JP4467478B2 - Transmission apparatus and time synchronization method - Google Patents

Description

  The present invention relates to a transmission apparatus and a time synchronization method, and more particularly, to a transmission apparatus and a time synchronization method that perform time synchronization.

  Although it has been conventionally performed to synchronize the time between a plurality of terminals connected to the network, it can be divided into three cases depending on the nature of the network.

  The first is a case where there is a medium for transmitting time information exclusively. For example, there are methods such as using a radio clock, using GPS, supplying a common clock using an ISDN line, and synchronizing with a commercial power source.

  The second case is a case where there is no medium for transmitting time information exclusively, and the time information can be received with a definite delay time. For example, the IEEE 802.11 standard wireless LAN has a function for time synchronization called TSF (Timing Synchronization Function).

  In this method, an AP (access point) on the wireless LAN serves as a timing master, and broadcast transmission is performed including time information in a beacon signal periodically transmitted. Each STA (terminal) receives a beacon signal and adjusts its own timer based on time information included therein. The delay time from when the AP transmits the beacon signal to when the STA receives the beacon signal is deterministic.

  The third is a case where there is no dedicated medium for transmitting time information and the delay time until the time information is received is not deterministic. The technology for adjusting the time in such a case is described in David L. Mills, “Network Time Protocol (Version 3) Specification, Implementation and Analysis”, [online], IETF RFC-1305, Mar. 1992, [June 2005]. 29 days search], Internet <URL http://www.faqs.org/rfcs/rfc1305.html> (Non-Patent Document 1), David L. Mills, “Simple Network Time Protocol (SNTP) Version 4 for IPv4, IPv6 and OSI ”, [online], IETF RFC-2030, Oct. 1996, [searched on June 29, 2005], Internet <URL http://www.faqs.org/rfcs/rfc2030.html> (non-patent Reference 2), David L. Mills, “Internet time synchronization: the Network Time Protocol”, IEEE Trans. Communications COM-39, 10 (October 1991), 1482-1493. [Search June 29, 2005], Internet <URL http://www.eecis.udel.edu/~mills/database/papers/trans.pdf> (Non-Patent Document 3) and JP-A-10-303987 (Patent Document 1) The

  David L. Mills, “Network Time Protocol (Version 3) Specification, Implementation and Analysis”, [online], IETF RFC-1305, Mar. 1992, [Search June 29, 2005], Internet <URL http: / /www.faqs.org/rfcs/rfc1305.html> (Non-Patent Document 1), and David L. Mills, “Simple Network Time Protocol (SNTP) Version 4 for IPv4, IPv6 and OSI”, [online], IETF RFC-2030, Oct. 1996, [Searched on June 29, 2005], Internet <URL http://www.faqs.org/rfcs/rfc2030.html> (Non-Patent Document 2) is an NTP ( This is a specification of Network Time Protocol) and SNTP (Simple NTP).

  NTP is a general protocol for adjusting time on an IP network. NTP refers to RFC (Request For Comment) -1305 or a simplified version of SNTP (Simple NTP) RFC-2030.

NTP is widely used for setting the time on the Internet.
FIG. 18 is a diagram for explaining a time adjustment method using NTP. Referring to FIG. 18, in the time adjustment method using NTP, processing is performed in the following order (1), (2), and (3). C ₁ and C ₂ indicate time stamps on the client side. S ₁ and S ₂ indicate time stamps on the server side.
(1) The client transmits a packet to the server (time C ₁ ).
(2) The server receives the packet from the client (time S ₁ ), and sends it back with the source address as the destination (time S ₂ ).
(3) The client receives a packet from the server (time C ₂ ).
Each packet in the processes (1), (2), and (3) includes a time stamp of C ₁ , S ₁ , and S ₂ .

The round trip time (round trip transmission delay time) δi is obtained by the following equations (101) to (103).
a = S ₁ -C ₁ (101)
b = S ₂ -C ₂ (102)
δi = a−b (103)
The clock offset (time shift) θi with respect to the server of the client is obtained by the following equation (104).
θi = (a + b) / 2 (104)
When the true clock offset is θ, the true clock offset θ can be evaluated by the following equation (105) (see Non-Patent Document 3). Therefore, the error of the clock offset θi does not become longer than the round trip time δi.
θi−δi / 2 ≦ θ ≦ θi + δi / 2 (105)
This means that if data having a large round trip time is used, the error of the clock offset θi may increase.

  For this reason, in NTP, in one correction, a packet may be transmitted and received a plurality of times and used as the clock offset θi using data when the round trip time δi is minimum. This is called a data filter.

  Further, in NTP, an autonomous local time may be generated by applying a clock filter θ to a loop filter and operating a PLL (Phase Locked Loop) based on the result.

  Since an error occurs only with one clock offset information, the error is smoothed using a loop filter.

  Further, if the local time value is corrected each time the clock offset θ is obtained, the local time may be skipped back and forth, but the local time can be continuously adjusted by forming a PLL. An NTP combining a data filter and a PLL is shown in FIG. Here, the contents of the loop filter are an example.

The technique disclosed in Japanese Patent Laid-Open No. 10-303987 (Patent Document 1) also repeats transmission / reception of a packet including a time stamp between two nodes, similarly to NTP. Specifically, the clock offset θ is calculated a plurality of times, and it is determined whether or not the result has a normal distribution. If it is determined that the distribution is not close to the normal distribution, statistical processing is performed by converting the distribution to a distribution close to the normal distribution by variable conversion. Thereby, since the statistical property of the clock offset θ can be grasped, it is possible to know how accurate the time synchronization is.
David L. Mills, “Network Time Protocol (Version 3) Specification, Implementation and Analysis”, [online], IETF RFC-1305, Mar. 1992, [Search June 29, 2005], Internet <URL http: / /www.faqs.org/rfcs/rfc1305.html> David L. Mills, “Simple Network Time Protocol (SNTP) Version 4 for IPv4, IPv6 and OSI”, [online], IETF RFC-2030, Oct. 1996, [searched June 29, 2005], Internet <URL http://www.faqs.org/rfcs/rfc2030.html> David L. Mills, “Internet time synchronization: the Network Time Protocol”, IEEE Trans. Communications COM-39, 10 (October 1991), 1482-1493. [Search June 29, 2005], Internet <URL http: //www.eecis.udel.edu/~mills/database/papers/trans.pdf> JP-A-10-303987

  Problems to be solved by the present invention are as follows. In the present invention, there is no medium for transmitting time information exclusively, and the case where the delay time until the time information is received is not definite.

  In the present invention, a general IP network is assumed. The IP network includes a wired LAN, a wireless LAN, a modem, a router, a switching hub, an optical fiber, and the like. There is a case where it does not depend on a specific medium and may extend over a plurality of media. In addition, the existence of a communication network other than the data communication network is not assumed.

  In such a configuration, the transmission delay time varies due to a change in the network state or a change in the state inside the terminal.

  The change in the network state means that the traffic (packet amount) flowing on the network changes every moment. Traffic is affected by other terminals and running applications. Traffic is also affected by relay devices such as routers. When the traffic is congested, it may take a long time for the packet to pass through the relay device, or the packet may be lost. The traffic is also affected by the buffer amount of the terminal's own network I / F. Traffic is also affected by the communication status of the medium in the case of a wireless LAN or the like.

  The state change inside the terminal is a change in the CPU usage rate, interrupt response time, time required for switching tasks of the operating system, and the like. Variations in the processing time of packet transmission / reception occur due to a change in the state inside the terminal.

  Further, not only does the transmission delay time vary, but the nature of the transmission path may become asymmetric between the client and the server on the way back and forth.

  For example, when video transmission is performed in one direction, the amount of traffic generated is unbalanced between going and returning. Even if the return and return are the same network path and the same transmission speed, the characteristics of the transmission path become asymmetric due to the deviation in the amount of traffic generated. In this case, the average value of the transmission delay time for going and returning is different.

  When the NTP obtains the clock offset θ, basically, information of one packet transmission / reception is used. NTP assumes that the transmission delay time is the same on the return and return when obtaining the clock offset θ, but in reality, the transmission delay time varies. In addition, the transmission delay time cannot be expected to be the same between the going and the returning because the properties of the transmission path may become asymmetrical between the going and the returning.

  Therefore, the clock offset θi calculated from the information of only one packet transmission / reception has a considerable error compared to the true clock offset θ.

  Therefore, in NTP, the data filter is performed, or the error is smoothed by a loop filter, so that the influence of the error of the clock offset θi per packet transmission / reception is minimized.

However, NTP is considered to have the following two problems.
(1) When video transmission is performed in one direction, the accuracy deteriorates.
(2) It takes time until the PLL operates stably.
A detailed examination of these issues is as follows.
(1) When video transmission is performed only in one direction, the amount of traffic generated is biased between going and returning. It is known that the average transmission delay time increases when the amount of traffic is large. The average transmission delay time in the direction with a large amount of traffic becomes larger than the average transmission delay time in the reverse direction.

