With the physical modulation scheme defined, the next layer is responsible for managing individual mobile unit use and access of the available radio spectrum. This is called Medium Access Control or MAC.
On the cellular network, each cellular call is assigned a pair of frequencies, typically called the RF channel, for the duration of the call. The RF channel remains dedicated for the user's conversation even if both parties are experiencing a long and awkward moment of silence. This is reasonable for voice communications systems. Human conversations suffer greatly if each party's voice is delayed in transit.5.3 If the cellular system released and re-acquired the channel through the duration of the call, the inconsistent delays would be unacceptable to the subscribers. The cellular telephone system therefore dedicates a RF channel to each call. This is extremely expensive use of the precious RF channel resource.
In contrast, CDPD is a data communications system and variable delays of individual data packets is quite acceptable. With this in mind, the CDPD system was designed with the Local Area Network (LAN) shared channel model of operation. The CDPD mobile units only transmits on the RF channel when it has data packets to deliver.
During periods when the mobile unit does not have any data packets to send, it turns off its transmitter and allows other mobile units to access the RF channel resource. In this manner, the precious RF resource can be shared by many devices. More efficient use of the RF channel can be accomplished.
There are many different ways to share a channel, RF or otherwise. Many of these methods have been used in the Local Area Network environment. In the following, a few of the common channel sharing methodologies are discussed briefly.
In a token passing system, a data packet of transmission authorization, the token, is transmitted from one unit to another. Only the device that possesses the token is allowed to transmit on the channel. Once the unit is finished with its transmissions, or if it has reached a predetermined maximum transmission time limit, it relinquishes the token and passes it to its "downstream" neighbor.
This mechanism provides a very orderly sharing of the channel resource. It relies on the ability of each device to unambiguously pass the token to the next appropriate unit. It works well within a network where the collection of units are well organized and the next token holder is correctly identifiable. Figure 5.3 illustrates two types of token passing networks.
The CDPD network architecture does not support this type of orderly passing of a token. The membership of the RF cell is not static. Mobile units enter and depart from the cell at will. This means that mechanisms must be added to allow new entrants into a cell to announce their presence. There must also be mechanisms for the controller of the token to recover from a mobile that departs with the token.
For these reasons, CDPD does not use token passing mechanisms to manage sharing of the RF resource.
Another approach to resource sharing is termed demand assigned with reservation scheme. Essentially, the channel is allocated on demand. When a mobile unit identifies itself to the network controller as requiring channel resources, the controller allocates a portion of the resource to that user. Typically, this is done on a time allotment basis, but it is also possible to assign the channel on a frequency assignment basis. Indeed, the cellular telephone system works on this scheme in terms of its assignment of channel pairs (frequencies) per call. Figure 5.4 and Figure 5.5 illustrate these two medium access control methods.
A common method of time slot based demand assigned multiple access scheme involves the partitioning of a channel resource into multiple time slots. Users that have data to transmit are allocated individual time slots within a larger time window. A unit that has been allocated a particular slot within the window will transmit its data during the set time slot and not any other time. This method allows the central controller to manage the transmissions from multiple units without fear of collisions.
To determine how to allocate the time slots, there must be a mechanism for the mobile units to indicate their need to access the channel. With a potentially large population of mobiles in a single RF coverage area, a polling mechanism would be too slow and inefficient. Typically, a contention based mechanism is used by the mobile units to indicate a need for a time slot. Even if polling were feasible, mobiles would have to use some contention mechanism to announce their presence in order to have time slots assigned.
A demand assigned with reservation approach requires the mobile units to contend for a "reservation" channel. It further requires the central controller to sort through the reservation traffic and formulate an assignment of slots for the next available transmission window. This process introduces a slot assignment delay. This delay is significant. If all the mobiles only have short transmissions, the delay may be several times the actual transmission period of the data. If all the mobiles have long transmissions, then the ability to demand continual use of a slot will reduce the effect of the reservation delay.
The CDPD system does not use the demand assigned with reservation scheme. The CDPD system is expected to support a traffic profile that contains a significant amount of short bursty transmissions from the mobiles. In addition, the movement of the mobile units may result in unusable statically-assigned slots when a unit moves out of the coverage area.
