e. an inherent characteristic of one signal. • An anisochronous signal can not be synchronous or plesiochronous with any other signal. • Example: Start/stopp-signals where the distance between to characters is random.

• ITU-T: ”The essential characteristic of a time-scale or a signal

8

4

Synchronization on many different levels in a network
• Bit level: All signals are isochronous.
Bit period (rate): Actual bit pattern:
time

• All «events» (i.e. bit level changes above) at time instants that are multiples of the bit period (in this case). • For certain types of actual coding on a channel, the period may be different, e.g. twice the bit period for Manchester coding:

9

Synchronization on many different levels in a network (2)
• Word level: (n x 8) bits, n=1 to 8 typically. - May or may not be isochronous, depending on network. - Some real-time services (speech and video) needs either synchronous transport (e.g. a circuit switched network) or reestablishment of an isochonous signal before playout at receiver. • Example for real-time speech/video streaming:
Synchonous operation
Circuit switched network:
Sender
Isochronous at word level Anisochronous at word level

*
Receiver Constant delays Clock/timing info

Packet switched network:
Sender Variable delays (queues) Re-timer Receiver

* Note: Adjustments at bit level may still be necessary, e.g. use of elastic store ()

10

5

Excercise:
• What about the effects from loss of information: - For a real-time speech conversation? - For a real-time video conversation? - For streaming video? - For transport of program code? - For backup of data, e.g. a database?

11

Synchronization on many different levels in a network (3) • Effects from loss of information: - for real-time services: a lost word from a speech or video coder must be replaced by something before playout, e.g. the previous word (often ok because of high autocorrelation in these streams) - for streaming services: same as above, but Forward Error Correction (FEC) could be used to increase quality. (Since buffering at receiver to handle delay variation - for general data services: retransmission (or FEC) necessary to recreate perfectly correct information at receiver.
12

gives time also for additional processing).

6

Synchronization on many different levels in a network (4)
Two alternative ways to organize information into frames for transport: • 1: Envelopes of constant length and constant rate (e.g. ATM):
Fixed period between headers, i.e. expected synchonization points

Information (any fixed length)

Header (length from a few bits to many bytes)

• 2: Envelopes of variable length: - Unique flag used to detect start of information unit, e.g. 01111110. - Transparency stuffing to avoid «simulation» of flag in information part. HDLC («High-level Data Link Control»): extra zeros are inserted and removed dynamically, after five «1»s.
Original:
13

011111101011111110000010100011111100101101111110

With stuffing: 01111110101111101100000101000111110100101101111110

Synchronization on many different levels in a network (5)
• Above the focus is synchronization between one sender and one receiver (at bit, word and envelope levels). • In a synchronous network we need to synchronize between more than two parties: i.e. multiparty synchronization:

Sender

Receiver

All of these should ideally be synchronized to each other

• Network elements must either have the same (global) clock or have mechanisms to handle the fact that they are plesiochronous with regard to each other.
14

7

Synchronization on many different levels in a network (6)

User/network Interface (UNI)

User/network Interface (UNI)

Codec

Codec

USER

Customer network part

Customer network part

USER/SERVICE PROVIDER

• Information units are stored only briefly (if needed) for switching inside the network. Storage time should be very close to constant, to maintain isochronism of signal. • Synchronization challenges (overview, return to each ): a) User equipment: Can usually adjust to the clock of its local switch. b) Switches: May be plesiochronous relative to each other. c) Multiplex systems: Solved by using bitrate recovery.
15

Digital transmission line
• Analog signals are transmitted, but denoted a digital line when representing digital values. • Isochronous signals.
Original digital signal Recreated digital signal

Coder

Transmission line
«Clipper»

Decoder

Bitrate recoverer

Constant bitlengths

Variable bitlengths

Clock to generate flanks (borders between bits)

