FIELD OF THE INVENTION
The invention relates to a wireless communication system. More particularly, the invention relates to initialization of a communication link between a base station (BS) and a user equipment (UE).
BACKGROUND OF THE INVENTION
During an initial cell search (ICS) or power-up of a UE, a training sequence of known symbols is used by the receiver to estimate the transmitted signal. In a time division duplex (TDD) signal, for example, the midamble of a TDD frameconventionally contains the training sequence of symbols. The conventional cell search process consists of a Step 1 algorithm which processes a primary synchronization code (PSC) on the primary synchronization code channel (PSCH) for synchronizationchannel (SCH) location determination. A Step 2 algorithm processes the secondary synchronization codes (SSC) for code group determination and timeslot synchronization, and a Step 3 algorithm performs midamble processing.
Variable control oscillators (VCOs) are commonly used at the end of an automatic frequency control (AFC) process to adjustably control the frequency of the receiver to achieve synchronization between a transmitter and a receiver. The input forthe VCO is a control voltage signal, which is typically generated by a control circuit that processes the amplitude and phase of the received symbols. A common problem during an AFC process is the initial fluctuations resulting from a potentiallysignificant frequency offset between the transmitter and the receiver.
SUMMARY
A method and apparatus for adjusting the frequency of a VCO at a receiver to synchronize the receiver with the transmitter by correlating a synchronization code channel with training sequences to estimate positive and negative offsets which areemployed to estimate an error, which is then filtered. The filter output provides a voltage controlling the VCO. The same technique may be employed to control a numeric controlled oscillator (NCO).
BRIEF DESCRIPTION OF THE DRAWINGS
The invention will be understood from the following description and drawings in which like elements are designated by like numerals and, wherein:
FIG. 1 is a block diagram showing the phase rotation approach for startup AFC.
FIGS. 2A and 2B, taken together, comprise a block diagram of the interaction between start-up AFC and algorithm Steps 1, 2 and 3 of cell search.
FIG. 2 shows the manner in which FIGS. 2a and 2b are arranged to create a complete block diagram.
FIG. 3 shows a process diagram for a PI filter.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT(S)
FIG. 1 is a block diagram of a start-up adaptive frequency control (AFC) 10 used to reduce the frequency offset between a base station (BS) and user equipment (UE) during initial cell search procedure. Start-up AFC uses a phase rotationapproach, which is based on the correlations of two sequences with the primary synchronization code (PSC). The stored PSC sequence 12 is rotated in opposing directions at 14, 14a, 16, 16a to respectively determine correlations with the received sequence18 at 20 and 22. The absolute values (a and b) are obtained at 24 and 26 and to obtain the value
.times..times..times. ##EQU00001## from circuit 27, where c is an arbitrary constant provided to prevent division by zero. The phase rotation at -3 kHz alternatively can be replaced by a conjugate of a rotated PSC sequence at 3 kHz since thePSC sequence can only have values of (1+j) and (-1-j).
During start-up AFC process, it is assumed that the PSC location provided is correct. Once Step 1 completes generation of the first outputs, the start-up AFC starts running. The Step 1 process and start-up AFC process run in parallel. Optimally, start-up AFC reduces the frequency offset from 6 kHz to less than 2 kHz in the least number of iterations. Table 1 shows a particular advantage of frequency correction which is an increase in allowable integrations. The number ofintegrations is limited, however, due to chip slip. The chip-slip upper bound is 0.5Tc since the maximum correlation is generated one sample later for a method utilizing twice the chip rate sampling. Table 1 summarizes the allowable number ofintegrations as frequency offset is reduced. Table 2 provides information on performance degradation for a coherent combining technique in the presence of carrier frequency offset.
TABLE-US-00001 TABLE 1 Frequency Offset vs. Number of Integration Allowed Frequency Offset Slip per frame Number of integrations allowed .+-.6 kHz = .+-.3 ppm 0.1152 Tc 4 .+-.4 kHz = .+-.2 ppm 0.0768 Tc 6 .+-.2 kHz = .+-.1 ppm 0.0384 Tc 13.+-.1 kHz = .+-.0.5 ppm 0.0192 Tc 26
TABLE-US-00002 TABLE 2 Frequency Offset vs. Code Length for Coherent Combining Length of the code Carrier frequency Offset Loss in dB integrated coherently Fc = 2 GHz 2.42 256 .+-.3 ppm 6 kHz 1.04 256 .+-.2 ppm 4 kHz 0.26 256 .+-.1 ppm 2 kHz0.06 256 .+-.0.5 ppm 1 kHz 12.62 512 .+-.3 ppm 6 kHz 4.53 512 .+-.2 ppm 4 kHz 1.04 512 .+-.1 ppm 2 kHz 0.26 512 .+-.0.5 ppm 1 kHz
The start-up AFC procedure includes a mechanism to realign the primary synchronization code (PSC) position that may shift during correction. The Step 1 procedure can be run to eliminate the need for the mechanism while the start-up AFC algorithmis running. The Step 1 procedure updates the peak location every 4th frame.
