Wireless Network Synchronization for Multichannel Multimedia ServicesSpeaker : Yi-Jie PanAdvisor : Dr. Kai-Wei KeDate : 2014/11/04

2

Outline Introduction

Wireless Multimedia Application Model

Precise Time Protocol (IEEE 1588)

Wireless LAN Synchronization Implementation

Timer Synchronization Performance

Conclusions

References

3

Introduction Broad band networking has become more widely supported by

wireless local area network (WLAN) technologies, especially by the IEEE 802.11 family.

Hence, WLAN is expected to develop their applications for multimedia contents distribution for multi-device audio and video playback.

Digital contents distribution over a network can reproduce high quality audio and video contents.

For example, the available bandwidth of IEEE 802.11n to the UDP client is as high as 90 Mbps, which can transmit several channels of high definition television (HDTV) contents by use of MPEG2 and MEPG4 formats.

4

Introduction (Cont.) A critical issue of high-fidelity (Hi-Fi) digital contents playback is

uncertainty in the play timing, for applications where audio and video devices are separated in space and thus digital reproduction of stereo sound and moving pictures are provided at different devices with respect to different time references.

In order to achieve such Hi-Fi digital multimedia playback, each device has to provide an accurate clock reference to the playback applications.

5

Introduction (Cont.) Such time reference can be used for a two-fold benefit

to control packet transmissions at a constant bit rate (CBR) in plesiochronous manner.

to control playback time of the audio and video applications.

Time and rate-guaranteed packet transmission over WLAN is a challenging technology because the air channel is very unreliable transmission medium especially when CSMA/CA is used.

6

Introduction (Cont.) Network synchronization is implemented in IEEE 802.11 inherently. In

a typical WLAN configuration, an AP broadcast time information in its beacon frames to its subordinate stations (STA).

In this way, STAs can be tuned to the AP within an accuracy of a few microseconds.

However, current IEEE 802.11 standard STA devices do not provide such time information to its client applications.

There are several prior works that report timing synchronization performance of IEEE 802.11 and IEEE 1588. However, performance analysis under high traffic volume, such as for multimedia streaming, has not been reported.

7

Introduction (Cont.) In this paper, we focus how accurate network synchronization can be

attained from commercially available standard IEEE 802.11n STA devices for multimedia services, when the synchronization is achieved by clients of the network.

Commercially available wireless routers are used to implement time synchronization by the use of IEEE 1588 precision time protocol (PTP).

Implementing PTP functions in Linux-based routing processes of WLAN devices, we achieve clock synchronization based on measurements of clock difference (offset) and delay time between a pair of an AP (as a clock master) and an STA (as a clock slave).

8

Introduction (Cont.) Statistic filtering is applied to increase accuracy of offset estimations

at STAs.

All implementation is accomplished without requiring a new hardware; the timer for synchronization is borrowed from the Linux processor clock.

Background traffic up to 72 Mbps is applied to make sure PTP measurements are accurate and acceptable under multimedia streaming with high traffic volume.

9

Introduction (Cont.) Our results suggest that IEEE 802.11n is a very strong candidate for

Hi-Fi wireless multimedia service with a typical maximum service throughput of 80 Mbps and a delay jitter less 430μsec, which will be possible to support stereo sound reproduction.

10

Wireless Multimedia Application Model – Wireless Multimedia Playback System In the multimedia system

model, each audio or video contents is streamed in a separate channel, thus requiring streaming synchronization.

Both the network streaming and playback application need synchronized timer at every device.

11

Wireless Multimedia Application Model – Contents Synchronization Requirements

12

Precise Time Protocol (IEEE 1588) – Definitions The packet delay of the -th packet of a given packet stream is the

time interval between the instant when the start of the packet is leaving end-system A and the instant when it is entering end-system B.

As will become clear later on, we also need to define the packet delay asymmetry of a pair of packets which are part of a simple bi-directional packet exchange

is the delay of the request packet, and is the delay of the response packet.

13

Precise Time Protocol (IEEE 1588) – Two-way Time Transfer In order to compensate packet

delays, PTP uses the well known Two-Way Time Transfer (TWTT) technique.

Figure 1 shows the TWTT packets which are exchanged between the PTP master clock and the PTP slave clock in order for the slave to determine and correct its time offset relative to the master.

14

Precise Time Protocol (IEEE 1588) – Two-way Time Transfer (Cont.) The diagram shows the two

time-critical packets typical of all TWTT-based protocols.

The slave exploits the measured transmit and receive times of the two time-critical packets in order to calculate the clock offset ( represents the real offset, represents the estimate).

15

Precise Time Protocol (IEEE 1588) – Two-way Time Transfer (Cont.) The calculation uses the well

known formula below, which is based on the assumption that the delays in both directions and are equal.