Multiple signals give an «eyecurve» and show variations in borders between bits

Clock to sample bits

16

8

Bitrate recoverer realized by using a Phase-locked loop
• ”Voltage Controlled Oscillator” (VCO) with adjustable frequency which is «locked» to the received signal.
Here: • Phase detector = Flank triggered flip-flop. • Positive flank on line => flip-flop to HIGH output. • Positive flank from VCO => flip-flop is clocked. • Manchester code: 1=high+low, 0=low+high

Line signal

Control voltage

Line signal

Phase detector

Low pass filter

17

Instability of the clock signal
• Oscillators (clock signals) usually have some instability. Two main types: 1: Wandering: Slow variations in rate. 2: Jitter: Quick/rapid variations in rate. • Possible sources of instability: a) Noise and interference: e.g. the control voltage to the VCO changes due to noise in the line signal => jitter for the flanks. May be counteracted by using a larger time constant in the low pass filter. However: a too large time constant => Impossible for the VCO to adjust to the received signal. => Compromise: Must allow some variation in the VCO frequency, but should be correct with regard to the mean of the received signal.
18

9

Instability of the clock signal (2)
• Possible sources of instability (cont.): b) Changes in length of link/connection and speed of signal. The length of a physical metal- or glass/fiber-link changes with temperature. Increase in length => room for more bits on the link. While the change is occuring more bits are sent into the link than exits the link. Normally small effect, but may be a source of instability if rapid change. Even the speed of the signal itself is influenced a bit by temperature. (And by humidity for radio signals). Usually ignorable. c) Dopplershift: (For radio signals). Movement gives change in length of transmission path => change in bitrate denoted doppler shift. Can be large enough that it needs to be taken into account, e.g. in communication with airplanes.
19

Elastic store
• • • • Example: Digital subscriber line. User must adjust to the bitrate of its local switch. Switch can not adjust its clock to the bitrate of the user. Signal from user must be adjusted to the internal clock of the switch before switching can be done => Elastic store.
Receiver
Phase align.

User equipment («Slave»)

Bitrate recoverer

Transmission line

Sender
Phase align.

Recovered rate

Recovered rate

Switch rate/clock Elastic store

Switch

Phase align.

Transmission line

Bitrate recoverer

Sender

Receiver Line equipment

(«Master»)

20

10

Elastic store (2)
Realisation of elastic store: • The most straightforward is to use a shift-register. • Shown with 8 bits, but other lengths are sometimes better. () • Starts in middle position when initialized, to cater for wandering in both directions.
Tin (rate to write in)

Databit in Count up Init to 4 3 bit counter

8 to 1 multiplexer
Databit out

Count down

Reset

Tout (rate to read out)

21

Elastic store (3)
• On long transmission lines a signal needs regeneration. • For example: 2.048 Mbit/s multiplex signal transmitted over a telephone quality copper pair: Retransmission is needed every 2000 meters. (For fiber transmission: ca. every 50 – 80 km). • Phase-locked loop is used in every regenerator: => However, these are not perfect, so jitter may add up!
Regenerator number Jitter free signal

22

11

Elastic store (4)
• Must sometimes use a regenerator with a larger stability (i.e. averaging over a longer time period) to remove sum of jitter. • Works by averaging the phase deviation over multiple bit intervals. • Can be realized with an elastic store and one extra VCO. The number of bits in the elastic store is now used as control signal (via a D/A converter) to this VCO.

Grade of filling

Low pass filter
Rate/clock

Line signal in

Bitrate recoverer

Rate/clock In Infobit Out

Line signal out
Infobit
Phase align.

Elastic store

23

Interconnection of plesiochronous network elements
• Some network elements will always be plesiochronous, e.g. switches in different countries with different clocks. • Interconnection is necessary, but care must be taken to minimize problems. • Clocks will wander relative to each other because they are plesiochronous => Information loss is unavoidable!
Clock 1 Elastic store Clock 2

