WO2023105705A1

WO2023105705A1 - Timestamp correcting system, sensor system, and timestamp correcting method

Info

Publication number: WO2023105705A1
Application number: PCT/JP2021/045305
Authority: WO
Inventors: 賢一松永; 利彦近藤
Original assignee: 日本電信電話株式会社
Priority date: 2021-12-09
Filing date: 2021-12-09
Publication date: 2023-06-15

Abstract

A timestamp correcting system (121) according to the present invention is a system for correcting a timestamp attached to a packet when receiving packets transmitted at a predetermined transmission time interval, and comprises: a delay circuit (1211) that delays a timestamp; a subtraction unit (1212) that calculates a difference in timestamp between a delayed packet and a subsequently received packet as an arrival time interval; a packet number estimating unit (1213) that calculates an average value of quantized numbers of arrival time intervals as a packet number estimation value; a computing unit (1214) that multiplies the transmission time interval by the packet number estimation value to calculate an arrival time interval; a moving average filter (1215) that executes moving averaging of the arrival time intervals; and an addition unit (1216) that adds an initial time to a moving average value.　In this way, the present invention can provide a timestamp correcting system that reduces an error of a timestamp attached to a packet.

Description

Time stamp correction system, sensor system and time stamp correction method

The present invention relates to a time stamp correction system, a sensor system, and a time stamp correction method in data transmission and reception.

In the IoT (Internet of Things) network, it is expected that various sensors will be connected, a large amount of data will be collected, and useful information will be extracted by analyzing that data. Therefore, terminals that accommodate sensors are required to meet various use cases and needs, and reduction of power consumption in long-term measurement is required (Non-Patent Document 1).

In particular, in the sensor, intermittent operation is achieved by appropriately sleeping the MPU (Micro-Processing Unit), which consumes more power than other elements, and only the Analog Front End (AFE) operates continuously. Therefore, cost and power consumption can be reduced (Non-Patent Document 2).

However, when timestamping is performed on the data collection terminal, the delay time associated with communication between the sensor and the data collection terminal cannot be canceled, and an error occurs in the timestamp.

Also, even if the interval at which the MPU is activated is accurate, there will be a constant error in the clock mounted on the AFE, fluctuations in the interval between packets to be transmitted, and an error in the time stamp.

Also, due to noise from the outside to the radio unit, the reception timing is shifted and an error occurs in the time stamp.

In order to solve the above-described problems, the time stamp correction system according to the present invention provides a time stamp that is given to each of a plurality of packets when a plurality of packets transmitted at a predetermined transmission interval time are received. A time stamp correction system for correcting stamps, comprising: a delay circuit for delaying the time stamp of one of the plurality of packets; a time stamp of another packet received following the one packet; a subtracting unit that calculates a difference between the delayed packet and the time stamp as an arrival interval time; an estimator for calculating the number of packets, an arithmetic unit for multiplying the transmission interval time by the estimated value of the number of packets to calculate the dequantized arrival interval time, and the inverse quantized arrival interval time A moving average filter that performs a moving average, and an addition unit that adds an initial time to the value obtained by the moving average.

Further, a time stamp correction method according to the present invention is a method for correcting time stamps given to packets when a plurality of packets transmitted at a predetermined transmission interval time are received, wherein the subtraction unit comprises a step of calculating the difference between the timestamps as an arrival interval time; a step of a packet number estimating unit quantizing the arrival interval time; calculating the average value of the quantization number as the estimated value of the number of packets; a moving average filter performing a moving average on the dequantized inter-arrival time; and an adding unit adding an initial time to the value obtained by the moving average. Prepare.

According to the present invention, it is possible to provide a time stamp correction system, a sensor system, and a time stamp correction method that reduce errors in time stamps attached to packets.

