Conference Paper: Random Access Preamble Format for Systems with Many Antennas

Random Access Preamble Format
for Systems with Many Antennas
Henrik Sahlin*, Stefan Parkvall*, Mattias Frenne**, Peter Nauclér**
*Ericsson Research, **Ericsson System and Technologies
Ericsson AB
Sweden
Abstract— The computational complexity of FFT
(Fast Fourier Transform) processing in an OFDM
(Orthogonal Frequency Division Multiplex) based receiver
is large with a large amount of receiver antennas. In LTE
(Long Term Evolution) release 8, FFTs of different size
are used for user data and random-access preambles [1]
requiring additional FFTs to be implemented for random-
access reception. Within the current paper, a 5G random-
access preamble format is proposed based on a short
sequence of the same length as the length of the OFDM
symbols that are used for other uplink physical channels,
such as user data, control signaling, and reference
signals. The preamble sequence is constructed by
repeating the short sequence multiple times. A
corresponding preamble detector in which FFTs of the
same size as for other uplink channels and signals are
used is also described. In this way, the amount of special
random-access related processing and hardware support is
significantly reduced for multi-antenna systems with
frequency-domain beamforming. This preamble detector
is also robust against inter-carrier interference from other
uplink channels and signals. Furthermore, the proposed
preamble detector scheme can be used in 5G scenarios
with a high amount of phase noise and frequency errors.
For time-domain beamforming, the beamforming weights
can be changed during preamble reception such that the
number of spatial directions is increased for which
preamble detection is done. Simulation results are used to
compare preamble formats for different lengths of the
sequences.
Keywords—Random Access, preamble, PRACH, LTE,
Massive MIMO, receiver imperfections, 5G
I. INTRODUCTION
With the emerging 5G technologies, the use of a
large number of elements is of great interest,
especially in conjunction with higher carrier
frequencies. At the same time, the cell radius at very
high frequencies is most likely smaller than for a
lower-frequency deployment due to the differences in
propagation conditions.
Random access is used for initial access for a UE
(User Equipment) including timing-offset estimation
at the base station, e.g. as in LTE (Long Term
Evolution) Release 8 [1]. A description of this
procedure is given in [2]. The random-access preamble
should be detected with high probability and low false-
alarm rate by the base station while at the same time
providing accurate timing estimates.
An illustration of the random-access preamble, as
specified for LTE Release 8 [1][3], is given in Fig. 1.
A sub-frame consists of 30720 samples and a random-
access preamble is constructed from a long sequence
of length 24576 samples, a cyclic prefix of length
3168 samples and a guard of 2976 samples. Several
additional formats are specified in LTE [1], with one
or two repetitions of the 24576 samples long sequence
and several different sizes of the cyclic prefix. In the
following, this format is referred to as the preamble
format with a long sequence.
An illustration of the PUSCH (Physical Uplink
Shared Channel) structure, used for data transmission,
is included in Fig. 1. Data transmissions are based on
DFTS-OFDM (Discrete Fourier Transform Spread
Orthogonal Frequency Division Multiplex) where each
DFTS-OFDM symbol has a length of 2048 samples
and a cyclic prefix of 144 or 160 samples.
Several detectors have been proposed for detecting
Fig. 1. User data and preamble with long sequence, with corresponding time windows (in red) and FFTs

