ITTO20120740A1 - METHOD TO SYNCHRONIZE THE KNOTS OF A NETWORK WITH A PACKAGE, NETWORK KNOT AND PACKAGE NETWORK - Google Patents

Description

DESCRIPTION

of the patent for industrial invention entitled:

"METHOD FOR SYNCHRONIZING THE NODES OF A PACKET NETWORK, NETWORK NODE AND PACKET NETWORK"

The present invention relates to a method of synchronizing nodes of a packet network, to a network node and to a packet network.

As is known, in the field of professional mobile radio communications ("Professional Mobile Radio", PMR), Simulcast communication systems are available, in which two or more nodes of a network simultaneously broadcast the same signal, at the same frequency; therefore, the two or more nodes transmit identical signals. In order to prevent the interference of these identical signals, as well as to allow the correct reception of the same by a node, the transmission times of the signals, and in particular of the corresponding envelopes in the baseband, must be aligned with high accuracy, and that the frequencies of the corresponding carriers ("carriers") are as equal as possible. In other words, the nodes are required to be synchronized with each other, both in frequency and in phase. In general, each node comprises a respective oscillator and a respective timer ("clock"), which is controlled by the oscillator; therefore, it is necessary that the oscillators of the network nodes operate at the same frequency, and that the timers of the network nodes are in phase with each other.

In order to synchronize the nodes, the use of so-called synchronization sources external to the network is known, such as the global positioning system ("Global Positioning System", GPS), which provides an absolute time reference to all nodes of the network.

Synchronization techniques are also known which make it possible to disregard the availability of an absolute time reference. For example, the article "Synchronization of Single-Frequency Simulcast Networks Using Network Time Protocol", by S. Bregni, L. Lacavalla, B. Propersi, F. Residori, IEEE International Conference on Communications, June 2007, describes a method of synchronization based on a hierarchical strategy of the so-called "master-slave" type, in which the time reference is represented by the timer of a master node, which is transmitted to all the other nodes (slaves) of the network.

In detail, according to the aforementioned article, the master node acts as a server of the so-called "Network Time Protocol" (NTP), while the slave nodes act as NTP clients. Periodically, the slave nodes send NTP request packets to the server node and await corresponding response packets from the server node. The response packets contain time stamps, on the basis of which the slave nodes determine estimates of the "round trip time"; on the basis of the estimates of the travel time per revolution, the slave nodes then determine the delays of their timers with respect to the timer of the master node, as well as the frequency "offsets" of their oscillators with respect to the oscillator of the master node.

In greater detail, each slave node implements a phase-locked loop (PLL), at the input of which there is a comparator, which, at each response packet, determines the time offset between the timer of the master node and the timer of the slave node. The time offset is input to a loop filter, which in turn controls the oscillator; the output of the oscillator is supplied in input to the comparator, in order to allow the determination of the aforementioned time offset. The loop filter is of the third order, therefore the phase locked loop is of the so-called second type; therefore, the phase locked loop is able to follow frequency steps with a zero residual phase error.

In practice, thanks to the use of the NTP protocol and the implementation of a phase locked loop of the type described, each slave node synchronizes in phase and frequency with the master node.

Although the method described in the aforementioned article allows the nodes of a packet-type Simulcast isofrequency network to be synchronized in phase and frequency, it makes use of the NTP protocol, which does not allow to calculate the time stamps with accuracy and precision of less than one microsecond. Furthermore, since packet networks introduce a not negligible fluctuation ("jitter") of travel times per revolution, the band of the loop filter must be quite narrow in order to properly filter this jitter. This implies that phase and frequency locking requires a long time; therefore, any drifts of the master node's oscillator are tracked slowly by the slave nodes.

The object of the present invention is therefore to provide a synchronization method which at least partially solves the drawbacks of the known art.

According to the present invention, a synchronization method, a network node and a packet network are provided as defined, respectively, in the attached claims.

For a better understanding of the invention, embodiments are now described, purely by way of non-limiting example and with reference to the attached drawings, in which:

Figure 1 symbolically shows a network formed by a plurality of nodes;

Figure 2 shows a block diagram of the internal structure of a master node, according to the present invention;

figure 3 shows a block diagram of the internal structure of a slave node, according to the present invention; And

Figures 4 and 5 schematically show time diagrams relating to the sending and receiving of messages by a master node and a slave node, according to the present invention; the times related to the sending / receiving of messages by the master node are related to a timer internal to the master node, while the times related to the sending / receiving of messages by the slave node are related to a timer internal to the slave node .

Figure 1 shows a network 1 formed by a plurality of nodes, generically indicated with 2.

In a per se known manner, one of the nodes, indicated with 4, acts as a master node, while the other nodes act as slave nodes; in figure 1 a slave node is indicated by 6. In the following of the present description reference is made, without any loss of generality, to the master node 4 and to the slave node 6. For the purposes of the present invention, it is irrelevant how each node 2 determines whether act as a master node or as a slave node; further details will however be provided later.

As described in more detail below, each of the nodes 2 is capable of executing the so-called "Precision Time Protocol" (PTP), as described in the IEEE 1588-2008 standard. In particular, before the roles (master or slave) of the nodes 2 of the network 1 are established, each node 2 is able to act both as a master node and as a slave node, in accordance with the IEEE 1588-2008 standard; following the definition of the roles of nodes 2, each node acts in accordance with its own role. Purely by way of example, the assignment of roles can take place according to the same IEEE 1588-2008 standard, ie the nodes 2 can perform self-configuration operations of the network 1, before carrying out the present synchronization method. In this regard, the IEEE 1588-20008 standard provides for the execution, by nodes 2, of a so-called "best master clock" algorithm. Alternatively, and again purely by way of example, the roles of the nodes 2 can be assigned by an external unit to the network 1.

