RU2369015C2 - Synchronisation of vodsl for dslam, connected to ethernet only - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: invention relates to POST synchronisation, transferred over DSL, i.e. for VoDSL, DSL, receiving data from a packet transmission network, more specifically ETHERNET. A retransmission unit, such as DSLAM, receives Ethernet packets and retransmits information in them over a digital subscriber line. To retransmit voice data of the received Ethernet packets, a reference clock signal is generated by a synchronising device and transmitted to digital subscriber line modems. The synchronising device comprises an extraction unit for selecting at least one stream of received Ethernet packets, and a unit for generating the clock signal, configured as an adaptive synchronisation unit for generating a reference clock signal in accordance with the arrival time of a packet in the selected stream or streams of packets. The reference clock signal can be transmitted to time referencing units in modems. The extraction unit can analyse reception of Ethernet packets so as to find packet streams from a single user transfer target, each packet in the stream carries information in real time, belonging to a service in real time, and then select one or more streams, which should be used by the adaptive synchronisation generator.

EFFECT: more accurate synchronisation.

14 cl, 19 dwg

Description

FIELD OF THE INVENTION

The invention relates to a method and apparatus for synchronizing POTS (plain old telephone service) carried on DSL (digital subscriber lines), i.e. for VoDSL (Voice over DSL), a DSL that receives data from a packet network, in particular an Ethernet network.

State of the art

The digital PSTN (Public Switched Telephone Network), intended primarily for transmitting voice information between two telephone sets, is a synchronous network, see FIG. 1. Analog-to-digital (A / D) and digital-to-analog (D / A) converters in two linear circuits at the ends of the connection, i.e. in subscriber lines to two telephones, use the “same” 8-kHz reference clock signal, see FIG. 2.

If a packet network using, for example, Ethernet, is used to transmit voice information over a portion of the route between two telephones, see FIG. 3, a common reference clock signal that can be used by A / D and D / A converters is not available. Voice samples are transmitted in packets; used packets are also called frames according to Ethernet. Packet arrival times vary with delay changes in the packet network. Then buffers of random fluctuations at the receiving ends are needed, see FIG. 4 to control delay changes. If clock signals that are not synchronized with each other are used in A / D and D / A converters at the ends of the established connection, random fluctuation buffers will sooner or later overflow or become empty. Voice samples should be deleted or inserted to avoid this, in cases where this correction involves deleting / inserting voice samples, this is usually called a shift. If a shift is inserted during the silence period of a voice broadcast, it will not be detected by the person receiving the voice message.

However, for example, some devices such as conventional telephone modems for transmitting data or facsimile messages are sensitive to shifts and jitter of clock pulses. Consequently, many modems will retrain when a shift occurs. In addition, DTMF signals are destroyed by shift.

POTS traffic i.e. conventional voice information over a physical subscriber line or a communication line also carrying ADSL (asymmetric digital subscriber line) for transmitting digital information is usually transmitted in its own frequency band obtained by dividing the available frequency band using analogue dividing filters. ADSL uses the frequency band from 25 kHz to 1.1 MHz or up to 2.2 MHz for the ADSL2 + standard, and POTS uses the usual POTS frequencies.

In VoDSL, POTS instead sends voice information encoded as digital data in the ADSL payload. Then the POTS codec (PCM codec-decoder) is located on the subscriber side of the ADSL line. The POTS codec needs an 8-kHz reference clock signal, such reference clock signals in this document are simply called “clock signals” or “clock signals”.

The advantages of VoDSL include that if POTS uses the digital range, conventional POTS frequencies for data can be used, and that frequency separation filters are not needed.

To provide a reference clock for decoding voice information at the receiving end of a communication route including a packet transmission network, a freely working generator may be used. However, the expected accuracy of a cheap generator is on the order of 100 ppm, which gives a bias rate of one bias per 104 samples, i.e. every 1.25 seconds. This is not acceptable for a high quality POTS system. IP networks use Network Time Protocol (NTP) to synchronize nodes, and it can be used to restore synchronization, see published international patent applications Nos. WO 00/42728 and 02/39630 for Telefonaktiebolaget LM Ericsson and related patents U.S. Nos. 6,532,274 and 6,711,411, inventor of Stefano Ruffini. The IEEE 1588 standard is a Layer 2 (Ethernet) method for providing a clock signal. It requires Ethernet all the way, i.e. routers cannot be connected in the connection route.