Even if a data filter is used for NTP, the clock offset is basically estimated by using information of one packet transmission / reception, so that the error of the clock offset tends to increase during video transmission. In this case, even if the error is smoothed by the loop filter, it is difficult to reduce the influence of the error.
(2) A PLL is required to set the time accurately with NTP, but in order to suppress the error small, the PLL parameters (KP and KI in FIG. 19) must be reduced. However, if the PLL parameter is decreased, it takes time until the PLL operates stably. There is a trade-off relationship between suppressing the error small and the time until the PLL operates stably.

  FIG. 20 is a diagram illustrating a configuration when video transmission is performed from the client to the server. Referring to FIG. 20, the direction from the client to the server is “going”. The direction from the server to the client is “return”. The measurement packet is reciprocated between the client and the server, and the round trip time is separated into a transmission delay time for the outgoing route and a transmission delay time for the return route. The results of measuring the transmission and return transmission delay times in the case of video transmission from the client to the server with the configuration of FIG. 20 are shown below.

FIG. 21 is a diagram illustrating a change in the outbound transmission delay time as time elapses.
FIG. 22 is a diagram illustrating a change in the return transmission delay time with time.

FIG. 23 is a diagram showing the average transmission delay time for each of going and returning.
From the results of FIGS. 21, 22 and 23, it can be seen that the average transmission delay time in the forward direction increases during video transmission, and the average transmission delay time in the return direction does not change much. During video transmission, the average transmission delay time increases only in the direction of travel, and the symmetry of the transmission path is lost on the way back and forth. In such a case, the accuracy of NTP deteriorates.

  In the technique disclosed in Japanese Patent Laid-Open No. 10-303987, the local time is adjusted by looking at the statistical property of the clock offset θ. However, as described in (1), when video transmission is performed only in one direction, since the symmetry of the transmission path is lost on the way back and forth, the error of the clock offset θi increases. In the meantime, the statistical properties of the clock offset change (the average value shifts), so it is considered that this method cannot cope well.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a transmission apparatus capable of accurately adjusting the time even when one-way video transmission is performed. It is to be.

  Another object of the present invention is to provide a time synchronization method capable of reducing the time until the error is suppressed and performing time adjustment with high accuracy.

  In order to solve the above-described problem, according to one aspect of the present invention, a transmission device connected to a network and operating at a first frequency and performing data communication with a device operating at a second frequency manages time. A time management unit, and a transmission unit that sequentially transmits the measurement packet to the apparatus. The apparatus receives the information about the first time when the measurement packet is received and the measurement packet as a reply packet each time the measurement packet is received. The reply packet including the second time information to be returned is sequentially returned, and the first and second time information included in the reply packet and the third time at which the measurement packet corresponding to the received reply packet is transmitted. And a first frequency calculating unit for sequentially calculating a time offset with respect to the apparatus based on time information composed of the information on the second time and information on the fourth time when the reply packet is received, And a correction means for correcting the time managed by the time management means based on a frequency ratio indicating a ratio with the second frequency, a time offset, and an arbitrary reference time managed by the time management means. .

Preferably, when the frequency ratio is α, the time before correction is C, the time to be an arbitrary reference is C ₀ , and the time offset is β, the correction means has F (C) = α × (C Based on the time F (C) calculated by −C ₀ ) + β, the management time is corrected.

Preferably, when the frequency ratio is α and the time offset is β, the calculation means calculates a minimum value αmin and a maximum value αmax of values that α can take based on a plurality of time information, and sets the plurality of time information. The minimum value βmin and the maximum value βmax of the possible values of β are calculated based on
The calculation means sets one value in a range of values that α can take as α as a frequency ratio, and sets one value in a range of values β can take as β as a time offset.

Preferably, the frequency ratio is α, the arbitrary reference time is C ₀ , the first minimum transmission delay time that is the minimum delay time that occurs when data is transmitted to the device, and the data is transmitted from the device to the transmission device Ε is the second minimum transmission delay time, which is the minimum delay time, and C _{1 is} the time when the transmission means transmits the n (natural number) measurement packet, and the time when the device receives the nth measurement packet. If S ₁ , the time when the device returned the nth reply packet is S ₂ , and the time when the transmission device received the nth reply packet is C ₂ , the calculating means is based on the nth time information. The minimum value αmin of α can be calculated by (S ₂ −β + ε) / (C ₂ −C ₀ ), and the maximum value αmax of α can be calculated based on the nth time information as (S ₁₋ β−ε) / (C ₁ −C ₀ The calculation means calculates the maximum value that αmin can take based on a plurality of pieces of time information, and calculates the minimum value that αmax can take.

Preferably, the frequency ratio is α, the time offset is β, the time when the transmission means transmits the n (natural number) measurement packet is C ₁ , the time when the device receives the nth measurement packet is S ₁ , and the device is n The time when the second reply packet is returned is S ₂ , the time when the transmission apparatus receives the nth reply packet is C ₂ , and the time indicating the delay time that occurs when data is reciprocated between the transmission apparatus and the apparatus. In the case of RTT, the calculating unit calculates the RTT based on the nth time information, the calculating unit calculates the minimum value minRTT that the RTT can take based on the plurality of time information, and the calculating unit The minimum and maximum values for changing β are the predetermined values S00 and minRTT, respectively, and the maximum value of αmin that can be taken while changing β within the range of changing β. Calculating the minimum value of the possible values of fine .alpha.max, based on a range of beta alpha can be present, to calculate a minimum βmin and maximum value βmax of possible values of beta.

  Preferably, when the time offset is β, the calculation unit sets β as a time offset to a value calculated by an arithmetic average of the minimum value βmin and the maximum value βmax of β.

  Preferably, when the frequency ratio is α, the calculation means sets a value calculated by arithmetic average of the minimum value αmin and the maximum value αmax of α as α as the frequency ratio.

  Preferably, the device has a function of playing back content, and the same content is played back simultaneously between the device and the transmission device.

  Preferably, the calculating means calculates at least one of a first transmission delay time that occurs when data is transmitted to the apparatus and a second transmission delay time that occurs when data is transmitted from the apparatus to the transmission apparatus.

Preferably, the frequency ratio is α, the time offset is β, the arbitrary reference time is C ₀ , the time when the transmission unit transmits the n (natural number) measurement packet is C ₁ , and the device receives the nth measurement packet. When the received time is S ₁ , the time when the device returns the nth reply packet is S ₂ , and the time when the transmission device receives the nth reply packet is C ₂ , the calculation means calculates the nth time. Based on the information, the first transmission delay time is calculated by S ₁ − {α × (C ₁ −C ₀ ) + β}, and the second transmission delay time is {α × (C ₂ −C ₀ ) + β}. It is calculated by the -S _2.

  According to another aspect of the present invention, a time synchronization method for synchronizing between a device operating at a first frequency connected to a network and a transmission device operating at a second frequency for data communication, A step of sequentially transmitting measurement packets to the device, and each time the measurement packet is received, the device transmits information on the first time when the measurement packet is received and a second time when the measurement packet is returned as a reply packet. The reply packet including the information is sequentially returned, and the first and second time information included in the reply packet and the third time information and the reply packet transmitted the measurement packet corresponding to the received reply packet A step of sequentially calculating a time offset with the apparatus based on time information composed of the received fourth time information, and a frequency indicating a ratio between the first frequency and the second frequency. The ratio and the time offset, based on the time at which any of the criteria process to manage to manage time, further comprising a step of correcting the time managed process that manages the time.

Preferably, when the frequency ratio is α, the pre-correction time is C, the arbitrary reference time is C ₀ , and the time offset is β, the correction step is F (C) = α × Based on the time F (C) calculated by (C−C ₀ ) + β, the management time is corrected.

  Preferably, when the frequency ratio is α and the time offset is β, the calculating step calculates a minimum value αmin and a maximum value αmax of α based on a plurality of time information, and a plurality of time information The minimum value βmin and the maximum value βmax of the values that β can take are calculated based on the above, and in the calculating step, one value in the range of values that α can take is set as α as the frequency ratio, and β can take One value in the range of values is β as a time offset.

Preferably, the frequency ratio is α, the arbitrary reference time is C ₀ , the first minimum transmission delay time that is the minimum delay time that occurs when data is transmitted to the device, and the data is transmitted from the device to the transmission device Ε is the second minimum transmission delay time, which is the minimum delay time, and C _{1 is} the time when the transmitting step transmits the n (natural number) measurement packet, and the time when the device receives the nth measurement packet. S ₁ , the time when the device returned the nth reply packet as S ₂ , and the time when the transmission device received the nth reply packet as C ₂ , Based on (S ₂ −β + ε) / (C ₂ −C ₀ ), a minimum value αmin of values that α can take is calculated based on the nth time information. _{(S 1 -β-ε) /} (C Calculated by -C _0), the step of calculating, based on the plurality of time information, calculates the maximum value of the possible values of .alpha.min, it calculates the minimum value of the possible values of .alpha.max.

Preferably, the frequency ratio is α, the time offset is β, the transmission step is C ₁ when the n (natural number) measurement packet is transmitted, the time when the device receives the nth measurement packet is S ₁ , and the device is The time at which the n-th reply packet is returned is S ₂ , the time at which the transmission apparatus receives the n-th reply packet is C ₂ , and the time indicating the delay time that occurs when data is reciprocated between the transmission apparatus and the apparatus , RTT is calculated based on the n-th time information, and the calculating step calculates a minimum value minRTT of possible values of the RTT based on a plurality of time information, In the calculating step, the minimum value and the maximum value of the value for changing β are set to a predetermined value S00 and minRTT, respectively, and while changing β within the range of changing β, That the maximum value of the values and to calculate the minimum value of the possible values of .alpha.max, based on a range of beta alpha can be present, to calculate a minimum βmin and maximum value βmax of possible values of beta.