That said, figure 2 shows a node, and in particular a master node 4.

The master node 4 comprises a processing unit 10 and a time-stamp unit 12 ("time-stamp unit", TSU), connected to the processing unit 10 and which is referred to hereinafter as the TSU 12 unit. .

The processing unit 10 implements a "protocol stack" 14, which, in a per se known manner and in accordance with the so-called "International Standard Organization - Open Systems Interconnect ion" (ISO-OSI) model, comprises a application layer, a transport layer, a network layer and a medium access control layer (MAC layer), which is referred to hereinafter as the MAC layer.

For example, the application layer can be adapted to allow the transmission of Simulcast signals on a packet network of the so-called "internet protocol" (IP) type in a per se known manner. Furthermore, the transport layer can conform to the so-called user datagram protocol (UDP); in this regard it should be noted that in the following, while assuming that the transport layer is actually compliant with the UDP protocol, we will refer to packets, rather than to datagrams. As regards the network connection layer, it conforms to the so-called "Internet Protocol" (IP). Finally, the MAC layer complies, for example, with the so-called Ethernet protocol.

Further details relating to the processing unit 10 will be described below.

The master node 4 also comprises an interface 16 of the so-called "Media independent interface" (MII) type, connected to the processing unit 10 and which is referred to hereinafter as the MII interface 16. The MII interface 16, known per se, it is suitable for receiving packets generated by the MAC layer of the protocol stack 14.

The master node 4 also comprises a physical interface 18 suitable for routing the packets passing through the MII interface 16 towards the network 1. The physical interface 18 depends, in a per se known way, on the physical characteristics of the transmission means used. to connect the nodes 2. For example, the physical interface 18 can be formed by a so-called PHY chip (also known briefly as "PHYceiver"), compatible with the Ethernet protocol. The master node 4 further comprises an oscillator 20, this oscillator being of the voltage-controlled type. In particular, the oscillator 20 is adapted to provide a local periodic signal, having a frequency which is controlled in the manner described below.

With regard to the aforementioned TSU 12 unit, it comprises a timer 22, which is controlled by the oscillator 20. In a per se known manner, the timer 22 therefore defines a local time ("local time"), relative to the master node 4, and whose unit of time is represented by a period of the local periodic signal supplied by the oscillator 20. In other words, the timer 22 periodically updates its local time, on the basis of the local periodic signal supplied by the oscillator 20.

The TSU 12 unit further comprises a detection unit 24, which is connected to the timer 22, in such a way that it has the local time. Furthermore, the detection unit 24 is electrically connected to the interface MII 16 and is adapted to detect the transit of packets through the interface MII 16; more particularly, the detection unit 24 detects the transit of packets through the MII interface 16 and associates, to each transit, a corresponding transit time, as provided by the timer 22. For the sake of brevity, hereinafter we also refer to detection, by the detection unit 24, of the transit time of a package, I mean the detection of the transit and the association of the corresponding time. Furthermore, in the following reference is made indifferently to packets or messages, implying that the messages mentioned in the present description each occupy a single packet, without any loss of generality.

Again with reference to the processing unit 10, it communicates with the timer 22 to receive the local time, in such a way that the local time is available to the protocol stack 14. Furthermore, the protocol stack 14 comprises a further application level, conforms to the aforementioned IEEE 1588-2008 standard, which has access to the local time provided by timer 22; later this further application level is referred to as the PTP application level. In addition, this further application level communicates with the detection unit 24, in order to receive the transit times determined by it, as described in greater detail below.

Figure 3 shows a slave node 6. Components of the slave node 6 already present in the master node 4 are indicated with the same reference numbers, unless otherwise specified; in particular, the oscillator and the timer of the slave node 6 are indicated respectively with 21 and 23. Further elements of the slave node 6 will be described below; furthermore, the slave node 6 will be described limitedly to the differences of the latter with respect to the master node 4 shown in figure 2. However, before further describing the slave node 6, operations carried out by the master node 4 and by the slave node 6 are described, with reference to figure 4.

In detail, the master node 4 transmits a first Sync message, in a per se known manner. In particular, the first Sync message is generated by the processing unit 10 of the master node 4, by implementing the corresponding protocol stack 14, subsequent transit through the interface MII 16 and input into the network 1, this input being carried out by the physical interface 18.

In greater detail, the first Sync message transits on the interface MII 16 of the master node 4 at a time ti <1-1>, detected by the detection unit 24 of the master node 4, which communicates this time t1 <i-1 > to the processing unit 10 of the master node 4, and in particular to the PTP application level of the protocol stack 14 of the master node 4. The time ti <i- 1> is relative to the timer 22 of the master node 4, i.e. it is measured on local time basis provided by the latter.

In the following, for the sake of brevity, reference is made to a node that transmits a generic message at a time tx to indicate the time in which the generic message passes through the interface MII of the node, this time being relative to the timer of the node and being detected by the node detection unit. In particular, as previously mentioned, the detection unit detects the transit of the message through the interface MII of the node and associates to this transit the corresponding time tx indicated by the timer. The time tx is communicated by the node detection unit to the node processing unit, and in particular to the PTP application level of the protocol stack; in the following the reference to the PTP application level is implied, unless otherwise specified. Furthermore, in the following the implementation of the respective protocol stack by the processing unit of the node, to generate the generic message, is implied.