The preceding two approaches were based in controlled access authorization mechanisms. Lest the reader think that all channel access mechanism rely on orderly exchange of access privileges, there are alternate approaches. One of such founding methods was developed in University of Hawaii and is fondly called the Aloha scheme.
In the Aloha network, all devices transmit on a single uplink frequency to a communications satellite. The satellite relays what it receives back towards the ground devices on a downlink frequency. However, instead of carefully managing the channel resource so that no device's transmission collides with another, the devices are allowed to access the channel on a free for all basis. Whenever a device has data to transmit, it does so without regard for condition of the channel. If a transmission collision occurs, the transmitting device learns of it through lack of acknowledgement from the peer entity. When such transmission loss is noticed, the transmitting device repeats the data transmission at a later time.
As the reader may postulate, the effectiveness of such a system is limited at best. If there are few devices, each with little data to transmit, then the probability of collisions is minimal. In this case, the low overhead approach to channel sharing is indeed very reasonable. However, if there was a large device population, or if the devices all have a large amount of data to transmit, then collisions could become problematic. Indeed, collisions result in retransmissions, which increases traffic load and result in greater probability of collisions. This snowballing effect can cause the eventual collapse of the channel. Theoretically, using Poisson arrival rates for the data and infinite retransmissions5.5 by devices, such a channel may only reach 18% of channel capacity before collapse occurs.
To address this low maximum throughput, the Aloha scheme was altered by forcing synchronized access. This means that all devices are still allowed to transmit whenever they have data, however, every device must start their transmissions at predefined times. This effectively reduces the probability of collision by half, which doubles the maximum effective channel capacity to 36% .
This type of medium access mechanism, though enjoying low protocol overhead, is unacceptably inefficient for a public data network. Much of the problem lies in the devices' lack of knowledge of the channel status prior to initiating data transmission.5.6 The approach discussed in the next subsection incorporates such assessment of channel status. CDPD uses neither Aloha nor slotted Aloha mechanisms.
The CDPD approach to media access control closely follows that of the more familiar Carrier Sense Multiple Access with Collision Detection (CSMA/CD) scheme. The CSMA/CD scheme is typified by the implementation in the ubiquitous Ethernet systems. In this scheme, a unit on the network may transmit on the channel whenever it has data to send and the channel is not already occupied by another unit.
In CSMA/CD, a unit on the channel wishing to transmit a block of data must first assess the state of the channel. If the channel is found to be unoccupied and available, the unit may start its transmission. If the channel is occupied, the device must wait an amount of time before attempting to access the channel again. This is the carrier sense portion of the methodology.
If two units find the channel unoccupied at the same time, they may both transmit, in which case, there would be a collision of their transmissions and the likelihood is that neither transmission would be successfully decoded by its intended recipient. In this case, a collision detection mechanism triggers both units to stop their current transmissions. Both units must then wait a random amount of time before attempting to reaccess the channel. This is the collision detection portion of the scheme.
The amount of random time a unit must wait before reaccessing the channel is the back-off time. This random value is chosen from an exponentially growing maximum value with each successive retry of the same transmission.
The CSMA/CD mechanism bases its performance on the ability of each unit to sense the transmission of another unit sharing the channel. While this is easy to achieve on a baseband transmission medium such as on a LAN, it is much more difficult and unreliable on the CDPD radio channel. The CDPD network uses two distinct frequencies for the forward and reverse channels. For a mobile unit to directly detect another mobile unit's transmission, it must configure it's receiver to capture the signal. This adds complexity, cost, size and power consumption to the mobile device.
Furthermore, even if every mobile unit was designed to receive transmissions by other mobile units, the performance would be unreliable. This is because most mobiles are low power units that operate at ground level. The transmissions from these units, while adequately received by base stations with high antennae and high gain circuitry, may not be detectable by other mobile units. These effects make CSMA/CD an inappropriate mechanism to deploy directly in the CDPD network.
However, it is recognized that the basic concept of CSMA/CD has much merit in an RF based multiple access network. Another media access mechanism based closely on CSMA/CD is indeed used in the CDPD network.