  Preferably, when the time offset is β, in the calculation step, a value calculated by an arithmetic average of the minimum value βmin and the maximum value βmax of β can be set as β as the time offset.

  Preferably, when the frequency ratio is α, in the calculating step, a value calculated by an arithmetic average of the minimum value αmin and the maximum value αmax of α can be set as α as the frequency ratio.

  The transmission apparatus according to the present invention includes information on the first time when the apparatus receives the measurement packet, information on the second time when the measurement packet is returned as a reply packet, information on the third time when the measurement packet is transmitted, and a reply packet. Based on the time information composed of the received fourth time information, the time based on the frequency ratio at which each of the device and the transmission device operates, the time offset, and the arbitrary time managed by the time management means The time managed by the management means is corrected.

  Therefore, there is an effect that the time until the error is suppressed can be reduced and the time can be accurately adjusted.

  The time synchronization method according to the present invention includes a first time information when the apparatus receives the measurement packet, a second time information when the measurement packet is returned as a reply packet, a third time information when the measurement packet is transmitted, and a reply packet. Based on the time information configured from the information on the fourth time at which the device and the transmission device are received, based on the frequency ratio at which each of the device and the transmission device operates, the time offset, and the arbitrary time managed by the time managing step Thus, the time managed by the process for managing the time is corrected.

  Therefore, there is an effect that the time until the error is suppressed can be reduced and the time can be accurately adjusted.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following description, the same parts are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed description thereof will not be repeated.

  FIG. 1 is a diagram showing an example system 1000 that constitutes the present embodiment. With reference to FIG. 1, a system 1000 includes a client 100, a network 110, and a server 120. The network 110 is a network such as the Internet, for example. The client 100 and the server 120 are connected to each other via the network 110.

  In this embodiment, it is assumed that the time of the client 100 (the value of the local counter) is set to the time of the server 120. Hereinafter, the value of the local counter is also referred to as a local counter value.

  The client 100 includes a CPU (Central Processing Unit) 101, a memory 102, a communication interface 103, an oscillator 104, a local counter 105, and a bus 109. The CPU 101, the memory 102, the communication interface 103, and the local counter 105 are connected to the bus 109. The local counter 105 may be incorporated in the CPU 101.

  The CPU 101 executes a program for causing the client 100 to operate as a communication device.

  The memory 102 stores a program executed by the CPU 101 and values necessary for the program. The communication interface 103 includes a transmission unit and a reception unit for communicating with other terminals connected to the network 110. In the local counter 105, the local counter value is incremented based on the oscillation frequency of the oscillator 104. The CPU 101 reads the local counter value of the local counter 105.

  The server 120 includes a CPU 121, a memory 122, a communication interface 123, an oscillator 124, a local counter 125, and a bus 129. The CPU 121, the memory 122, the communication interface 123, and the local counter 125 are connected to the bus 129. The local counter 125 may be incorporated in the CPU 121.

  The CPU 121 executes a program for causing the server 120 to operate as a communication device.

  The memory 122 stores programs executed by the CPU 121 and values necessary at that time. The communication interface 123 includes a transmission unit and a reception unit for communicating with other terminals connected to the network 110. In the local counter 125, the local counter value is incremented based on the oscillation frequency of the oscillator 124. Note that the local counter 125 may be synchronized with UTC (Coordinated Universal Time), which is Coordinated Universal Time, by a configuration not shown in FIG.

The CPU 121 reads the local counter value of the local counter 125.
Note that the internal configurations of the client 100 and the server 120 in the system 1000 of FIG. 1 are shown only as a minimum necessary configuration, and other components are not shown.

  The client 100 and the server 120 are, for example, general PCs (Personal Computers). In the case of a PC, the local counter uses an API (Application Program Interface) provided by an OS (Operating System). For example, when the OS is Microsoft Windows (registered trademark), QueryPerformanceCounter can be used as a counter.

  Further, when the OS is Linux (registered trademark), gettimeofday can be used as a counter. Since both of these accuracies are 1 μsec or less, they can be sufficiently used in consideration of the accuracy of target time adjustment. Note that the frequency of QueryPerformanceCounter varies depending on the hardware. The frequency can be acquired with QueryPerformanceFrequency.

  This embodiment can also be configured with a general PC. However, each of the client 100 and the server 120 may include dedicated hardware (circuit).

  The dedicated hardware (circuit) is, for example, a PCI (Peripheral Component Interconnect) standard hardware board. The hardware board is connected to each of the client 100 and the server 120 as a PC through an expansion bus such as a PCI (Peripheral Component Interconnect) bus.

  The dedicated hardware (circuit) is desirably loaded with a stable oscillator with high accuracy (having a resolution of at least 1 μsec or less). Dedicated hardware (circuit) provides a local counter based on this oscillator, and there may be a circuit for correcting how the local counter value advances.

  FIG. 2 is a diagram illustrating a configuration of a client 100A as an example on which dedicated hardware 108 is installed. Referring to FIG. 2, dedicated hardware 108 is accessed from CPU 101 through bus 109. The dedicated hardware 108 includes at least an oscillator 104, a local counter 105, a correction circuit 106, and a register 107.

  In the local counter 105, the local counter value is incremented based on the oscillation frequency of the oscillator 104. The correction circuit 106 corrects the local counter value based on the parameters set in the local counter 105 and the register 107. The correction circuit 106 provides the corrected local counter value to the CPU 101 and the like.

The register 107 stores parameters necessary for correction. Specifically, time C ₀ , clock frequency ratio α (estimated value), and time offset β (estimated value) at C = C ₀ are stored as arbitrary references. These parameters are calculated and set by the CPU 101 by the process described below. Since the technique taken by the present invention is vulnerable to clock frequency drift, it is desirable that the oscillator 104 and the oscillator 124 have stable oscillation frequencies. It is desirable to use a temperature-compensated crystal oscillator or a rubidium or cesium oscillator as the oscillator.

FIG. 3 shows the relationship between the local counter on the client 100 side and the local counter on the server 120 side. Referring to FIG. 3, client 100 transmits a measurement packet PK for time adjustment to server 120. A local counter value (C ₁ ) on the client 100 side at the time when the client 100 transmits is added to the measurement packet PK.

  Upon receiving the measurement packet PK, the server 120 creates a reply packet PKR with the transmission source address of the measurement packet PK as the destination address. Then, the server 120 sends back a reply packet PKR to the corresponding destination.

In addition to C ₁ , the reply packet PKR includes a local counter value (S ₁ ) on the server 120 side when the server 120 receives and a local counter value (S ₂ ) on the server 120 side when the server 120 transmits. Is added.

The client 100 receives the reply packet PKR from the server 120. The local counter value on the client 100 side at which the client 100 receives the reply packet PKR and _{C 2.}

The present invention obtains time information (C ₁ , S ₁ , S ₂ , C ₂ ) related to packet transmission between the client 100 and the server 120 by repeating the above protocol at an appropriate time interval. And the characteristic technique which can perform time adjustment between both using them is disclosed. The specific method will be described below.

First, the client 100 obtains N pieces of time information (C _1k , S _1k , S _2k , C _2k ) by transmitting N measurement packets PK to the server 120 (k = 0, 1). , ..., N-1). (C _1k , C _2k ) indicates a local counter value on the client 100 side. (S _1k , S _2k ) indicates a local counter value on the server 120 side.

Here, it is assumed that the local counter value of the client 100 can be corrected to the local counter value of the server 120 by the following equation (1).
F (C) = α × (C−C ₀ ) + β (1)
In Expression (1), F (C) represents the local counter value of the client 100 after correction. C indicates the local counter value of the client 100 before correction. α represents a clock frequency ratio between the oscillation frequency of the oscillator 124 of the server 120 and the oscillation frequency of the oscillator 104 of the client 100. β represents the local counter value of the server 120 at C = C ₀ . C ₀ indicates a local counter value as an arbitrary reference in the client 100.

  The condition for satisfying Expression (1) is that the oscillation frequencies of the oscillators included in the client 100 and the server 120 are sufficiently stable. Actually, the oscillation frequency of the transmitter may change due to temperature change, humidity change, and other influences. In addition, the oscillation frequency may change due to the performance of the oscillator itself.

  The oscillation frequency of the transmitter cannot be said to be stable for a long time, but is assumed to be stable during the measurement period (for example, about several tens of minutes). If it is about several tens of minutes, there is no problem even if it is assumed that the oscillation frequency is stable.

Assuming that the local counter value of the client 100 can be corrected to the local counter value of the server 120 by the equation (1), the following equations (2) and (3) are established.
S ₁ −F (C ₁ )> ε (2)
F (C ₂ ) -S ₂ > ε (3)
The left side of Expression (2) represents a packet transmission delay time from the client 100 to the server 120 (this direction is “going”). The left side of Expression (3) represents a packet transmission delay time from the server 120 to the client 100 (this direction is “return”).