The first Sync message is received by the slave node 6. In particular, the first Sync message is received by the physical interface 18 of the slave node 6, passes through the MII interface 16 of the slave node 6 and is processed by the processing unit 10 of the slave node 6 by implementing the respective protocol stack In greater detail, the first Sync message passes through the interface MII 16 of the slave node 6 at a time t2 <i-1>, detected by the detection unit 24 of the slave node 6 , which communicates this time t2 <i-1> to the processing unit 10 of the slave node 6, and in particular to the PTP application level of the protocol stack 14 of the slave node 6. The time relates to the timer 23 of the slave node 6 .

In the following, for the sake of brevity, reference is made to a node that receives a generic message at a time ty to indicate the time in which the generic message passes through the interface MII of the node, this time being relative to the timer of the node and being detected by the node detection unit. In particular, as previously mentioned, the detection unit detects the transit of the message through the interface MII of the node and associates the corresponding time tyindicated by the timer to this transit. The time ty is communicated by the node detection unit to the node processing unit, and in particular to the PTP application level of the protocol stack; in the following the reference to the PTP application level is implied, unless otherwise specified. Furthermore, in the following the implementation of the respective protocol stack by the processing unit of the node, in order to receive the generic message and possibly extract the information contained therein, is understood.

After sending the first Sync message, the master node 4 transmits a first follow-up message, at a time t1 <i-1 *>. The first follow-up message is generated by the processing unit 10 so as to contain a time stamp equal to the time ti <1-1>.

The first follow-up message is received by the slave node 6 at a time t2 <i-1 *>; after receipt, the processing unit 10 of the slave node 6 extracts, in a per se known manner, the time stamp contained in the first follow-up message.

Following the extraction of the time stamp contained in the first follow-up message, the processing unit 10 of the slave node 6 determines a first delay from master to slave which is equal to t2 <i-1> - t1 <i -1>. In other words, in the present description the first delay, as well as the delays described below, are intended as propagation delays.

Subsequently, at a time t1 <i>, the master node 4 transmits a second Sync message, which is received by the slave node 6 at a time t2 <i>. The time t2 <i-1> is communicated by the detection unit 24 of the slave node 6 to the processing unit 10 of the slave node 6.

Subsequently, at a time t1 <1 *>, the master node 4 transmits a second follow-up message, which is generated by the processing unit 10 of the master node 4 so as to contain a time stamp equal to the time t1 <i>.

The second follow-up message is received by the slave node 6 at a time t2 <i *>; after reception, the processing unit 10 of the slave node 6 extracts, in a per se known manner, the time stamp contained in the second follow-up message.

Following the extraction of the time stamp contained in the second follow-up message, the processing unit 10 of the slave node 6 determines a second delay from master to slave ΔM-to-si which is equal to t2 <i> - t1 <i>.

In other words, on the basis of each pair formed by a Sync message and the corresponding follow-up message, the slave node 6 is able to determine, in a per se known manner, a corresponding delay from master to slave AM-to -s <1>

As shown, by way of example, again in Figure 4, the slave node 6 is also configured to transmit a first delay request message, which is received by the master node 4 at a time t4. Purely by way of example, in Figure 4 the transmission of the first delay request message by the slave node 6 occurs at a time t3, subsequent to the time t2 <i>; however, in a per se known manner, the transmission of the first delay request message can take place at different times, such as between the transmission of the first follow-up message and the transmission of the second Sync message.

Upon receipt of the first delay request message, the master node 4 transmits a first delay response message, at a time t4 <*>. The first delay response message is received by the slave node 6 at a time t3 <*>.

Upon receipt of the first delay response message, the processing unit 10 of the slave node 6 determines a delay from slave to master which is equal to t4-t3.

In other words, on the basis of each pair formed by a delay request message and a delay response message, the slave node 6 is able to determine, in a per se known manner, a corresponding delay from slave to master.

The operations shown in Figure 4 are repeated over time by the master node 4 and the slave node 6, for example periodically, so that the processing unit 10 of the slave node 6 has a succession of samples of the delay to be master to slave and a succession of samples of the delay from slave to master

In particular, such sequences of samples are determined as previously described by the PTP application level of the protocol stack 14 implemented by the processing unit 10 of the slave node 6.

Having said all this, and again with reference to Figure 3 and to the characteristics of the slave node 6, the timer 23 has a first and a second input, the first input being connected, for control purposes, to the output of the oscillator 21, the second entrance being described below. Furthermore, the processing unit 10 of the slave node 6 implements a statistical processing stage 30 and a calculation stage 34.

In detail, the statistical processing stage 30 receives the samples of the delay from master to slave and the samples of the delay from slave to master; moreover, the statistical processing stage 30 stores the temporal duration of a temporal observation window. More particularly, given an observation time window, the statistical processing unit 30 processes the master-to-slave delay samples and the slave-to-master delay samples received during that observation time window, as described below. For example, assuming that the master node 4 transmits, every second, one hundred twenty-eight triples each consisting of a Sync message, a followup message and a delay response message, and assuming that the observation time window has the same duration at five seconds, the statistical processing stage 30 processes, subject to losses, six hundred and forty samples of the delay from master to slave and six hundred and forty samples of the delay from slave to master

In the following, the operations carried out by the statistical processing stage 30 are described with reference to a steady state situation, and in particular with reference to a generic j-th observation time window, preceded by a j-1-th observation time window. , and assuming that the operations carried out by the statistical processing stage 30 are the same during each observation time window.