There are various methods for generating a synchronization signal, for example, from received packets, some of which require specially designed packets or non-standard Ethernet equipment. There is a need for methods that have high performance and can use existing packet streams in an Ethernet transmission network.

“Adaptive synchronization” is a generic term for methods that regenerate a sync pulse from packet synchronization. Some of these methods are point-to-point methods that operate transparently through various types of network equipment such as switches and routers. The quality of the regenerated clock is highly dependent on changes in packet delays. For example, in a switch, a packet carrying voice information may wait until a long data packet is transmitted. The time that a packet should wait is proportional to the length of the data packet and the speed, for example, in bit / s, of the communication line. The maximum length of Ethernet packets is 1500 bytes. In a 100 Mbps Ethernet link, this corresponds to 120 μs, while in Gigabit Ethernet it corresponds to 12 μs. On an ADSL link, the speed is much lower, for example, 0.5-8 Mbit / s. Thus, for example, a speed of 0.5 Mbps corresponds to 24 ms for an Ethernet packet of maximum length. Obviously, in this case, delays due to long data packets are too high. This means that the change in packet delay after the DSL link is increasing and that ADSL adaptive synchronization methods are less useful.

ITU-T G.992.1 ADSL describes the transmission of an 8-kHz clock reference signal. In the case of voice over DSL, the reference clock available in DSLAM can be connected to an NTR device (network reference clock) and transmitted to the subscriber. ADSL was originally created for the ATM interface and in this case the 8-kHz reference signal is usually accessible from the ATM network.

In NTR devices, a phase difference is created between 8 kHz synchronization and synchronization of ADSL samples, which is obtained by dividing the transmission time of 2.208 MHz by 276 to obtain an 8-kHz signal, see FIG. 5, and its value is transmitted to the subscriber. In the case where the DSLAM (DSL access multiplexer) used at the end of the ADSL uplink is only connected via Ethernet in which such reference synchronization is not available, this is the case discussed in this document.

SUMMARY OF THE INVENTION

The aim of the invention is to provide a relay device, such as DSLAM, which receives Ethernet frames and creates a reference clock signal for use in a digital subscriber line, in particular, ADSL.

Another objective of the invention is to provide a method of creating a reference clock signal from a received stream / received streams of Ethernet frames for use in a digital subscriber line, in particular, ADSL.

Typically, the method uses the arrival time of the RTP packets to form a reference clock signal.

In the case of voice over DSL, voice information is transmitted digitally in the ADSL bitstream, the reference clock signal is generated using the adaptive synchronization method and communicated to the NTR input of the ADSL modem. Thus, an 8-kHz reference clock signal can be regenerated on the subscriber side. Once, for example, a voice call that initiates packets useful for synchronization is available, the synchronization process can be initiated, and the reference synchronization signal is transmitted to the client or subscriber using NTR.

ADSL DSLAM examines the packet flow going to the ADSL subscriber. As soon as a stream of voice packets or other real-time service that is synchronized with the PSTN, such as video conferencing, is identified, it is used in the synchronization process. Such data streams can be identified, for example, using VLAN tags, source IP addresses, port numbers of their packets or frames, or by identifying their RTP content. DSLAM can also be managed from another system whose packages are to be used. If multiple streams are available, one or more streams may be used. Preferably, the stream with the least delay variability is used.

Typically, a relay device, such as DSLAM, is configured to receive Ethernet packets and relay them or the information contained in them over digital subscriber lines. For relaying voice information contained in received Ethernet packets, a reference clock signal is generated by a synchronization device and provided to modems for digital subscriber lines. The synchronization device includes an extraction unit for selecting at least one stream of received Ethernet packets and a synchronization signal generation unit, configured as an adaptive synchronization unit or generator, for generating a reference synchronization signal according to the moments of arrival of the packet in the selected stream or packet streams. A reference clock signal may be provided to time reference units in modems. The extraction unit can analyze the received Ethernet packets to find the packet flows from one destination to one user, the packets of each stream carry real-time information belonging to the service in real time, and then select one or more streams that should be used by the adaptive synchronization generator .

Brief Description of the Drawings

The invention will now be described by non-limiting embodiments with reference to the accompanying drawings, in which:

FIG. 1 is a diagram for transmitting voice information between two telephone sets over a public switched telephone network.

FIG. 2 is a diagram similar to FIG. 1, also illustrating A / D and D / A converters located in local telephone exchanges at the ends of a connection between two telephone sets.