  Let ε be the minimum value of the transmission delay time for going and returning (hereinafter also referred to as the minimum transmission delay time). Since the transmission delay time is not physically smaller than 0, ε ≧ 0.

First, F (C ₁ ) is calculated by substituting C ₁ into the variable C in equation (1). F (C ₂ ) is calculated by substituting C ₂ for the variable C in equation (1). Then, the calculated F (C ₁ ) is substituted into the equation ( ₂ ), and F (C ₂ ) is substituted into the equation (3). Then, when α is solved, the following equation (4) is obtained.
(S ₂ −β + ε) / (C ₂ −C ₀ ) <α <(S ₁ −β−ε) / (C ₁ −C ₀ ) (4)
Formula (4) shows a range where α can exist. On the right side of Equation (1), C ₀ may be arbitrarily determined, but α and β are unknowns. Therefore, if β is a fixed value, the existence range of α can be narrowed down from Equation (4).
Assuming that the lower limit value of α is αmin and the upper limit value of α is αmax, αmin and αmax are expressed by the following equations (5A) and (5B) from equation (4).
αmin ← (S ₂ −β + ε) / (C ₂ −C ₀ ) (5A)
αmax ← (S ₁ −β−ε) / (C ₁ −C ₀ ) (5B)
Here, “←” represents substitution into a variable.

Equation (4) shows the range to be satisfied by α in certain specific data {C ₁ , S ₁ , S ₂ , C ₂ }, but equation (4) should be satisfied for all the series of data. is there.

  FIG. 4 shows a plot of αmin and αmax using actual data. Note that the value of β is temporarily determined. Referring to FIG. 4, it can be seen that the range between the upper limit value αmax of α and the lower limit value αmin of α is narrowed with time. Further, it can be seen that α satisfying αmin <α <αmax exists in all the series of data.

  Looking at FIG. 4 in more detail, there are a point where αmax is plotted greatly deviating upward compared to the previous and subsequent points, and a point where αmin is greatly deviating downward compared to the previous and subsequent points. I understand. The point where αmax is plotted in the upward direction is due to the fact that the outgoing transmission time is much longer than the previous and subsequent points. The point where αmin is plotted in the downward direction is due to the fact that the return transmission time is much longer than the previous and subsequent points.

  In order to determine whether or not α satisfying αmin <α <αmax exists in all the series of data, the maximum value is taken for the lower limit value αmin of α, and the minimum value is taken for the upper limit value αmax of α. To go.

Each time the client 100 obtains the k-th data set {C _1k , S _1k , S _2k , C _2k } by transmitting and receiving the measurement packet, the expressions (5A), (5B), and (6A) and Equation (6B) are calculated (hereinafter also referred to as calculation processing A).
MAX_αmin ← MAX (αmin, MAX_αmin) (6A)
MIN_αmax ← MIN (αmax, MIN_αmax) (6B)
MAX_αmin on the left side of Equation (6A) indicates the maximum value of αmin at the current time point. MIN_αmax on the left side of Expression (6B) indicates the minimum value of αmax at the current time point. MAX (a, b) in Expression (6A) is a function that returns the larger value of {a, b}. MIN (a, b) in Expression (6B) is a function that returns the smaller value of {a, b}.

By repeating the calculation process A, α becomes a numerical value around 1. Therefore, the initial value of MAX_αmin may be set to 0, and the initial value of MIN_αmax may be set to 2, for example. In this case, the possible range of α is as shown in the following formula (7).
MAX_αmin <α <MIN_αmax (7)
If β is determined temporarily, {MAX_αmin, MIN_αmax} is obtained from the series of data from Equation (5A), Equation (5B), Equation (6A), and Equation (6B).

  Here, when MAX_αmin> MIN_αmax is established, there is no α satisfying αmin <α <αmax in all the series of data in the temporarily determined β. The absence of α means that the local counter value of the client 100 cannot be corrected to the local counter value of the server 120 by the equation (1). This is probably because the setting of β in equation (1) is strange.

  Next, {MAX_αmin, MIN_αmax} is calculated for each β while changing β as much as possible.

The initial possible range of β may be determined as shown in the following formula (8) using a single measurement data {C _1n , S _1n , S _2n , C _2n }.
C ₀ = C _1n , β = [S _1n −RTT, S _1n ] (8)
In Formula (8), [A, B] indicates that β is in the range from A to B. RTT in equation (8) represents the round trip time. RTT is calculated by the following equation (8A).
RTT = (C _2n −C _1n ) − (S _2n −S _1n ) (8A)
It is preferable to select measurement data having a round trip time RTT as small as possible. This is because the initial possible range of β becomes narrower. Specifically, the round trip of the measurement packet is repeated about 100 times, and a method in which the round trip time RTT is selected is selected.

  As a method of changing β, for example, β is changed at regular intervals within the initial possible range of β. This interval is determined by the accuracy of target time adjustment.

As a specific method of changing β at regular intervals, for example, βi obtained from the following equations (9A) and (9B) is used where β is the initial possible range of β and β is the target accuracy. .
N = INT ((βe−βs) / R) (9A)
βi = βs + (βe−βs) / N × i, (i = 0 to N) (9B)
INT (X) in Expression (9A) is a function that rounds to an integer.
MAX_αmin and MIN_αmax are calculated from Equation (5A), Equation (5B), Equation (6A), and Equation (6B) for each of a plurality of βs obtained by Equation (9A) and Equation (9B).

  If MAX_αmin <MIN_αmax, the calculated data of MAX_αmin and MIN_αmax is plotted. If MAX_αmin> MIN_αmax, the calculated data of MAX_αmin and MIN_αmax is not plotted. Once the contradiction has occurred (MAX_αmin> MIN_αmax), β can be excluded from the calculation because it is meaningless to calculate any more. By excluding β when contradiction occurs, the possible range of β is narrowed. Hereinafter, the above-described process of narrowing the possible range of β is also referred to as a β-existing range narrowing process.

With the above processing, a β-α plot is created as shown in FIG. 5 with β on the horizontal axis and α on the vertical axis. Referring to FIG. 5, the upper curve in FIGS. 5A and 5B is MIN_αmax, and the lower curve is MAX_αmin. A plot when data of MAX_αmin and MIN_αmax is started (when the data is small) is as shown in FIG. As the data of MAX_αmin and MIN_αmax increases, it gradually becomes as shown in FIG. A range surrounded by the upper curve and the lower curve indicates a possible range of {α, β}. When the measurement time is lengthened and the data of MAX_αmin and MIN_αmax is increased, the range of α is considerably narrowed, but β is narrowed only to a certain extent.
The range of α is narrowed for the following reason. Substituting Equation (5A) and Equation (5B) into (αmax−αmin) and assuming S ₂ = S ₁ + δ, the following Equation (9) is obtained.
αmax−αmin = (C ₂ −C ₁ ) × (S ₁ −β) / (C ₁ −C ₀ ) / (C ₂ −C ₀ ) −δ / (C ₂ −C ₀ ) −ε / (C _{1 -C 0) -ε / (C 2} -C 0) ··· (9)
δ is a time until the server 120 receives the measurement packet and transmits it to the client 100. That is, since it is a time for communication processing in the server 120 and depends only on the server 120, δ falls within a certain range. “Fit within a certain range” means that δ <M always holds, that is, δ has an upper limit value. Also, the minimum transmission delay time ε falls within a certain range. On the other hand, (C ₁ -C ₀ ) and (C ₂ -C ₀ ) increase as much as the measurement time is increased.

Therefore, from the second term on the right side of Equation (9), if the measurement time is lengthened,
δ / (C ₂ -C ₀ ) → 0
ε / (C ₁ -C ₀ ) → 0
ε / (C ₂ −C ₀ ) → 0
As good as Here, “→” indicates that the value approaches “0” infinitely (when the measurement time is lengthened). Here, C ₀ is a local counter value as an arbitrary reference in the client 100 as described above. As described above, β is a local counter value of the server 120 at C = C ₀ .

Here, C ₀ represents the time of the client at the reference time (considered as the time when the measurement was started (that is, the time when the first measurement packet was transmitted)), and β represents the time of the server at the reference time. C ₁ and C ₂ represent the client time at the current time. S ₁ and _{S 2} represents the time of the server at the current time. Therefore, if the measurement is continued, (S ₁ -β) and (C ₂ -C ₀ ) increase in the same manner as the time from the reference time to the current time. Therefore, since each of (S ₁ -β) and (C ₂ -C ₀ ) may be regarded as a measurement time,
(S ₁ -β) / (C ₂ -C ₀ ) → 1
As good as (In actuality, the ratio between the server and client clock frequencies is around 1.)

Therefore, it can be estimated as (αmax−αmin) to (C ₂ −C ₁ ) / (C ₁ −C ₀ ). ( "~" Is generally indicating that it is the same _value.) (C 2 _-C 1) is a round trip time including the processing time of the server 120, fit into a certain range. Further, (C ₁ -C ₀ ) increases in the same manner as the time from the reference time to the current time, as in (S ₁ -β) and (C ₂ -C ₀ ). Therefore, (C ₁ -C ₀ ) may be regarded as the measurement time. Therefore, if the measurement time is lengthened, (C ₂ -C ₁ ) / (C ₁ -C ₀ ) → 0
As good as From the above, if the measurement time is lengthened, (αmax−αmin) approaches zero.

