You might be tempted to call these the "services" provided by the phone company, but you have to be careful using the word service with ISDN, because it means things like audio, video, etc. - higher level services. What you can get from the phone company in terms of service are varying data rates, and various combinations of separate channels for data and signaling. These are access interfaces.
ISDN was designed around the notion of separate channels at 64Kbps. This number springs from the fact that that is essentially the data rate at which the analog lines are sampled at (8000 samples per second, 8 bits per sample) for the phone company's IDN. ISDN is essentially combinations of these channels, and also slower 16Kbps channels used only for signaling. The 64Kbps channels are called B channels. The 16Kbps channels are called D channels.
The names of the channels allegedly spring from analog circuits being called A-channels (A for analog). The next type of channel to come along got labeled B, which also happily can stand for binary (some also say it is the Bearer channel). The D channels were at one time called delta channels, because of their relationship to the B channels, but that particular greek symble being hard to type, it became D.
There are two main interfaces, Basic Rate, and Primary Rate. The Basic Rate Interface is intended for home use, and Primary Rate is intended for businesses.
The Basic Rate Interface (BRI) is designed to carry the most data you can possibly send to the home through existing copper phone lines. It turns out that they found you could reasonably squeeze about 160Kbps into those lines. With this, the phone company can provide two B channels, one D channel, and still have 16Kbps for the overhead (data framing, maintenance, and control) of communicating with your house's phone network.
In practice in the U.S, you might find the ISDN service available to you includes only a single B channel, plus a D channel, with the second B channel costing extra. One could assume this is only the quest for profit, but it also may be due to poor equipment that isn't capable of providing Basic rate ISDN without being upgraded. In addition, many locations within the U.S. Offer B channels that are only 56Kpbs. This is because much of the older equipment phone companies are using assumes that there is only analog data. Under purely analog days, extra bits were pulled out from the higher frequencies of the audio in order to do out-of-band signaling. This signaling now belongs on the D channel, but it will take some time for all of the phone equipment to catch up.
The Primary Rate Interface is designed for businesses with larger data needs, or with the need to set up their own local phone system. It is generally just a much faster connection to the phone company, with several B channels. In the U.S. the most common Primary Rate Interface (PRI) is designed for 23 B channels and 1 D channel, which is the equivalent of a U.S. DS1 service. In Europe, the most common PRI is 30 B channels, and one D, which is the equivalent of their E1 service.
With a PRI, you also have the option of combining several B channels into one bigger fatter channel called an H channel. There are several different speeds of H channel. The most common, H0, is 384Kbps, or 6 B channels. H11 is 24 B channels, or the equivalent of DS1 service. H12 is 30 B channels, or European E1 service. Above that, H21 provides 32Mbps (512 B channels); H22 provides 44Mbps (690 B channels); and H4 provides 135Mpbs (2112 B channels), and is anticipated for use with compressed HDTV.
In practice, the phone company will probably be able to provide any combination of B, D, and H channels that it thinks it can make a buck off of.
Figure 1. Common reference configurations
Here's a quick glossary of some of the things shown:
For instance, you could conceivably buy a card that plugs into your computer and utilizes the device drivers for ethernet, and the card would convert the ethernet requests into an ISDN data stream. In this case, the card would be a TA, and your computer would be a TE2. The card has to worry about converting one protocol to another.
Note the letters, R, S, T, U, and V in the diagram. These are reference points that everyone uses to talk about each of these parts of the network. For instance, the R reference point is the interface between an old-style telephone and Terminal Adaptor equipment. Since most homes won't have any NT2 equipment, the S and T reference points are usually one and the same, and are sometimes called the S/T bus.
The point to all of this is that different things happen in different parts of the network. What goes on along reference point U is completely different that at the S/T reference point - different wiring requirements, different data speeds, different encoding, etc.
Notice that reference point V, and the LT and ET equipment are in the phone company's domain. I lied when I told you that ISDN defines only the customer's part of the phone network, but I only lied a little. This portion of ISDN is seldom discussed, and still largely left up to the telephone companies.
ISDN uses a phone jack that looks just like the standard phone jacks in use today, except that it is a bit wider. Instead of the older 4-pin jacks (which only used two wires), ISDN uses an 8-pin jack (which only uses four wires). The CPI is based on a four wire scheme, two wires for transmitting, and two for receiving (which means you'll probably have to rewire your house). These wires are typically copper wiring of some sort, and can be longer than most users will ever need.
Figure 2. Typical CPI
(Note that each connection shown is a two-wire pair.)
