What is ISDN?

Most broadcast engineers will have heard of ISDN (the Integrated Services Digital Network) and will have an idea of its potential uses. Put simply, it is the digital equivalent of the ordinary telephone network and, being digital, data can be transferred reliably at high speeds not supported by modems on the PSTN.

This gives you the capability of sending high quality audio via dial-up telephone lines, replacing music circuits and other dedicated or pre-arranged links. Being dial-up, the system is more flexible than fixed point-to-point links, and being an extension of what telecommunications carriers do for everyone every day, it is inexpensive.

For the purposes of this article, ISDN gives you two 64kbit/sec data channels (called bearers), usually with a single telephone number. This is called Basic Rate ISDN, or ISDN-2, or 2B+D (the D signifying the separate 16kb/s data channel used for call handling). There is also Primary Rate ISDN, which is typically presented as a fibre-optic connection for large PABXs - the service is the same, but presentation is quite different and is less suited to audio applications.

What Does It Do?

ISDN supports three types of call: Voice, Asynchronous Data and Synchronous Data.

A voice call (sometimes called G.711 after the protocol used) is a conventional telephone call, accessed via ISDN. You can make a voice call to any number, PSTN or ISDN, and likewise receive a call, so long as you have an ISDN telephone or a suitably equipped terminal adapter. The benefits over the ordinary telephone service are the additional clarity and fast call handling. Fast call handling (typically connection is made almost immediately upon dialing) is significant, as a number of broadcasters have already found out. With suitable audio interfaces at each end (such as our VX-2039 Dual 4-Wire ISDN interface), ISDN voice calls can extend a talkback/intercom system to a remote site with near instantaneous connection and low call charges.

Asynchronous data (frequently described as V.24 on spec sheets) is the kind sent between computers. The network simply transfers data from one point to another, letting the user's hardware or software sort out any transmission errors (though these are few compared with a modem). The main application for this type of connection is, of course, computers and the internet. If you have an audio file stored on a computer and want to send it to another, this type of link will work fine - but it can't be used for live or real-time audio links.

Synchronous data transfer (look for X.21 on specs) is quite different. Here, the network provides a clock to which data transfer is synchronized. If a piece of data is lost, then it is ignored and the network (and the equipment attached to it) carries on regardless. A synchronous data link is used for most live and real-time audio links, as you don't want some equipment interrupting the audio flow while it retries for some suspect data.

To make use of ISDN in a broadcast context, two items of equipment are required: a codec and a terminal adapter.

Digital Audio Codecs

A digital audio codec (COder/DECoder) performs two functions: it turns analogue audio into data, then it reduces the volume of data to manageable proportions, reversing these processes for incoming data. The data reduction is performed according to one of several algorithms, the same algorithm being used for the encoding and decoding processes. These can result in anything from a fourfold reduction in data up to sixty-fold or more, allowing high-quality broad bandwidth audio to be sent over a small number of ISDN bearers.

If you simply digitized an audio signal prior to sending it via ISDN, you'd need several bearers just to get a mono voice signal transferred. For example, a signal sampled at 16kHz into 8-bit words would require a data flow of 256kb/s for a 7kHz duplex link. Thus, ordinary speech contributions would require four 64kb/s bearers. The cost of the equipment required would be considerable, and call charges would soon mount up. However, a simple algorithm like G.722, described below, will enable the same duplex signal to be transmitted on a single ISDN bearer, resulting in lower capital costs and call charges.

There are several algorithms in common use amongst broadcasters, each with its own advantages and disadvantages. The principal differences are between low compression and high compression algorithms. The former tend to be more robust, allowing the audio signal to be processed several times without noticeable degradation, and offering very fast processing (and thus a short delay to the audio). Higher compression algorithms give more audio bandwidth for any given data rate (giving as much as 20kHz stereo on 2 ISDN bearers), and may offer additional functionality such as a serial data channel embedded within the audio (useful for controlling remote routers, etc). The downside is that a signal cannot repeatedly be processed via such a system without developing artifacts, and the encoding delay is sufficiently long to make such a system unsuitable for some live audio contributions - particularly two-way interviews.

The two most popular types of data reduction techniques for audio are Adaptive Pulse Code Modulation (ADPCM) and psycho-acoustic modeling. Here is a brief description of what these two types of system do:

ADPCM coding exploits the fact that an audio signal does not vary greatly from moment to moment, and that variations can be predicted. Using identical processes, the encoder at the sending end and the decoder at the receiving end predict what a sample will be, based on the previous sample. The encoder compares its prediction with the actual sample, and transmits the difference to the decoder. The decoder applies this correction to its prediction to reconstruct the original sound. This typically achieves around 4:1 data reduction, and processing is almost instantaneous (typically less than 5ms). CCITT Recommendation G.722 and the proprietary apt-X 100 algorithm are examples of ADPCM coding.