The reason why the range of β is not narrowed is as follows. First, F (C ₁ ) is calculated by substituting C ₁ into the variable C in equation (1). F (C ₂ ) is calculated by substituting C ₂ for the variable C in equation (1). Then, the calculated F (C ₁ ) is substituted into the equation ( ₂ ), and F (C ₂ ) is substituted into the equation (3). Then, when β is solved, the following equation (10) is obtained.
S ₂ −α × (C ₂ −C ₀ ) + ε <β <S ₁ −α × (C ₁ −C ₀ ) −ε (10)
Here, α is an unknown number, but is determined temporarily. Assuming that the lower limit value of β is βmin and the upper limit value of β is βmax, βmin and βmax are expressed by the following equations (11A) and (11B) from equation (10).
βmin ← S ₂ −α × (C ₂ −C ₀ ) + ε (11A)
βmax ← S ₁ −α × (C ₁ −C ₀ ) −ε (11B)
{Βmin, βmax} in Equation (11A) and Equation (11B) indicates a value at a specific α. If α is different, {βmin, βmax} is also different.
If Expression (11A) and Expression (11B) are substituted into (βmax−βmin), Expression (12) is obtained.
βmax−βmin = α × (C ₂ −C ₁ ) − (S ₂ −S ₁ ) −2 × ε
(12)
Α × (C ₂ −C ₁ ) − (S ₂ −S ₁ ) in Expression (12) indicates a round trip time viewed from the server 120 side clock. Since the round trip time is a time required for packet transmission, it takes a certain time or more, and there is no property that the measurement time is lengthened and becomes as small as possible.

  The minimum transmission delay time ε is a fixed value. Therefore, (βmax−βmin) does not have a property of becoming small as long as the measurement time is lengthened, and is reduced only to a certain value.

  FIG. 6 shows a plot of βmin and βmax using actual data. Note that the value of α is temporarily determined. Referring to FIG. 6, it can be seen that the range between the upper limit value βmax of β and the lower limit value βmin of β is not narrowed.

  For reference, even if the measurement time is increased, the possible range of β does not narrow, but as is clear from equation (12), the larger the minimum transmission delay time ε is, the more β exists. The possible range can be narrowed.

  The minimum transmission delay time ε can be set as follows. The transmission delay time includes not only the time for flowing through the network but also the time for processing in the terminal. The processing time at the terminal includes OS context switching, interrupt response time, and the like. If it is known in advance that the processing time of the terminal takes at least 40 μs, for example, ε = 40 μs may be set. The processing time of the terminal can be obtained by transmitting / receiving a measurement packet for time adjustment on the same machine.

  Also, if it is known that a specific medium exists on the network path and the minimum transmission delay time of the medium is known in advance, the time can be included in the processing time of the terminal. In particular, if the minimum transmission delay time ε cannot be set, ε = 0 may be maintained.

  Although the possible range of {α, β} has been obtained by the above description, the actual {α, β} should be in the β-α plot diagram of FIG. Above this, {α, β} cannot be determined unless some assumption is made.

In this embodiment, an example of estimating {α, β} is shown. First, β is estimated. In the β-α plot diagram of FIG. 5, when the possible range of β is β = [β1, β2], an estimated value βest of β is obtained by the following equation (13). In FIG. 5, the lowest lower limit value of β where the plot points exist is β1, and the highest upper limit value of β is β2.
βest = (β1 + β2) / 2 (13)
Equation (13) takes the arithmetic mean of the lowest and highest limits in the possible range of β, but is not limited to this. For example, it is a weighted average like Equation (13A). May be.
βest = β1 × ω + β2 × (1−ω), 0 <ω <1 (13A)
Next, α is estimated. The estimated value αest of α is obtained by the following equation (14).
αest = (MAX_αmin [βest] + MIN_αmax [βest]) / 2 (14)
Here, {MAX_αmin [βest], MIN_αmax [βest]} represents {MAX_αmin, MIN_αmax} calculated with β = βest, respectively, according to the above formula.

In the equation (14), αest is obtained by the arithmetic average of MIN_αmax and MAX_αmin. However, the present invention is not limited to this, and for example, a weighted average like the equation (14A) may be used.
αest = MAX_αmin [βest] × ω + MIN_αmax [βest] × (1-ω), 0 <ω <1 (14A)
Equation (13) assumes that the minimum transmission delay time for going and returning is the same. This assumption is based on the premise that symmetry is physically maintained between the outgoing transmission line and the return transmission line, but it is not always necessary to maintain the symmetry of the transmission line.

  In the present embodiment, both a packet whose outbound transmission delay time is close to the minimum value and a packet whose return transmission delay time is close to the minimum value during a series of measurement periods during which measurement packets are transmitted and received are It only has to be observed.

  In this embodiment, a packet having a transmission delay time close to the minimum value is observed even during one-way video transmission. FIG. 21 described above is also a diagram showing a change in the transmission delay time in the course of time in the present embodiment. For example, in FIG. 21, data sufficiently close to the minimum transmission delay time (about 100 μsec) is obtained even during video transmission in which the symmetry of the transmission path is lost between going and returning.

By substituting the estimated value {αest, βest} of {α, β} obtained in this way into equation (1), the local counter value at an arbitrary point on the client side is calculated by the calculation based on the following equation (1A): Can be adjusted to the server side.
F (C) = αest × (C−C ₀ ) + βest (1A)
The outbound transmission delay time and the return transmission delay time can be obtained by the following equations (2A) and (3A), respectively.
Outgoing transmission delay time = S ₁ − {αest × (C ₁ −C ₀ ) + βest} (2A)
Return transmission delay time = {αest × (C ₂ −C ₀ ) + βest} −S ₂ (3A)
With the above calculation, the transmission delay time in one direction can be measured.

Hereinafter, a simple example will be described in order to facilitate understanding of the features of the present invention.
FIG. 7 is a diagram illustrating an example of data obtained by reciprocating the measurement packet. Here, in order to simplify the explanation, it is assumed that the time advance in the server 120 and the client 100 is the same. That is, α indicating the clock frequency ratio between the server 120 and the client 100 is “1”. Substituting α = 1 into equation (1), the following equation (1B) is obtained.
F (C) = (C−C ₀ ) + β (1B)
Referring to FIG. 7, it is assumed that the time of client 100 is exactly one hour ahead of the time of server 120. The client 100 transmits the measurement packet PK1 at the time 11:00 of the client 100. The server 120 receives the measurement packet PK1 at time 10:00 of the server 120. The server 120 transmits the measurement packet PKR1 at the time 10:00 of the server 120.

  The client 100 receives the measurement packet PKR1 at the time 11:10:00 of the client 100. As described above, the measurement packet reciprocates between the client 100 and the server 120.

The client 100 performs {C ₁ , S ₁ , S ₂ , C ₂ } = {11: 00: 00: 00, 10:00:00, 10: by the round trip of the first (first) measurement packet (PK1, PKR1). Data of 05:00, 11:10:00} is obtained. The client 100 obtains similar data by reciprocating the 2, 3,..., N th measurement packet.

In Formula (1B), C ₀ may be any time, but C ₀ = 11: 00: 00: 00. β is the time of the server 120 at the time of C = C ₀ , but the client 100 does not know. β is an unknown.

  The following shows what the possible range of β will be each time data is obtained. First, the round trip time in the first data is 9 minutes in total, 4 minutes for going and 5 minutes for returning.

  It should be noted that the client 100 only knows that the total time for going and returning is 9 minutes. I don't know how long it took to go and go back. However, it is certain that the range is {bound time = 9 minutes, return time = 0 minutes} to {bound time = 0 minutes, return time = 9 minutes}.

  Since the measurement packet transmitted at the time 11:00 on the client 100 side is received at the time 10:00:00 on the server side, the server time β at the time of C = 11: 00: From 10:04:00 to 00: 09: 00 = 9: 55: 00, it should be in the range of 9:55:00 <β <10:04:00 (hereinafter also referred to as Formula A).

  Next, the round trip time for the second data is 9 minutes in total, with 3 minutes going and 6 minutes returning. Therefore, from 10: 23: 00-00: 09: 0 = 10: 14: 00, 10:14:00 <β + (11: 20: 00-11: 00: 00) <10:23:00, 9 : 54: 00 <β <10:03:00 (hereinafter also referred to as Formula B).

  From Formula A and Formula B, 9:55:00 <β <10:03:00 (hereinafter also referred to as Formula C) is obtained. Therefore, it can be seen that the possible range of β is narrowed by adding the second data.

  Next, the round trip time in the third data is 5 minutes for going and 2 minutes for returning, totaling 7 minutes. Therefore, from 10:45:00 to 00: 07: 00 = 10: 38: 00, 10:38:00 <β + (11: 40: 00-11: 00: 00) <10:45:00 9:58:00 <β <10:05:00. From Formula A, Formula B, and Formula C, 9:58:00 <β <10:03:00 (hereinafter also referred to as Formula D) is obtained. Therefore, it can be seen that the possible range of β is further narrowed by adding the third data.