FIG. 3 is a diagram for transmitting voice information from / to a telephone, where a packet network using, for example, Ethernet, is used to transmit voice information over a portion of a transmission route.

FIG. 4 is a block diagram of a local telephone exchange connected between a packet transmission network and a telephone network and including a local generator, A / D and D / A converters, and a random fluctuation buffer to control the changing delays of arriving packets.

FIG. 5 is a block diagram of an NTR unit for use in DSL modems.

FIG. 6 is a diagram similar to FIG. 3, also illustrating a digital telephone connected through an ADSL communication line to a public switched telephone network.

FIG. 7 is a block diagram of a DSLAM located for transmitting voice information over digital subscriber lines.

FIG. 8 is a protocol diagram or protocol processing diagram.

FIG. 9 is a diagram illustrating delays of packets received from a packet network.

FIG. 10 is a diagram illustrating a direction of delays with respect to a reference clock signal for delays of packets received from a packet transmission network.

FIG. 11 is a block diagram of the circuits required for relaying digital information including voice information over an ADSL link, where the generated clock reference signal is used as a network time reference signal.

FIG. 12 is a block diagram of an adaptive synchronization unit including a PLL and a tunable generator.

FIG. 13 is a diagram illustrating delay meter values and packet arrival times.

FIG. 14 is a block diagram of a minimum value calculating unit using a sliding window algorithm and grouping or blocking received values.

FIG. 15 is a schematic diagram of a PLL filter.

FIG. 16 is a block diagram similar to FIG. 12 but illustrating a generator having a fixed frequency.

FIG. 17 is a block diagram similar to FIG. 11, where the generated reference timing signal is used to control transmission timing.

FIG. 18 is a flowchart of steps performed by an extraction unit.

FIG. 19 is a block diagram of an extraction unit.

DETAILED DESCRIPTION OF THE INVENTION

In FIG. 6 schematically illustrates a system in which voice information is transmitted over an Ethernet and ADSL network. The digital telephone set 1 is connected via an ADSL modem 3 and a line 5 carrying an ADSL line to a special interface unit 7, also containing an ADSL modem, not shown. Node 7 and its ADSL modem are connected via an Ethernet network, symbolically illustrated by Ethernet switch 9, with PSTN 11. The PSTN is connected to an Ethernet network at the local telephone exchange / switch 13 and TAG (telephony access gateway) 15. TAG places voice samples into Ethernet frames and extracts voice samples from received Ethernet frames. A plurality of digital telephones can typically be connected to the same interface node 7 corresponding to the DSLAM 8 included in this document. Analog telephones 15 can also be connected to a special provided node 7, which has the necessary functions, such as, for example, A / D and D / A converters.

In the block diagram of FIG. 7 illustrates a DSLAM 8 that is connected to an Ethernet network through a network port 21. A DSLAM includes many separate DSL modems 23, each has an NTR unit 25 and is connected to subscriber lines 27. In addition, DSLAM 8 includes features not shown, for routing Ethernet frames received from network port 21, to adjust one of the DSL modems and relay digital information received from subscribers from DSL modems to the network port.

DSLAM 8 includes a synchronization device 28 comprising a retrieval unit 29, also referred to as a retrieval unit, for retrieving, preferably, a single packet stream, typically selected from a plurality of packet streams included on the network port 21. For this purpose, the block retrievals can be connected to suitable DSLAM protocol control units, not shown. Typically, the extraction unit 29 continuously examines packet flows to ADSL subscribers. As soon as the stream of voice packets or other services in real time is determined, which is or is supposed to be synchronous with the PSTN network 9, for example, a stream of video conference packets, it can be used in the synchronization process.

Choosing the right packet stream can be done in several ways. The choice is mainly made by examining information such as IP addresses, port numbers, and VLAN tags of received packets. Some examples are given below.

1. POTS traffic can be transmitted in a separate VLAN in order to be able to prioritize voice traffic and data traffic. A VLAN tag is part of the Ethernet packet header, and it can be used to distinguish and select voice traffic.