Psycho-acoustic modeling uses the fact that a loud signal at a particular frequency will mask a quiet signal close to the same frequency, rendering it inaudible, and that the cumulative effect of masking is to produce a noise threshold below which all audio is effectively masked. The encoding process removes this redundant information, sending a scaling factor with the audio so that the decoder can rebuild the signal at the correct volume levels. This process takes some time (typically 250ms for a stereo signal) but achieves data reduction of up to 16:1, allowing greater audio bandwidth over the same ISDN link. ISO/MPEG Layers 1, 2 and 3 are all examples of psycho-acoustic algorithms.

The codec is an important part of your ISDN link, and there are many different types and models to choose from, ranging from studio models with multiple operating modes to a portable G.722 codec that will fit in your pocket. Here are some questions to ask when selecting a codec, either for a particular application or for general use:

What audio signals will it carry?

Some algorithms are designed for mono only; stereo algorithms are implemented as mono in certain codecs; the bandwidth may be restricted either by the algorithm or by the number of ISDN bearers which can be connected to the codec; the quality of the signal may be poor at low data rates.

How is the audio presented?

Generally, codecs will give you line level balanced inputs and outputs on XLR connectors. Some offer mic input in addition to line, and headphone output, in which case there should be level controls (and the codec may be in a portable or desktop housing, rather than the usual rackmount arrangement). AES/EBU digital connections may be available, perhaps as an option, as may SPDIF. Some codecs are available as PC plug-in boards, where generally the physical restriction of the format prevents the use of XLR connections. (If the board has a feature connector then it can be joined to a broadcast audio interface module such as the CI-525, which fits neatly into a drive bay giving conventional line/mic/headphone connections.)

Taking the PC's involvement a step further, it is now possible to buy a software version of an ISO/MPEG codec, using the processing power of a PC to perform the encoding and decoding and a sound card for the audio interface. This can offer a tremendous cost saving over a conventional hardware codec, whilst still offering the same quality and real-time and live duplex operation.

How is the codec operated?

Some codecs are fully automatic, offering no controls whatever. Others require you to make a choice between different operating modes, and perhaps different algorithms. In such cases the controls are usually front panel mounted, and there may be an optional remote control. Serial control is useful, particularly if the codec is shared among several studios. If the codec does have several modes, look for the facility whereby it will automatically adapt to another codec calling in.

Is the processing fast enough?

A live interview or two-way conversation requires a fast algorithm to avoid the delays which can so easily throw an inexperienced contributor. Similarly, an STL should not delay the signal significantly, otherwise your time checks will be wrong. A music feed from a concert might be delayed by half a second or more without anyone being aware, so long as the presenter has a suitable cue. Generally, low compression algorithms are fast and high compression slow, but circuit design is also a factor - don't assume that all codecs provide the same processing speed for the same algorithm.

Simplex or duplex?

If you have an application where the traffic is one-way, such as a studio-transmitter link, it may pay to use separate encoder and decoder modules, one at each end, avoiding the cost of a full codec at each site. (Most codecs are duplex, allowing two-way traffic.) At least one codec on the market offers asymmetrical duplex operation, sending ISO/MPEG broadband audio in one direction while receiving fast G.722 audio as a cue - ideal when you require the maximum quality in a two-way remote.

Does it manage the synchronization of multiple bearers?

Synchronization (more usually described as inverse multiplexing) is required whenever you make a connection over multiple ISDN bearers, and can be performed either in the codec or in the terminal adapter (see below). The purpose is to ensure that several signals which may arrive via very different routes stay in sync. Some standards actually define how inverse multiplexing should work - ISO/MPEG Layer-II includes recommendation J.52 for the inverse multiplexing protocol.

Is it compatible with other codecs?

There may be codecs within your organisation already, or you may have some specific connections you wish to make. Either way, not just the algorithm but the way it is used must be compatible. Most codecs support a single algorithm, but some support two or three. There are variants and subsets of some algorithms, so simply knowing the algorithm's name is not a guarantee of compatibility. (For instance, there is a little-used variant of G.722 which uses the H.221 protocol to sync the data, rendering it incompatible with standard G.722.) Remember that more links fail due to misunderstandings over compatibility than any other issue.

In short, the more you know about how you will use the codec, the better informed your choice will be. Talk to the manufacturer or your supplier about the application, as they will probably have come across a similar requirement before and can give you the benefit of that knowledge.