  As shown in the individual data examples, if the transmission delay time of “going” is updated with the minimum value so far, the upper limit value of β is updated (see Formula B obtained from the second data). Further, if the transmission delay time of “return” is updated with the minimum value so far, the lower limit value of β is updated (see Expression D obtained from the third data). By repeating the round trip of the measurement packet in this way, the possible range of β can be narrowed to some extent.

  Considering how far it can be narrowed, it can be narrowed to “minimum transmission delay time for outbound” + “minimum transmission delay time for return”.

  This is not obtained by reciprocating one measurement packet. Among the plurality of measurement packets that are reciprocated, the measurement packet that has the minimum transmission delay time is not necessarily the same as the measurement packet that has the minimum transmission delay time. Rather, there are many things that do not agree.

  The range in which β can exist can be narrowed by reciprocating a plurality of measurement packets and considering all the series of data. Then, after the possible range of β is narrowed, the value of β is estimated. The method of estimating the value of β is the arithmetic average of the upper limit value of β and the lower limit value of β as described in Expression (13).

  In this example, since the upper limit value of β = 10: 03: 00 and the lower limit value of β = 9: 58: 00, βest = (10: 03: 00 + 9: 58: 00) / 2 from equation (13). = 10:00:30 is obtained. That is, it is estimated that the client 100 has advanced 11: 00: 00: 00: 30 = 0: 59: 30 as compared to the server 120.

  If the transmission path is physically symmetric, the round trip of the measurement packet is repeated to some extent, so the “minimum transmission delay time” and “minimum transmission delay time” are considered to be almost the same value. The estimation method described in 13) is reasonable.

  As described above, the time of the client 100 can be set to the time of the server 120.

  Actually, since the time advance does not match between the server 120 and the client 100, α is also an unknown number. In the present invention, assuming that {α, β} are both unknown, a method of narrowing the possible range of {α, β} is adopted.

  Hereinafter, operations of the CPU 101 on the client 100 side and the CPU 121 on the server 120 side that execute the program will be described in detail.

  FIG. 8 is a diagram illustrating a flowchart of a program executed by the server 120. Referring to FIG. 8, in step S101, the network is initialized. More specifically, the CPU 121 opens a port for transmitting / receiving packets so that packets can be transmitted / received. Thereafter, the process proceeds to step S102.

  In step S102, the CPU 121 generates a thread A1 and a thread A2. Each of the thread A1 and the thread A2 is processed independently. In the thread A1, a counter rd_cnt whose initial value is set to 0 is used. In the thread A2, a counter wr_cnt whose initial value is set to 0 is used.

  In the thread A1, the CPU 121 executes steps S105, S106, S107, and S108. In the thread A2, the CPU 121 executes S103 and S104.

  In step S103, the CPU 121 determines whether a measurement packet has been received. If YES in step S103, the process proceeds to step S104. On the other hand, if NO at step S103, the process at step S103 is performed again.

In step S104, the CPU 121 acquires a local counter value (time) on the server side and substitutes it into the variable S1. Since the acquisition of the local counter (time) depends on the hardware configuration, a method corresponding to that is taken. It contains variable C ₁ is in the payload area 1 of the measurement packet. Then, the CPU 121 increments the counter wr_cnt by 1. Thereafter, the process of step S103 is performed again.

In the thread A1, first, the process of step S105 is performed.
In step S105, the CPU 121 determines whether there is a measurement packet received in step S103. This determination is made based on whether or not the counter rd_cnt is different from the counter wr_cnt incremented by 1 in step S104.

  If YES in step S105, the process proceeds to step S106. On the other hand, if NO at step S106, the process proceeds to step S108 described later.

FIG. 9 is a diagram illustrating an example of the configuration of the measurement packet and the reply packet.
FIG. 9A shows an example of the configuration of a measurement packet. The measurement packet includes a transmission source address, a destination address, and a payload area 1.

  FIG. 9B is a diagram illustrating an example of the configuration of a reply packet. The reply packet includes a transmission source address, a destination address, a payload area 1, a payload area 2, and a payload area 3.

  Note that the packet configuration shown in FIG. 9 is the minimum necessary for the description, and a header and a payload area may be included in addition to that. Further, since the payload areas of the measurement packet and the reply packet are sufficiently small, they may be transmitted / received by being included in an arbitrary packet exchanged between other servers and clients without being transmitted / received alone.

Referring to FIG. 8 again, in step S106, a reply packet creation process is performed. In reply packet creation process, CPU 121 obtains the local counter value server 120 (time) is substituted into the variable _{S 2.}

Next, the CPU 121 sets C ₁ , S ₁ , and S ₂ in the payload area 1, payload area 2, and payload area 3 of the reply packet, respectively. Then, the source address of the received measurement packet is set as the destination address of the reply packet. Thus, a reply packet is created. Then, the CPU 101 increments the counter rd_cnt by 1. Thereafter, the process proceeds to step S107.

  In step S <b> 107, the CPU 121 transmits the generated reply packet to the client 100. Thereafter, the process proceeds to step S108.

  In step S108, the CPU 121 waits for 50 ms. Note that the waiting time is not limited to 50 ms and may be changed as appropriate. Thereafter, the process of step S105 is performed again.

  Note that the process of receiving the measurement packet (S103) and the process of transmitting the reply packet (S107) may be performed in the same thread.

  FIG. 10 is a diagram illustrating a flowchart of a program executed on the client 100. Referring to FIG. 10, in step S201, the network is initialized. More specifically, the CPU 101 opens a port for transmitting / receiving packets so that packets can be transmitted / received. Thereafter, the process proceeds to step S202.

  In step S202, the CPU 101 generates a thread B1 and a thread B2. Each of the thread B1 and the thread B2 is processed independently. In the thread B1, the CPU 101 executes steps S203, S204, and S205. In the thread B2, the CPU 101 executes S206, S207, and S208.

In step S203, a measurement packet creation process is performed. In the measurement packet creation processing, CPU 101 obtains the local counter value of the client 100 a (time) is substituted into _{C 1.} Next, CPU 101 sets the _{C 1} in the payload area 1 of the measurement packet. Then, the CPU 101 sets the address of the server 120 as the destination address of the measurement packet. As described above, a measurement packet is created. Thereafter, the process proceeds to step S204.

  In step S <b> 204, the CPU 101 transmits the generated measurement packet to the server 120. Thereafter, the process proceeds to step S205.

  In step S205, the CPU 101 waits for 100 ms. Note that the waiting time is not limited to 100 ms and may be changed as appropriate. Thereafter, the process of step S203 is performed again.

  In step S206, the CPU 101 determines whether a reply packet has been received. If YES in step S206, the process proceeds to step S207. On the other hand, if NO at step S206, the process at step S206 is performed again.

In step S207, CPU 101 obtains the local counter value of the client 100 a (time), is substituted into _{C 2.} Next, the CPU 101 acquires C ₁ , S ₁ , and S ₂ included in the payload area 1, the payload area 2, and the payload area 3 of the reply packet, respectively. Thereafter, the process proceeds to step S208.

In step S208, process A is performed using {C ₁ , S ₁ , S ₂ , C ₂ } obtained in step S207 as an argument. When the process A ends, the process of step S206 is performed again.

Through the above processing, data {C ₁ , S ₁ , S ₂ , C ₂ } can be obtained by reciprocating the measurement packet between the client 100 and the server 120.

FIG. 11 is a flowchart of processing A. The process A includes an algorithm that is a key of the present invention. Process A receives {C ₁ , S ₁ , S ₂ , C ₂ } as an argument. Referring to FIG. 11, in step S <b> 301, CPU 101 determines whether or not the number of calls for process A is within 100 times. If YES in step S301, the process proceeds to step S302. On the other hand, if NO at step S301, the process proceeds to step S303 described later. In step S302, a part of a program written in C language is described in order to ensure the accuracy of the contents.

By the process of step S302 is repeated, CPU 101 is to determine _{C 0,} select the data round-trip time is minimized among the first 100 data, the _{C 1} at that time _{C 0} .

  In step S302, the CPU 101 calculates a round trip time (RTT) using the above-described equation (8A). Each time the process of step S302 is repeated, the CPU 101 compares the previously calculated RTT with the currently calculated RTT, and stores the smaller value in the memory 102.

  When the number of executions of the process in step S302 reaches 100, the CPU 101 secures an array and sets an initial value using the minimum data of the round trip RTT so far. The imin and imax of the program in S302 correspond to βmin and βmax, respectively. In the initial value setting, initial values of the lower limit value βmin and the upper limit value βmax of β are set. Moreover, 100 times is a standard and may be changed as appropriate. Thereafter, when the process of step S302 ends, the process A ends, the process returns to the process of FIG. 10, and the process of step S206 is performed again.

In step S303, process B is performed.
FIG. 12 is a diagram showing a program in which process B is described in C language. Referring to FIG. 12, the processing of the first for statement of the program is performed by changing the upper limit value MIN_αmax and the lower limit of α while changing β (i in the program) within the range of β set in step S302 described above. A value MAX_αmin is calculated. Specifically, in the process B, the CPU 101 calculates MAX_αmin and MIN_αmax based on the above formula (5A), formula (5B), formula (6A), and formula (6B). Next, the CPU 101 narrows the possible range of β by performing the above-described β existence range narrowing process.