If you are using ISDN with a single device (for instance, your computer is hooked up to ISDN, and your phones are still hooked up the old way), then you can have up to a kilometer (thereabouts) in your home for typical copper wiring. This is called a point-to-point configuration. But in most cases, you'll be using ISDN to hook up several devices, as shown in Figure 2, above. This is a multipoint configuration. With the standard ISDN equipment, up to eight different devices can be hooked up to the S/T bus. With this configuration the total length can be about 200 meters, and each device can be connected to the bus with up to 10 meters of wire. Devices can be placed anywhere on the bus under this setup.
This can also be modified somewhat, to extend the S/T bus up to about 500 meters. To do this, all of the devices must be connected close to the bus termination end of the bus. Further, each device on the bus must be 25-50 meters apart.
Eight devices might seem a bit low if you have an active imagination, but some of these devices could actually be brokers for other things -- for instance it is more likely that you'd have a single device that could simoultaneously control your microwave, furnace, A/C, alarm clock, and house lights. Even though you can only hook up eight devices, you have an almost unlimited number of addresses (i.e. phone number extensions) for each of those devices, so it is likely that one ISDN TE1 would be used for several different purposes. On the other hand, you can't simoultaneously use more devices than the available number of B-channels; for most customers this means only 2 devices can be in operation at once. In fact, with some ISDN provider's switches, you can only hook up two devices period, one assigned to each B-channel. This isn't the way things are supposed to work, but that's how a particular piece of phone company equipment works (specifically, the DMS-100 switch)(actually, it's more complicated than that - DMS-100s can work (almost) correctly with the right software, but sometimes they still use older software).
Also, those other four unused wires in the 8-pin ISDN jack are specified in the standard to be used for alternate power supplies. Whether these will actually be used remains to be seen, but it is possible that a UPS (uninterruptible power supply) could be added to your NT1, and it could use these auxilliary lines to provide guarenteed power. Note that one of these alternate power supplies is designed to go from the TE to the NT.
If you are outside of North America, and your power DOES go out, you are still covered though. The phone company will still provide the same power levels they used to. This should be sufficient to keep at least one TE1 device in operation. The assumption is that this would be your telephone, so that you could still call the power company and complain about your loss of power. The NT1 notifies all devices on the S/T bus of the power failure by reversing the polarity of the receive and transmit line pairs. All non-essential devices are supposed to respond by shutting themselves off. As I implied, this standard has not been used in North America - if your power goes out here, you have no phone. Don't ask me why.
The encoding on the S/T bus is a pseudoternary line code, known as modified alternate mark invert (MAMI). In this encoding, ones are represented by a zero voltage, and zeros are represented by a pulse, which is alternately either positive or negative:
Figure 3. MAMI encoding
In the U.S., however, it was decided that the NT1 equipment should be the customer's responsibility. This meant that what happens at the U reference point must be carefully defined in the U.S. so that the different vendor's products will all properly talk to the phone company.
Once your DSS1 signal makes it to the phone company, their own signaling system takes over to pass the call information within their system, and between other phone companies. Signaling System #7 (SS7) is supposed to be used for this. SS7 defines a communications protocol, and formats similar to DSS1, however SS7 is designed in a broader, more general way. DSS1 is specific to ISDN, however SS7 will handle the signaling needs of ISDN as well as other older signaling systems and (hopefully) will adapt well to future needs.
One important featuer of SS7 is providing CCS. This makes it harder for malicious users of the phone network to put one over on the phone company. It also improves the service, for instance by offering faster connection establishment. However, the phone companies haven't yet fully converted their equipment to use CCS. Older equipment still looks for the signaling information in the same channel as the voice, in the eighth bit of each piece of voice data. This is why many parts of the country only offer 56Kbps B-channels - they've lost 1/8 of their bandwidth to the older in-band signaling system.
If you are using CSV, they are free to route your call through any type of switch, even the old analog switches (there are a few left here and there). Your digital channel may also be shared with other channels, in the moments when there is silence on your phone line. And the digital parts of a CSV call can go through noisy switches that might create an undetected error here or there - it's only voice and you won't hear it.
For CSD, they can't do any of these things - your call must be routed only on pieces of equipment that will give dependable full time data channels. So even though the service in ISDN is supposed to be transparent, for the time being you have to tell the phone company how you are going to be using your B-channels. This seems to be more of a problem in the U.S. than in Europe.
Typically, each B-channel is setup for only one of these types of data. There are actually a standard set of combinations defined for setting up BRIs. These are called National ISDN Interface Groups (NIIGs), so there will be a limited menu of offerings available. Typically you can get both B-channels for data, or one for voice and the other for data, or one for voice and the other for either voice or data.