In detail, the statistical processing stage 30 determines, on the basis of the samples of the delay from master to slave, one or more of the following statistical quantities: the minimum value, the average value and the most probable value, the latter being also known as fashion. Similarly, on the basis of the samples of the delay from slave to master, the statistical processing stage 30 determines one or more of the following statistical quantities: the minimum value, the average value and the mode.

The statistical processing stage 30 then selects one of the statistical quantities calculated on the basis of the samples of the delay from master to slave

hereinafter referred to as the statistical quantity of delay from master to slave.

Purely by way of example, the statistical processing stage 30 can select the statistical quantity with the lowest variance, as calculated on the basis of previous values of the same statistical quantity, relating to previous observation time windows, i.e. to a set of n-th time windows. observation, with n <j.

Alternatively, and again purely by way of example, the statistical processing stage 30 can select the mode, provided that a relevant fraction (for example, at least equal to 80%) of the samples of the delay from master to slave such around being centered in the mode and having an amplitude for example equal to ten microseconds.

Similarly, the statistical processing stage 30 selects one of the statistical quantities calculated on the basis of the samples of the delay from slave to master

hereinafter referred to as the statistical quantity of delay from slave to master. For example, the criterion for selecting the statistical quantity of delay from slave to master can be the same as the criterion used to select the statistical quantity of delay from master to slave.

In the following, for simplicity, it is assumed that, given the jesima observation time window, the statistical quantity of delay from master to slave is equal to the minimum among the samples of the delay from master to slave

relating to this observation time window. Similarly, in the following it is assumed that, again given the jth observation time window, that the statistical quantity of delay from slave to master is equal to the minimum among the samples of the delay from slave to master.

relating to this observation time window.

The statistical processing stage 30 communicates to the calculation stage 34 the statistical quantity of delay from master to slave, which in the following for brevity is referred to as the quantity G_MS <j>, and the statistical quantity of delay from slave to master, which in the following for the sake of brevity we refer to the quantity G_SM <j> j Furthermore, the statistical processing stage 30 selects the Sync message corresponding to the quantity G_MS <j> and communicates to the calculation stage 34 the times and relative, respectively, to the transmission and upon receipt of the selected Sync message.

In turn, the calculation stage 34 calculates a time difference Δφ <j>, equal to:

(1)

Furthermore, the calculation stage 34 calculates a frequency deviation Δf <j>, equal to:

(2)

in which the times and are the instants of transmission and reception of the Sync message corresponding to the minimum between the samples of the delay from master to slave relating to the j-1-th observation time window, i.e. to the Sync message selected by the statistical processing stage 30 during the j-1-th time window of observation. In practice, according to equation (2), the frequency difference Δf <j> is proportional to the difference between the quantity G_MS <j> and the quantity G_MS <j- 1> relative to the j-1-th time window of observation .

In use, the calculation stage 34 therefore provides a succession of values of the time difference Δφ <j> and a succession of values of the frequency difference Δf <j>. In particular, the calculation stage 34 has a first and a second output; the values of the frequency difference Δf <j> are provided on the first output, while the values of the time difference Δφ <j> are provided on the second output.

It should also be noted that, if the quantities G_MS <j> and G_SM <3> are equal to the mode or to the average value of the corresponding delay samples, the operations relating to the calculation of the differences Δf <j> and Δφ <j> do not change, provided that, in the case of selection of the average value, the times

are the instants of transmission and reception of the Sync message such that the corresponding delay from master to slave is closest to the average value of the samples of the delays from master to slave

That said, the processing unit 10 of the slave node 6 also implements a frequency control branch 40 and a phase control branch 42.

The frequency control branch 40 comprises a numerical filter 44 of the low-pass type, an amplifier 46 of the variable gain type, a differentiator 48 and an integrator 50; purely by way of example, the numerical filter 44 is of the second order, with polynomial y (n) = x (n) 1.820 y (n-1) - 0.828 y (n-2), where x (n) indicates l 'nth sample in input to the numerical filter 44 and y (n) indicates the n-th sample in output. The phase control branch 42 comprises a phase tracking stage 52.

In detail, the numerical filter 44 is connected to the calculation stage 34, and in particular it is connected to the first output of the calculation stage 34, so as to receive the frequency deviations Δf <j>. The output of the numerical filter 44 is connected to the input of the amplifier 46; moreover, a control terminal of the amplifier 46 is connected to the first output of the computing stage 34, so as to also receive the frequency deviations Δf <j>. The amplifier 46 varies its own gain as a function of the frequency deviations Δf <j>, as described below.

The output of the amplifier 46 is connected to the input of the differentiator 48, which is of the discrete type and implements, purely by way of example, the polynomial y (n) = x (n) - 0.9 x (n-1) .

The output of the differentiator 48 is connected to an adder 60, which is interposed between the differentiator 48 and the integrator 50. The adder 60 has a first and a second input; the first input is in fact connected to the output of the differentiator 48. The output of the adder 60 is connected to the input of the integrator 50, which is of the discrete type and has a transfer function for example equal to y (n) = x ( n) y (n-1).

