[For a copy of this article in PDF format, which displays figures and equations, click here. Requires Adobe Acrobat Reader, free download.]

Convergence: not products…not technology…not people…but synergy! Convergence promises to be all thing to all people, but can it? Will it? Do you really want it to be? The term convergence still reminds me of the planetary alignment thing that the ‘60 generation claimed happened when, “the moon is in the 7^th house, and Jupiter aligns with Mars”…aka the age of Aquarius.

Today, convergence is being played as the final frontier — the end result of the impending integration of voice, data, video, music, the Internet, home automation, satellite delivery systems, instant messaging, and global positioning systems (GPS), just to scratch the surface.

Terms like Bluetooth, HomeRF, wireless Internet, the wireless access protocol (WAP), broadband, asychronous transfer mode (ATM), various flavors of digital subscriber lines (xDSL), wireless local loop (WLL) and others yet to be defined, all will likely play a part in the upcoming convergence.

Many “adapt” the term convergence to benefit them as they see fit. One of the key issues with this term is that it is relational. It can relate to whatever product, technology or “solution” it is being applied to.

3G players are using it to bring video, stock quotes, music, messaging and other services to the mobile communicator.

Computer players use it to add wireless features to wired devices. In a recent advertisement I noticed that Bluetooth-based wireless access points and computer add-in cards are being listed right next to the wired Ethernet devices.

Bluetooth, HomeRF, 802.11x and other unlicensed wireless players are using the term to integrate home automation, Internet appliances, computer command centers and wireless Internet in a convergence that promises complete control over your lifestyle.

And still other players (like the entertainment industry) are integrating digital cameras with MP3 players (can wireless interfaces be far behind?)

Yet trying to make sense of convergence is overwhelming. Convergence should not stack after stack of protocols patching disparaging technologies and applications together. Rather, it should be a seamless, synergistic melding of the technologies and applications — one where they complement each other…where they are universally available…where they are reliable and dependable, no matter what, where or when.

Promising technologies exist at all levels. From the global pipeline referred to as “broadband” to basic modulation schemes like orthogonal-frequency division multiplexing (OFDM), designers are seeking ways to implement these technologies on a system-wide platform.

But perhaps the most important of all is that ubiquitous pipeline that delivers the content. It doesn't matter if it's the Internet protocol (IP), ATM, DSL, or cellular digital packet data (CDPD). And it doesn't matter if its 802.11x Bluetooth, WAP, Wi-Fi (an acronym for wireless high fidelity). And it doesn't matter if it's spread spectrum, OFDM, code-division multiple access (CDMA) or any other coding scheme. It matters that the end user can get and send the content over one, single megabit-wide channel — the broadband.

Of late, one cannot open a technology magazine, view a technology advertisement or carry on a conversation with a generation “d” individual without hearing the term “broadband.”

Like convergence, many players morph the term broadband to suite their needs and applications. Internet service providers claim broadband to be anything that is faster than current dial-up technologies. Satellite providers use it to consolidate the gamut of signals being beamed down to us. Computer players use it to tie together wireless and wired Internets and Intranets and ship voice, data and video within the loop. And wireless and wireline providers are finding that broadband offers them the opportunity to partner, buyout and cross-platform.

Yet broadband will open a plethora of opportunities, once the bugs get worked out and everyone is on the same page.

Broadband can be delivered by any number of platforms. Some are: Integrated services digital network (ISDN — to 128 kb/s), various flavors of digital subscriber lines (ADSL, HDSL, IDSL, MDSL, RADSL, SDSL, VDSL — to 25 Mb/s), microwave (to 1.5 Mb/s), T1/T3 (to 45 Mb/s), fiber (OC-48 — to 2 Tb/s), personal satellite systems (currently download only speeds to 2 Mb/s), and cable modem (to 4 Mb/s). And there are others as well. Most are derivations of the above technologies deployed in various standards around the world.

A broadband or “fat” data pipeline requires that all the components be digitally savvy. Content must be digitally codeable. The transmission medium must be digitally capable. And the destination site must be digitally-decode capable.

Although many options are available for broadband delivery, essentially all of them implement either packetized data or frame relay data. And, while there are almost innumerable derivations for packet and frame transmission technologies, this article will focus on the most common.

Today, common broadband technology involves wrapping data in “packets.” Each packet contains a certain amount of data. Packets can be either fixed or variable in length. Each packet can contain all of the data, or just part, depending on contents. Packet switching is based on the premise of all of the bandwidth, part of the time (see Figure 1). As packets are placed into the pipeline, bandwidth is allocated as needed. This give all transmissions as much bandwidth as they need, a chunk at a time. This allows for a fast, bursty pipeline.

Currently, packet data is implemented in three major protocols: Internet protocol (IP), frame relay and ATM. All packets begin with a delimiter, followed by a header, the payload (contents), the trailer or delimiter (see Figure 2).

The major differences among them is that ATM packets are of a fixed length while frames and IP are variable. Each has its advantages and disadvantage.

ATM packets are easier to process. There are no variables when it comes to size and overhead. This makes it easier to predict system performance and to design the overall system. ATM and frame relay packets are called connection-oriented. This means that a virtual circuit must be established for the packet to traverse. However, once established, such a circuit reduces the need for overhead.