Switch

Transmission line

Switch

Transmission line
24

Elastic store

12

Interconnection of plesiochronous network elements (2)
• The elastic store handles short term variations in bitrate (jitter). • Long term variations (wandering or «drift»): Rate in > Rate out => Store overflows! • Store with 8 bit positions gives simultaneous loss of 8 bits (= 1 word, usually). Better than losing 1 bit at a time 8 times as often! «Slip» is often used to denote loss of 1 bit. • A slip => Loss of synchronization at word and higher levels and extra loss of information while getting back into a synchronous state. • This extra loss is avoided if one or more whole words are lost, i.e. synch at word level is then maintained (and possibly higher levels). Also works in opposite direction, i.e. better to repeat a full word twice, than only 1 bit twice. • «Controlled slip» is used to denote this.
25

Interconnection of plesiochronous network elements (3)
Realization 1: • Length of elastic store = one word. • Assume word length 8 bit og Tin > Tout => Sometimes two Tin-pulses between two Tout-pulses.
Tin (rate to write in)

Databit in Count up Init to 4 3 bit counter

8 to 1 multiplexer
Databit out

Count down

Reset

Tout (rate to read out)

• Change in read-out position: ....33...3344...44555....5566... ..66777 (loss of word) 000...00111...112... etc.. • Would be OK, we loose a whole word. • But in reality, because of jitter: ...33...33434.. ..44555...5556566...66676....6777...77 (loss of word) 0 (Reads bits from previous frame:) 77000...etc. • I.e. bits from two different words are mixed together. (But synchronism at word level is kept).
26

13

Interconnection of plesiochronous network elements (4)
• Better realization 2: Length of elastic store = two words. • At both underflow or overflow the read-out counter is reset to the middle position of the elastic store. • I.e. we use an elastic store with 16 bits, but word length is still 8 bits. • Now: change in read-out position : ...15 15 (loss of word) 8 7 7 8 8 ... etc. (position 7 contains next bit in same word, i.e. OK) • Thus, no mix of information from two different words any more. • Normally only half the elastic store will be in use to control the slips since clocks have the same wandering relative to each other over a long period of time.

27

Multiplexing of plesiochronous signals
• Relatively easy to realize. • Introduce extra bit positions to cater for (small!) differences in bitrate between the signals. Is called bitrate adjustment. • Leads to increase in bitrate. • Adjustment bit: Potential bit which has to be transmitted in addition. • Control bit: A flag indicating if an adjustment bit is in use or not, in a given frame. (More important that the control bit is correct than the actual value of the adjustment bit!)
Channel 1 1 frame

Channel 2 Empty bit Adjustment bit
28

Control bit

14

Excercise
• See any problems with the implementation on the previous foil? • What happens if the Adjustment bit is wrong? • What happens if the Control bit is wrong?

29

Multiplexing of plesiochronous signals (2)
• European 2. order multiplexsystem, i.e. a combination of four european 1. order signals (each transporting 2,048 Mbit/s):
212 bit subframe Frame synchronization 200 channel bits

One adjustmentbit pr. channel. Three controlbits pr. channel, distributed across the frame. Majority decision between controlbits.

208 channel bits

208 channel bits

204 channel bits

4 adjustment bits Frame = 848 bits, of which 820 ordinary channel bits + 4 adjustment bits

30

15

Multiplexing of plesiochronous signals (3)
• Each channel exiting the demultiplexer is not isochronous. • Reasons: pauses in bit stream due to frame synchronization, control bits and «empty» bits (i.e. unused adjustment bits).
Channel 1 Elastic store

Rate/clock

Multiplexed signal

Rate/clock

Elastic store

• Isochronism must be Rate/clock restored by use of an elastic store and Channel 2 phase-locked loop pr. channel.
Rate/clock

31

Synchronisation of switches
• Within the domain of one operator: - Most used solution: One ultra-stable master clock (Drift: 1 second in 10 000 years) and local good quality clocks synchronized to this in a master-slave fashion. • An alternative solution: Phase-locked loops used between all master clocks. Would be more reliable, but difficult to make stable. • Communication between switches belonging to different operators is normally between plesiochronous systems.
32

(Shown earlier).

