Selecting Compression Technologies

Compression and decompression processes are found in most forms of electronic communications. Compression, of course, involves taking input data and generating a shorter representation of the data composed of a fewer number of bits compared to that of the original data. Fortunately, practically all forms of communication (even words on a printed page) are made up of redundant components. Data compression is thus essentially a redundancy reduction technique. The reverse process, decompression, takes the compressed data and generates or reconstructs the original data. The compression (coding) and decompression (decoding) systems together are often called a “codec”. Such codec software keeps the IP Communications industry profitable. The greater number of voice, video and multimedia “conversations” that can be squeezed in the same bandwidth “pipe”, the more revenue the pipe generates. The trick is to do the compression efficiently and without distorting or losing information contained in the transmission.

When an ordinary PSTN voice call with a frequency range of 200-3400 Hz is digitized for packetization using the G.711 codec (which merely digitizes and doesn’t actually do any compression), 8,000 samples are made of the analog waveform each second at 8 bits of “resolution” or information per sample. 8000 x 8 = 64,000 Kbps of bandwidth. Since a signal must be sampled at twice its highest frequency in order to represent an analog waveform (such as sound) digitally with reasonable accuracy (as dictated by a neat theorem devised by a fellow named Nyquist), G.711 can thus encode frequencies between 0 and 4 kHz — there is more than enough bandwidth, since analog channels in the U.S. are clipped at 3.3 kHz and European analog systems reach up only a little more, to 3.4 kHz. The analog-speech waveform, once having been encoded as binary words, is then transmitted serially, at digital bit rates of 48, 56, or 64 Kbps. ISDN channels and digital phone sets on digital PBXs use G.711.

When we add varying levels of compression to the mix, the results in conserving bandwidth can be dramatic. The G.723.1 codec is a dual-rate speech coder for multimedia communications transmitting at either 5.3 and 6.3 Kbps. G.723 was designed for sending compressed digital audio over ordinary analog lines as part of the H.324 videoconferencing standard. The even more interesting G.729 codec has a Conjugate-Structure Algebraic-Code-Excited Linear Prediction (CS-ACELP) speech compression algorithm that supports 3.4 kHz speech at a mere 8 Kbps.

No Video Codec Cornucopia?

When we approach the problem of digital video compression for things such as IPTV, the algorithms get more complicated but the choices for a service provider are reduced. Originally, MPEG-2 was pretty much the standard in the industry, an encoding and compression system for digital multimedia content defined by the Motion Pictures Expert Group (MPEG). In recent years however, two competitors have emerged: Microsoft took its Windows Media Video 9 codec and managed it through the SMPTE (Society of Motion Picture and Television Engineers) standardization process to become 421M, now better known as VC-1. (You can cram about three times as many video channels of the same quality using Windows Media as you can in that same amount of bandwidth when using MPEG-2.) The other major alternative is MPEG-4 AVC (Advanced Video Coding), better known as H.264.

Vendors have devised equipment for telco and cable vendors that support one or more of these compression schemes.

For example, Comtrend (www.comtrend.com) designs, manufactures, and markets telecom and datacom equipment for both the central office and CPE (Customer Premise Equipment) side. Comtrend offers products for xDSL, wireless LANs and VoIP, such as IP DSLAMs, VDSL switches, WAN access routers, CSU/DSUs, multiplexers, and digital cross-connect systems.

Andrew Morton, Comtrend’s General Manager for the North and South America, says: “We’re involved in the telecom side. These days it’s no secret that telcos are focused squarely on offering video and multiple services including voice. Specifically on the telecom side we have both copper and fiber equipment for physical layer termination, so we’re more on the customer premise equipment side. Also, we make software to run on that side. Additionally, we offer video set-top boxes; last year we shipped about 2 million units in the DSL space alone, to give you some scope.”