Because of the presence of a packet delay asymmetry the calculated offset is affected by the following error

16

Precise Time Protocol (IEEE 1588) – Mitigating Packet Delay Asymmetry How does PTP mitigate the detrimental effects of packet delay

asymmetry? The basic idea is to somehow remove or compensate those delay components which are likely to be asymmetrical.

The delays generated by the links are approximately symmetrical.

The node delays, however, are often asymmetrical because of the queues and because of asymmetrical traffic loads.

PTP uses a combination of techniques in order to remove or compensate these delay components.

17

Precise Time Protocol (IEEE 1588) – Mitigating Packet Delay Asymmetry (Cont.) The first technique consists in

measuring the four timestamps of a TWTT interrogation (T1, T2, T3 and T4) as close as possible to the physical port, as shown in Figure 2.

The time-stamps are measured by the ‘Precise Time Stamp Generator’ when the packet traverses the Physical Layer of the protocol stack.

18

Precise Time Protocol (IEEE 1588) – Mitigating Packet Delay Asymmetry (Cont.) Any delay generated while the

packet is moving up or down the protocol stack, including queuing, is singled out, since the TWTT exchange takes place virtually between the physical layers of the master and the slave.

19

Precise Time Protocol (IEEE 1588) – Mitigating Packet Delay Asymmetry (Cont.) The second technique is

illustrated in Figure 3. The figures show a PTP path with two end-systems and two network nodes.

The two time-critical packets SYNC and DELA_REQ experience both the link delays and the node delays.

20

Precise Time Protocol (IEEE 1588) – Mitigating Packet Delay Asymmetry (Cont.) The node delays are measured

by the network nodes and the measured values ( for SYNC, and for DELAY_REQ) are transferred to the slave via the additional packets FOLLOW_UP and DELAY_RESP.

21

Precise Time Protocol (IEEE 1588) – Mitigating Packet Delay Asymmetry (Cont.) With this information the slave

calculates an estimate of the asymmetry and uses it for the calculation of the slave clock’s offset

22

Precise Time Protocol (IEEE 1588) – Mitigating Packet Delay Asymmetry (Cont.) The remaining error is now half the difference between the actual

and the estimated asymmetry

The formula above gives the error for a single TWTT interrogation.

PTP slaves use sophisticated processing and filtering techniques applied to consecutive TWTT interrogations in order to further reduce the residual clock offset.

23

Wireless LAN Synchronization Implementation In general, micro processor unit (MPU) clocks and hence the kernel

timers are free running independently causing timer skews between different device units.

Time synchronization compensates such timer skew by measuring offset and compensates timers for the offset.

To implement a time-synchronization module in IEEE 802.11 n, there are following possible considerations that are wireless communications characteristics such as CSMA/CA, automatic retransmission with a random back-off delay, packet loss and so on.

24

Wireless LAN Synchronization Implementation (Cont.) We adopt IEEE 1588 precise time protocol (PTP) as the base line of

our research. PTP exchanges time stamping information that denotes a PTP control message's arrival or transmission time.

In order to know these time stamp information, it needs a time-stamping mechanism.

25

Wireless LAN Synchronization Implementation (Cont.) Application layer time stamping (ATS) approach stamps a time mark

at application layer.

In ATS approach, it includes uncertainties of processing delay time or process scheduling delay when measures a timer offset between two devices.

To achieve more precise time synchronization in PTP, we need to know the actual timer value when a packet is transmitted or arrived.

But typically, commercial wireless AP devices or MAC chips do not support these functions and moreover, we cannot modify them.

26

Wireless LAN Synchronization Implementation (Cont.) For these reasons, we implement a time-stamping module at the

device driver layer as close as possible to MAC or PHY level time-stamping. We called it device driver layer time stamping (DTS).

By this approach, we can minimize the processing delay to measure the timer offset value between two devices at existing commercial AP devices without any PHY or MAC modifications.

27

Wireless LAN Synchronization Implementation (Cont.) For example, master sync

device saves a SYNC message's transmission time stamp information when a SYNC message is sent to the device driver from the network layer.

At that time, the DTS module saves current time information as actual transmission time of SYNC message, and later, kernel informs to application layer when an application sync module requests.

28

Wireless LAN Synchronization Implementation (Cont.) At the slave side, when the

SYNC message is received by slave's wireless interface, it is transferred from device driver to network layer.

At that time, DTS module detects the PTP message and saves current time information as SYNC arrival time at the kernel.

And also this information is used at the application sync module by a system call request.

29

Timer Synchronization Performance Using IEEE 1588 PTP messaging, we measure the timer skew

between an AP and STA device pair.

Since the time stamps are based on MPU clock, the time stamping references are limited to the MPU clock accuracy and the embedded Linux kernel performance.

Time stamping accuracy is first investigated in order to understand the IEEE 802.11n device characteristics.

