PCR and MDI Measurements for VOD Applications
Introduction
Transport of MPEG2 encoded video over Ethernet data networks has emerged as the architecture of choice for at least a part of many digital video distribution systems such as cable-based Video on Demand (VoD) and broadcast systems infrastructure. ISO/IEC standard 13818-1 details how MPEG encoded digital video streams and digital audio streams should be multiplexed, packetized and encoded into Transport Streams suitable for storage and transmission. The timing parameters that are required to guarantee product interoperability are also included but this specification does not address systems that then encapsulate the Transport Streams using Internet Protocol (IP) over Ethernet data networks.
This technical brief compares two key measurements, PCR characteristics and MDI (Media Delivery Index) which are critical in MPEG encoded video Transport Streams and for Ethernet data networks carrying video, respectively.
It is recommended that measurement of the video Transport Steam's PCR parameters generally be restricted to a program's network source location and then MDI be used to characterize the Ethernet distribution network as Ethernet networks can cause misleading PCR measurements.
MPEG Transport Stream Encoding Process and PCR
The Program Clock Reference (PCR) is defined by ISO/IEC 13818-1 and consists of samples of the encoder system time clock.
They provide a resolution of 1 in 27,000,000 and are required to appear in intervals of 100 ms in a Transport Stream (every 40 ms for DVB compliance). The video encoder and multiplexer which generates the Transport Stream packets use a common 27 MHz program clock reference and this clock must be reconstructed exactly at the receiver to assure proper regeneration of video at its output. The receiver's chroma sub-carrier, pixel clock, and frame rate may all be derived from a phase locked loop PLL) which is locked to the system clock as determined by updates arriving via the PCR in the Transport Stream. Since the receiver PLL state is updated from received PCR samples, the accuracy of the updates are required to be very precise.
Thus, the PCR must arrive at the time it was intended to as dictated by the PCR value itself (i.e. the system clock sample). Too much variation between when it was intended to arrive and when it did arrive will cause the PLL to lose lock, affecting the video output quality.
Reconstructing the encoder's clock at the receiver depends on the PCR value and the PCR position in the Transport Stream as the receiver determines its PLL correction value from these parameters. PCR impairments at the encoder may be due to the fixed PCR resolution of 37 ns available from the 27 MHz reference, drift or inaccuracies in the 27 MHz reference clock, inaccurate calculation of the PCR value, and errors in PCR position within an Ethernet packet. PCR tolerance as specified in ISO/IEC 13818-1 is 500 ns but specifically excludes the transport layer; that is, errors in packet arrival times due to network jitter or other network impairments.
Typically, in Transport Stream transmission over IP on Ethernet networks, seven Transport Stream Packets are combined in a single IP/Ethernet packet for network efficiency reasons. Seven 188 Byte Transport Stream packets result in 1362 Byte long Ethernet packets when combined with UDP/IP header information.
The result fits comfortably within the 1518 Byte maximum size IEEE 802.3 Ethernet packet thus minimizing the amount of total header information that must be transmitted for a stream and reducing the possibility of frames being fragmented by the network. Since each Ethernet packet is transmitted at the media rate of 1 Gigabit/second (typically), a receiver must buffer the seven Transport Stream packets for proper decoding and video presentation. This burst of Transport Stream packets can be extended as a result of network delays due to network component buffering resulting from localized traffic congestion or internal queue management algorithms.
The PCRs may then be sent several packets earlier or later than desired due to the Ethernet packet assembly algorithm. For these reasons the measurement of PCR impairments within a network can be misleading when referenced to ISO/IEC specifications not intended to account for network impairments. PCR jitter measurements made within an Ethernet network will typically include large amounts of jitter when compared to lower bandwidth, common clock interconnects such as DVB-ASI. This is normal. At a receiver, the packets are buffered and smoothed resulting in good quality video with ISO/IEC compliant PCR jitter as long as such buffers do not underflow or overflow. Preventing underflow and overflow is key to smoothing the PCR jitter as seen by the decoder.
Ethernet and MDI
To characterize the impairments that the network may induce, the MDI provides a relative indicator of needed buffer depths at a network consumer node due to packet jitter and network latency as well as an indication of lost packets. MDI provides the necessary information to detect all network induced MPEG video display impairments. By probing a streaming media network at various nodes and under varying load conditions, it is possible to quickly identify devices or locales which introduce significant jitter or packet loss to the packet stream. By monitoring a network continuously, deviations from nominal jitter or loss behavior can be used to indicate an impending or ongoing fault condition such as excessive load.
