The television industry is quickly adopting digital networking technologies for content distribution. As a consequence, systems are being organized around the concept of multilayered protocol stacks. There are three basic layers encountered in today’s digital video: the program, the protocol and the transmission layers. Each layer type has distinct measurements associated with it.
Equipment failures or other operational problems that occur at one layer will have a direct effect on neighboring layers. The discovery and analysis of such situations is called cross layer test and monitoring. This article will examine various types of problems that can occur in a layer and discuss some of the implications on neighboring layers. It also will describe some of the measurements that can be made at each layer and various dependencies between these measurements.
Cross-layer monitoring systems
In cross-layer monitoring systems, probes are divided into three broad categories: transmission probes, protocol probes and program probes. Each probe sends results to one or more measurement servers, typically via high-speed IP connection such as SNMP.
This system provides top-down testing and bottom-up testing. Top-down testing identifies a particular fault at the program layer and makes various inferences, which result in tests at the protocol layer. Suppose, for example, the picture quality drops with a corresponding increase in blockiness. This situation is often the result of bit-rate starvation at the protocol layer, which can be easily tested. It is important to make this cross-layer inference and to verify it with bit-rate measurements to correctly identify the initial source of the problem.
Bottom-up testing infers from the transmission layer to the protocol layer or from the protocol layer to the program layer. An example of this is the identification of excessive time stamp jitter in MPEG transport streams, which can have serious effects on the final decoded video. Many decoders, however, can correctly decode high-jitter streams with no loss in picture quality. Testing in the transport stream is necessary to show whether picture quality measurements at the program layer are needed.
It can be used to detect loss of picture quality due to the following:
Intermittent signal loss. Many times signal errors are obvious at many layers. The Tektronix PQM300 measures picture quality at the program layer in order to expose picture noise, block-iness and frozen frames in a graphical form, especially when errors are severe enough to be obvious and annoying to most viewers.
As the PQM reports blockiness artifacts, a Tektronix MTM300 looks at the protocol layer for a number of different protocol and timing errors. As with the PQM, the MTM reports on errors such as program clock reference(PCR), drift rate errors, missing presentation time stamp/decode time stamp (PTS/DTS) errors and continuity counter errors in a graphical display.
These errors are each accompanied by a time stamp, allowing correlation by a measurement server. With these two sets of readings, an intelligent measurement server will infer that the RF layer is likely to show signs of errors as well.
From this collection of multilayer artifacts, a report can be generated.
Packet loss artifacts at the program layer. This common problem is the result of decoding with missing packets at the protocol or transmission layer. The artifacts created depend highly on whether the lost packet comes from an I frame, P frame or B frame.
A packet loss in the B frame will have a negligible effect on the picture quality and will affect only the single frame. A packet loss in an I frame will result in a propagation of bad blocks across the entire group of picture (GOP) structure. An example of a subtle packet loss in the B frame is shown in Figure 1.
It is possible to measure this packet loss at the program layer. The graph in the inset of Figure 1 shows two superimposed graphs of blockiness levels — before packet loss and at the receiving end of a transmission that encountered a single packet loss. A single packet loss will result in a slight elevation of the blockiness level for the single B frame. The impact is shown in the graph as a red line where the two measurements are no longer the same.
Figure 2 on page 63 shows a much more severe example of packet loss. Three frames of video are shown.
The last frame shows severe packet-loss artifacts caused by a packet loss in a neighboring I frame. This packet loss is barely discernible in the I frame, but its effects are severe in the B frame. Notice that the severe blockiness that occurs in the third frame takes place in the background blocks but not the foreground blocks.
Systems elements to consider when choosing a monitoring system include:
Clock synchronization. Comparing measurements from separate probes requires a coordination and synchronization of time.
Inference engine. Although experienced operators can draw inferences from the results of various probes, a more automated solution would use an inference engine. This is a software module that maintains a collection of rules and actions.
No system can test for every condition all the time. Therefore, it is necessary to draw conclusions and to make inferences from a subset of measurements.
It also is important to acknowledge that most errors or measured artifacts have a root cause associated with a particular layer. Measurements from neighboring layers clarify system problems and help determine what further tests might be applied or what solutions might be available.
Greg Hoffman is a product marketing manager for Tektronix.