2. DSLAM 8 can know the IP address of TAG 15 and then filter out packets coming from that IP address.

3. POTS traffic to a specific phone has a unique PORT number, and it can be used to distinguish between voice traffic.

A combination of several criteria may be required to find packet streams suitable for synchronization. This process can be called “filtering the incoming streams of Ethernet packets”. To find filtering criteria, three main ways can be mentioned:

1. In transparent mode, the extraction unit 29 must itself determine the filtering criteria to be used. This can be done in two ways:

a. Signal Tracking. In this case, the extraction unit examines the signal transmission packets, i.e. packets containing call setup information. This kind of information is usually transmitted using, for example, SIP, H.323 or H.248. Signal transmission is analyzed and suitable POTS streams are selected.

b. Blind search. In this case, a blind search for RTP packets is performed. RTP packets do not have a unique header, and therefore it is not possible to detect them in a simple way. However, it is possible to find the RTP stream by looking for some special features, such as:

i. Packet size. Ethernet packets carrying voice information have a specific length, usually 40, 80 or 160 bytes, plus fixed overhead.

ii. Incremental time stamps. The RTP header contains a timestamp that increases by 40, 80, or 160 for each packet, depending on the number of bytes in each packet. This can be used to find possible RTP streams.

2. In proxy mode, DSLAM 8 takes part in signaling and knows about different streams. Client equipment such as ADSL modem 3 speaks with DSLAM, and DSLAM speaks with TAG 15.

3. In slave mode, DSLAM 8 is informed of the available flows from the observing system, not shown. This system may use a suitable protocol, such as SNMP, to inform the DSLAM of the available packet streams and / or filtering criteria that should be used.

If several suitable streams are available, one or more streams may be used. Preferably, however, a stream with the least delay variability is used.

The protocols supposed to be used for Ethernet network packets are (bottom-up) Ethernet, IP, UDP and RTP, see FIG. 8. RTP provides information regarding the serial number and timestamp as inserted by TAG 15. The timestamp is used to associate the packet with a real time. The timestamp is increased by the packetization time interval with the sampling rate. For example, for voice packets containing 20 ms, which corresponds to 50 Hz, speech sampled at a frequency of 8000 Hz, the time stamp for each block of speech samples is increased by 160, even if the block is not sent due to the suppression of pauses. This document assumes that there is one RTP packet in each Ethernet frame. Therefore, in this example, the RTP frames of one stream should arrive at a frequency of 50 Hz. From the arrival times, it is determined whether the reference clock signal should have a frequency selected among values, usually close to the nominal value of 8000 Hz.

Ethernet frames in a stream containing RTP packets can ideally arrive at equal intervals. However, due to the variability of the delay in the transmission network, random delays are added to the expected arrival times. Such random delays can be considered as a noise source in the synchronization recovery system, see FIG. 9. The noise is always positive, but direction can still be estimated, see FIG. 10. An oblique line designated as “Direction due to improper synchronization” in FIG. 10, indicates in the case shown that the reference clock used has a too high frequency for a short time. The line should ideally coincide or parallel with the abscissa axis of the circuit.

The extraction unit 29 controls the synchronization pulse generation unit, which is configured as the adaptive synchronization pulse generator 31 in the synchronization device 28, see FIG. 7 and 11, and provides a common clock signal to the NTR units 25 of the DSL modems 23. The adaptive clock pulse generator generates an 8-kHz reference signal having an adapted frequency or phase. This can be created in various ways in the next device using both analog and digital functions, and a device using only digital functions will be described.

In the block diagram of FIG. 12 shows the main functions of an adaptive clock pulse generator in which a PLL (phase locked loop) is used to regenerate an 8-kHz clock sampling signal. In the clock generator VCXO (variable frequency crystal oscillator) 33 is used together with the D / A converter 35. The generated clock signal is provided to the NTR units 25 of the DSL modems. As seen in FIG. 12, the PLL includes a VCXO 33, a frequency pre-divider 36, an N-bit counter 37, an N-bit register 39, an adder 41, a minimum calculation unit 43, a PLL filter 45, and a D / A converter (DAC) 35. Reference the synchronization signal to the DSL modems is provided by the frequency reduction unit 46 receiving pulses from the generator 33.

The arrival time of each received Ethernet frame is measured using the internal timer in the extraction unit 29. The timing for the timer is a reference signal provided by an oscillator 33, which in this embodiment is a controlled or custom oscillator. Arrival time is saved for later analysis to determine if the packet is useful for synchronization. In fact, the arrival time is defined as the deviation calculated in some clock pulses from the expected time, the expected time is provided by the internal timer.