random-access preambles, see e.g. [4], including
frequency-domain and hybrid time-frequency-domain
approaches. In the frequency-domain approach, the
received signal is processed with an FFT (Fast Fourier
Transform) corresponding to the length of the
preamble. Hence, an FFT of length 24 576 is thus
required for each receive antenna, see illustration in
Fig. 1. Dedicated hardware might be needed for the
sole purpose of random-access preamble detection.
After this large FFT, the preamble bandwidth is
extracted, which is an output from this large FFT.
Alternatively, a hybrid time-frequency approach can
be used where a low-pass filter is first used in the time
domain in order to extract the preamble bandwidth.
This low-pass filter is followed by an FFT of a size
much smaller than 24 576. One such low-pass filter
has to be applied to receive each antenna signal.
The random-access preamble as illustrated in Fig.
1 covers a time interval which is much longer than the
length of OFDM symbols used for other transmissions
such as user data. The preamble receiver is thus
designed with the assumptions of propagation
conditions not varying over time during the length of
the preamble.
This paper is organized as follows. A random-
access preamble format suitable for 5G systems with
smaller cell sizes and extensive use of beamforming is
proposed and described in section IIA with a
corresponding detector described in section IIB.
Section IIC provides a statistical analysis to be used
for designing preamble detection thresholds. A
discussion of frequency-domain and time-domain
beamforming is found in Sections IIIA and IIIB,
respectively. The paper is concluded with simulations
results in section IV, a discussion in section V and
conclusions in section VI.
II. PREAMBLE FORMAT
A. Transmitter
A short random-access preamble sequence can be
constructed by using Zadoff-Chu sequences. The u
root Zadoff-Chu sequence is defined [1] as
  10, ZC
)1(
ZC



Nnenx N
nun
j
u

(1)
where the length ZCN of the Zadoff-Chu sequence is
a prime number. The use of a Zadoff-Chu sequence is
beneficial due to its constant amplitude properties
leading to low PAPR (Peak to Average Power Ratio).
Also, Zadoff-Chu sequences have good cyclic auto-
and cross-correlation properties [4].
The preamble sequence depends on the frequency
allocation, such that the number of sub-carriers,
denoted by seqN , allocated for preamble transmission
equals the maximum number of symbols. For
example, with LTE nomenclature, 6 resource blocks
are allocated to random access, which correspond to
72 sub-carriers. For an allocation of 72 sub-carriers,
the sequence length ZCN can be set to 71.
A time-continuous short random-access signal can
be formulated as
   
 







1
0
2
1
0
2
short
seq
0
seq
seq
)(
N
k
tfkkj
N
n
N
nk
j
u eenxts


 (2)
where short0 Tt  , 71ZCseq  NN ,  is an
amplitude-scaling factor in order to conform to the
transmit power of preamble, f is the sub-carrier
spacing and 0k is used to control the position of
preamble in the frequency domain, see [1] for a similar
notation. A short sequence of the same length as the
OFDM symbol is achieved by fT  /1short .
The preamble to be transmitted is constructed by
repeating this short sequence as illustrated in Fig. 2
and expressed as
    short0short mod)( TTtsts  (3)
where seq0 Tt  , and 0T is an optional parameter
for designing start of the first short sequence. By this
repetition of the short sequence, each short sequence
will act as a cyclic prefix for the next short sequence.
In Fig. 2 the short sequence is repeated 15 times, and
succeeded by a smaller part of the short sequence. This
last part of the short sequence is inserted in the end
such that the preamble covers the whole length of the
last receiver FFT window.
Fig. 2. User data and proposed random access preamble with time windows (in red) and FFT processing

B. Preamble detector
The radio transmission from mobile to base-
station, for receiver antenna a , is modelled as the
output from an L tap FIR filter




1
0
),()(),(),(
L
m
anwdmnxamhanr (4)
where )(nx is transmitted sequence, ),( anw is
interference and additive white Gaussian noise with
variance )(2
aw , and d corresponds to a round-trip
delay for the current UE. This round-trip delay is
limited by the cell radius, i.e. 10  Dd where
 cFRD scell2 , cellR is the cell radius in meters,
sF is the sampling rate, c is the speed of light, and
 x denotes rounding towards nearest lower integer.
The time-domain signals are inputs to Fast Fourier
Transforms (FFTs) as illustrated in Fig. 3. The FFT
window positions correspond to the distance in time
between the start of the subframe and each DFTS-
OFDM or OFDM symbol in uplink. For example, in
LTE Release 8, the cyclic prefix for the first symbol in
each slot is 160 samples, while the remaining cyclic
prefixes are 144 samples. Each DFTS-OFDM or
OFDM symbol is 2048 samples such that the positions
)(shift pn of each FFT window p equals






13,..,7for)2048144(176
6,..,0for)2048144(160
)(shift
pp
pp
pn . (5)
For each antenna 1,...,0  aNa and FFT
window p , calculate a DFT or FFT over FFTN
samples:





1
0
/2
FFT
FFT
FFT
)),((
1
),,(
N
n
Nknj
s eapnnr
N
apkR 
(6)
for 1,...,0 FFT  Nk and ,...,0pp  10  Pp
where 0p is the index of the first, out of P , FFT
windows included in the preamble detector. See e.g.
Fig. 3 for which 10 p and 12P .
The preamble in the frequency domain is obtained
by extracting sub-carriers corresponding to those sub-
carriers used for the preamble, i.e. seqN samples,
where FFTseq NN  and
  ),,,(,, 0PRACH apkkRapkR  (7)
for 1,...,0 seq  Nk .
The matched filters (of seqN coefficients) in Fig. 3
can be expressed as
seq
PRACH
,MF
),,(),(
),,(
N
apkRpkP
apkC v
v



. (8)
These matched filters are constructed from the
DFT of the known short sequence and the cyclic shifts
of the short sequence. The cyclic shift corresponds to a
frequency-domain rotation with the shift )(shift pn i.e.
 




1
0
/2
seq
)(
2 seq
seqFFT
shift
1
),(
N
n
Nknj
u
N
pn
kj
u enx
N
epkP


.(9)
The matched filter outputs can be coherently added
as




1
,MF
cc0
c0
),,(),,(
NNcp
Ncpp
uu apkCcakC (10)
where cN denotes the number of coherent windows,
 cnc NPN  is the number of non-coherent FFT
windows, and 1,...,0 nc  Nc .
Now, in order to detect preamble and estimate
round-trip time, the output from the IFFT are
transformed to the time domain. Calculate an IDFT, of
size IFFTN , resulting in a correlation vector of length
IFFTN :




1
0
/2
IFFT
seq
IFFT
),,(
1
),,(
N
k
Nkmj
uu ecakC
N
camc 
. (11)
Fig. 3. Preamble detector with long coherent accumulation

Selecting seqIFFT NN  corresponds to an
interpolation, which can be done in order to increase
the resolution of the timing estimation.
A Neyman-Pearson criterion [5] is used as decision
variable, which for deterministic signals in additive
white Gaussian noise equals the normalized absolute
square of the matched filter output, formulated as
 





1
0
1
0
2
2
ncc
nc
)(ˆ
),,(1
)(
N
c
N
a w
u
a
u
a
a
camc
NNN
m

 (12)
where a simple estimator of the noise variance can be
formulated as
 






1 1
0
2
,MF
seq
2
0
0
seq
),,(
1
)(ˆ
Pp
pp
N
k
uw apkC
NP
a . (13)
Preamble number u is detected if the absolute
squared value of this autocorrelation exceeds a
threshold
Th)(  mu (14)
for at least one value of m , within the search window
of size D . A preamble detector and round-trip time
estimator can now be formulated as searching for the
maximum value in this vector of decision variables
and comparing this maximum value with a threshold.
In other words, the preamble with index u is detected
if there is an  1,0  Dm such that Th)(  mu .
This preamble detector threshold Th should be
selected with care such that the false detection rate is
low without causing a too low detection rate, see
section IIC.
A timing estimate follows as the value of m which
corresponds to the maximum value of )(mu , i.e.
 )(maxargˆ mm um  (15)
such that a timing error estimate in seconds equals
)/(ˆˆ
IFFTerr NfmT  . (16)
C. Detection threshold design
Denote the preamble false detection rate, for each
value of m , as fdP such that the false alarm rate FAP
equals
fd
1)1(1 fdFA
DPD
ePP 
 . (17)
Modelling the received signal ),( anr as noise
only, then the decision variable )(mu is a chi-square
distributed random variable with unity mean and
nc2 NNa degrees of freedom. The false detection
probability can thus be formulated as
)2,2(1 ncThncχfd 2 NNNNFP aa  . (18)
where the chi-square Cumulative Density Function
equals
 