FIG. 1 is a block diagram showing the configuration of a sensor system according to the first embodiment of the invention. FIG. 2A is a diagram for explaining the operation of the timestamp correction system according to the first embodiment of the present invention. FIG. 2B is a diagram for explaining the operation of the time stamp correction system according to the first embodiment of the present invention; FIG. 3 is a block diagram showing the configuration of the time stamp correction system according to the first embodiment of the invention. FIG. 4 is a flow chart showing the time stamp correction method according to the first embodiment of the present invention. FIG. 5 is a diagram for explaining the operation of the time stamp correction system according to the first embodiment of the present invention. FIG. 6A is a diagram for explaining the operation of the time stamp correction system according to the first embodiment of the present invention; FIG. 6B is a diagram for explaining the operation of the time stamp correction system according to the first embodiment of the present invention; FIG. 6C is a diagram for explaining the operation of the time stamp correction system according to the first embodiment of the present invention; FIG. 6D is a diagram for explaining the operation of the time stamp correction system according to the first embodiment of the present invention; FIG. 7 is a diagram for explaining the operation of the time stamp correction system according to the first embodiment of the present invention. FIG. 8 is a diagram for explaining the operation of the time stamp correction system according to the first embodiment of the present invention. FIG. 9 is a block diagram showing the configuration of a time stamp correction system according to the second embodiment of the invention. FIG. 10 is a flow chart showing a time stamp correction method according to the second embodiment of the present invention. FIG. 11 is a schematic diagram showing the configuration of a sensor system according to the third embodiment of the invention. FIG. 12 is a schematic diagram showing the configuration of a sensor system according to the fourth embodiment of the invention.

<First Embodiment>
A timestamp correction system, a sensor system, and a timestamp correction method according to a first embodiment of the present invention will be described with reference to FIGS. 1 to 8. FIG.

<Configuration of sensor system>
A sensor system 10 according to the present embodiment includes a sensor 11 and a receiver 12, as shown in FIG. Signals are transmitted and received between the sensor 11 and the receiver 12 wirelessly or by wire.

The sensor 11 operates intermittently and includes an AFE 111 , a memory 112 , an MPU 113 and a transmitter 114 . It also has an AFE clock 115 connected to the AFE 111 and a packet clock 116 connected to the MPU 113 .

The AFE 111 samples and quantizes the measurement signal 1 at time _TAFE counted by the AFE clock 115 .

The memory 112 stores the quantized measurement signal 1 as sensor data 2.

Here, one packet 3 is composed of a predetermined number of sensor data 2.

The MPU 113 is activated at a predetermined interval T _packet counted by the packet clock 116 and checks the memory 112 . When sensor data 2 to be stored in one packet 3 is accumulated in memory 112 , a wireless circuit (for example, BLE, Bluetooth Low Energy) is activated and packet 3 is transmitted from transmission unit 114 . Thus, the packet transmission interval time is T _packet .

The receiver 12 includes a time stamp correction system 121 according to the present embodiment, a time stamp adding unit 122, and a receiving unit 123. It also has a timestamp clock 125 .

After the packet is received by the receiving unit 123, the time stamping unit 122 according to the present embodiment adds a time stamp T _stamp , which is counted by the time stamping clock 125 and provided via the OS 124 of the receiver, to the packet. do. Here, the time stamps are synchronized by GPS (Global Positioning System), NTP (Network Time Protocol), NITZ (Network Identity and Time Zone), or the like.

The time stamp correction system 121 corrects the time stamp given to the packet input from the time stamp giving unit 122, as will be described later.

<Configuration and Operation of Timestamp Correction System>
The configuration and operation of the time stamping unit 122 according to this embodiment will be described below.

First, the case without a time stamp correction system will be described. In this configuration, the _TAFE at the AFE clock 115 has a large error due to low power consumption and cost reduction, and may deviate by a maximum of several minutes per day. This error is a problem for biosensors that measure over a day. Details are described below.

2A and B show examples of packet transmission modes 131_1 and 132_1 and packet reception modes 131_2 and 132_2 when timestamp correction is not performed. Here, one packet length Lp corresponds to eight pieces of sensor data 2 . White circles 3_1 to 3_4 indicate packets, and T1 to T4 indicate packet arrival intervals.

If the clock 115 of the AFE is faster than the startup interval of the MPU 113, as shown in FIG. 2A, it may hold two or more packets of data (for example, packets 3_2 and 3_3) at startup (131_1). In this case, since the packets 3_2 and 3_3 are continuously transmitted, the packet arrival interval time T2 on the receiver side becomes very short (132_1).

Also, if the AFE clock is slower than the MPU startup interval, the amount of data in the packet to be transmitted at startup is insufficient (dotted line white circle 3'), as shown in FIG. 2B. At this time, no packet is transmitted (131_2). In this case, the packet arrival interval time T4 becomes longer on the receiver side (132_2).