The slave node 6 also comprises a digital-to-analog converter (DAC) 62, whose input is connected to the output of the integrator 50, and whose output is connected to the input of the oscillator 21, which is therefore controlled by the DAC 62 converter.

With regard to the phase tracking stage 52, it has an input, connected to the second output of the calculation stage 34, so as to receive the values of the time difference Δφ <j>; furthermore, the phase tracking stage 52 has a first and a second output. The first output is connected to the aforementioned second input of the timer 23, while the second output is connected to the second input of the adder 60, for reasons which will be described below.

In greater detail, the phase tracking stage 52 operates in the following manner.

For each observation time window, the phase tracking stage 52 compares the corresponding time difference Δφ <j> with a threshold Δφ_th.

If the time difference is higher, in absolute value, than the threshold, the phase tracking stage 52 generates on its first output a value equal to the time difference without modifying its second output. In practice, this value contributes to controlling the timer 23 of the slave node 6, together with the oscillator 21, as described below.

If the time difference is lower, in absolute value, to the threshold, the phase tracking stage 52 generates on its second output a value equal to the time difference without modifying its first output.

For practical purposes, the frequency control branch 40 forms, together with the protocol stack 14, the statistical processing stage 30, the calculation stage 34, the DAC converter 62, the oscillator 21 and the timer 23, a latching loop. frequency which operates in such a way as to minimize the difference between the frequencies of the oscillators of the master node 4 and of the slave node 6. The input of this frequency locked loop is precisely defined by the difference between the frequencies of the oscillators of the master node 4 and of the slave node 6. Furthermore, purely by way of example, the frequency locked loop can operate at a sampling frequency for example equal to 200 mHz and can have a cut-off frequency of the order of millihertz.

The phase control branch 42 instead carries out an open loop control of the timer 23 of the slave node 6. In greater detail, as previously mentioned, the timer 23 periodically updates its local time, on the basis of the local periodic signal supplied by the oscillator. 21 on the first input of the timer 23; in other words, the timer 23 updates its local time synchronously with the oscillator 21, in a manner known per se. Furthermore, whenever the phase tracking stage 52 generates a value on its first output, the timer 23 updates its local time, asynchronously with respect to the oscillator 21. For example, the timer 23 adds (algebraically) to its own time local the value

In practice, by means of the aforementioned closed-loop control, the slave node 6 tends to make the frequency of its own oscillator 21 equal to the frequency of the oscillator 20 of the master node 4. Furthermore, by means of the aforementioned open-loop control, the slave node 6 tends to cancel phase errors between itself and the master node 4, therefore it tends to cancel the difference between the absolute values, i.e. between the local times, of the timer 22 of the master node 4 and of the timer 23 of the slave node 6. The local time of slave node 6 is then linked to the local time of master node 4.

As regards the connection between the second output of the phase tracking stage 52 and the second input of the adder 60, it is optional. This connection allows to implement, inside the slave node 6, a further control of the phase of the timer 23 and in particular allows to implement a closed-loop control loop. In fact, as previously mentioned, for small phase differences, the phase tracking stage 52 does not update its first output, and therefore does not cause any direct update of the timer 23; on the contrary, the phase tracking stage 52 generates a new value which involves an indirect update of the local time, this update being mediated in particular by the integrator 50. In other words, small phase deviations between the master node 4 and the slave node 6 are slowly canceled, by generating an additional frequency offset, which makes it possible to recover the phase shift

In general, the differentiator 48 allows to follow relatively fast frequency variations, which can occur between the master node 4 and the slave node 6. Furthermore, the integrator 50 allows the frequency deviations to be integrated over time, so as to provide input to the oscillator 21 is a signal such that the frequency of the oscillator 21 is effectively equal to the frequency of the oscillator 20 of the master node 4.

Again with reference to the amplifier 46, as previously mentioned, its gain is proportional to the frequency deviations.Therefore, high frequency deviations correspond to high gain values, so that the frequency locked loop has a band wide, and therefore has a greater speed in following the frequency. On the contrary, reduced frequency deviations (for example, lower than a limit threshold) correspond to reduced values of the gain, so that the frequency locked loop has a limited band, and therefore is characterized by a greater capacity of filter jitter. However, it is still possible to use a constant gain amplifier, instead of the amplifier 46.

According to a different embodiment, the phase tracking stage 52 can perform, in a per se known manner, a moving average filter operation on the time deviations Δφ <j>. In particular, the tracking stage of phase 52 can calculate, for each observation time window, a corresponding weighted time difference equal to a weighted sum of the time difference relating to the observation time window, and one or more time differences relating to previous observation time windows. In this case, the operations described above for comparing with the threshold and generating the values are modified accordingly, by using the weighted time difference instead of the time difference.