IP and frame relay packets are variable in length. Their main advantage is that they use bandwidth more efficiently because they can be adapted to available bandwidth readily. IP has the additional advantage in that it is “connectionless.” Basically what this means is that the packet has a unique address in a universal system. It can find the address regardless of the path. However, this universal system requires a much larger address space because it has to search the entire universe, as opposed to following an ad-hoc established virtual circuit.

Static data and video are fairly easy to process into packets. Broadcasters have been doing it for years, pretty much the same for computer graphics and other data.

Voice, on the other hand has some unique issues.

First of all, we are used to a certain grade of acceptable quality — that of the plain old landline. Second, society as a whole is not particularly tolerant of voice system delay (a significant reason satellite phones aren't extensively deployed). Such is not the case for data, which isn't sensitive to delay.

Of late, voice over the Internet (wireline or wireless in various forms) has become a significant industry. But early attempts to get voice into ones and zeros was dismal. It didn't come anywhere close to the quality we have come to expect. There were (and still are) issues with application and equipment incompatibilities, bandwidth, and legal and political issues as well.

Broadband has given some relief to this problem, but the main improvement has come in processing power.

Now, combine wide bandwidths with high-speed digital signal processors (DSPs) and suddenly, voice as data becomes attractive.

Today's protocols for voice traffic are comprised of voice over Internet protocol (VoIP), voice over frame relay (VoFR) and voice over ATM (VoATM). All use algorithms based on mixed excitation linear prediction (MELP) and, more commonly, code-excited linear predictive coding (CELP).

CELP coders using complex algorithms to provide good-quality analog-to-digital conversions at low bit rates have been around for more than 15 years. But only recently have they been integrated into large scale implementation because they extract such a high price in terms of processing power.

Typical CELP coders require hundreds of MIPs (millions of instructions per second) to produce acceptable quality. This has only been practical of late, with the development of high-speed DSPs.

Current algorithms for CELP are based on the International Telecommunications Union's specifications. For VoIP the standard is G.723.1, which calls for 5.3 and 6.3 kb/s. For VoFR, the standard is G.729 at 8.0 kb/s. So far, it seems that general opinions tends to notice degradation when the frame loss rate hits about 3%(see figure 3). Overall, VoXX systems work well if packets aren't too long, transmission speeds are kept up and voice packets are given priority.

The reality of convergence and broadband is that DSP is the basic hardware that will make it happen. DSPs are now reaching speeds and integration levels that make it both practical and possible to digitize information from all sources. DSP technology allows coding and compression of megabit information quickly and efficiently for transmission over broadband systems.

The next generation of convergence products will have embedded DSPs as the core enabler. One currently available device is an integrated DSP that includes a plethora of hardware-implemented protocols such as hypertext transfer protocol (HTTP), simple mail transfer protocol (SMTP), file transfer protocol (FTP), terminal connection point (TCP), IP, and point-to-point (PPP) protocol. It also integrates a programmable memory chip (PMC) that can have various software functions programmed in. Such a device is ideal for Internet broadband systems.

In the near future, expect to see high-performance real-time embedded controllers addressing a “virtual multi-processor-based environment.” Multi-tasking will be the key with highly sectionalized processors that offer multiple interfaces so system designers can embed multiple hardware-based functions into systems.

The next generation of DSPs will be also be much more real-time capable. They will be 32-bit, superscalar, embedded RISC controllers utilizing high levels of integration, reliability and parallel processing. These functions will not only apply to the DSP, but control functions as well. Bandwidth will be in the hundreds of MIPs, and addressable memory will be in the gigabytes.

Furthermore, refinements in DSP techniques within embedded control loops will be the integrator among applications in the convergence of consumer, industrial, computer, automotive and communications markets. Expect to see devices such as third-generation mobile telephones with auto-location and preemptive artificial intelligence so your theater tickets will be waiting when you arrive, paid for, with your preferred seats reserved. Compact disk, MP3 and other various flavors of music players will be capable of wireless downloads. Multifunction communications devices will integrate music, digital video, messaging, voice and data, seamlessly. DSP-based, wireless-enabled in-car navigation systems will render the gene that prevents men from asking for directions, obsolete.

New DSPs will eliminate the concerns current generation systems have with benefits and drawback of floating point software support within an embedded control environment.

Given the advancements in sub-micron processes that will allow extremely packed semiconductor densities, software-based precision floating-point performance will become mainstream.

Within the next couple of years, DSPs will routinely achieve speed of 200 MHz and higher and 500 MIPs and up. Increased implementation of superscalar architecture, along with 64- and 128-bit wide code and data pipelines and enhanced multithreading capabilities will become the mainstays of convergence and broadband technology. Power consumption will be reduced to less than 2 VDC and RFI will be minimized.

In short, the dream of worldwide, seamless application-independent content capable of traversing any channel, medium or platform will be realized. It likely won't be tomorrow, or the day after, but certainly, once the technical, political, and infrastructure issues are finally addressed, and the lure of financial opportunity is identifiable, progress will astound us.