The MDI is composed of two components: Delay Factor and Media Loss Rate. The Delay Factor (DF) is the maximum difference, observed at the end of each media stream packet, between the arrival of media data and the drain of media data, assuming the drain rate is the nominal constant traffic rate of media stream packet data (i.e. stream bitrate). The minimum DF will be an Ethernet packet's worth of media data as that is the minimum amount of data that must be buffered.
This buffered media data in bytes is expressed in terms of how long it would take to drain (or fill) this data at the nominal constant traffic rate in milliseconds to obtain the DF. The Delay Factor indicates how long a data stream must be buffered or delayed at its nominal constant bit rate to prevent packet loss. Thus, it is proportional to latency.
The Delay Factor gives a hint of the size of the buffer required at the next downstream node. As a stream progresses, the variation of the Delay Factor indicates packet bunching and packet spacing (jitter). The DF is comprised of a fixed part based on packet size and a variable part based on the various network component switch elements' buffer utilization that comprise the switched network infrastructure.
The Media Loss Rate (MLR) is the number of lost or out of order Media packets counted over a selected time interval, where Media packets are packets carrying media information -- typically Transport Stream packets. There may be 0 or more media packets in a single layer 2 network packet such as Ethernet. Recalling that it is common to carry seven 188 Byte MPEG Transport Stream packets in a 1362 Byte Ethernet frame, a single Ethernet frame loss would result in 7 lost Media packets.
Combining these quantities for presentation results in the MDI made up of DF and MLR expressed as: DF:MLR, where DF is the Delay Factor and MLR is the Media Loss Rate shown in Figure 1-1.
DF difference indications between one node and another in a data stream can indicate local areas of traffic congestion or possibly misconfigured QOS flow specification(s) leading to greater filling of measurement point local device buffers, resultant flow rate deviations, and possible data loss. A high DF may cause downstream buffer overflow or underflow or unacceptable latency even in the absence of lost data as illustrated in Figure 1-2.
Due to transient network failures or DF excursions, packets may be lost within the network. The MLR component of the MDI shows this condition.
Figure 1-2 Multiple MDI Measurements across a Distributed Network. The broadcast media streamed from the Headend to New York show and MDI value of 4.37:0 at the headend. Downstream (New York), the same streaming media shows a DF of 7.18 and MLR of 2, which is an indication of a corrupted video stream cause by excessive IP packet delay and loss of MPEG data.
Figure 1-3 Simplified Block Diagram of a VoD Network
In a typical VoD system, as shown in Figure 1-3, a server farm provides hundreds or thousands of video streams to a network interconnect infrastructure in response to requests from viewers' set top boxes (STBs). Program content is MPEG-encoded off line prior to being loaded on the servers. The servers provide cost effective, industry standard Gigabit Ethernet network connections to the distribution network for delivery to a STB.
PCR verification along with other MPEG Transport parameters are important measurements for stored content. Since stored content is pre-encoded, off-line analysis of the content will ensure program encoding integrity before transmission.
A low jitter transfer to Gigabit Ethernet from the server with MDI of less than 8:0 for 3.75 Mb/s streams minimizes network component buffer demand keeping latency low and helps to minimize the chance of downstream device buffer overflow.
If a server is re-encoding or modifying the Transport Stream as it is being sent, real time PCR analysis at the source may be desired but network time distortions must still be taken into account for PCR measurements as discussed above. Otherwise, if the Transport Stream has no impairments when entering the network, no further MPEG analysis is required or useful until the stream is buffered and decoded at a receiver. The MPEG Transport Stream characteristics are not changed by the distribution network.
Missing or out-of-order packets or network induced jitter is best tracked throughout the network distribution infrastructure by MDI -- the measure specifically tailored for detecting streaming media impairments caused by network infrastructure issues such as accumulating jitter or loss due to faulty or miss-configured components.
PCR measurements and other MPEG Analysis may be performed reliably using Deferred Time Analysis (DTA) by recording a stream at any location and post processing. By recording a stream under test, real time network-induced jitter will be removed just as it would be at a decoder/receiver. The content's MPEG transport parameters may then be accurately analyzed.
Conclusion
Streaming Video transported by packet switched networks is an emerging technology that requires an appropriate selection of quality measurements. Assuring high quality at the user points of such networks requires measuring and monitoring the MPEG quality of the program sources via PCR and other ETSI TR 101 290 parameters and network component performance using MDI. PCR jitter measurements are not recommended within the Ethernet interconnect due to the normal network induced jitter from packetization and switch queuing but may be used at ASI or QAM interfaces. Measurement and monitoring of MDI at several strategic locations in the network is recommended to detect and rapidly locate the cause of network induced impairments. Download |
Ken Chiquoine, SeaChange International 
James Welch, IneoQuest Technologies, Inc.