“Where I see compression going depends on the service provider and what they’re capable of delivering to their customers,” says Morton. “Our customers, being specifically telecom providers, focus on competing with the cablecos, and they’re all scrambling to find a combination of the best and most available technology – that’s almost an oxymoron. There’s always something better coming down the road, after all. H.264, which is sort of a derivative of MPEG-4, appears to be the ‘holy grail’ for a lot of the non-Microsoft based telecoms out there, and they’re all moving to adopt it. But the ‘reality check’ here concerns the availability of the devices that are capable of supporting VC-1 or H.264. Being a manufacturer of devices, we’re forced to pick and choose silicon in our devices capable of encoding and decoding here and on the customer premise side. Our biggest challenge is developing device that are cost-effective while at the same time capable of delivering these services at these high compression rates. Certainly the price points are improving but it’s still a challenge to get devices out to carriers that are going to hit the price points you need to be able to find a profitable model. And you can’t be writing a mortgage on hardware over several years and still be competitive against cable companies.”

“Obviously, Moore’s Law applies,” says Morton, “which is an overused term in the technology business, but things are certainly getting cheaper and faster. This year [2007] we at Comtrend are able to offer to carriers devices both in the video set-top box or IP set-top box world cheaper and faster than we could even a year ago. It’s gotten to the point where we’re now offering devices regarding – and you’ll probably laugh at this one – ‘karaoke on demand’ for one specific Asian market we supply. We actually have quite a good market for it. To a westerner, it might seem silly, but when you think about it, it proves out that these compression technologies do work, once you’re able to get the price and the performance into one box, which is a challenging thing for a hardware vendor to do. It also proves that we’re able to deliver video content on demand to an audience that is looking for that content. The model does work, since karaoke is basically just a bunch of music videos. We’re sending this over MPEG-4.”

“Carriers will either offer Microsoft VC1, their concept of video compression within the Windows Media version 9 world, or MPEG-4 or its derivative, H.264,” says Morton. “So we see carriers as either being ‘Microsoft’ or ‘non-Microsoft’.”

“The earlier MPEG-2 compression scheme is really for the crowd that is not as competitive with cable,” says Morton. “In the European markets we’re seeing a lot of MPEG-2 deployments, because cable is not necessarily a major competitor there and telecoms are able to offer video with less expensive devices and the competition is a bit less sophisticated than in the U.S., where we have some serious competition from the cablecos.”

“We have 41 IPTV customers worldwide which includes people such as SureWest in the U.S. and a division of Telfonica in Europe,” says Morton. “We see a mix of MPEG-2 and MPEG-4, but in the U.S. the first wave of adoption consisted of customers using MPEG-2. That’s because the technology on the hardware side had not yet matured, which we’re now finally seeing in 2007. That maturation is coming from better silicon, better processors and less expensive components, all of which equals more affordable next-gen service deployments for carriers.”

“The compelling phenomenon for MPEG-4 is HDTV,” says Morton. “In the case of karaoke-on-demand, it’s based on a lot of downloads, and since bandwidth is at a premium, you’re always looking for better ways of dealing with how bandwidth is laid out, so one naturally gravitates to such higher compression schemes.”

“Looking at things from another perspective, depending on the carrier’s infrastructure, MPEG-2 or MPEG-4 really helps them out,” says Morton. “For example, imagine a situation where you have fiber to the home. Since that offers so much bandwidth, MPEG-2 is an acceptable compression scheme under such a scenario. However, when there’s limited bandwidth available, say over copper, a user may have to rely on ADSL2+ or VDSL2 if they’re lucky, or they may have bonded two copper pairs together to gain greater bandwidth. MPEG-4 will maximize a 30 Mbps line where the users can get multiple streams into a sophisticated home. Some homes in America have 10 televisions. In such a case you’ve got to stuff a heck of a lot of streams down a 30 Mbps pipe to accommodate that, and only MPEG-4 will give you that ability.”

Comments Morton, “Several other things come into play. One is obviously the type of content, how sophisticated the competition is in the neighborhood or in the general territory, and then the flip side is what your infrastructure looks like from a carrier standpoint.”

Enough of a Choice?

Morton elaborates: “When faced with the bandwidth-hungry applications that are starting to appear, carriers have a few other choices: They can lay down some bigger pipes by going to a faster transport technology per pipe involving fiber, or they may bond existing copper pairs — ADSL2+ bonded has gained a lot of interest through a number of phone companies, as they utilize the availability of existing copper to at least maximize what can be done today with MPEG-2. At the same time, to make the most of an ADSL2+ pair, which at most yields about 24 Mbps, better compression is the next best way to get HDTV to those customers. The infrastructure comes into play. MPEG-4 helps carriers maximize what they have without having to spend a fortune on doing fiber, for example.”

