CDG : Technology : Welcome to the World of CDMA

Handoff

The act of transferring support of a mobile from one base station to another is termed handoff (handover, in some parts of the world). The implementation of handoff is radically different between the narrowband standards and the CDMA standards.

AMPS handoffs fail frequently, causing dropped calls. This contributes to perceptions of poor service quality. Moreover each handoff is preceded and followed by long intervals of poor link quality, resulting in annoying noise and distortion. By comparison, CDMA is specifically designed not only to reduce handoff failures but also to provide (pardon the cliché) seamless service. CDMA maintain good voice quality at all times: before, during, and after handoff. The handoffs are undetectable, even by skilled listeners.

Handoff Rate

How much of a problem is this? After all, if handoffs are rare, doing them poorly is a less important consideration.

The number of cells transited during a call can be crudely estimated with the following assumptions:

Cells are circular with a uniform radius R.
Mobile station trajectory is a straight line.
Mobile enters each cell at a random location on the periphery, with a random direction within ±90 degrees from radial.

With these assumptions the average path within a cell is easily shown to be about pi*R/2. The number of cells transited during a call of duration T, by a vehicle with speed v is thus crudely estimated at about 0.65*v*T/R. Some representative cases are shown in the table for three minute calls.

Case	Speed (mph)	Speed (kmph)	Cell Radius (km)	Handoffs
Freeway, rural	65	104	16	0.33
Freeway, urban	65	104	1.6	3.25
Surface streets, urban	30	48	1	2.40
Pedestrian, urban	1.5	2.4	1	0.12
Pedestrian, microcell	1.5	2.4	0.1	1.20

The number of handoff is seen to be anywhere from perhaps one every 8-10 calls to perhaps 3-4 per call. Although this is only a crude estimate, it does indicate that handoff is frequent enough that good performance is important.

It is perhaps noteworthy that CDMA cell radii can be considerably larger than AMPS cell radii for any particular load distribution. This reduces handoff rates simply on the basis of geometry, independent of the many other performance advantages that acrue to CDMA.

Steps in a Handoff

Regardless of technology, the following steps are part of handoff of any call.

Starting in a state where only one cell is supporting the call in question.
Determining that over-the-air link conditions between the mobile and the old serving cell are deteriorating, and that there is a potentially better link to a new, candidate cell.
Informing the candidate cell of the imminent handoff, including parameters needed to identify the mobile and execute the handoff.
Signaling the mobile to begin executing the handoff.
New cell beginning to service the mobile
Mobile beginning to use the new cell
Entering the mid-handoff state (prolonged only in CDMA)
Mobile discontinuing use of the old cell
Old cell stopping service to the mobile
Ending in a state where one cell, the new one, is supporting the call in question

Real network exchanges are much more complex, but these are the major events. Some steps entail different activities in AMPS versus CDMA. For example, CDMA handoffs do not normally require frequency tuning; instead they require change of the code channel in the Forward CDMA Channel. No tuning, either frequency or code channel is required in the Reverse CDMA Channel at any time.

Handoff - The Old Way

Handoff in AMPS has several significant aspects that create trouble:

It is "hard", meaning that communication is interrupted briefly while control is transferred. Mobiles can tune to only one channel pair at a time, and a switch of pairs is required because adjacent cells never use the same frequency sets.
There is no diversity during handoff in the sense that all communication is with one base station or the other. At no time is there simultaneous communication with more than one station.
The base stations, not the mobiles do the signal quality measurements that lead to the decision to do a handoff.

The handoff trigger in an AMPS system can be several things:

Absolute received signal level as measured by the current serving cell receiver. This can be misleading because the signal strength measurement does not distinguish between the desired signal and interference; the latter can sometimes be large. If it is, the measurement can be misleading and will fail to initiate the handoff.
Signal power difference between the current serving cell and a candidate cell. This tends to work better than absolute threshold because it tends to remove the effects of variations in mobile antenna gain, placement, and power output.
Receiver quieting as measured, for example, by the post-detection signal-to-noise ratio of the SAT. This does not seem to be used much in practice.

The trigger mechanism must anticipate the need for handoff with enough margin for execution. It is not possible to do this with perfect reliability based only on signal strength measurements. If the trigger is too sensitive, then handoffs will be initiated when they are not needed. Mobiles transiting areas of low signal strength well within a cell may trigger a handoff, followed immediately by a handoff back. Spurious handoffs will reduce system capacity somewhat, and will increase the likelihood of a failure for other reasons, such as lack of resources in the target cell.

The concept of an abrupt handoff is fundamentally flawed in the sense that static signal strength measurements cannot foretell the future. Even if one could make assumptions about the dynamics of the mobile station, extracting information about that motion from the communication signal is all but impossible. Moreover, calculating propagation loss from the station's location is an extremely difficult problem, even in principle. Doing such a calculation "on the fly" in a fraction of a second is not remotely practical.

The solution that has generally been adopted by the infrastructure manufacturers is to monitor signal strength (RSSI) on the current serving cell. When the RSSI drops below a threshold, then measurements are requested from predetermined handoff candidate cells. After the measurement reports are collected the messaging to carry it out is initiated. Even when this works, it is unsatisfactory because the initial RSSI threshold has to be set so low that the voice quality has generally deteriorated badly before the process is even triggered.

The ensuing handoffs often fail, even if they were warranted. Some of the most frequent causes are:

There was no channel available in the target cell due to traffic loading.
The RSSI measurement in a candidate was erroneous due to interference, leading to the incorrect target being selected.
The link quality deteriorated so badly prior to the need for handoff that the signaling needed to carry it out failed.
The 5 second fade timer expired due to low signal before the handoff could be executed.
The mobile was not really at a handoff boundary but rather was in a coverage hole within a cell so no handoff target could be found.

And the voice quality just before and just after the handoff, even when it is successful, is often poor because the handoff boundary is the point where there is large loss to both cells. Nothing in the system design can make up for large loss at the handoff boundaries.

Handoff - The CDMA Way

CDMA handoff differs from AMPS in several key aspects:

It is "soft", meaning that communication is not interrupted by the handoff. This is sometimes called "make before break." But it is more than that.
The handoff is not abrupt, but rather it is a prolonged call state during which there is communication via two or more base stations. The multi-way communication diversity improves the link performance during the handoff. The diversity gain partially compensates for the large path loss at the cell boundary.
The signal measurement that triggers the handoff is performed by the mobile stations, not the base stations.

There is no handoff boundary in CDMA but rather a handoff region. The distributed handoff alleviates most of the shortcomings of the AMPS-style hard handoffs. For example, a decision that handoff should be initiated means only that another base station is added to the active set of base stations for this mobile. The handoff can be completed either by the mobile moving completely into the new cell, or by the mobile returning to the original serving cell. In either case the call is never in jeopardy due to link failure.

The diversity during handoff improves link performance to the point where not only are the handoffs not disruptive, they are not even detectable in the voice quality, even by skilled observers. This also very much reduces the likelihood of dropped calls due to signaling failures that disrupt handoff coordination.

Handoff Details