In this way, in a configuration without a time stamp correction system, the arrival interval time of packets fluctuates greatly depending on the AFE clock, and errors occur due to transmission errors during wireless transmission, so the time stamp at the time of reception is used as it is. Then the error increases.

Next, the time stamp correction system 121 and the time stamp correction system according to this embodiment will be described. FIG. 3 shows the configuration of the time stamp correction system 121 according to this embodiment. Also, FIG. 4 shows a flow chart of the time stamp correction method according to the present embodiment.

The time stamp correction system 121 includes a delay circuit 1211, a subtraction section 1212, a packet number estimation section 1213, a calculation section 1214, a moving average filter 1215, and an addition section 1216, as shown in FIG.

Here, the time stamp correction system 121 acquires and retains the transmission interval time T _packet in advance. The transmission interval time T _packet may be pre-stored in the time stamp correction system 121 or transmitted from the sensor 11 .

The time stamp correction system 121 receives the arrival time T _arrival given to the packet. Here, the _arrival time T_arrival is the same as the time stamp _{T_stamp} .

The delay circuit 1211 is composed of a one-stage delay circuit, and delays the arrival time T _arrival (time stamp T _stamp ) of the received packet (one packet), for example, T _arrival [i−1].

The subtraction unit 1212 calculates the arrival time T _arrival (for example, T _arrival [i]) of the subsequently received packet (another packet) and the T _arrival (for example, T _arrival [i−1]) delayed by the delay circuit. (eg, T _arrival [i]-T _arrival [i-1]) is calculated (step S11). The difference between the arrival times T _arrival (time stamp T _stamp ) is the arrival interval time T _interval [i].

Packet number estimation section 1213 estimates the number of packets from arrival interval time T _interval [i].

Here, as shown in FIG. 5, the arrival interval time is distributed with noise with respect to integral multiples (n) of the packet transmission interval time T _packet from the sensor. Therefore, in order to cancel this noise, the estimated value of T _packet is set to n=1, and n=0, 1, 2, . ).

Quantization by the number of packets will be described with reference to FIGS. 6A to 6D, taking as an example packets received when the AFE clock is faster than the MPU activation interval.

Packets P(i) to P(i-5) are received and their respective arrival interval times are T _interval (i) to T _interval (i-4) (FIG. 6A).

First, let the time between packets to be received be the transmission interval time T _packet .

Each of the actually received packet arrival interval times T _interval (i) to T _interval (i−4) is compared with T _packet , and the packet arrival interval times T _interval (i) to T _interval (i−4) are calculated. When is equivalent to n times T _packet , let the quantization number be n (FIG. 6B).

Specifically, since T _interval (i) is equivalent to T _packet , the quantization number is "1". Similarly, T _interval (i−1) is equivalent to T _packet , so the quantization number is “1”. Next, since T _interval (i-2) is shorter than T _packet , the quantization number is "0". Next, T _interval (i-3) is equivalent to T _packet , so the quantization number is "1". Similarly, T _interval (i-4) is equivalent to T _packet , so the quantization number is "1".

In this way, the quantized inter-arrival time varies from 1→1→0→1→1 (FIG. 6C).

Next, the quantization number of the quantized arrival interval time, that is, the total number of packets is calculated (step S13). In this case, 1+1+0+1+1=4. Next, the average of the quantized arrival interval times is calculated as an estimated value of the number of packets (step S14). In this case, since the number of arriving packets is 5, the average value is 4/5. In this way, the quantized inter-arrival time is flattened and the variation of the inter-arrival time is suppressed (Fig. 6D).

As a result, the real value of the arrival interval time is converted to an integer value, and noise can be reduced.

Next, the calculation unit 1214 restores (converts) the quantized arrival interval time (integer value) to a real value. That is, inverse quantization is performed. Specifically, the transmission interval time T _packet is multiplied by the quantized arrival interval time (integer value) (step S15). In the above example, (4/5)*T _packet is calculated.

Next, the inversely quantized arrival interval time is input to the moving average filter 1215 to perform moving average (step S16). Here, a moving average value is calculated using the average value of T _intervals that is subsequently obtained as the average value of T _intervals .

