TSG-RAN Working Group 1 meeting #4 3GPP/TSGR.

1#4(99)-384
Yokohama, Japan
April 19-20, 1999

TITLE:

Comparison of Differential and Coherent RACH Preamble Detection

SOURCE:

Motorola

ABSTRACT:

This contribution compares conventional coherent detection of the RACH preamble with a differential
approach as a function of Doppler spread. Coherent detection is found to be superior at speeds up to about
200 km/h at a carrier frequency of 2 GHz..
1.0 Introduction
Two methods of RACH preamble detection have been studied. The first consists of coherently
summing the preamble symbols, squaring the result, and comparing with a threshold. In [1] an
alternative approach based on differential detection is proposed. This approach is claimed to have
significantly superior performance in channels with multipath fading and frequency offset. The
performance of these approaches was compared in [2]. This contribution presents simulation
results which further compare these approaches. The comparison is based on the probability that a
transmitted preamble is detected for a fixed false alarm probability. This differs from [2] where
the preamble with the largest metric is compared with the transmitted preamble to determine a
detection probability.

2.0 Signal Model

Figure 1: illustrates the model used for the comparison. The transmitted signal consists of the
T
length 16 preamble sequence, p = p 0 p 1 … p 15 spread by a length 256 spreading code. The
channel consists of multipath fading and additive noise. The fading is assumed to be constant over
the duration of a single preamble symbol, 256 chips, and distributed as complex Gaussian random
variables, i.e., independent real and imaginary Gaussian distributed components with equal vari-
T
ance. Denote the random vector of fading values across the preamble as f = f 0 f 1 … f 15 .
The correlation between these values is assumed to follow the classical Jakes spectrum:

512πvT c l
E { f k f k∗– l } = J 0  ------------------------ (1)
 λ 

where v is the vehicle speed and λ is the carrier wavelength.The additive thermal noise has two-
sided density N 0 ⁄ 2 The T c ⁄ 2 spaced received signal r ( k ) is first matched filtered with the
T
length 256 spreading code. The result is the 16 element vector y = y 0 y 1 … y 15 :

y = M Ecf • p + n (2)

T
where • denotes Schur or element by element product and n = n 0 n 1 … n 15 is a vector of
independent complex Gaussians random variables with variance (real + imaginary part) M N 0 ,
E c is the energy per chip, and M = 256 .

Coherent Detection
The coherent detection statistic, decision z c , is calculated as

(3)
 15 2
zc = ∑ y l p∗l
l=0
(4)
15
= ∑M E c f l + p∗l n l
l=0

Differential Detection
The differential detection decision statistic, z d , is calculated as

14
zd = ∑ Re { y∗l yl + 1 pl p∗l + 1 }
l=0
(5)
14
= ∑ Re { ( M E c f ∗l + p l n∗l ) ( M E c f l + l1 + p∗l + 1 n l + 1 ) }
l=0

For both methods the decision statistic is compared with a threshold to determine the presence of

pl

2
⋅ zc
Preamble matched
y p l p∗l + 1
Sequence r filter

n Re zd
f
Spreading
Code

z-1 ∗
Figure 1: Signal Model

the preamble.

3.0 Simulation Results

The signal model above was simulated to obtain the probabilities of detection and false alarm.
Figures are plots of detection probability vs. E c ⁄ N 0 (SNR) for various speeds. The false alarm
probability is fixed at 2.3 × 10 –5 for both methods. Figure 2 shows the expected superiority of
coherent accumulation at slow speeds: differential detection is approximately 2 dB worse. As a
point of reference, the -21 dB Ec/Ior value corresponds to an Eb/No of 5 dB for SF=128
( data rate ≈ 10kbps ). Figures 3 through 5 show the degradation of coherent accumulation with
increased speed. Note that coherent detection outperforms differential detection through 200 km/
h. Only at 300 km/h does differential detection offer improved detection. At approximately 380
km/h coherent detection suffers the same 2 dB degradation that differential detection incurs at
slow speeds.

4.0 Conclusion
The simulation results presented in this contribution show little reason to modify the preamble
sequences for differential detection. The 2 dB degradation in performance at slow speeds could
only be justified if even greater degradation would occur at high speeds with the coherent
approach. Although such degradations do occur at speeds greater than 380 km/h, the occurrence
of such speeds in fully scattered environments, as was assumed here, would be quite rare. In addi-
tion it is likely that the channel estimation algorithms used on the preamble message will be opti-
mized for speeds much lower than 380 km/h; the result being that the message error rate will tend
to be high at speeds above 380 km/h anyway. The reduced sensitivity of coherent detection at
these speeds will compensate for this effect.
[1] Tdoc SMG2 UMTS-L1 620/98, “Random Access Preamble Detection in the Presence of Dop-
pler”, Interdigital
[2] Tdoc SMG2 UMTS L1 2x99-011, “Random Access Preamble Detection in Doppler; Perfor-
mance in ITU Channel Model”. Interdigital.

1 antenna, One User, One Ray, false alarm = 2.3e−05

512 search bins Carrier frequency = 2.0 GHz
0
10
Detection Probability

speed = 3 km/h Differential

speed = 3 km/h Coherent
−1
10

−2
10
−30 −29 −28 −27 −26 −25 −24 −23 −22 −21
Ec/Io (dB)

Figure 2: 3 km/h
 1 antenna, One User, One Ray, false alarm = 2.3e−05
512 search bins Carrier frequency = 2.0 GHz
0
10

Detection Probability

−1
10

speed = 120 km/h Differential

speed = 120 km/h Coherent
speed = 160 km/h Differential
speed = 160 km/h Coherent
−2
10
−30 −29 −28 −27 −26 −25 −24 −23 −22 −21
Ec/Io (dB)

Figure 3: 120 and 160 km/h

1 antenna, One User, One Ray, false alarm = 2.3e−05

512 search bins Carrier frequency = 2.0 GHz
0
10
Detection Probability

−1
10

speed = 200 km/h Differential

speed = 200 km/h Coherent
speed = 300 km/h Differential
speed = 300 km/h Coherent
−2
10
−30 −29 −28 −27 −26 −25 −24 −23 −22 −21
Ec/Io (dB)

Figure 4: 200 and 300 km/h

 1 antenna, One User, One Ray, false alarm = 2.3e−05
512 search bins Carrier frequency = 2.0 GHz
0
10
Detection Probability

−1
10

speed = 400 km/h Differential

speed = 400 km/h Coherent
speed = 500 km/h Differential
speed = 500 km/h Coherent

−2
10
−30 −29 −28 −27 −26 −25 −24 −23 −22 −21
Ec/Io (dB)

Figure 5: 400 and 500 km/h