Digital Switch Basics

Early switching systems revolved around telephone operators who used plug-ended cords to connect telephone calls often while on roller skates. In 1889 Almon T. Strowger invented the first automatic telephone switching system called "step-by-step". Mr. Strowger, an undertaker by profession, developed the switch because of loss in business. He suspected his competitors of bribing central telephone operators in order to learn about the latest gun fights and where the bodies could be found. The step-by-step switch was design to stop this problem by routing calls without the use of central telephone operators.

Some of the first telephone switches were single-stage space switch (S) formed out of a rectangular array of crosspoints. Each input in a single stage space switch can be connected to any of the outputs. Early space switches used relays to make the crosspoint connections. Newer digital switches use digital crosspoints. If the space switch is non-blocking then every free input has access to every free output. The problem with single-stage space switches is that a large number of crosspoints is required to perform the switching. For example, if we had a 20 input/output switch, 400 crosspoints would be needed. The number of crosspoints is quickly prohibitive because of the limited number of pins on digital integrated circuits. The crosspoints in this switching design are also very inefficiently utilized since only one crosspoint in each row or column can be used at one time. In order to make a space switch more efficient extra stages need to be employed. In a three-stage space switch (SSS) two extra stages are utilized to form a switching matrix. By breaking a 20 input/output single stage space switch into three separate stages you can dramatically lower the number of corsspoints. This is accomplished even while keeping the switch fully non-blocking.

If we needed a 131,072-port switch and used a single stage switch matrix 17 billion crosspoints would be required. In using a three-stage space switch only 268 million crosspoints would be needed. This is still several factors higher then what is possible with today’s digital electronics.

The enormous number of crosspoints needed in a large switch brings us to consider the time slot interchanger (T) also known as a "time switch". A time slot interchanger functions by switching in the time dimension instead of the space dimension. Before we can examine the time slot interchnager we need to do a little reviewing. In modern switching systems analog phone calls are converted to digital signals using Pulse Code Modulation (PCM). This function is preformed by sampling the analog signal 8,000 times a second. Each sample is then converted into an 8 bit digital value yielding a 64kbps digital signal, this process is repeated every 125 micro seconds. A time slot interchanger works by multiplexing many 64kbps digital signals into one digital frame. Each 64kbps signal is referred to as a time slot and is matched to one of the lines on the switch. The switching is performed by writing the multiplexed frame into digital memory in sequential order. The frame is then read out of the memory depending on the switching information in the control memory, this process is repeated every 125 micro-seconds. This switch is 100% non-blocking because any free time slot can be switched to any other free time slot. This process of writing time slots into digital memory and then reading them out in the order required to be switched allows us to switch a large number or ports; but, we are constrained by memory speed. In order for a time slot interchanger to function, the entire frame needs to be written into memory and then read out according to the control store every 125 micro-seconds. If we needed to switch a T1 (24 channels) our memory would need to be able to write and read each time slot in 5208 nano-seconds (125/24). If we needed to switch 2048 time slots we would need memory that could be written to and read from every 61 nano-seconds (125/2048), and so on. With toady’s memory the largest switch that can be built using just a time stage is around 3000 ports.

So how can we create large non-blocking switches? One of the ways this can be accomplished is by utilizing the space switch and the time switch to form a time-space-time switch (TST). Using these two forms of switching we can greatly expand the size of a switch while keeping the switch non-blocking. If we take a 40 input/output single stage space switch and connect a 2048 time switch to each input and output we will have a switch with the total capacity of 81,920 (2048 * 40) ports. This is the architecture of many modern (well, that is within the last 10 years) switches. Not all switches are TST, some popular configurations are TS, TSSSST, and STS.

When selecting a switch for an ISP application we need to make sure that the switch will be non-blocking or at least have a very low blocking probability. Most of the smaller switches you see on the market are just time switches. This is due simply to the fact that building a time switch without a space stage is less expensive. The downside of course, as mentioned above, is that we are limited by memory speed to about 3000 ports. Note that in most ISP/CLEC applications a switch that supports twice the number of ports intended for modem use is required. So, if we need to support 1500 modems a switch with at least 3000 ports is needed, 1500 connecting to modems and 1500 connecting to the ILEC. If we need to support only a few modems a small time switch is fine, but if we need to support a large number of modems we must step up to a larger switch. Some vendors of smaller switches suggest the use of several small switches in parallel, but there are many disadvantages to this and this course should be avoided.