The internal timer is an N-bit timer that performs an entire count at the same time as the time interval between two packets, i.e. as the time intervals between the start of consecutive packets, i.e. the timer will cycle once for each frame containing the RTP packet of the stream. The actual packet arrival time may be anywhere in the interval, comparing the circuit in FIG. 13.

In the embodiment of FIG. 12 VXCO 33 provides high-frequency pulses, such as 32.768 kHz in the example, the frequency of the pulses is reduced by a preliminary frequency divider 36 by dividing by the number M, for example 10, the preliminary frequency divider provides pulses to the counter 37, in the example to a 15-bit counter, so that they counted, and the counter provides values such as those shown in the graph in FIG. 13. The frequency division coefficient of the preliminary frequency divider should be selected so that the N-bit counter 37 makes a cyclic transition every period of the packet, i.e. 10 ms if the packet size is 10 ms. The time of arrival of frames should be at the same position in the interval of the count, i.e. on a line parallel to the abscissa axis. When packets arrive, the actual value of counter 37 is transmitted through an adder 41, where a constant is added, to the minimum calculation unit 43 and stored. This unit calculates the minimum value that is provided to the PLL filter 45, which, in turn, outputs a value for controlling the VCXO 33, this value is first converted to an analog value by means of the DAC 35.

To get the same thing, for example, the zero conclusion from the determination of the arrival time in a steady state, i.e. when the system is in exact synchronization, the synchronization offset by the adder 41 is added to the value calculated by the counter 36, so that the estimated arrival time will have the same value, for example, will be zero in the example, i.e. at the beginning of the reference interval. If, for example, the timer value at the estimated time of arrival of the packet is 0x7000, an offset value of 0x9000 is used. The sum of these values makes a cyclic transition and gives the desired result in 0x0000. The synchronization offset is calculated once in the first synchronization operation, as well as in each synchronization operation, when a new stream is used to form the reference synchronization signal.

The minimum calculation unit 45 may, for a block of consecutive packets, for example, for all stream packets received over a significant period of time, typically in one or more seconds, use the packet that arrived with the least delay as an estimate of the true moment of arrival, i.e. . he can use the minimum value algorithm to find the smallest value from the set of values to find the packet with the least delay. The algorithm can be implemented as a traditional minimum value search algorithm with a sliding window, see FIG. fourteen.

PLL filter 45 is a low-pass filter and performs its calculations once for each new packet, for example, 100 times per second if 10-ms packets are used and packets are not lost, see FIG. 15. When starting, the PLL filter 45 contains zero, this corresponds to the upper extreme frequency, and the frequency of the VXCO 33 is set to an average value, for example, the frequency corresponding to the generated synchronization reference signal, exactly 8 kHz. The minimum value algorithm is allowed to work, for example, for about 1 second, and the minimum arrival time delay is calculated. The synchronization bias value is calculated and used as a constant added in summing circuit 41.

In the resynchronization operation, the PLL filter 16 is not reset, i.e. it retains the same filtering characteristics, and a new synchronization bias value is calculated to be used as a constant added in the summing circuit 41 by running the minimum value algorithm in the minimum value calculating unit 43, for example, for about 1 second.

When the RTP stream used for synchronization is stopped, the PLL filter 45 is fixed, i.e. it retains the same filtering characteristics as before, and a new frame stream is selected for synchronization, and the re-synchronization operation is performed to find a new synchronization offset. Also, if pauses are suppressed, the RTP stream may disappear for a certain period of time, and then it may be useful to change the packet stream used for synchronization.

When all the RTP streams disappear, the PLL filter 45 and the clock frequency of the VXCO 33 are locked, i.e. remain unchanged until a new RTP stream is found and selected. When a new RTP stream is encountered, a re-synchronization operation is performed.

Several conditions must be met in order to approve a package for use in a synchronization algorithm.

1. This should be an Ethernet frame containing an RTP packet belonging to the stream that is currently being used for synchronization.

2. The package must belong to the correct period of time, i.e. the time stamp of the RTP packet must match the local time stamp counter (TSC), as described below.

Each RTP packet contains a timestamp in its header. This timestamp is increased by TAG 15 by the number of voice samples in the packet for each new packet. The local time stamp counter is incremented in the same way once in each period of the expected packet. It can be implemented as a hardware or software counter connected to a timer unit, which includes a generator 33 and an N-bit counter 37.

First, the timestamp from the received RTP packet of the selected stream is loaded into the TSC. Each time a packet arrives, the time stamp in the RTP header is compared with the TSC value, and if they are equal, the packet is approved for use in the synchronization algorithm.