  Next, the CPU 101 calculates βest based on the above equation (13). Further, the CPU 101 calculates αest based on the above-described equation (14).

Next, CPU 101 calculates the outbound and return transmission delay times based on the above-described equations (2A) and (3A), respectively. It is optional to calculate the outbound and return transmission delay times. Thereafter, the process proceeds to step S304.

In step S304, the CPU 101 causes the memory 102 to store the C ₀ acquired by the process A and the αest and βest data acquired by the process B in a format that can be used by other programs. Thereafter, the process A ends, the process returns to the process of FIG. 10 described above, and the process of step S206 is performed. The data of C ₀ , αest, and βest may be stored in the register 107 when the client includes the dedicated hardware 108 of FIG.

  If the process A is performed for a long time, the oscillation frequency of the oscillator 104 of the client 100 or the oscillator 124 of the server 120 may change, so that the possible range of β may be lost. If the oscillation frequency changes as described above during the measurement period, the local count value may not be corrected by equation (1A).

As described above, when the possible range of β disappears, the static variables defined in FIG. 11 are returned to the initial values, and the processes of steps S301, S302, S303, and S304 may be performed again. Alternatively, the above-described calculation may be performed by changing C ₀ that is the reference time of the client at regular intervals regardless of whether the possible range of β disappears or does not disappear.

  Next, processing for correcting the local counter value (hereinafter also referred to as counter correction processing) performed by the client 100 will be described. The counter correction processing is performed simultaneously with the processing of the program executed by the client 100 in FIG. The counter correction process is, for example, a process performed by an application program that reproduces content (hereinafter also referred to as a content reproduction program) that requires time synchronization between the server 120 and the client 100. In this case, in each of the server 120 and the client 100, the same content is simultaneously reproduced by the content reproduction program.

  FIG. 13 is a diagram illustrating a flowchart of the counter correction process. Referring to FIG. 13, in step S405, CPU 101 obtains local counter value C indicating an arbitrary time. Thereafter, the process proceeds to step S410.

In step S410, CPU 101 is, Arufaest stored in the memory 102, βest, the data of _{C 0} is read. Thereafter, the process proceeds to step S420.

In step S420, CPU 101 is a local counter value C, αest, βest, the _{C 0,} by substituting the equation (1A) described above, calculates the F (C) is a local counter value of the client 100 after the correction .

The corrected local counter value F (C) is time synchronized with the server 120.
Then, the counter correction process ends. Through the counter correction process described above, the local counter value of the client 100 can be matched with the local counter value of the server 120.

The counter correction process is repeatedly performed in the application program. If the dedicated hardware 108 shown in FIG. 2 is provided, the corrected local counter can be autonomously run. A local counter value C that is the current time is obtained from the local counter 105. From the register 107, C ₀ , αest, and βest are obtained. The process set in the register 107 is periodically performed by the process A described above. The correction circuit 106 corrects the local counter value by substituting these values into the above equation (1A). The corrected local counter value can be read from the CPU 101.

The present invention is different from the conventional NTP in the following three points.
1) The point that the local counter value of the client 100 is corrected to the local counter value of the server 120 by the equation (1A) (equation (1)).
2) The point where the possible range of {α, β} is narrowed by using all of a series of data obtained by a round trip of measurement packets.
3) The point that {α, β} is estimated by focusing on the fact that the minimum transmission delay time for going and returning approaches approximately the same value after multiple rounds of measurement packets.

  Although it is considered that it is desirable to use the equation (13) for the estimated value of β, the estimated value of β largely deviates when the minimum transmission delay time on the way and the minimum transmission delay time on the return are greatly different. is there.

  For example, in FIG. 7, it is assumed that the last reply packet PKR3 transmitted from the server 120 to the client 100 arrives at the server 120 in 0 minutes (11:46:00) instead of 2 minutes. In this case, the round trip time is 5 minutes.

  10: 45: 00-00: 05: 00 = 10: 40: 00 <β + (11: 40: 00-11: 00: 00) <10:45:00, 10: 00: 0 <β <10: 05:00 is obtained. Then, 10: 00: 0 <β <10:03:00 is obtained based on all the changed data in FIG.

  In FIG. 7, the minimum transmission delay time for outbound is 3 minutes, while the minimum transmission delay time for return is 0 minutes. In this case, from equation (13), it is estimated that βest = (10: 03: 00 + 10: 00: 0) / 2 = 10: 0: 30, but the error becomes larger than the correct answer 10: 00: 00: 00. .

  On the other hand, when the minimum transmission delay time for outbound and the minimum transmission delay time for return are equal, the estimated value of β becomes extremely close to true.

  For example, in FIG. 7, it is assumed that the last reply packet PKR3 transmitted from the server 120 to the client 100 arrives at the server 120 in 11 minutes (11:49:00) instead of 2 minutes. In this case, the round trip time is 8 minutes.

  10: 45: 00-00: 08: 0 = 10: 33: 00 <β + (11: 40: 00-11: 00: 00) <10:45:00, 09:57:00 <β <10: 05:00 is obtained. Based on all the changed data in FIG. 7, 09:57:00 <β <10:03:00 is obtained.

  In this case, the minimum transmission delay time for the outbound route and the minimum transmission delay time for the return route are both 3 minutes. Further, from equation (13), it is estimated that βest = (10: 03: 00 + 09: 57: 00) / 2 = 10: 00: 00, but this estimated value is a correct answer.

  If the measurement is continued for a while, the minimum transmission delay time for going and returning should be almost the same (if the communication path is physically symmetric), so the method for estimating β in the present invention is reasonable.

  Next, a diagram showing a simulation result of time adjustment of an NTP (including data filter and PLL) technology (hereinafter also referred to as Conventional Technology A) and a diagram showing a simulation result of time adjustment of the present invention are compared. The following simulation results are obtained when video is transmitted from the client 100 to the server 120.

  FIG. 14 is a diagram showing a simulation result in which time adjustment is performed by the prior art A.

FIG. 15 is a diagram showing a simulation result of time adjustment according to the present invention.
Referring to FIG. 14, it can be seen that the error is large in the simulation result in which the time is adjusted from the prior art A. On the other hand, with reference to FIG. 15, it can be seen that the error is suppressed small in the simulation result of time adjustment according to the present invention. That is, according to the present invention, it is possible to perform time adjustment with high accuracy even when the symmetry of the transmission path is lost on the way back and forth as in the case where video transmission is performed only in one direction.

  Next, a diagram showing a simulation result of time adjustment according to the related art A in a case where video transmission from the client 100 to the server 120 is not performed is compared with a diagram showing a simulation result of time adjustment according to the present invention.

  FIG. 16 is a diagram illustrating a simulation result in which time adjustment is performed by the related art A.

FIG. 17 is a diagram showing a simulation result of time adjustment according to the present invention.
Referring to FIG. 16, in the simulation result in which the time is adjusted by the conventional technique A, the error is suppressed small, but it can be seen that a certain amount of time is required until the error is suppressed small. On the other hand, with reference to FIG. 17, it can be seen that the error is suppressed to a small value in the simulation result of time adjustment according to the present invention and that the time is short until the error is suppressed to a small value.

  Examples of specific systems to which the present invention can be applied are described below. An example of the system is a system including a plurality of display devices connected to a network. Another example of the system is a system that includes a plurality of devices each having a speaker. In this case, sound is output simultaneously from the speakers included in each of the plurality of devices.

  In addition, the present invention can be applied to a device that measures a one-way delay time at two points in a remote place and other application programs in which time synchronization is important.

  Processing performed in applications where time synchronization is important has a wide variety of processing such as transaction processing, distributed computing, and network processing.

The element constituting the present invention may be a general PC. However, if dedicated hardware (circuitry) provides a local counter based on a highly accurate and stable oscillator, and there is a circuit that corrects the value and progress of the local counter, {αest, βest, C ₀ } Based on this, the client-side local counter may be corrected by hardware.

In the present invention, the measurement packet and the reply packet are transmitted and received between the server and the client. In the step of obtaining {C ₁ , S ₁ , S ₂ , C ₂ }, a general NTP framework is used. Also good.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

It is a figure which shows the system as an example which comprises this Embodiment. It is a figure which shows the structure of the client as an example which mounts exclusive hardware. The relationship between the local counter on the client side and the local counter on the server side is shown. The figure which plotted (alpha) min and (alpha) max using actual data is shown. It is a figure which shows the beta-alpha plot figure in this invention. The figure which plotted (beta) min and (beta) max using actual data is shown. It is a figure which shows an example of the data obtained by reciprocating a measurement packet. It is a figure which shows the flowchart of the program performed with a server. It is a figure which shows an example of a structure of a measurement packet and a reply packet. It is a figure which shows the flowchart of the program performed with a client. It is a flowchart of the process A. It is a figure which shows the program which described the process B in C language. It is a figure which shows the flowchart of a counter correction process. It is a figure which shows the simulation result which performed time adjustment by the prior art A. It is a figure which shows the simulation result which performed time adjustment by this invention. It is a figure which shows the simulation result which performed time adjustment by the prior art A. It is a figure which shows the simulation result which performed time adjustment by this invention. It is a figure for demonstrating the method of the time adjustment which uses NTP. It is a figure which shows the structure which combined the data filter and PLL in NTP. It is a figure which shows a structure when transmitting a video from a client to a server. It is a figure which shows the change of the transmission transmission delay time with time passage. It is a figure which shows the change of the return transmission delay time with progress of time. It is a figure which shows the average transmission delay time of each going and returning.