For example, as shown in FIG. 7, for an arrival interval time from an arbitrary (i-th) arrival interval time T _interval [i] to an arrival interval time T _interval [i−N+1] preceding N-1 times, Calculate the moving average.

Here, a simple moving average, a weighted moving average, an exponential moving average, etc. can be used as the moving average.

The effect of the processing (moving average) of the moving average filter 1215 will be explained below.

A cutoff frequency Fc of the moving average filter 1215 is designed to be lower than the Nyquist frequency. Here, the Nyquist frequency is a frequency represented by f _packet /2 when 1/T _packet =f _packet .

As shown in FIG. 8, quantization noise 151 is distributed over the entire frequency range in the process of quantization and inverse quantization. The quantization noise 151 is noise caused by information (analog value) lost during quantization.

Also, when sensor data is actually transmitted in a packet, the packet has a high frequency signal 152 such as a measurement signal and a low frequency signal 153 including information on time stamps. Here, the high frequency signal 152 varies in seconds and the low frequency signal 153 varies in minutes.

By setting the cutoff frequency Fc of the moving average filter 1215 to be lower than the Nyquist frequency, the moving average filter 1215 functions as a low-pass filter, blocking the high frequency signal 152 and allowing the low frequency signal 153 to pass therethrough. can correct timestamps contained in .

Here, since the low frequency signal 153 fluctuates in minutes while the high frequency signal 152 fluctuates in seconds, it is effective to set the cutoff frequency Fc to 1/60 or less of the frequency of the high frequency signal 152.

Also, since the quantization noise 151 is uniformly distributed with respect to frequency, the noise can be reduced by f _packet /2Fc.

Finally, the adder 1216 adds the initial time _T0 , which is an offset, to the calculated moving average value of the arrival interval time to obtain the time stamp correction value _{T_correct} (step S17).

Here, regarding the determination of the initial time _T0 , whether the first arrival time is used or the method of least squares is used for estimation, only an absolute time error of about several tens of ms occurs. The effect is small compared to before the application where the deviation occurs.

Further, in the time stamp correction system 121, by configuring the moving average filter 1215 in multiple stages, the slope 154 of the low-pass filter characteristic of the moving average filter 1215 becomes steeper, the jitter component can be reduced, and the jitter component can be reduced with higher accuracy and stability. Can block high frequency signals. On the other hand, there is a delay in tracking the variation of T _packet , but this delay is not a problem in a biological data measurement system that cuts DC components.

In addition, since the speed of change of T _packet is very slow depending on temperature change and aging deterioration, there is little effect of narrowing the bandwidth of the filter.

Further, in the time stamp correction method, the output of the moving average filter 1215 may be fed back to the step of quantizing the arrival interval time (step S12) to calculate the moving average value again (dotted line arrow in FIG. 4). . As a result, the T _packet changes with the passage of time, so by performing the calculation again using the updated T _packet , it is possible to correct the time stamp with higher accuracy.

According to the timestamp correction system according to the present embodiment, the moving average filter can remove noise, mitigate delays, and reduce timestamp errors.

Furthermore, when a packet transmission error occurs, two or more packets are continuously transmitted immediately afterward, so the T _interval contains temporally correlated noise. Therefore, this noise can be easily removed by the moving average filter 1215 . That is, rather than using the arrival time including noise and delay as it is, using the transmission interval time T _packet as a reference can reduce jitter and obtain a time stamp with high precision.

It is also useful when real-time correction is required in a system where packets arrive on the stream. In addition, it is useful in that it can be applied to a wider range of applications than a system that performs batch processing on a server.

In addition, since signal processing basically consists of delay and sum-of-product operations, acceleration processing by a DSP (digital signal processor) can be applied, making it suitable for real-time data processing.

In addition, since correction is performed with clocks synchronized by GPS, NIP, NITZ, etc., correction can be performed with high accuracy.

<Second Embodiment>
A timestamp correction system, a sensor system, and a timestamp correction method according to a second embodiment of the present invention will be described with reference to FIGS. 9 and 10. FIG.

<Configuration and Operation of Timestamp Correction System>
A sensor system 20 according to the present embodiment has substantially the same configuration as that of the first embodiment, except for the configuration of a time stamp correction system 221 .

FIG. 9 shows the configuration of the time stamp correction system 221 according to this embodiment. Further, FIG. 10 shows a flowchart of the time stamp correction method according to this embodiment.

