October 2000

BY BRIAN CARR

A lot of attention has been focused on the needs of large-scale, carrier-class voice-, fax-, and data-over-packet gateways and networks where there are specific requirements for service and billing management, quality of service targets, and system maintenance. When considering system architectures for infrastructure grade gateways, developers have tended to concentrate on the limitations of the signal processing capabilities, but this is only one element of the entire processing chain of a media gateway.

OPEN SYSTEM MEDIA GATEWAY DESIGN
With commonly available hardware and software, equipment developers no longer need to invest heavily in proprietary technology. Instead, they can focus on gaining a competitive advantage through enhancing functionality -- such as configuration, management, and billing -- and rely on OEMs to provide standardized building blocks. For infrastructure grade systems, CompactPCI technology is increasingly accepted as the standard.

This can reduce the time to market and development risk of any new technology, and ultimately lower the total cost of ownership.

An open-system based media gateway comprises a number of parts:

• A single-board computer (SBC) such as Motorola's CPV5350 Pentium class SBC or Sun's CP1500 SPARC SBC;
• One or more digital signal processor (DSP) resource boards; and
• A network line interface, trans-coding software, and a board enclosure.

The basic building blocks for a voice-over-packet system are a media gateway, media controller, and a signaling gateway. A media gateway is required to convert between the digitized analog signals found on the telephone network and the stream of data packets on the packet network.

Internally, the media gateway transcodes voice, fax, and data between circuit-switched and packet-switched protocols, essentially translating T1/E1 voice circuits to IP packets over Ethernet, frame relay, or ATM. The transcode -- often involving voice compression, echo cancellation, tone detection, etc. -- is a complex task that is best performed by a digital signal processor.

DESIGNING AN OPEN SYSTEM MEDIA GATEWAY
The best configuration for a media gateway depends on your overall system requirements, as determined by your application. When considering the system architecture, you must consider factors such as channel density, call quality, and functionality, as they all have a bearing on the size and the complexity of the completed media gateway. For example, for reasons of cost efficiency we may wish to maximize the system's channel density, giving us as many simultaneous call channels to use as possible.

In theory, the limitation on the number of channels is the number of T1/E1 lines present to carry the call data, or the amount of channels each DSP resource board is physically able to process. However, the reality is slightly more complex. While attempting to achieve maximum channel density it is also necessary to maintain a minimum call quality standard as well as to provide added-value features designed to differentiate a product. It is also essential to consider the distribution of processing tasks across the functional modules in order to get the most from the system.

A Question Of Balance
Ultimately, there is little point in being able to increase the performance of one processing stage or the other, in isolation, because performance will always be limited by the slower of the two sides in the system. Ideally, the designer's goal should be to completely balance performance.

A balanced system means equalling out the demands placed on both processing modules by careful choice of hardware, system architecture, and functionality, while bearing in mind the demands of your application. The optimum system configuration occurs when the two sides -- DSP and microprocessor -- are close enough in performance so that neither has a negative impact on channel density. Thanks to its flexibility, the open system approach enables you to do this.

To assist you in selecting the best architecture for your application, the following sections outline three approaches to open system media gateway design, highlighting some of the situations to which they are best suited.

Simple And Scalable Gateway
In this architecture all media gateway functionality is performed on the DSP resource board. This is a typical architecture used where the signaling interface is relatively simple and can easily be associated with the channels that are directly terminated by the DSP resource board.

PSTN lines are terminated directly on the DSP resource board. The DSP board performs all the telephony signaling and network protocol processing (including H.323 call signaling) as well as any voice/fax processing. The CompactPCI host CPU is responsible for initial set-up and configuration of the system, and monitors each individual board for errors or alarms. It can run element manager software that allows this gateway to be managed from a central management system across its own private IP connection. Increasing the channel density of this system is straightforward, since the DSP board performs all channel dependent processing. It's simply a case of increasing the number of DSP boards.

However, the disadvantage of this approach is that the channel density is limited to the number of channels terminated on each board, i.e., the number of T1/E1 lines the DSP resource board is capable of supporting.

If, for a particular voice and fax configuration, the DSP resource board is capable of more channel capacity than can be provided on the direct line termination, then one way to improve overall density is to utilize external line termination capability.

External Signaling And Line Termination Board
To beat the T1/E1 line limitation, an additional open systems line interface board is provided to handle the line termination function. The voice channels are passed to the DSP board via a backplane timeslot bus (ECTF H.110).

This architecture allows for more complex telephony signaling such as SS7, where signaling for all the channels can be carried on a few resilient link sets, or alternatively it can allow for very-high-density termination (such as T3 or E3). The DSP board still performs all the voice and network protocol processing (including H.323 signaling). Now the CompactPCI host single-board computer performs additional functions, for example acting as the signaling gateway and the Media Gateway Controller (MGCP) call agent for the system.

One advantage of this approach is that the number of channels directly terminated on the board no longer limits the channel handling capacity of the DSP board. This enables the amount of DSP power required to be more accurately matched to the number of channels terminated on accompanying line interface board(s). The on-board line termination capability could be used to supplement the external line termination capability, if required.