Explanation of symbols

  100 client, 101, 121 CPU, 102, 122 memory, 103, 123 communication I / F, 104, 124 oscillator, 105, 125 local counter, 106 correction circuit, 107 register, 108 dedicated hardware, 110 communication network, 120 server.

Claims (17)

A transmission device connected to a network and operating at a second frequency and communicating with a device operating at a first frequency;
Time management means for managing the time;
Transmission means for sequentially transmitting measurement packets to the device;
Each time the measurement packet is received, the apparatus sequentially receives the reply packet including information on a first time when the measurement packet is received and information on a second time when the measurement packet is returned as a reply packet. Reply,
The first and second time information included in the reply packet, the third time information at which the measurement packet corresponding to the received reply packet is transmitted, and the fourth time information at which the reply packet is received. Calculation means for sequentially calculating the time offset with the device based on the time information to be performed;
The time managed by the time management means based on the frequency ratio indicating the ratio between the first frequency and the second frequency, the time offset, and an arbitrary reference time managed by the time management means. And a correction unit for correcting the transmission. When the frequency ratio is α, the time before correction is C, the arbitrary reference time is C ₀ , and the time offset is β, the correction means has the calculation means F (C) = α × The transmission apparatus according to claim 1, wherein the management time is corrected based on the time F (C) calculated by (C−C ₀ ) + β. When the frequency ratio is α and the time offset is β, the calculation means calculates a minimum value αmin and a maximum value αmax of a possible value of α based on a plurality of the time information, and a plurality of the times Based on the information, the minimum value βmin and the maximum value βmax of the values that β can take are calculated,
The calculation means sets one value in a range of possible values of α as α as the frequency ratio, and sets one value in a range of possible values of β as β as the time offset, The transmission apparatus according to claim 1 or 2. The frequency ratio is α, the arbitrary reference time is C ₀ , the first minimum transmission delay time that is the minimum delay time that occurs when data is transmitted to the device, and the data from the device to the transmission device Ε is the second minimum transmission delay time that is the minimum delay time that occurs during transmission, C _{1 is} the time when the transmitting means transmits the n (natural number) measurement packet, and the device determines the nth measurement packet. When the received time is S ₁ , the time when the device returns the nth reply packet is S ₂ , and the time when the transmission device receives the nth reply packet is C ₂ , the calculation means , Based on the nth time information, a minimum value αmin that α can take is calculated by (S ₂ −β + ε) / (C ₂ −C ₀ ), and α is calculated based on the nth time information. Maximum possible value The .alpha.max, calculated by _{(S 1 -β-ε) /} (C 1 -C 0),
The said calculation means calculates the maximum value of the value which can take (alpha) min based on the said some time information, and calculates the minimum value of the value which can take (alpha) max. Transmission equipment. The frequency ratio is α, the time offset is β, the time when the transmission unit transmits the n (natural number) measurement packet is C ₁ , the time when the device receives the nth measurement packet is S ₁ , and the device S _{2 is} the time at which the nth reply packet is returned, and C _{2 is} the time at which the transmission apparatus receives the nth reply packet, when data is reciprocated between the transmission apparatus and the apparatus. When the time indicating the generated delay time is RTT, the calculation means calculates the RTT based on the nth time information,
The calculation means calculates a minimum value minRTT of possible values of RTT based on a plurality of the time information,
The calculation means sets the minimum and maximum values for changing β as predetermined values S00 and minRTT, respectively, and changes the maximum value of αmin while changing β within the range of changing β. And a minimum value βmin and a maximum value βmax of the possible values of β are calculated based on a range of β in which α can exist. The transmission apparatus in any one of. When the time offset is β,
The transmission according to any one of claims 1 to 5, wherein the calculating means sets a value calculated by an arithmetic average of a minimum value βmin and a maximum value βmax of β to be β as the time offset. apparatus. When the frequency ratio is α,
The transmission according to any one of claims 1 to 6, wherein the calculation means sets a value calculated by an arithmetic average of a minimum value αmin and a maximum value αmax of α to be α as the frequency ratio. apparatus. The device has a function of playing back content,
The transmission apparatus according to claim 1, wherein the same content is simultaneously reproduced between the apparatus and the transmission apparatus.   The calculation means calculates at least one of a first transmission delay time that occurs when data is transmitted to the device and a second transmission delay time that occurs when data is transmitted from the device to the transmission device. The transmission apparatus according to claim 8. The frequency ratio is α, the time offset is β, the arbitrary reference time is C ₀ , the time when the transmission unit transmits the n (natural number) measurement packet is C ₁ , and the device is the nth measurement. When the time when the packet is received is S ₁ , the time when the device returns the nth reply packet is S ₂ , and the time when the transmission device receives the nth reply packet is C ₂ , the calculation is performed. The means calculates the first transmission delay time by S ₁ − {α × (C ₁ −C ₀ ) + β} based on the n-th time information, and calculates the second transmission delay time by {α The transmission device according to claim 1, wherein the transmission device is calculated by × (C ₂ −C ₀ ) + β} −S ₂ . A time synchronization method for synchronizing between a device operating at a first frequency connected to a network and a transmission device operating at a second frequency for data communication,
A process for managing time;
Sequentially transmitting measurement packets to the device,
Each time the measurement packet is received, the apparatus sequentially receives the reply packet including information on a first time when the measurement packet is received and information on a second time when the measurement packet is returned as a reply packet. Reply,
The first and second time information included in the reply packet, the third time information at which the measurement packet corresponding to the received reply packet is transmitted, and the fourth time information at which the reply packet is received. Sequentially calculating a time offset with the device based on the time information to be performed;
The step of managing the time based on a frequency ratio indicating a ratio between the first frequency and the second frequency, the time offset, and an arbitrary reference time managed by the step of managing the time. And a step of correcting the time managed by the time synchronization method. When the frequency ratio is α, the time before correction is C, the arbitrary reference time is C ₀ , and the time offset is β, the correction step is F (C) = The time synchronization method according to claim 11, wherein the management time is corrected based on the time F (C) calculated by α × (C−C ₀ ) + β. When the frequency ratio is α and the time offset is β, the calculating step calculates a minimum value αmin and a maximum value αmax of a possible value of α based on the plurality of time information, and Based on the time information, the minimum value βmin and the maximum value βmax of the possible values of β are calculated,
In the calculating step, one value in a range of possible values of α is α as the frequency ratio, and one value in a range of possible values of β is β as the time offset. The time synchronization method according to claim 11 or claim 12. The frequency ratio is α, the arbitrary reference time is C ₀ , the first minimum transmission delay time that is the minimum delay time that occurs when data is transmitted to the device, and the data from the device to the transmission device Ε is the second minimum transmission delay time that is the minimum delay time generated during transmission, C _{1 is} the time when the transmitting step transmits the n (natural number) measurement packet, and the device is the nth measurement packet. Is calculated when S ₁ is the time when the device has received the reply, S ₂ when the device has returned the nth reply packet, and C ₂ is the time when the transmission device has received the nth reply packet. The process calculates a minimum value αmin that α can take based on the n-th time information by (S ₂ −β + ε) / (C ₂ −C ₀ ), and based on the n-th time information α can take Maximum value αmax of, calculated by _{(S 1 -β-ε) /} (C 1 -C 0),
14. The calculation according to claim 11, wherein the calculating step calculates a maximum value that αmin can take based on a plurality of pieces of time information, and calculates a minimum value that αmax can take. The time synchronization method described. The frequency ratio is α, the time offset is β, the transmission step is C ₁ when the n-th (natural number) measurement packet is transmitted, and the time when the device receives the n-th measurement packet is S ₁ . When S _{2 is} the time when the device has returned the n-th reply packet, and C _{2 is} the time when the transmission device has received the n-th reply packet, when data is reciprocated between the transmission device and the device. When the time indicating the delay time occurring in RTT is RTT, the calculating step calculates RTT based on the nth time information,
The step of calculating calculates a minimum value minRTT of possible values of RTT based on a plurality of the time information,
In the calculating step, the minimum value and the maximum value of the value for changing β are set to a predetermined value S00 and minRTT, respectively, and the maximum value that αmin can take while changing β in the range of changing β. The minimum value βmin and the maximum value βmax of β can be calculated based on the range of β in which α can exist, and the minimum value of the value and αmax that can be taken. The time synchronization method according to any one of 14. When the time offset is β,
16. The calculating step according to any one of claims 11 to 15, wherein a value calculated by an arithmetic average of a minimum value βmin and a maximum value βmax of β can be set as β as the time offset. Time synchronization method. When the frequency ratio is α,
The calculation step according to any one of claims 11 to 16, wherein a value calculated by an arithmetic average of a minimum value αmin and a maximum value αmax of α can be set as α as the frequency ratio. Time synchronization method.

Images