The aforementioned two time-stamping functions in the application layer and in the kernel device driver layer are used to compare to see if the application layer implementation is acceptable.

30

Timer Synchronization Performance (Cont.) IEEE 1588 PTP gives a means of measuring time difference of the

timers at two devices.

At the slave device implemented at the STA, the measured time offsets are used to correct the local timer value.

However, a typical transmission delay performance of a IEEE 802.11n channel is very nondeterministic due to the nature of CSMA/CA, and automatic retransmission on frame losses.

In addition, the MPU kernel and application process uncertainty adds up the delay uncertainty in the frame transmission.

31

Timer Synchronization Performance (Cont.) Such time stamping uncertainty and delay jitter affects the timer

offset measurement between two devices, resulting in measurement errors.

In order to minimize the impact of such measurement error, we introduce statistical filtering before applying timer compensation at the slave STA device.

The statistical filtering is accomplished by the following steps Conduct n number of IEEE 1588 PTP offset measurements. Number is

widely varied between 1 and 100. Calculate the mean Po and standard deviation of the offset samples. Delete offset samples outside an interval defined by the boundaries at Calculate a new mean with filtered samples.

32

Timer Synchronization Performance (Cont.) Synchronization accuracy estimation: calculate the offset standard

deviation of m's after a large number of repetitions of Steps 1 to 4 in the above.

We find empirically optimized value in the range of , and we choose as it is not a sensitive value against the offset standard deviation performance metric.

In our characterization experiments, background traffic is added to the air medium in order to investigate the impact of extensive traffic.

33

Timer Synchronization Performance (Cont.)

34

Timer Synchronization Performance – Application Layer Time Stamping As discussed in the previous section, application layer time stamping

is the simplest form of implementing IEEE 1588 PTP.

In this case, the propagation delay consists of application scheduling delay by the kernel at the master, kernel process scheduling delay at the master, MAC access delay at the master, PHY and air propagation delay, MAC access delay at the slave, kernel process scheduling delay at the slave, and application scheduling delay by the kernel at the slave.

35

Timer Synchronization Performance – Application Layer Time Stamping (Cont.) In this regard, the nondeterministic process of automatic

retransmission by IEEE 802.11 is not included, as most of such cases are eliminated by the aforementioned statistical filtering.

The synchronization accuracy is measured under various traffic load conditions.

36

Timer Synchronization Performance – Application Layer Time Stamping (Cont.) There are two important

features to understand from this data First, statistical filtering is highly

effective, which can improve the performance by an order of magnitude.

Second, the impact of background traffic causes a large penalty, so the worst synchronization accuracy is 210ms.

37

Timer Synchronization Performance – Device Driver Layer Time Stamping As discussed in the previous sections, device driver layer stamping

can eliminate time delay jitter in the kernel's application scheduling.

We observe that this is a critical improvement when the traffic is highly offered.

In this case, the propagation delay consists of kernel process scheduling delay at the master, MAC access delay at the master, PHY and air propagation delay, MAC access delay at the slave, and kernel process scheduling delay at the slave.

38

Timer Synchronization Performance – Device Driver Layer Time Stamping (Cont.) There are three important

observations in this data First, overall synchronization

accuracy is improved by at least an order of magnitude.

Second, statistical filtering is highly effective, and even more efficient algorithm is anticipated to enhance the synchronization accuracy under a large traffic volume.

Third, at light volume this system can support tightly coupled audio applications with about 50μs accuracy as well as all other services.

39

Timer Synchronization Performance (Cont.) As the average sample size

increases the overall trend show reduction of offset jitter .

However, too large value of requires a long intervals between compensations, thus the performance can get worse.

In our filtering scheme n=25 or 50 can be chosen as an optimum.

40

Conclusions Multichannel multimedia wireless service feasibility is studied based

on IEEE 802.11n and IEEE 1588 technologies.

We investigate the time synchronization accuracy in the proposed system to report a potential of multichannel wireless service.

When the network is lightly load the system can provide tightly couple audio reproduction, such as stereo sound, with about 50μs timing accuracy in the network layer.

At the high traffic limit up to 72 Mbps application throughput, the timing accuracy of 430 μs can be provided when the time stamping is achieve at the interface of MAC/PHY hardware by a device driver level implementation.

41

References Yonghwan Bang, Jongpil Han, Kyusang Lee, et al., ” Wireless Network

Synchronization for Multichannel Multimedia Service,” ICACT 2009. 11th International Conference on Advanced Communication Technology, vol. 02, pp 1073 – 1077, 15-18 Feb. 2009

André Vallat, Dominik Schneuwly, ” Clock Synchronization in Telecommunications via PTP (IEEE 1588),” IEEE International Frequency Control Symposium, 2007 Joint with the 21st European Frequency and Time Forum, pp 334 – 341, May 29 2007-June 1 2007

42

Thanks for Listening