FIG. 2 depicts the parallel processing relationship among start-up AFC and Steps 1, 2 and 3 of cell searching. Of particular concern is the relationship between Step 1 and start-up AFC. Since Step 1 works in parallel with the startup AFC, thereis no need for a code tracker circuit to follow a given path. Each time Step 1 updates an output that is based on the largest detected value, start-up AFC uses the new peak location to estimate the new frequency offset.
The frequency estimator block (FEB) 31 of the start-up AFC comprises a Sequence Locator and Splitter 32, frequency estimators 34 38, a proportional plus integral (PI) filter 42, and a voltage controlled oscillator (VCO) or numeric controlledoscillator (NCO) 46 coupled to PI filter 42 through the sign flop 44. The input 32a to the Sequence Locator and Splitter 32 includes the PSC peak location chip-offset provided by Step 1. Start up AFC 30 is an open loop gain control block that stepsthrough pre-defined gain levels in order to set proper input power level before digitizing the input. The main input to both Step 1 and the Sequence Locator and Splitter 32 is sampled at twice the chip rate with a length of 76,800 complex elements. Since the chip-offset points to the peak location, the beginning of the PSC is 511 samples before the chip-offset. The outputs of the Sequence Locator and Splitter 32 are generated by the following general equation: Output=input[i-511]i Eq. (1)
Accordingly, the three particular outputs of the Sequence Locator and Splitter 32 are represented by the following equations for early (32b), punctual (32c) and late 32(d) estimates: Early[i]=input[i-511]i=offset-1, offset, offset+1, . . . ,offset+510 Eq. (2) Punctual[i]=input[i-511]i=offset, offset+1, offset+2, . . . , offset+511 Eq. (3) Late[i]=input[i-511]i=offset+1, offset+2, offset+3, . . . , offset+512 Eq. (4)
Although the Locator and Splitter 32 in the example given in FIG. 2, is a PSC locator, it should be understood the same approach can be used with any received sequences other than PSC.
The input samples to the Sequence Locator and Splitter are taken at twice the chip rate.
The frequency estimators 34, 36 and 38 each receive one of the three inputs provided by Equations (2) (4). The frequency estimators estimate a different frequency offset, summed at 40, for each input sequence in accordance with FIG. 1. Thefrequency offset, summed at 40, is the summation of early, punctual and late estimates.
The sum of the estimates is passed through a proportional plus integral (PI) filter 42 with coefficients alpha and beta, respectively as shown in detail in FIG. 3. The PI filter bandwidth has two settings. Initially, alpha and beta arepreferably 1/12 and 1/256, respectively as shown in detail in FIG. 3. The loop gain k is set at (k=-1.0). During steady state, alpha and beta are set to 1/16 and 1/1024, respectively. FIG. 3 depicts such a PI filter structure 42. The preferablesettings for coefficients alpha and beta are summarized in Table 3. However, other filters may be substituted for the PI filter.
TABLE-US-00003 TABLE 3 PI Filter Coefficients as a Function of Operating Conditions. Condition alpha beta initial 1/2 1/256 steady state 1/16 1/1024
Steady state condition is established when:
the startup AFC completes at least ten (10) iterations;
while the last eight (8) outputs (inputs to VCO) are put into a buffer of length eight (8); the difference between the absolute value of the average of the first half and that of the second half is within .+-.1 kHz; and
the current output to the VCO is within .+-.1 kHz of the absolute value of the average of the second half.
For digital applications, a numerically controlled oscillator (NCO) is used in place of the VCO.
The start-up AFC algorithm relies on PSC location update to estimate the carrier frequency offset. Step 1 runs during frequency correction to update the PSC location. As such, it is preferable that start-up AFC is begun immediately following asuccessful Step 1 process, with Step 1 running in parallel. Step 1 continues to provide updated PSC locations once every N1 frames as per the Step 1 algorithm, where N1 is the maximum number of frames for averaging. Start-up AFC is run in this mannerfor a duration of L frames, with L=24 as the preferred value. The Step 1 FLAG 61 from controller 60 is set when a sequence is detected. The FEB 31 runs when the controller 60 provides an enable condition to FEB 31 at 62. Since the peak locations shiftleft or right in time, the Step 1 algorithm is run constantly. At the end of L frames, the startup AFC reduces the frequency offset to about 2 KHz in many cases, which provides considerable enhancement to the Step 2 performance. The inclusion of Lframes contributes to the overall cell search delay budget and hence is chosen conservatively to be L=24.
PSC processing block 66 correlates against the primary synchronization code in (synchronization channel) (SCH) over frames. The SCH location is not known.
SSC extractor block 68 utilizes the SCH location and extracts only the SCH portion, which is then passed to SSC processing block 70.
SSC processing block 70 correlates against the secondary synchronization code in synchronization channel over SCH.
Midamble Extractor block 72 utilizes the SCH location and SSC processing results and extracts the midamble portion to pass to midamble processing block 74.
Midamble processing block 74 correlates against possible midambles given by SSC processing and picks the one with the highest energy.
Periodic Cell Search block 76 performs a process which constantly searches for the best base station for the given period.
Controller 60 coordinates among stages to synchronize to a base station.
Layer 1 Controller 80 coordinates all layer 1 related hardware and software in order to maintain proper operation in the receiver.
* * * * *