If K consecutive timestamps have values less than the corresponding TSC value, the TSC is synchronized with the new value. If L consecutive timestamps have values less than the corresponding TSC value, the TSC also synchronizes with the new value. K, L are predefined by suitably selected numbers.

The synchronization algorithm contains a number of parameters. Some of them should be installed according to the current situation. Other parameters can be used to change the behavior of the synchronization algorithm.

The window size used by the minimum calculation unit 43 in determining the minimum value determines the number of packets that are evaluated in order to find the packet arriving with the least delay. A large window size increases the stability of the PLL, but reduces the possibility of a quick response. There is a relationship between the window size and the time constant of the PLL filter 45. If the window is too large compared to the time constant, the cycle may become unstable. The window size must be saved such that the time span to which it corresponds is much smaller than the time constant of the PLL filter.

The clock generator can be implemented either as a voltage controlled crystal oscillator, VCXO, as described above, or using a generator having a fixed frequency and digital frequency control, i.e. digital controlled generator (DCO).

The adaptive clock generator includes in a possible implementation using DCO, see FIG. 16:

1. The clock generator 71, which is a crystal oscillator with a fixed frequency that is a multiple of 8 kHz, for example 6-32,768 MHz.

2. Counter 73, a counter that divides the signal from the clock 71 into a settable number x. Thus, the counter has an adjustable maximum value, such that when this value is reached, the counter outputs a pulse, and the counter starts counting from the starting value, the starting and maximum values are 0 and (x-1), respectively, in the traditional way. Therefore, it can be said that the frequency of the output clock signal is externally controlled by dividing it by slightly different coefficients, thus forming small frequency adjustments of the reference clock signal. In the example, the number x can take the values 5, 6, 7, where 6 is a “normal” or “nominal” value.

3. Block 75 division provides the desired reference clock signal, nominally equal to, for example, 8 kHz, it is a signal provided to the blocks NTR. The division block 74, which is essentially a counter, splits the output signal of the counter 73, dividing it by a suitable number P, in the example, this block is a 16-bit counter, which means P = 65536.

4. A frequency divider 36, which divides the signal output from the counter 73 by the number M, in the example M = 10.

5. N-bit counter 37, receiving a signal from the frequency divider 36, the number N is determined so that the counter performs a cyclic transition at a packet frequency of, for example, 50, 100 or 200 Hz.

6. Arrival time measurement unit 39, which is coupled to receive the value of counter 37, and has the function of reading the counter each time a packet of the selected stream arrives, so that the same value will be output every time a packet arrives at exactly the expected time.

7. The minimum calculating unit 43, which is connected to the arrival time measuring unit 37 and is used to sort incoming packets with low delays and to create a control signal indicating the “best” expected packet delay in the selected stream, for example, the smallest delay in the sliding window.

8. The PLL filter 45 ', which is connected to the minimum computing unit 43 to receive a control signal, and is a low-pass filter for smoothing the control signal.

9. A phase accumulator 77, which accumulates phase errors, as received from the PLL filter 45 ', as long as there is time for tuning the frequency output from the counter 73, i.e. value of its cyclic transition. The phase accumulator includes an adder 79, a register 81 and a function block 83 modulo. The modulo function — the module of some suitable number z — creates a “cyclic transition” of the phase accumulator during overflow or negative overflow, while at the same time adjusting the phase of the adjustable counter 73, i.e. changing the currently used number x to (x + 1) or (x-1), respectively.

In the resynchronization operation, the N-bit counter 37 may be reloaded with a new calculated value, or the offset value may be added to its contents.

As seen in the block diagram of FIG. 17, instead of the case described above in which the NTR clock pulses are tuned according to the clock pulse recovered from the packet stream, the ADSL clock pulses 85 can be tuned independently. In the latter case, NTR is not used, but the ADSL clock will be tuned to be synchronous with the 8-kHz clock. The adjustments of the ADSL clock generator must be very small and slow so as not to affect the performance of the ADSL, this is achieved by selecting parameters, for example, PLL filters 45, 45 ', respectively.