Terminal Adapters

Your codec cannot be connected directly to the ISDN network - a terminal adapter (TA) is required to provide call handling, data protocol management and the like. Sometimes the TA is built into the codec, other times it is a separate desktop, rackmount or even PC-mounted item. Once you have selected a particular codec, choosing the appropriate terminal adapter is quite straightforward. Here again is a checklist of questions:

Does it support synchronous data calls?

Fairly basic, but most terminal adapters are designed for use with computers and don't need this function. You may also want to use the TA to place voice calls, either directly from the TA (if it has a handset) or via a telephone connected to it.

How many bearers?

Typical TAs support two ISDN bearers. If you need more, then you'll have to use multiple TAs. And a point related to the last question - make sure the TA supports the call types you want on each of the bearers - it may be a dual port TA offering synchronous and voice calls, but only one of each.

How is it operated?

As mentioned before, the big application for ISDN (and therefore TAs) is computing. Because of this, most TA manufacturers assume that you'll have a computer or serial device handy to provide the control for dialing and mode selection. That's often not the case in broadcast environments, so you'll need an on-board or remote keypad. One TA actually allows you to plug in an ordinary telephone and use the dialing pad to make a data call. If the TA is built into the codec, you may find all the controls - TA and codec - integrated on the front panel or with common serial control.

Sub-addressing and MSN?

No, not the Microsoft Network. Sub-addressing is a flexible feature (you have to pay for it) which allows you to identify individual TAs sharing the same ISDN-2 circuit, and route an incoming call to the chosen TA. Sub-addressing works by adding digits at the end of the phone number you are assigned, and matching those with the digits you program into each of the devices attached to the line. With Multiple Subscriber Numbering (again, chargeable), you are allocated a block of nine numbers, and each TA can be programmed to respond to one of the nine. It may not be high on your list of priorities, but the TA should support both of these facilities for when your network expands. MSN is now the more widespread method of the two.

Incidentally, you can no longer have two numbers and map one to each of the bearers - this service, whilst still available to existing subscribers, falls foul of Euro-ISDN and has not been available to new customers in the UK since April '96.

Does it support BONDING?

This is a feature incorporated in some terminal adapters to overcome a lack of inverse multiplexing in codecs. Say you have a codec with an inverse multiplexer for two bearers, but you want to make a 256kb/s call requiring four. With BONDING terminal adapters at each end, the codecs are presented with two 128kb/sec data streams and the problem is overcome. BONDING stands for Bandwidth ON Demand INteroperability Group, and is a technique shared by several manufacturers. Others have chosen to provide custom multiplexing solutions, in which case the same type of TA must be used at each end of the line.

Will it work elsewhere?

EuroISDN is by far the most common service you will encounter, being adopted in most countries outside of the USA and Canada. EuroISDN is a rigorous standard and a device built for one market should operate in another without problems. Do bear in mind that there are some residual systems (like the pre EuroISDN services in the UK), and that sometimes you'll need to spend a little time getting things to work. The USA and Canada are quite different - whilst Switched-56 is fast disappearing, ISDN has itself been implemented in an entirely different way, requiring assorted parameters to be programmed into the TA for each different location in which it is used. Confusion between the different carriers and service providers can render this a somewhat taxing issue, though doubtless it will improve over time.

The easiest way to get the right terminal adapter is to talk to a broadcast equipment supplier about your application(s). The TA manufacturer is mainly interested in computer applications and sees the broadcast market in a small light. For the codec supplier, though, knowing about terminal adapters is vital to the sale of the codec itself!

How do I get hold of ISDN?

The service itself must be ordered from your telephone company. In the UK, there are few areas where ISDN can be obtained from anyone other than BT, although this situation is changing. Currently, the installation charge is £400 although there are now several options (the "highway" bundles are examples) which allow you to build part of this charge into the rental. It can take anything from a couple of days up to several weeks, depending on where you are and what the availability of circuits is like. Please note that the "highway" bundles may come with terminal equipment not suited to your requirement, and you should investigate what is being offered carefully.

It comes on a copper pair, so disruption is minimal. A small box is mounted on the wall, offering you two RJ-45 sockets. These do not relate to one or other bearer, they are simply parallel connections. The TA handles bearer information, automatically or via MSN or sub-addressing.

In some countries and in some bundled solutions a terminal adapter is provided with the circuit. If so, remember that the telephone company knows very little about the your application for ISDN, and will probably supply a TA for asynchronous (computer-type) connections, these being the most popular.

Once you have made your decisions, it is a simple matter to connect the TA to the line, the codec to the TA, and your audio equipment to the codec. There are variations in the types of cable and connector used, and hence there is an argument for getting all the equipment together from a single source, but a good supplier will make sure that you get the right cables anyway.

I want to get a link from Timbuktu - what can I do?