According to a still different embodiment, the statistical processing block 30 is absent, in which case the quantities are respectively equal to the samples of the delay from master to slave and to the samples of the delay from slave to master In other words, this is equivalent to assuming an observation time window of such duration that, during the same observation time window, a single Sync message, a single follow-up message, a single delay request message and a single delay response message are transmitted. In other words, this means that, for each observation time window, only one sample of the master-to-slave delay and only one sample of the slave-to-master delay are available For practical purposes, in this embodiment the time window of observation is therefore absent, in the sense that statistical operations are not performed on the samples of the delays. It follows that, given two samples of the delay from master to slave and one sample of the delay from slave to master, the frequency difference Af <3> is proportional to the difference between the two samples of the delay from master to slave while the time difference Acp < 3> is proportional to the difference between the sample of the delay from slave to master and one between the two samples of the delay from master to slave

According to a different embodiment, also conforming to the IEEE 1588-2008 standard, the transmissions of the first and second Sync message by the master node 4 are not respectively followed by the transmissions of the first and second follow-up messages. In particular, according to this embodiment, the master node 4 does not transmit any follow-up message. In this case, the transmission times of the first and second Sync message are still detected by the detection unit 24 of the master node 24, but, instead of being communicated by the latter to the processing unit 10 for subsequent insertion into the corresponding follow-up messages are inserted by the detection unit 10 within the same first and second Sync message. This occurs in a per se known manner and allows to reduce the number of messages transmitted by the master node 4.

According to a different embodiment, it is also possible that in order to determine the magnitude

and the size relative to a given window

observation time, the statistical processing stage 30 discards the delay samples from master to slave and the delay samples from slave to master

deemed unreliable. For example, the stage of

statistical processing 30 can discard the samples of the delay from master to slave which deviate for more than a time limit (for example, equal to 5μμ or 10ps) from the magnitude relating to the previous observation time window; similarly, the statistical processing stage 30 can discard the samples of the delay from slave to master which deviate for more than this limit time from the magnitude relating to the previous observation time window. In this way, it is probable that the rejected samples relate to moments in which the network 1 has undergone appreciable changes, such as for example load changes ("load"), or changes due to path modifications.

It should also be noted that, for practical purposes, the previous embodiments provide that each frequency offset Af <3> is calculated on the basis of at least two delays from master to slave and that each time offset is calculated on the basis of at least one delay from master to slave and a slave to master delay

However, embodiments are possible in which

each frequency offset is calculated on the basis of at least two slave-to-master delays each of the two slave-to-master delays is determined as described with reference to Figure 4.

In practice, assuming for simplicity not to make use of observation time windows, the processing unit 10 of the slave node 6 can calculate a frequency deviation and a time deviation also having, as previously mentioned, only two delays from master to slave and one slave-to-master delay, or one master-to-slave delay and two slave-to-master delays. In the first case, as previously mentioned, the frequency difference is a function of the two delays from master to slave, and the time difference is a function of the delay from slave to master and one of the two delays from master to slave; in the second case, the frequency difference is a function of the two delays from slave to master, and the time difference is a function of the delay from master to slave and one of the two delays from slave to master.

In this regard, and purely by way of example, Figure 5 shows, in addition to the transmissions and receptions of the first Sync message, the first follow-up message, the first delay request message and the first delay response message, the transmissions and receptions of further messages, for the purpose of determining a second delay from slave to master.

In detail, Figure 5 shows the transmission, at an instant t5e by the slave node 6, of a second delay request message, which is received by the master node 4 at a time t6. Upon receipt of the second delay request message, the master node 4 transmits a second delay response message, at a time t6 <*>. The second delay response message is received by the slave node 6 at a time t5 <*>. Upon receipt of the second delay response message, the processing unit 10 of the slave node 6 determines the second delay from slave to master, which is equal to t6-t5.

The advantages that the present synchronization method allows to obtain emerge clearly from the preceding description. In particular, it allows to synchronize the repeaters (nodes) of a Simulcast isofrequency network, whose nodes are interconnected through a packet network, without using external synchronization sources. Even more particularly, thanks to the use of a frequency locked loop, it is possible to filter the jitter introduced by the packet network, and at the same time promptly track any drifts of the master node's oscillator. In this regard, it should also be noted that the frequency locked loop is of a lower order than a corresponding phase locked loop, and therefore is characterized by greater stability, with the same band.

As regards, on the other hand, the statistical processing stage 30 it allows to filter the contribution caused by samples of the delay from master to slave and by samples of the delay from slave to master affected by jitter, limiting the negative effects on the achievement of the frequency and phase.

Finally, it is evident that modifications and variations can be made to the present synchronization method, without thereby departing from the scope of the present invention, defined by the attached claims.

For example, one or more components of the processing units of the master node and / or of one or more of the slave nodes can be implemented by means of dedicated hardware, rather than by means of the described software implementation.

Furthermore, the shunt 48 and / or the integrator 50 can be absent, or they can be replaced by corresponding filters; also the amplifier 46 can be absent.

With regard to the Sync messages and the follow-up messages, it is also possible that, in a per se known manner, they contain respective sequence numbers, suitable for allowing the slave node to determine, for each follow-up message, to which Sync message matches. Furthermore, each Sync message can contain an indicator suitable to notify the slave node of the subsequent sending of the corresponding follow-up message.

Finally, to determine the delays from master to slave and the delays from slave to master and therefore in order to determine the time and frequency deviations, the master node 4 and the slave node 6 can implement a different protocol than the PTP protocol, such as the NTP protocol.

