WIRELESSAPPLICATIONS

IN

spread-spectrum systems, such as wideband code-division multiple access (WCDMA) [under the Third-Generation Partnership Project, or 3GPP] and cdma2000 (the Third-Generation Partnership Project 2, or 3GPP2), the handset can transmit multiple channels at different amplitude levels. Modulation schemes such as OQPSK or GMSK are no longer suitable, creating a need for a modulation format or a spreading technique that can accommodate multiple channels at different power levels while producing signals with low peak-to-average power ratios. Hybrid-phase-shift keying (HPSK), also known as orthogonal-complex quadrature phase-shift keying (OCQPSK), has been proposed as the spreading technique for WCDMA and cdma2000. HPSK is a complex spreadA COMPLEX SPREADING TECHNIQUE ing scheme that is very different from the modulation formats commonly used. THAT IS VERY DIFFERENT FROM OTHER BASIC 3G STRUCTURES MODULATION FORMATS, HYBRID-PHASEIn 3G systems, the mobile station can SHIFT KEYING HAS BEEN PROPOSED FOR transmit more than one channel. The different channels are used for control purWCDMA AND cdma2000. poses or to send voice and/or high-speed data. For example, for WCDMA, the baB Y M A R T A I G L E S I A S - X A M A N I sic uplink signal comprises one dedicated physical-data channel (DPDCH) and one dedicated physical-control mobile communications, battery fier can operate more efficiently. However, in third-generation (3G) channel (DPCCH). Optionally, more life is one of the most important characDPDCHs may be added to supteristics of the handset, and efport higher data rates. ficiency in the handset power In the case of cdma2000, the amplifier (PA) is key to maxiIchip To DPDCH Gain baseband or mobile always transmits a remizing battery life. Amplifiers scale filter R-Pilot verse pilot (R-Pilot) channel. are typically most efficient Additional channels, such as when they operate close to their Walsh/OVSF generator the reverse fundamental chansaturation level. Therefore, Gain nel (R-FCH) and reverse supmobile-station amplifiers Other scale channels plemental channels (R-SCHs), should ideally be designed so are used to send voice and that the average power level of Walsh/OVSF generator high-speed data, respectively. the signal is as close as possible For WCDMA and cdma2000, to this saturation level. To Qchip Gain DPCCH baseband the channels are in phase/quadThis works well for secondscale or R-FCH filter rature (I/Q) multiplexed. In the generation (2G) modulation case of transmitting only two formats, such as offset-quadraWalsh/OVSF generator channels, one of the channels ture-phase-shift keying Gain Other (DPDCH or R-Pilot) is applied (OQPSK) or Gaussian miniscale channels to the I path and the other chanmum-shift keying (GMSK). nel (DPCCH or R-FCH) is apFor these formats, the headWalsh/OVSF generator plied to the Q path. Additional room required in the amplifier high-data-rate channels are to prevent compression of the signal and interference with the 1. This shows the basic reverse-channel structure of 3G systems combined alternatively on the I adjacent frequency channels is (WCDMA and cdma2000) without scrambling. Without spreading or Q paths. Each channel is small (or, in the case of GMSK, or scrambling, the I and Q signals would be directly filtered and spread by a different orthogonal code. non-existent). Thus, the ampli- applied to the I/Q modulator.
WIRELESS SYSTEMS DESIGN / JANUARY 2000

Hybrid-PhaseShift Keying Proposed For 3G Systems

24

WIRELESSAPPLICATIONS
The different channels can be at different power levels. If there was no other spreading or scrambling applied (Fig. 1), the I and Q signals would be directly filtered and applied to the I/Q modulator. In that scheme, channels with different power levels result in different amplitudes for I and Q, which can produce strange constellations. For example, in the case of only two channels, unequal power levels result in a rectangular 4-quadrature-amplitude-modulation (QAM) constellation. In general, equal distribution of powers between the axes is desired, especially at the receiver. COMPLEX SCRAMBLING In the reverse link of cdmaOne systems, the I and Q signals are scrambled with a pseudorandom-noise (PN) sequence prior to modulation. Instead of traditional scrambling, complex scrambling has been proposed for the reverse link (uplink) of WCDMA and cdma2000 systems. In addition to providing differentiation among users, complex scrambling fixes the unequal distribution of powers by continuously rotating the constellation. Figure 2 shows a block diagram of the reverse link (uplink) with basic complex scrambling. Mathematically, complex scrambling performs the multiplication of two complex signalsthe complex data signal which has already been spread into chips
Ichip + + I + jQ 90 deg. Qchip + + Complex scrambling Q +

Is Qs

3. Mathematically, complex scrambling performs the multiplication of two complex signals. The expression for the final I and Q signals can be derived from this diagram.