Extraction unit 29 may perform the procedure illustrated in the flowchart of FIG. 18. Initially, the frequency of the reference clock signal is set to a nominal, average value, as indicated in block 101. In the case of FIG. 16, this means setting the counter 73, which should be an x-bit counter, where the integer x is selected so that the output synchronization reference signal as represented by the frequency division device 75 has a frequency, for example, equal to exactly 8 MHz. When a frame is received, in block 103, it analyzes whether it can be used for synchronization in block 105. If it cannot be used, another block is expected in block 103. If the received frame can be used, block 107 is executed in which the length of the voice information in the frame is determined, i.e. number of voice samples. The N-bit counter 37, 73 is adapted accordingly so that it performs a cyclic transition at the same rate as packets from the same stream are expected. By sending the appropriate commands to the minimum calculation unit 43, storing all the parameters in the PLL in the initial values, the smallest delay of the received packets in the same stream is determined for a relatively long period of time, typically in one or several seconds. Based on this, re-synchronization is performed, so that the appropriate correction value is calculated from this smallest delay and is provided to the adder 41 or N-bit counter 37. It is also possible to set other parameters, such as the window size of the minimum calculation unit 43 and the time constant of the PLL filter 45 , 45 '. The stream is finally selected by setting an entry for this stream and setting a pointer to this entry in memory, see FIG. 19.

A new frame is expected in block 109, and when it arrives, it is determined in block 111 whether it belongs to the selected stream. If it belongs, it is determined in block 113 whether the timestamp of the frame is equal to the TSC value for this stream. If true, block 115 is executed in which the actual time is sent to the adaptive clock pulse generator, register 39. Then, the actual time is saved as a record for this stream, as indicated in block 117. If the time stamp does not correspond to the TSC value, block 119 is executed, in which stores the timestamp and the actual TSC value. Then block 117 is executed.

In block 121, it is asked if there are K consecutively received packets for which timestamps are less than the corresponding TSC values by analyzing the values as stored in block 119. If this is true, the TSC value is corrected in block 123, and then a new frame is expected in block 109. If this is not the case, block 125 asks if there are K consecutively received packets for which timestamps are greater than the corresponding TSC values. If this is the case, block 123 is executed, and a new frame is expected in block 109. If this is not the case, a new frame is expected in block 109.

If it was determined in block 111 that the frame does not belong to the selected stream, it is determined in block 127 whether a timeout exists for packets of this stream, i.e. if a period of time greater than some predetermined value has passed from the last reception of the frame of the selected stream by analyzing the time stored in block 117. If this is not the case, a new frame is expected in block 109. If this is true, it is determined in block 129 whether there is another RTP stream that can be used for synchronization. If such a stream exists, it is taken as a new stream, which should be used in the case when there is more than one such stream, one of them is taken in block 131. Then a new frame is expected in block 109. If the RTP stream does not exist, as defined in block 129, the PLL parameters are fixed, i.e. the command is not to be changed by sending the appropriate signals to the minimum value calculating unit 43 and the PLL filter 45, 45 '. Then, in block 109, a new frame is expected.

The main blocks included in the extraction unit 29 are visible in the block diagram of FIG. 19. The extraction unit includes a general frame analysis unit 201 used in selecting suitable packet streams. For this purpose, it may, in turn, include analyzer modules, such as a VLAN tag analyzer 203, a source IP address analyzer 205, a destination port analyzer 207, and a packet length analyzer 209. The stream selection section 211 selects a suitable stream, for example, as prescribed by the results obtained from the general frame analysis section 201. Sample length determining section 213 determines the length of voice information in packets of the selected stream, and counter length setting section 215 attached to it sends suitable commands to N-bit counter 37 in accordance with the determined packet length. Block 217 controls the minimum calculation unit by sending appropriate commands, for example, to initiate, fix, or count, to set the window size and to allow it to work for a relatively long time, finding only the smallest value, i.e. least delay. Resynchronization block 219 performs the calculations necessary to create parameters for resynchronization operations, receiving information from various other blocks and sending commands and values, such as an offset value, to adder 41.

In addition, block 221 is provided for analysis when there is a selected stream of packets, packets received in this stream. In particular, this block includes a comparison block 223 for comparing the time stamp of the received packet with the TSC value. Block 225 is configured to receive pulses from the PLL, from its N-bit counter 37, and then increase the TSC of the selected stream by the value valid for that stream.

Memory 227 includes a memory cell 229 for storing a pointer to a selected stream. It also includes a table or a small database of 231 current RTP stream data. For each stream, thus, memory cells exist, not shown, for storing some identification of the stream, so that its packets can be recognized in the stream of incoming packets. For each stream, there is a memory cell for its TSC and cells for storing the timestamp values of the most recently received packets along with associated TSC values are not shown. Also, for each stream, there is a memory cell that stores the time when the last packet of the stream arrived.