For this architecture it is important to attain the right balance between DSP-based voice packet module functionality, and microprocessor-based network protocol module functionality, in order to optimize channel density. In some situations, however, this approach may not be satisfactory. For example, assume that the system has been optimized for compressed voice data. If an application calls for the gateway to make use of uncompressed data instead, this results in an increase of available DSP processing power as the DSPs are no longer required to run the compression algorithms.

However, the microprocessor may not be able to handle the increased packet load. Although the extra DSP capability should mean extra channels throughout the system, in practice there is no increase in channel count. The upper limit of channel density is now dictated solely by the microprocessor -- the system is now unbalanced. One way to solve this imbalance is to provide an external network protocol engine, effectively increasing the available microcontroller resource.

External Network Protocol Engine
Here, both the T1/E1 line termination and the network protocol functions are performed away from the DSP resource board. Voice data channels are fed to/from the DSP resource board across the H.110 bus. The DSP board still performs voice packet processing, but the resulting packets are sent to the external network protocol engine for onwards transmission.

The effect is to re-distribute the network protocol processing, reducing the demands on the microprocessor and restoring the balance to the system. The Network Protocol Engine can be chosen according to application needs, again taking advantage of an open systems approach.

CONCLUSION
Packet networks are increasingly being deployed by both new and existing telecom operators. Voice gateways in the system are enabling these operators to provide efficient and function-rich services. The introduction of high-performance, industry standard hardware, and flexible, efficient software has dramatically improved the ability of equipment vendors to quickly develop high-density, manageable solutions meeting the needs of service providers. There are, however, challenges that still need to be met, and it is important to ensure the balance of the system, matching voice/fax/data processing and packetization. This must remain a primary consideration during the design process. 

Brian Carr is a product manager with Blue Wave Systems, Inc., one of the world's leading suppliers of high-channel Digital Signal Processing (DSP) subsystems used in telecommunication infrastructure equipment, such as voice-over-packet gateways, digital wireless communications, and intelligent peripherals. Blue Wave Systems' open system approach to media gateway design provides the flexibility and performance required to achieve an effective solution. 

[ Return To The October 2000 Table Of Contents ]

BY DENNIS R. GATENS

With ADSL technology firmly established as a preferred remote access technology, value-added services such as Voice over DSL (VoDSL) are quickly moving from the trial to deployment phase. VoDSL provides two key benefits to service providers: Lower cost for deployment of additional voice lines and an ideal technology for bundling voice and data services. There are several choices of implementation in terms of the transport mechanism: IP/ATM/ADSL, PCM(ADPCM)/ATM/ADSL, and the emergent PCM/ADSL. Each transport mechanism has distinct advantages serving different market segments. There are also degrees to which voice can be compressed for transport over an ADSL loop through the utilization of ADPCM and low-bit-rate (LBR) codecs such as G.729ab and G.723.1a.

As part of an embedded hardware and software VoDSL solution, digital signal processor (DSP)-based platforms are enabling VoDSL services. As a programmable signal processing solution, they enable platform providers to offer flexibility to their service provider customers in two fundamental ways. First, the DSP enables a platform provider to support a wide range of voice features, which can be configured to meet the end users' specific requirements. Second, DSPs extend the life cycle of their hardware design while continuing to offer new software features and enhanced performance.

A case in point will be the migration to the utilization of LBR codecs. Today's VoDSL deployments utilize PCM or ADPCM, consuming 64 Kbps or 32 Kbps of symmetrical bandwidth respectively. LBR codecs reduce the amount of bandwidth per voice channel from 64 Kbps with PCM to as low as eight Kbps with G.729ab. DSPs allow a platform provider to offer service providers a roadmap from PCM/ADPCM to LBR codecs on the existing platform.

The type of service that can be offered over a local loop, and whether a service can be offered at all, is highly dependent on the length and quality of the loop because both affect the amount of bandwidth that can be provisioned to the end user. Additionally, with ADSL technology, the ratio of downstream to upstream bandwidth is approximately 10:1, thus the upstream bandwidth becomes the limiting factor for VoDSL services. Therefore, with as much as an eight times reduction in the amount of bandwidth required per voice channel, LBR codecs offer a number of advantages to the service provider. These advantages can be viewed from both a market and an operational perspective. From a market perspective a service provider can: Expand the reach for voice services effectively expanding his addressable market; offer a greater number of voice channels (thus expanding his service offering portfolio); and provide the ability to offer greater flexibility to provision voice and data bandwidth. Operationally, a service provider can: Reduce the amount of bandwidth required for existing voice services; and, as the volume of voice traffic becomes significant with respect to data services, reduce the amount of bandwidth required in the service provider backbone.

VoDSL is a key technology of the voice/data convergence evolution. As is true with any emerging technology and service, flexibility and scalability of a platform -- whether an IAD or a gateway -- has potential to determine a platform's degree of success. As a result, programmable DSPs are the embedded processor of choice for VoDSL technology which enables platform providers to offer service providers future-proof solutions, including a migration path from PCM/ADPCM to LBR codecs.

Dennis R. Gatens is product management director at Telogy Networks, A Texas Instruments company, which develops integrated silicon and software solutions, leveraging TI's DSPs and Telogy software products for voice, fax, and data over IP. 

[ Return To The October 2000 Table Of Contents ]