(Ichip +jQchip) and the complex scrambling signal (IS + jQS). The expression for the final I and Q signals can be derived from Fig. 3: I = Ichip . I S Qchip . QS (1)
Q = Ichip . QS + Qchip . I S
(2)

final signals can be expressed as complex signals:

I + jQ = ( Ichip . I S Qchip . QS ) + j ( Ichip . QS + Qchip . I S ) = ( Ichip + jQchip ) . ( I S + jQS ) = Achip . AS . e
j ( chip + S )

Since the final I and Q signals are later I/Q modulated, the data, scrambling, and

(3)

DPDCH or R-Pilot Walsh/OVSF generator

Gain scale

Ichip +

To baseband filter

Other channels Walsh/OVSF generator

Gain scale

DPCCH or R-FCH Walsh/OVSF generator Other channels Walsh/OVSF generator

Gain scale

Qchip

+ +

Q To baseband filter

Gain scale Is Qs

2. This is the reverse link (uplink) for 3G systems (WCDMA and cdma2000) with basic complex scrambling.
26

where: Achip and ej chip = the amplitude and the phase of the Ichip +jQchip signal, and AS and e j chip = the amplitude and the phase of the IS +jQS signal. Therefore, the amplitude (A) of the resulting I+jQ signal is the product of the amplitudes of both signals. Its phase is the sum of their phases. The original complex data signal spread into chips is referred to as the chip signal and to its I and Q components as Ichip and Qchip, respectively. The resulting complex signal will be referred to as the final signal and its I and Q components simply as I and Q, respectively. For simplicity, a signal with only two channels (one in the I path and the other one in the Q data paths) is used to explain this concept in the I/Q plane (Fig. 4). In the case of two channels with the same amplitude (same amplitude for Ichip and Qchip), the chip signal maps onto a QPSK constellation. The scrambling signal also maps onto a QPSK constellation (since the scrambling Is and Qs signals have values of 1 or -1).
WIRELESS SYSTEMS DESIGN / JANUARY 2000

WIRELESSAPPLICATIONS
In the case of two channels with different amplitudes, the chip signal maps onto a rectangular 4-QAM constellation. The scrambling signal still corresponds to a QPSK constellation. What is the final constellation for these two cases? Figures 5a to d illustrate what happens for a single-chip point in the original QPSK constellation (case 1). The amplitude of the final signal is the product of the amplitudes of the chip and scrambling signals. The phase of the scrambling signal is added to the phase of the original chip signal (45 deg.) Any point from the original QPSK chip constellation is rotated by 45 deg., -45 deg., 135 deg., or -135 deg., depending on the values for Is and Qs at that time. For example, in Fig. 5a, the phase of the scrambling signal is +45 deg., so 45 deg. is added to the phase of the original chip point. TWO CHANNELS Therefore, in the case of two channels with equal amplitudes (QPSK constellation), the result is a QPSK constellation that is rotated 45 deg. from the original constellation. In the case of two channels with unequal amplitudes, the resulting constellation has eight points at constant amplitude. The angular distribution of the points is determined by the relative amplitudes of the two channels (Ichip and Qchip). Therefore, complex scrambling is used in 3G spread-spectrum systems, because it distributes the power evenly between the I and Q axes (Fig. 6). Second-generation systems use modulation formats that limit transitions through zero in the reverse link (uplink). For example, cdmaOne uses OQPSK. This reduces the peak-to-average power
Case 1: two channels at equal amplitude Ichip + jQchip 2 1, 1 45 deg. Is + jQs 2 1, 1 45 deg. 0, 2 2 +45 deg. I + jQ 45 deg.

(a)

Ichip + jQchip 2

1, 1 45 deg.

Is + jQs

I + jQ 2 45 deg. 2 1, 1 +135 deg. 2 +135 deg. 2, 0 45 45 deg. deg. 2, 0

(b)

Ichip + jQchip

Is + jQs 1, 1 1, 1 2 45 deg. 2

I + jQ 45 deg.

(c) Ichip + jQchip 1, 1 2 45 deg. 2 (d) 1, 1 0, 2 135 deg. +135 deg. Is + jQs I + jQ 2 45 deg.