Claims (14)

1. The relay unit, receiving Ethernet packets and relaying them and / or information in them via a digital subscriber line, characterized by a device for creating a synchronization signal, including:
a selection unit for selecting at least one stream of received Ethernet packets,
a synchronization signal generating unit for generating a reference synchronizing signal according to the moments of arrival of the packet in the selected at least one stream, and
a modem for communication with a digital subscriber line, the modem is connected to a synchronization signal generating unit that receives a reference synchronization signal for use in transmitting voice information to a digital subscriber line, in particular, a reference synchronization signal is provided to a time reference unit in a modem or a transmission synchronization unit. 2. The relay unit according to claim 1, characterized in that the selection unit includes a general analysis unit for analyzing received Ethernet packets to find a stream of packets from one target to one user, the packets carry real-time information belonging to the service in real time , a selection block is set to select such a stream. 3. The relay unit according to claim 2, characterized in that the general analysis unit is adapted to analyze the VLAN tag, source IP address and / or destination port address of the received packets in order to find suitable packet streams. 4. The relay unit according to claim 2, characterized in that the general analysis unit is adapted to analyze the length of the received packets in order to find suitable packet streams. 5. The relay unit according to claim 1, characterized in that the synchronization signal generating unit includes a delay determination unit defining delays of the arriving packets with respect to the reference synchronizing signal. 6. The relay unit according to claim 5, characterized in that the delay determination unit includes:
a signal generator providing a first clock signal having a frequency,
a delay measurement counter receiving a first clock signal or a signal obtained by dividing it by a predetermined number, and counting the pulses of the received signal modulo the first number, and
a measurement unit connected to a counter for determining, as an estimate of the packet delays relative to the reference clock signal, a counter value for the delay times of arrival of each packet of the selected stream. 7. The relay unit according to claim 6, characterized in that the selection unit is configured to set the number of cyclic transitions of the delay measurement counter, in particular its length, i.e. the number of bits in it, so that the times for the cyclic transition are in accordance with the estimated arrival times of the packets of the selected stream, i.e. occur at the same frequency, block adaptation, in particular, based on the number of voice samples in packets of the selected stream. 8. The relay unit according to claim 6, characterized in that the synchronization signal generating unit includes a frequency division unit receiving a first synchronization signal to provide a reference synchronization signal. 9. The relay block according to claim 5, characterized in that the block for generating a synchronizing signal includes:
a calculation unit for calculating from certain delays the correction of the frequency or phase of the reference clock signal, in particular, for calculating the minimum delays during the last period of time of the set length, and
an adjustment unit for adjusting the frequency or phase, respectively, of the reference clock signal according to the calculated correction. 10. The relay unit according to claim 9, characterized by a PLL filter connected between the calculation unit and the adjustment unit for filtering the calculated correction, the PLL filter, in particular, has a time constant that is significantly longer than the period of packet relay on a digital subscriber line . 11. The relay unit according to claim 2, characterized by a second analysis unit for analyzing packets of the selected stream by time stamp to determine whether the time stamp of the arrived packet corresponds to the time stamp counter increasing by the signal command from the delay measurement counter, the synchronization signal generating unit with the ability to generate a reference clock signal according to the moments of arrival of only those packets for which such a match is defined as existing. 12. A method of obtaining a reference clock signal in a unit receiving Ethernet packets and relaying them on a digital subscriber line, characterized in stages in which:
analyze the received Ethernet packets to search for a stream of packets from one destination to one user, the packets carry real-time information belonging to the service in real time, and select such a stream,
determining from the arrival times of packets of said selected delay stream with respect to the clock signal,
calculating from certain delays the correction of the frequency of the clock signal, and
adjust the frequency or phase, respectively, of the reference clock signal according to the calculated correction. 13. The method according to p. 12, characterized in that at the stage of adjustment, the correction is first provided to the PLL filter, the PLL filter, in particular, has a time constant, which is significantly longer than the period of packet relay on a digital subscriber line. 14. The method according to p. 12, characterized by an additional step in which the packets of the selected stream are analyzed by time stamp to find whether the packets arrive in sequential order, and in the determination step only the arrival times of packets arriving in sequential order are determined.

Images