16

Synchronization of constant length envelopes
• Fixed length information units («frame») and (normally) fixed synchronization words («flag»). • Acquisition, i.e. getting into synchronism: • a) Scan for & find
No flag





candidate flag. b) Check if flag also at pos. + one frame. - If YES: repeat N times and then go to c). - If NO (in any of the N above): go to a). In Synch: c) Check flag in every frame. If NO flag M times, assume lost synch. and go to a).

33

Synchronization of constant length envelopes - ATM
• Asynchronous Transfer Mode (ATM) is isochronous at cell level, but used for asynchronous communication: • 1 byte Header Error Control (HEC) in ATM header is used for acquiring and keeping cell stream in synch. • HEC is found from 4 other bytes in header by use of shift register 1 + x + x2 + x8 : • Procedure is same as previously, but «flag» = HEC value changes for each cell. • Match between bytes 14 and HEC value (byte 5) => SYNCHRONISM.

34

17

Synchronization of radio systems
• In general for TDMA systems:
Carrier synch. Clock synch. Unique word Payload (burst of information - variable length/duration)

Receiver:

• Every burst must be synchronized independently
« »

• Unique word correlator can e.g. be a shift register with same length as UW. • With +1 and -1 denoting the two possible binary states: «Correlation» is (UWbit*Candbit)
35

«

»

Or: -1

1

-1

3?

Synchronization of radio systems - GSM
• Different frequency bands to and from BS. • Only concerned with synchronization of bursts. • Frequency bands are organized as TDMA timeslots , difference of 3 bursts between sending and receiving for a MS («Time division duplex»). • «Timing advance procedure»: must take into account the distance of each MS to arrive at correct time(-slot) at BS.
Mobile Stations (MS) GSM Base Station (BS)

• => Largest GSM cell is 35 km.
• (Distance from BS is also used to adjust power levels in communication between MS - BS).

36

18

Synchronization of radio systems – GSM (2)
• Normal «Timing advance procedure» used to adjust the timing for moving MS’ shown in (a) below. • BS keeps track of changes in how well a received burst from a MS hits its timeslot and instructs it to adjust, ~ 1 bit length at a time. • MS is tuned to frame of BS when not in a call. • (b) shows sending of random access burst to initiate call. • BS estimates distance rel. to frame start.
37

Synchronization in Wireless LAN
• No master frame synchronization. • Terminals are plesiochronous, with local clock not synched to any master clock. • IEEE 802.11 frame (SFH-CDMA):
Synchronization
80 bits

Start frame delimeter
16 bits

PLCP header
32 bits

Payload
variable

• Synchronization: 0101010... Used to synch both carrier frequency and digital clock of receiver to the sender. • Start frame delimeter is Unique Word (UW). • Physical Layer Convergence Protocol (PLCP) contains length of payload and data rate (1 Mbps or 2 Mbps). • Payload contains the Media Access Control (MAC) layer.
38

19

Synchronization of satellite systems
• Geostationary satellite system: • Signal from one specific earth station used as reference. • Other stations adjust positions of own bursts relative to this by using PLLs with large loop delays • Initial synch more difficult
39

Synchronization of satellite systems (2)
• One method: Spread sprectrum with large spreading factor. • Initial synch signal ~ 1 Mbps Unique Word. • Using CDMA and 100 chips per bit => 100 Mchips per second. • Can be sent at level 20dB below ordinary signal. • New earth station estimates sending time, attempts to send and adjusts based on result. May be repeated. • Fine adjustment by sending a short (normal) burst in middle of allocated burst position.
40

20

Synchronization of satellite systems (3) • Dynamic burst length allocation:

41

Scrambling and interleaving
• Used to avoid problems created (e.g. for phase-locked loops) from long sequences of only 1’s or only 0’s in bit streams. • Data («duty») signal is combined with a pseudorandom signal. • Pseudorandom signal can be generated by a shift register. • Sender and receiver must have identical and synchronized scramblers. Example: Scrambler used in SDH (1 + x6 + x7).

• Alternative: Block interleaving: Use buffer, write in data row by row, but read out column by column. => Delay.
42

21

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close