5. This figure shows (a) scrambling a signal at +45 deg., (b) scrambling a signal at -45 deg., (c) scrambling a signal at +135 deg., and (d) scrambling a signal at -135 deg.

ratio of the signal. However, in complex scrambling, if random PN signals are assigned to Is and Qs, transitions from any point to any point in the final constellation are possible. This results in a high peak-to-average ratio when compared to existing 2G formats for the reverse I + jQ link. The spreading Ichip + jQchip Is + jQs technique that uses 1, 1 1, 1 basic complex ? Case 1: Two channels scrambling and PN at equal signals for I s and amplitude Q s is known as 1, 1 1, 1 pseudo-noise complex-phase-shift 1, 1 1, 1 ? Case 2: keying (PNCPSK). Two channels Third-generation at different systems use HPSK amplitude 1, 1 1, 1 to reduce the peakto-average power ratio of the signal. 4. For simplicity, a signal with only two channels is used to HPSK is a variaexplain this concept in the I/Q plane.

tion of basic complex scrambling that ensures 90-deg. transitions for every second chip point. HOW HPSK WORKS Basically, HPSK uses complex scrambling with a fixed repeating function as the scrambling signal. This function is known as the Walsh rotator, and it is defined as W0 = {1, 1} for Is, and W1 = {1, -1} for Qs. Figure 7 illustrates the concept in the I/Q plane. In this case, two consecutive chip points fall at the same place on the I/Q plane. The repeating Walsh rotator sequence (Is = W0 = {1, 1}; Qs = W1 = {1, -1}) is used as the scrambling signal. For two consecutive identical chip points, the first one is rotated by +45 deg., and the second one is rotated -45 deg., which ensures that they are 90 deg. apart in the final constellation and the transition between them does not go through zero.
WIRELESS SYSTEMS DESIGN / JANUARY 2000

28

WIRELESSAPPLICATIONS
This technique assumes pairs of consecutive identical chips. This can be achieved by using only even-numbered Walsh functions to spread the data from the different channels. Even-numbered Walsh functions consist of pairs of identical bits. For example, W0 = {1, 1, 1, 1, 1, 1, 1, 1}, W2 = {1, 1, -1, -1, 1, 1, -1, -1, and so forth. SCRAMBLING SIGNALS Figure 8 shows a generic block diagram where the Walsh rotator is used as the scrambling signal. To simplify the diagram, only two channels at the same amplitude are transmitted. The orthogonal functions chosen in this case for each channel are 8-b Walsh codes 0 and 2, respectively. For example, if the data signal is ID = 1 and QD = -1, the chip signal is Ichip = 1, 1, 1, 1, 1, 1, 1, 1 and Qchip = 1, -1, 1, 1, -1, -1, 1, 1. The constellation of this signal consists of the complex points Ichip +jQchip = {1-j1, 1-j1, 1+j1, 1+j1, 1-j1, 1-j1, 1+j1, 1+j1}. Therefore, the chip signal consists of pairs of identical consecutive points. Each pair is multiplied with the scrambling signal formed by the Walsh rotator Is + jQs = W0 +jW1 = {1+j1, 1-j1}. Therefore, the first point in the pair is phase shifted by +45 deg., and the second one by -45 deg. In the real system, the length of the orthogonal code for each channel depends on that channel's data rate. HPSK limits the choice of available orthogonal spreading codes. However, a single mobile does not need to support a large number of traffic channels. Therefore, in most cases, there are sufficient orthogonal codes to handle most channel configurations for cdma2000 and WCDMA.
I + jQ Ichip + jQchip Case 1: Two channels at equal amplitude 1, 1 Is + jQs 1, 1

1, 1 1, 1 Case 2: Two channels at different amplitude 1, 1

1, 1 1, 1

1, 1

6. Complex scrambling is used in 3G spread-spectrum systems because it distributes the power evenly between the I and Q axes.

Ichip + jQchip

Is + jQs = W0 + jW1 = {1, 1} + j{1, 1}

I + jQ 1st

+45 deg. 45 deg. 45 deg. 2nd

1st

2nd

1st 1, 1 45 deg. 45 deg. 2nd 1, 1

45 deg.

7. This illustrates how the Walsh rotator works in the I/Q plane.

In HPSK, zero crossings are eliminated for every other signal transition. Therefore, an HPSK signal has fewer zero crossings than a QPSK signal without HPSK. In addition to minimizing zero cross-

ings, HPSK eliminates 0-deg. phaseshift transitions for every other chip point. A 0-deg. phase transition occurs when two consecutive points are at the same place on the final constellation. This causes overshooting trajectory,

1 channel

Ichip + W0 I Peak caused by 90-deg. phase shift Peak caused by 0-deg. phase shift

{1, 1, 1, 1, 1, 1, 1, 1} Qchip

1 channel

+ +

W2 {1, 1, 1, 1, 1, 1, 1, 1} W {1, 1} W {1, 1}

8. This is a generic diagram where the Walsh rotator is used as the scrambling signal. To simplify the diagram, only two channels at the same amplitude are transmitted.
30