“That’s how we at Comtrend look at things as a hardware provider because we’re always interested in how we can help carriers be more successful in their broadband deployments, particularly of IPTV,” says Morton.

Morton concludes with, “Fiber is of course the ideal solution for the future. Carriers at this moment are laying fiber all the way to home, and they are certainly structuring their business models toward that type of deployment. Other carriers don’t have it in their respective budgets, at least for the foreseeable future.”

Making Transport as Flexible as Your Choice of Codecs

Vendors must consider that, increasingly, multiple services will be traveling over their pipes, and sometimes TDM services must be accommodated in a next-gen network along with IP Communications and its various codecs.

RAD Data Communications (www.rad.com) offers TDM over IP (TDMoIP®) a “pseudowire emulation” that helps both carriers and enterprises slash network expenses without compromising the functionality or scope of services. RAD’s pseudowire emulation offerings include the IPmux TDMoIP gateways, Vmux voice trunking gateways, Megaplex multiservice access multiplexers, and Kilomux subrate multiplexer.

RAD’s Eitan Schwartz, Vice President, Pseudowire and Ethernet Access, says, “Everybody throws a lot of the different codecs into their equipment. Ours complies with the pseudowire [pseudowires enable the transport of legacy TDM, Frame Relay and ATM services over an IP/Ethernet network] standards, allowing you to trunk T1s and E1s and analog voice and so forth, across IP and MPLS networks. On top of that, there’s the matter of how you optimize it once you can trunk it. So you can trunk it as a pure circuit-emulated type of pseudowire where there is no compression, but you’ve obviously got very nice latency figures and you’re not limited to VoIP. The big thing you can accomplish in such an environment is cellular backhaul. And then you’ve got a type of pseudowire that is more AAL2 based [AAL2 specifies the bandwidth-efficient transmission of low-rate, short and variable packets in delay-sensitive applications and supports both variable and certified bit-rates]. It has the capability of optimization and silence suppression and so forth, and there’s a standard for that. That’s what we’ve incorporated into one of our product lines, the Vmux family of trunking gateways. They allow you to trunk and you can choose whatever codec you like. It could be G.711, G.729, G.723, or whatever. We also support such things as fax relay, modem relay and silence suppression. All of these features allow you to use this bandwidth in various ways.”

“When you compare pseudowire with VoIP,” says Schwartz, “VoIP gives you a lot of flexibility, with each ‘channel’ supporting pretty much any-to-any connectivity. With TDM-over-IP pseudowire, you’re dealing with nailed-up circuits traveling over an IP infrastructure. What IP brings to this is greater efficiency because all of the channels travel over one pipe. So, not only don’t you need any classic signaling, but your packets can be larger and therefore a lot more efficient. One packet can share traffic for 20, 30 or 60 voice channels, so you can fill up a packet quickly and reduce latency. Also, by using large packets, you’re eating up less bandwidth overall. On top of all this, of course, are the codecs, but these are the same codecs that you see in VoIP.”

“Different applications requiring different codecs,” says Schwartz. “You can look at military applications where they secure voice by coding it with quite a high compression codec so that they can then encrypt it and still carry it over a modem transmission on a regular telephone line. It would be difficult to do that even with G.729; they normally have to go with G.723. So, different applications force you to use different codecs, and even different hardware can force you to use different types of codecs because of performance reasons.”

No More Fiber Glut?

As bandwidth-hungry triple and quad-play services bundles proliferate, one wonders how long the “fiber glut” will continue. Certainly video and audio streams are gobbling up available bandwidth at an increasing rate, and service providers will have to rely on the latest compression schemes to tame these rather large beasties. In any case, sophisticated compression is necessary because most users communicate over measly copper pairs in their “last mile,” at least until fiber reaches everybody (don’t hold your breath on that one). Compression schemes will no doubt be ensconced as part of the basic representation or coding formats for future multimedia applications.

Richard Grigonis is Executive Editor of TMC’s IP Communications Group.