Claims (15)

CLAIMS 1. Method of synchronizing a master node (4) and a slave node (6) of a packet network (1), the master node comprising a first timer (22) controlled by a first oscillator (20) so as to updating a first local time ("locai time") with a first frequency, the slave node comprising a second timer (23) controlled by a second oscillator (21) so as to update a second local time with a second frequency, said method comprising the steps of determining at least a first delay from master to slave, at least a first delay from slave to master and at least one further delay between a second delay from master to slave and a second delay from slave to master; said method comprising, to determine each delay from master to slave, the steps of: - transmit, by the master node, a packet from master to slave (Sync) in a respective transmission instant, referred to the first local time; - receiving, by the slave node, the packet from master to slave in a respective receiving instant , referring to the second local time; - communicating to the slave node, by the master node, said respective instant of transmission; And - calculating, by the slave node, said delay from master to slave on the basis of the respective instants of transmission and reception of the packet from master to slave; said method further comprising, to determine each delay from slave to master, the steps of: - transmit, by the slave node, a packet from slave to master (DELAY REQUEST) in a respective transmission instant (t3; t5), referred to the second local time; - receiving, by the master node, the packet from slave to master in a respective reception instant (t4; t6), referred to the first local time; - communicating to the slave node, by the master node, said respective instant of reception; And - calculating, by the slave node, said delay from slave to master, on the basis of the respective instants of transmission and reception of the packet from slave to master; said method further comprising, on the part of the slave node: - determining a frequency difference indicative of the difference between the first and second frequency, on the basis of at least two delays from master to slave, or on the basis of at least two delays from slave to master; And - determining a time difference indicative of the difference between the first and the second local time, on the basis of at least one delay from master to slave and of at least one delay from slave to master; said method also comprising the steps of: - controlling the second oscillator as a function of the frequency difference, so as to lock the second oscillator in frequency to the first oscillator; And - updating the second local time on the basis of the time difference, asynchronously with respect to the second oscillator and in such a way as to lock in phase the second timer to the first timer. 2. Synchronization method according to claim 1, wherein the slave node (6) comprises a slave interface (16) towards the network (1), of the "media independent" type, and the master node (4 ) comprises a master interface (16) towards the network, independent of the medium; and in which, for each packet from master to slave (Sync): - the respective transmission instant refers to the transit of said packet from master to slave through the master interface (16); And - the respective receiving instant refers to the transit of said packet from master to slave through the slave interface (16); and in which, for each packet from slave to master (DELAY REQUEST): - the respective transmission instant (t3; t5) refers to the transit of said packet from slave to master through the slave interface (16); And - the respective reception instant (t4; t6) refers to the transit of said packet from slave to master through the master interface (16). 3. Synchronization method according to claim 1 or 2, further comprising the step of controlling the second oscillator (21) as a function also of the time difference (Δφ <j>), so as to cause corresponding variations of the second local time, synchronous with respect to the second oscillator. Synchronization method according to claim 3, further comprising the steps of, by the slave node (6): - compare the time difference (Δφ <j>) with a threshold; And - if the time difference is greater, in absolute value, than the threshold, algebraically add said time difference to the second local time; - if the time difference is lower, in absolute value, to the threshold, control the second oscillator so as to vary the second frequency as a function of the frequency difference (Δf <j>) and the time difference (Δφ <j>). Synchronization method according to any one of the preceding claims, wherein said step of controlling the second oscillator (21) as a function of the frequency deviation comprises numerically filtering the frequency deviation with a low-pass type digital filter (44) of the second order, and control the second oscillator on the basis of the filtered waste. 6. Synchronization method according to any one of the preceding claims, wherein each delay from master to slave is proportional to the difference between the receiving and transmitting instants of the corresponding master-to-slave packet, e each delay from slave to master (As-to-M <1>) is proportional to the difference between the reception and transmission instants (t4, t6; t3, t5) of the corresponding packet from slave to master. The synchronization method according to claim 6, wherein the frequency offset is proportional to the difference between two of the determined master-to-slave delays, or is proportional to the difference between two of the determined slave-to-master delays; and in which the time difference is proportional to the difference between one of the determined master-to-slave delays and one of the determined slave-to-master delays. 8. Synchronization method according to any one of claims 1 to 6, further comprising the steps of: define a succession of time windows; determining a plurality of delays from master to slave (Aw-to-s <1>), said step of determining a plurality of delays from master to slave comprising transmitting by the master node, during a first time window, a corresponding plurality of packets from master to slave (Sync); - determining a plurality of delays from slave to master, said step of determining a plurality of delays from slave to master comprising transmitting by the slave node, during the first time window, a corresponding plurality of packets from slave to master (DELAY REQUEST ); said method further comprising the steps of, by the slave node: - determining at least a first statistical quantity (G_MS <j>) between: the minimum value, the average value and the mode of said plurality of delays from master to slave; - determining at least a second statistical quantity (G_SM <j>) between: the minimum value, the average value and the mode of said plurality of delays from slave to master; And - selecting a packet alternatively between said plurality of packets from master to slave or between said plurality of packets from slave to master, so that the delay calculated on the basis of the transmission and reception instants of the selected packet alternately corresponds to the first, or to the second statistical quantity; and in which said step of determining a time difference (Δφ <j>) comprises calculating the time difference so that it is proportional to the difference between said first and second statistical quantities; and in which said step of determining a frequency deviation (Δf <j>) comprises calculating the frequency deviation on the basis of the transmission and reception instants of the selected packet and, if said packet has been selected among the packets from master to slave , of the instants of transmission and reception of a packet from master to slave transmitted during a previous time window, or, if said packet has been selected among the packets from slave to master, of the instants of transmission and reception of a packet from slave a master transmitted during the previous time window. 