Even though Alec Reeves of Britain had developed Pulse Code Modulation (PCM) techniques in 1937 for digitizing audio signals, and Bell labs had invented the tran-sistor in 1948, which was required for the large-scale implementation of digital tech-niques, it would take more than a decade to make digital transmission a reality (and longer still before the advent of digital switching could make the full signal path dig-ital outside the local loop). 1963 brought the introduction of the T1 or Transmission One digital carrier using revolutionary signal manipulation techniques that would forever change telephony.
Unlike all previous carriers, the T1 started in an all-digital format, meaning that it was structured as a series of bits (193 per frame to be exact, 8 bits per channel, 24 channels, plus the framing bit—moving at the rate of 8,000 frames, or 1,544
Megabits per second) that by design could be completely regenerated again without data loss over long distances (see Figures 4.3 and 4.4).This provides a 64-kilobit-per-second digital bitstream for each of the 24 channels, using Time Division
Multiplexing (TDM).
96 Chapter 4 • PSTN Architecture
372_PRAC_VoIP_04.qxd 3/1/06 5:31 PM Page 96
Figure 4.3A T1 Frame*
*Eight bits in each channel capture a 125µs slice of each associated analog audio signal.
Figure 4.4Time Division Multiplexing
TDM as introduced in the T1 is the multiplexing workhorse of the telecommu-nications world and will be the base multiplexing environment for the rest of our discussion of the PSTN.Yet for the T1 to be successful, it is just as important to have a foolproof way of converting an analog signal to digital bits that would make or break the new form of digital transmission.This is the job of a codec. Although today in the era of digital media we take for granted the engineering required to create the first effective PCM codec—now commonly known as G.711—it was no small feat in its day.Yet, even today as debate rages over what codec is best to use for VoIP, G.711 is still considered the “toll quality” standard that others must beat, and is especially good at preserving modem and FAX signals that low-bandwidth codecs can break.
www.syngress.com
PSTN Architecture • Chapter 4 97
N
OTEAlthough we’re not going to do a deep dive on digital/analog conversion here, it is worth pointing out that slight differences between U.S. and European stan-dards will mean that some conversion needs to take place even withina stan-dard G.711-encoded channel in order for that channel to move from a T1 to an E1 or vice versa. Specifically, slight differences in PCM encoding algorithm (µ-law vs. A-(µ-law) may require conversion when voice or VoIP streams cross interna-tional boundaries. Of course, on a data circuit, that conversion is not going to happen automatically (if it did, it would scramble the data). But it can cause problems across a VoIP if you’re not careful.
Similarly, when using a T1 circuit for data, it’s important to make sure the circuit is properly configured since some signaling modes can use what’s called
“robbed-bit” signaling, which is fine for circuit-based voice but will corrupt data running on it. For this reason, only 56K of the 64K channel could be used for data on early data circuits. Today, clear channel data can be provisioned that uses a full 64K channel.
Back to the codec issue, however. It’s worth pointing out that very complex trade-offs exist in codec selection and they’re not as simple as quality vs. band-width. Some codecs require much more processing, others work poorly with modems, faxes, and other nonvoice applications (particularly low bandwidth codecs: it’s not hard to imagine the problems inherent with sending a 56 Kbps modem signal through a 4Kbps voice-optimized codec. Even the best compres-sion algorithms would struggle to represent that much information in so few bits, not to mention the inherent distortion present in D/A-A/D conversion.
Starting with the introduction of the T1, timing became an important considera-tion for the PSTN. Digital circuits like the T1 must be plesiochronous, meaning that their bit rate must vary only within a fairly limited range or other problems can be created within the PSTN. In comparison, analog circuits are completely asyn-chronous.This requirement has forced a hierarchy of master clocks to be incorpo-rated into its infrastructure.
With the advent of SONET, a fully synchronous solution to the timing problem has arrived, along with massive bandwidth that can be further enhanced with Wavelength Division Multiplexing (WDM—basically the use of different colored light on a single optical fiber to increase capacity). Pointers and bit-stuffing in SONET and SDH are used to minimize the impact of clock drift between digital circuits, though the advent of VoIP has created some challenges because VoIP is asynchronous. VoIP is also a packet technology (since it runs on packet networks), so it is subject to variations in latency and jitter and packet loss that are simply not
sig-98 Chapter 4 • PSTN Architecture
372_PRAC_VoIP_04.qxd 3/1/06 5:31 PM Page 98
nificant issues in circuit networks because timeslots are guaranteed. On the other hand, the PSTN’s circuit network is far less efficient overall than any packet network because of the excess capacity it reserves.