9. A 0-deg. phase transition occurs when two consecutive points are at the same place on the final constellation. This causes an overshooting trajectory, which increases the peak-to-average power ratio of the signal.
WIRELESS SYSTEMS DESIGN / JANUARY 2000

WIRELESSAPPLICATIONS
100 + W0 {1, 1, 1, 1, 1, 1, 1, 1} PN 1 channel Qchip + W2 {1, 1, 1, 1, 1, 1, 1, 1} W {1, 1} P W {1, 1} Decimate PN(2) +
(1)

1 channel

Ichip I Percentage

Delta marker: 1.05 dB 10 1 0.1 0.01 0.001 0.0001 0 Meas BW 5 MHz 20 dB 1 2 Signal with basic complex scrambling

Signal with HPSK spreading

10. This shows a complete block diagram for HPSK.

11. This compares the CCDF for two signals with basic complex scrambling and with HPSK spreading.

which increases the peak-to-average power ratio of the signal (Fig. 9). HPSK forces 90-deg. transitions between pairs of consecutive points. This minimizes 0-deg. phase transitions, which further reduces the peak-to-average power ratio of the signal. PN FUNCTIONS The real HPSK block diagram is more complex (Fig. 10). After complex multiplication with the Walsh rotator, a primary PN spreading code PN(1) is applied to the final I and Q signals to enable identification of the mobile and correlation at the receiver. The PN(1) sequence does not affect the number of 90-deg. transitions. Also, a decimated secondary PN spreading code (P) is multiplied with the Q component of the Walsh rotator (W1 = 1, -1}). The secondary PN spreading code minimizes multiaccess interference (MAI), thereby improving reception at the receiver. P is a decimated version of the real chip-rate sequence PN(2). For example, for a decimation factor of two, P holds its value for two-chip time periods, which effectively makes its rate half the chip rate. P randomizes the direction of the phase rotation while keeping the 90-deg. phase shift for every other transition. When PN(1) and P are included, the total spreading (or scrambling) function for HPSK is:
I S + jQS = PN
(1)

Since W0 = {1, 1}, this function can be removed from the equation and does not need to be implemented:
I S + jQS = PN (1) + j PN (1) P W1
(5)

(Wo + jP W1 )

(4)

where: W0 = {1, 1}, and W1 = {1, -1}.

32

WCDMA and cdma2000 may implement this function in different ways. The probability of zero crossing for a regular QPSK or PNCQPSK signal with two channels at the same amplitude is 1/4. For HPSK, the probability of zero crossing is limited to every other chip point and, therefore, is reduced in half (1/8). The probability of 0-deg. phaseshift transitions is also reduced from 1/4 to 1/8. All of this improves the peak-toaverage power ratio of the signal by 1 to 1.5 dB, approximately. HPSK spreading is still advantageous when multiple channels at different amplitudes are transmitted. Figure 11 shows the complementary cumulative distribution function (CCDF) for two signals. One is a signal with basic complex scrambling. This signal has been generated from a cdma2000 SR1 forward-link signal (which uses basic complex scrambling) with a single traffic channel. The baseband filtering has been modified to remove equalization. The other is a signal with HPSK spreading. This signal is a cdma2000 SR1 reverse-link signal with two channels (R-Pilot and R-FCH) at the same power level. The CCDF curve provides the distribution of particular peak-to-average ratios versus probability. In this case, for a

probability of 0.1 percent, the peak-toaverage ratio of the reverse-link signal is approximately 1 dB lower than that of the forward-link signal. The plot is indicative of the performance of HPSK spreading versus basic complex scrambling. Signals with high peak-to-average power ratio may saturate the PA, causing higher interference in the adjacent channels and a reduction of system capacity. To minimize this, the amplifier must be designed with a larger back-off, which, in turn, reduces amplifier efficiency. Therefore, high peak-to-average power ratios reduce battery life, one of the critical characteristics of the mobile phone. In WCDMA and cdma2000 systems, the mobile phone can transmit multiple I/Q multiplexed channels at different power levels. Complex scrambling facilitates this by distributing the power evenly between the axes. HPSK is a variation of complex scrambling that uses a Walsh rotator and specific orthogonal (or Walsh) spreading functions to minimize zero crossings and 0-deg. phase-shift transitions. This improves the peak-to-average power ratio of the signal. For this reason, HPSK has been selected as the spreading technique for the reverse link of these 3G systems. WSD
MARTA IGLESIAS-XAMANI, Wireless Industry Marketing Engineer, Agilent Technologies, Inc., Microwave Instruments Division, 1400 Fountaingrove Parkway, Santa Rosa, CA 95403; (707) 577-4620, FAX: (707) 577-5834, e-mail: Marta_Iglesias@agilent.com. WIRELESS SYSTEMS DESIGN / JANUARY 2000