In order to facilitate this, North American phone companies use an optional part of the ISDN standard to identify each TE1 or TA you use. The phone company assigns a Service Profile IDentifier (SPID) to each of these devices, and you have to manually enter them into each device you use. The phone company then stores this data somewhere, and when you connect your machine to the network, it sends its SPID to the nearest phone company switch which identfies what type of connection the device needs and (therefore) how to route its calls. Presumably, the SPIDs have to refer to a configuration that matches one of the two B-channels you have.
By the way, the SPIDS are arbitrary numbers that refer to data stored by the phone company. The phone company often includes the phone number in the SPID for their own convenience, but in general you won't get anywhere trying to find significance in the patterns of SPIDs.
One older type of phone company switch, a DMS-100, was improperly designed with respect to the standards relating to SPIDs (the standards may not have been complete when the DMS-100 was designed). This switch misguidedly assigns one SPID to each B-channel that you use, rather than to each device. Therefore if your nearest switch is a DMS-100, you will only be able to hook up two devices to your CPI, rather than eight.
If you are only going to be hooking up a single device to your ISDN (i.e. setting it up in a point-to-point configuration, you might not need a SPID at all, as the phone company can identify your ISDN line as one particular type, full time. This depends on what equipment they have - the old DMS-100 switch will still require you to have a SPID.
In addition you could also use PSD on the B-channels, although this is generally only used for X.25 or something similar.
These bearer services are defined in terms of a number of attributes, which include mode (circuit or packet), structure (bit-stream or octet-stream), transfer rate (e.g. 64Kbps), transfer capability (basically, the content, for instance speech, 7Khz audio, video, or unrestricted), and several other attributes that specify protocols to use and other things.
The attributes of the bearer service are encoded into a Bearer Code, or BC, that is sent everytime a new connection is being set up. In theory, this allows the switches to dynamically choose from a variety of different switching paths and techniques depending on requirements. In practice, as discussed before, the SPID is used to determine what services are needed for switching, as this greatly simplifies things for the telephone companies. The BC will not be completely ignored, however there are certain bearer services that will be unavailable on your B-channels, based on how they are configured
It is important to note that the BC is sent to the switch every time a connection is established. However, the SPID is only sent to the switch when you physically attach your equipment to your phone line. At this time the switch gives your device a Terminal Equipment Identifier (TEI) which is used from then on to identify all connection requests from that piece of equipment. This allows the switch to look at the TEI and BC, determine the SPID, and see if the BC and the SPID match up.
Finally, there is a feature in some TAs that allows you to use a CSV bearer service to carry data (perhaps because it is cheaper, or possibly CSV is all that is available), which is called Data Over Speech Bearer Service (DOSBS). This works by providing additional end-to-end data guarentees that can't be relied upon from the speech Bearer Service.
Because this is an end-to-end issue, it will only work if both end points can speak the same protocol. Two common rate adaption standards that are emerging as the most popular are V.110 and V.120.
V.110 is the earlier standard of the two, and is mainly concerned with synchronous transmissions. It was designed for putting low-rate (2400 or 9600 bps, for instance) synchronous data onto 56Kbps channels prior to the development of ISDN. V.110 does also support asynchronous data up to 19.2Kbps. It does not have any error correction.
V.120 is a frame-oriented protocol based on LAPD, and also supports both synchronous and asynchrounous data streams. Because of its use of LAPD, it provides error correction.
Both V.110 and V.120 support the multiplexing of several lower rate data streams onto a single channel, although this feature isn't currently found in many products. V.110 is an easier protocol to implement and is better suited to synchrounous data, while V.120 is more suited to asynchronous communications, and is more complicated to implement, especially if all V.120 features are included in the implementation.
V.110 is more commonly implemented, and V.120 is gaining popularity. It is likely that for the near future vendors will try to support both protocols in their products, but eventually one will win over the other.
The most common method, BONDING (for Bandwidth ON Demand INteroperability Group), is implemented by most vendors. The standard is still developing, and some vendors may have features that others lack, so interoperability could still be a problem. BONDING is implemented outside of the ISDN architecture, so only the end points know it is a single connection - the ISDN thinks it is just several separate phone calls. It is able to support up to 63 combined 56 or 64Kbps B-channels.
The second method, Multilink PPP, is used only when routing IP over ISDN. Under the PPP standard, it is possible to have a single logical connection multiplexed across several physical connections, and this method is widely implemented. As with BONDING, this works entirely outside of the ISDN architecture.
The third method, Multirate Service (sometimes called Nx64 service), is a more expensive service, part of the primary rate interface service. Under this service, you get a single channel, of whatever size you need (in multiples of 64Kbps), on a per-call basis. This has the advanatage that only one single call is made, therefore connection setup is much faster. Multirate service is only now becoming available from the telephone companies.