As shown in FIG. 9, the time stamp correction system 221 includes a multi-stage delay circuit 2211, a subtraction section 2212, a packet number estimation section 2213, a calculation section 2214, a moving average filter 2215, and an addition section 2216. .

The time stamp correction system 221 receives the arrival time T _arrival (time stamp T _stamp ) given to the packet.

The delay circuit 2211 is composed of M stages of delay circuits, and delays the input T _arrival (for example, T _arrival [i−1]). Here, _{T_arrival} is the same as _{T_stamp} .

The subtraction unit 2212 uses the outputs of the M-stage delay circuits to calculate the difference between the arrival times, that is, the arrival interval time (step S21). As a result, M inter-arrival times are obtained.

The packet number estimation unit 2213 quantizes the arrival interval time, and calculates the average value of this quantization number as the estimated number of packets (steps S22 to S24).

Next, the calculation unit 2214 multiplies the transmission interval time T _packet by the quantized arrival interval time, inversely quantizes it, and converts it into a real value (step S25).

Here, as described above, since the arrival interval time is multiplied by M by the M stages of delay circuits, the quantized arrival interval time is multiplied by the T _packet value and divided by M (step S26).

Next, the arrival interval times that have been inversely quantized and divided by M are input to the moving average filter 2215 to perform moving averaging (step S27).

Finally, the adder 2216 adds the initial time _T0 , which is an offset, to the calculated moving average value of the arrival interval time to obtain the time stamp correction value _{T_correct} (step S28).

Here, as in the first embodiment, the output of the moving average filter 2215 may be fed back to the step of quantizing the arrival interval time (step S22) to calculate the moving average value again (FIG. 10). middle, dotted arrow).

In addition, multi-stage delay circuits can remove correlated noise before the number of packets is quantized. In addition, since the data multiplied by M is multiplied by 1/M in inverse quantization, the noise itself due to inverse quantization can be reduced.

In addition, since the time stamp correction system requires M stages of delay in the process of estimating the number of packets, noise is reduced as the number of stages (M stages) of the delay circuit increases, but the real-time performance is reduced. Therefore, it is necessary to adjust the number of stages (M stages) of the delay circuit depending on the application. However, since the T _packet to be tracked fluctuates on a minute-by-minute basis, a delay of about 10 seconds is considered to have no effect on real-time performance.

<Third Embodiment>
A timestamp correction system and sensor system according to a third embodiment of the present invention will be described with reference to FIG.

<Configuration of sensor system>
As shown in FIG. 11, the sensor system 30 according to the present embodiment includes sensors 31_1 to 31_M and mobile information terminals 32_1 to 32_M such as smartphones. A time stamp correction system according to the form of and a time stamp applying unit. Here, a time stamp correction system according to the second embodiment may be provided.

In the sensor system 30, data acquired by the sensors 31_1 to 31_M are collected by the mobile information terminals 32_1 to 32_M. The collected data is given a time stamp in the portable information terminals 32_1 to 32_M, the time stamp is corrected, transmitted to the network system 4, and handled in a form such as a cloud.

According to the sensor system according to the present embodiment, the time stamp correction system can remove noise, mitigate delays, and reduce time stamp errors. In addition, since the correction is performed with a clock synchronized by GPS, NIP, NITZ, etc., the correction can be performed with high accuracy.

Furthermore, for example, in Android and iOS, which are typical OSs for smartphones, the resolution of timestamps is limited to 1 ms, so the achievable jitter value may be limited. According to the sensor system according to the present embodiment, the problem is reduced to the number of packets arriving at an arbitrary time interval, and the precision of the time stamp itself is not a problem. Therefore, the sensor system according to the present embodiment is effective when performing software processing on the OS of the smartphone.

Further, in the sensor system according to the present embodiment, one smartphone may be configured to connect to multiple sensors and provide time stamps. In this configuration, the effect of delay is large, and the effect on other applications is also large. Therefore, a configuration in which a single smartphone is connected to a single sensor is more desirable than a configuration in which a plurality of sensors are connected to a single smartphone.

<Fourth Embodiment>
A timestamp correction system and sensor system according to a fourth embodiment of the present invention will now be described with reference to FIG.