As T1 and other digital trunks were deployed in the PSTN, digitized voice ser-vices in 64Kbps increments, each called a Digital Signal 0(DS0) —became the basic switchable unit of the PSTN. A single DS0 is a 64Kbps channel equivalent to an analog line converted to digital via G.711. With the advent of TDM-based digital switching, the DS0s were aggregated by digital access and cross-connect systems (DACS) for transport or presentation to the switch via DS1 (1.5 Mbps) or DS3 (45 Mbps) interfaces.These digital switches communicate over T1 and other digital trunks to access and toll tandem switches, sending calls across the telephone network to destination switches.The DS0 voice channels are then split back out to their orig-inal 64Kbps state and converted back to analog signals sent onward to the destina-tion local loop.
In fact, there is now a full hierarchy to the T carrier system in North American and the E carrier system in Europe (as well as the more recent SONET/SDH optical carrier system). Aggregation of voice and data channels at many levels can take place, and knowing how these systems can interact is essential.Table 4.1 roughly defines the capacity and equivalency of the various North American, Japanese, and European digital signal hierarchies in a single chart. I’ve never been able to find this information in one place, so I created a single chart to cover the whole range of PSTN transport solutions in use today.
www.syngress.com
PSTN Architecture • Chapter 4 99
Table 4.1Digital Signal Hierarchy (North America and Europe)
100 Chapter 4 • PSTN Architecture
372_PRAC_VoIP_04.qxd 3/1/06 5:31 PM Page 100
In Table 4.1, dark bands are for the circuits most commonly provisioned for business customers. Bolded items are used most commonly in wide area networks overall. Note: Although SONET and SDH are directly equivalent to each other, the process of mapping between them and their T or E-carrier counterparts requires the use of SONET Virtual Tributaries (VTs) and Virtual Tributary Groups (VTGs) or SDH Virtual Containers (VCs).
As you can see from Table 4.1, 24 DS0 channels make up a T1 circuit, 28 T1 cir-cuits make up a T3 or OC-1 link, and so forth. An OC-12 link can support up to 7936 DS0 channels if it’s broken out into E4 circuits or 8064 if it’s broken out by T3 circuits through a DACS or Add Drop Multiplexer (ADM). 10 Gigabit Ethernet can run over an OC-192 SONET ring, and so on.These mappings are essential to understanding capacities for Internet access circuits as well when sizing for VoIP, since upper limits on Speed (left column) cannot be physically exceeded (note that actual throughput will be at least 10% lower because of overhead).
Perhaps you have ordered and provisioned a voice or data T1 for your company or clients. Have you ever thought why only one voice T1 is needed for a company of 100 employees with a PBX, knowing that only 24 channels can be used at any one time? The answer is that not everyone will be on the phone, receiving a fax, or otherwise using an available channel at once. Normally you can count on a six-to-ten ratio when calculating how many DS0s are needed.Those in the sales and ser-vice industry may go as low as four-to-one because they are on the phone more and need higher channel availability. Even with VoIP, sizing access circuits is important, since there are hard limits on the amount of data that can be pushed through that circuit network, even if the number of channels isn’t so important. Less bandwidth might be required if G.729 was used in place of G.711, but more would be required if the link also supported Internet access, especially if Quality of Service (QoS) limi-tations weren’t set up on the corresponding routers.
In Figure 4.5 we see that the DACS can be used to combine a wide variety of digital signal inputs and present them through a single interface to the next hop, which might be a switch, SONET multiplexing equipment, enterprise routing equipment, or something else. Keep in mind that although both voice and data traffic of any flavor can run over SONET, timing requirements won’t allow some-thing like a T1 to run over somesome-thing asynchronous like Gigabit Ethernet.
www.syngress.com
PSTN Architecture • Chapter 4 101
Figure 4.5DACS Channel Aggregation
N
OTET1 links in particular have a lot of nuances not discussed here in detail, from dif-ferent framing and superframing formats like D4 and Extended Super Frame (ESF) to special line coding like Bipolar 8 Zero Substitution (B8ZS) used to ensure byte synchronization without losing data or bandwidth.
Other framing considerations come into play for different digital carriers such as E1, T2, T3, STS-1, STM-2, and so on. There are excellent books on the topic for those that need more details, but in general none of these formatting issues require any security consideration.