9. Slave node for a packet network (1) including a master node (4) comprising a first oscillator (20) and a first timer (22) controlled by the first oscillator so as to update a first local time with a first frequency, said slave node comprising: - a second oscillator (21) and a second timer (23) controlled by the second oscillator so as to update a second local time with a second frequency; - first receiving means (10,14,16,18) configured to receive one or more packets from master to slave (Sync) in respective reception instants referred to the second local time, said packets from master to slave being transmitted by the master node in respective transmission instants referred to the first local time; - first learning means (10,14) configured to learn from the master node the instants of transmission of the packets from master to slave; - first processing means (10,14) configured to determine, on the basis of the transmission and reception instants of the packets from master to slave, corresponding delays from master to slave; - first transmission means (10,14,16,18) configured to transmit to the master node one or more packets from slave to master (DELAY REQUEST) in respective transmission instants (t3, t5) referred to the second local time; - second learning means (10,14) configured to learn from the master node the reception instants (, t6) in which said one or more packets from slave to master are received by the master node; second processing means (10,14) configured to determine, on the basis of the transmission and reception instants of the packets from slave to master, corresponding delays from slave to master; - third processing means (10,14) configured to determine a frequency difference indicative of the difference between the first and second frequency, on the basis of at least two delays from master to slave, or on the basis of at least two delays from slave to master ; - fourth processing means (10,14) configured to determine a time difference indicative of the difference between the first and second local time, on the basis of at least one delay from master to slave and of at least one delay from slave to master; - first control means (30,34,44,46,48,50,60,62) configured to control the second oscillator as a function of the frequency deviation, so as to lock the second oscillator in frequency to the first oscillator; and - updating means (30,34,52) configured to update the second local time on the basis of the time difference, asynchronously with respect to the second oscillator and in such a way as to lock in phase the second timer to the first timer. 10. Slave node according to claim 9, further comprising a slave interface (16) towards the network (1), of the independent type from the medium, and in which said first receiving means (10,14,16,18) are further configured in such a way that the instants of reception of said one or more packets from master to slave (Sync) refer to corresponding transits of said one or more packets from master to slave through the slave interface; and in which said first transmission means (10,14,16,18) are further configured in such a way that the transmission instants (t3, t5) of said one or more packets from slave to master (DELAY REQUEST) they refer to corresponding transits of said one or more packets from slave to master through the slave interface. 11. Slave node according to claim 9 or 10, further comprising second control means (30,34,50,52,60,62) configured to control the second oscillator (21) also according to the time difference (Δφ <j> ), so as to cause corresponding variations of the second local time, synchronous with respect to the second oscillator. Slave node according to claim 11, further comprising comparison means (52) configured to compare the time difference (Δφ <j>) with a threshold; and in which the updating means (30,34,52) are configured to add algebraically if the time difference is greater than the threshold in absolute value, said time difference to the second local time; and in which the first and second control means (30,34,44,46,48,52,60,50,62) are configured to vary, if the time difference is lower than the threshold in absolute value, the second frequency of the second oscillator (21) on the basis of the frequency deviation and the time deviation. 13. Slave node according to any one of claims 9 to 12, further comprising: - first statistical means (10,30) configured to determine at least one first statistical quantity between: the minimum value, the average value and the mode of a plurality of delays from master to slave formed by the delays from master to slave determined by the first processing means (10,14) on the basis of the instants of transmission and reception of packets from master to slave (Sync) transmitted during a first time window; - second statistical means (10,30) configured to determine at least a second statistical quantity between: the minimum value, the average value and the mode of a plurality of delays from slave to master formed by the delays from slave to master determined by the second processing means (10,14) on the basis of the instants of transmission and reception of packets from slave to master (DELAY REQUEST) transmitted during said first time window; And - selection means (30) configured to select a packet alternately between said plurality of packets from master to slave or between said plurality of packets from slave to master, in such a way that the delay determined on the basis of the instants of transmission and reception of the selected packet corresponds alternatively to the first, or to the second statistical quantity; and in which said fourth processing means (10,14) are configured in such a way that said time difference (Δφ <j>) is proportional to the difference between said first and second statistical quantity; and in which said third processing means (10,14) are configured in such a way that said frequency difference (Δf <j>) is a function of the transmission and reception instants of the selected packet and, if the selected packet is a packet from master to slave, of the instants of transmission and reception of a packet from master to slave transmitted during a previous time window, or, if said packet selected is a packet from slave to master, of the instants of transmission and reception of a packet from slave to master transmitted during said previous time window. Packet network comprising the slave node (6) according to any one of claims 9 to 13 and said master node (4), and wherein the master node further comprises: - second transmission means (10,14,16,18) configured to transmit said one or more packets from master to slave (Sync); - first communication means (10,14) configured to communicate the transmission instants to the slave node of said one or more packets from master to slave; - second receiving means (10,14,16,18) configured to receive said one or more packets from slave to master (DELAY REQUEST); And - second communication means (10,14) configured to communicate to the slave node the receiving instants (14,16) of said one or more packets from slave to master. Network according to claim 14, wherein said master node (4) comprises a master interface (16) towards the network (1), of the media-independent type, and wherein said second transmission means (10,14,16 , 18) are configured in such a way that the transmission instants of said one or more packets from master to slave (Sync) refer to corresponding transits of said one or more packets from master to slave through the master interface; and in which said second reception means (10,14) are configured in such a way that the reception instants (t4; t6) of said one or more packets from slave to master (DELAY REQUEST) refer to corresponding transits of dictate one or more slave-to-master packets through the master interface.