<Configuration of sensor system>
A sensor system 40 according to the present embodiment includes sensors 41_1 to 41_M, data collection terminals 42_1 to 42_N, and a server 43, as shown in FIG. Here, the data collection terminals 42_1 to 42_N are provided with time stamping units, and the server 43 is provided with the time stamp correction system according to the first embodiment. Here, a time stamp correction system according to the second embodiment may be provided.

In the sensor system 40, the data acquired by the sensors 41_1 to 42_M are collected by the data collection terminals 42_1 to 42_N and time stamped. The time-stamped data is transmitted to the server 43, the time stamp is corrected by the server 43, transmitted to the network system 4, and handled in a form such as a cloud.

In this way, the sensor system 40 is configured such that the data collection terminals 42_1 to 42_N only apply time stamps, and the server 43 corrects the time stamps.

According to the sensor system according to the present embodiment, the time stamp correction system can remove noise, mitigate delays, and reduce time stamp errors. Furthermore, since the data collection terminal can allocate resources only to giving time stamps, it is possible to prevent degradation of time stamp accuracy due to calculation in an environment where a plurality of sensors are connected to the data collection terminal.

In a normal sensor network, higher servers (closer to the network system) have higher computing power, and lower sensors (farther from the network system) have lower computing power. is useful in terms of resource concentration.

In particular, in the sensor system according to the second embodiment, when a plurality of sensors are connected, packet transmission fails more frequently. It becomes correlative and can maintain accuracy.

In the embodiment of the present invention, an example of the structure, dimensions, etc. of each constituent part has been shown in the configurations of the time stamp correction system and the sensor system, the time stamp correction method, etc., but the present invention is not limited to this. Any configuration of the time stamp correction system and the sensor system, as long as the function of the time stamp correction method is exhibited and the effect is produced.

The present invention relates to a time stamp correction system, a sensor system, and a time stamp correction method, and can be applied to systems and communication systems that transmit and receive data acquired by sensors.

121 time stamp correction system 1211 delay circuit 1212 subtraction unit 1213 packet number estimation unit 1214 calculation unit 1215 moving average filter 1216 addition unit

Claims

A time stamp correction system for correcting a time stamp given to each of a plurality of packets when a plurality of packets transmitted at a predetermined transmission interval time are received,
a delay circuit for delaying the timestamp of one of the plurality of packets;
a subtraction unit that calculates, as an arrival interval time, the difference between the time stamp of another packet received following the one packet and the time stamp of the delayed one packet;
a packet number estimation unit that quantizes the arrival interval time and calculates an average value of the quantized number as an estimated value of the number of packets;
an arithmetic unit that calculates an inversely quantized arrival interval time by multiplying the transmission interval time by the estimated value of the number of packets;
a moving average filter that performs a moving average on the dequantized inter-arrival times;
and an addition unit that adds an initial time to the value obtained by the moving average.
the packet includes a signal acquired by a sensor and a signal with a time stamp;
2. The time stamp correction system of claim 1, wherein the moving average filter blocks signals acquired by the sensor and transmits signals bearing the time stamp.
3. The time stamp correction system according to claim 1, wherein the delay circuit has multiple stages.
a sensor that transmits a packet generated from a measurement signal at a transmission interval time;
a time stamping unit that receives the packet and adds a time stamp;
A sensor system comprising: a timestamp correction system according to any one of claims 1-3.
the sensor;
5. The sensor system according to claim 4, further comprising: a portable information terminal that includes the time stamping unit and the time stamp correction system, and that transmits the corrected time stamp to a network.
the sensor;
a data collection terminal comprising the time stamping unit;
5. The sensor system of claim 4, comprising: a server comprising the timestamp correction system and transmitting the corrected timestamp to a network.
A method for correcting timestamps given to packets when a plurality of packets transmitted at a predetermined transmission interval time are received, comprising:
a subtraction unit calculating the difference between the timestamps as an arrival interval time;
a packet number estimating unit quantizing the arrival interval time;
a step in which the packet number estimating unit calculates an average value of the quantization number of the quantized arrival interval times as an estimated value of the number of packets;
a step in which the calculating unit multiplies the transmission interval time by the estimated value of the number of packets to obtain an inversely quantized arrival interval time;
a moving average filter performing a moving average on the dequantized inter-arrival times;
A time stamp correction method comprising: adding an initial time to the value obtained by the moving average.