12
0
χ
1),(2
r
k
kz
zerzF (19)
if 2r is an integer. The threshold can now be
calculated as the value of Th in (18) resulting in a
specified false alarm rate FAP according to (17) which
can be formulated as
)2),1ln(
1
1(
2
1
ncFA
1
χ
nc
Th 2 NNP
D
F
NN
a
a
 
 . (20)
III. BEAMFORMING
A. Frequency domain beamforming
A frequency domain beamforming before
preamble detection is illustrated in Fig 4a. Here, the
antenna signals are first received in a Radio Unit (RU),
and then sampled and quantized in an Analog-to-
Digital Converter (ADC). A transformation from time
to frequency domain is done using an FFT, after which
the preamble detector is applied. Here, an FFT is
typically calculated for each antenna or subset of
antennas, such that different users and channels in
different sub-bands of the received signal can be
extracted before further signal processing.
The beamforming (BF) is used in order to increase
received signal strength before the preamble detector.
Beamforming aims at combining received signals
from several antennas such that more signal energy is
received in specific spatial directions.
When beamforming is done in the frequency
domain one or several specific beamformers can be
applied to those sub-carriers which are used for
random access. By these beamformers, the preamble
detector is sensitive in several spatial directions.
Fig. 4. Frequency domain and time domain beamforming before preamble detectors.

B. Time-domain beam-forming
For time-domain beamforming, the beamforming
scaling and phase shifts are applied before the FFT,
see illustration in Fig. 4b. Time-domain signals from
many antennas are thus combined in the beamforming.
In this way, the beamforming is the same for all sub-
carriers. This time domain beamforming might be
done on an analog signal (i.e., before ADC) as in Fig.
4b, or on a digital signal (i.e., after the analog-to-
digital converter, ADC) but before the FFT (no
illustration included for brevity).
At initial access, the base station has limited
knowledge of the position of the UE. The preamble
receiver must therefore evaluate several beamformers
in order to be able to detect the preamble. With time-
domain beamforming, this requires one sequence of
processing from FFT to preamble detector per
beamformer. An illustration is given in Fig. 5 of an
approach in which the beamforming is changed
between each FFT window. Here the outputs from
each beamforming, followed by an FFT, are
individually processed in a matched filter, an IFFT, an
absolute square calculation, and finally a preamble
detector. If the hardware supports several
simultaneous beamformers in the same time-window,
then several spatial directions can be processed in
parallel. Each such beamformer is referred to as one
receiver baseband (BB) port.
See an example in Fig. 5 where one BB port is
used with a fixed time-domain beamforming for all
time windows of a sub-frame. A second BB port is
used to switch beamformers between each window.
Thus, one “beam 0” is applied to several FFT
windows, such that more energy can be accumulated
for preamble detection as compared to “beam 1” to
“beam 12”, in Fig. 5 for which only one FFT window
is used in each preamble detector.
For UEs with high SNR (typically close to the base
station), a reliable detection can thus be done with a
small number of FFT windows, while UEs with a low
SNR (typically further away from the base station),
can in most cases only be done if many, or all, FFT
windows are included. By combining detectors with
few FFT windows included (with many different
beamforming directions) and detectors with many FFT
windows (but few beamforming directions), a balance
can be achieved between fast preamble detectors with
high SNR and slow detection for UEs with low SNR
and the access latency for these UEs will be longer.
That is, many random-access occasions might be
needed for UEs with low SNR. This since the
baseband ports with many FFT windows included do
not search all directions during each preamble
occasion.
IV. SIMULATIONS
For the simulations of this section, the preamble
thresholds are designed in order to get a false alarm
rate of 0.1% using the approach given in section IIC.
Simulation results of false detection rate (excluded due
to brevity) with these thresholds indicate good
agreement between simulations and theory. A similar
receiver structure is used for preambles both with short
sequences and one long sequence, where the timing
estimates are limited to a value between 0 and shortT
seconds.
Performance simulations of the proposed preamble
format are given in terms of missed detection rate in
Fig. 6. Here, the performance results are similar for the
preamble formats with long and short sequences with
coherent accumulation of all FFT windows ( 12c N ,
1nc N ). This is an expected result since the
accumulated energy is the same for both approaches.
An increased missed detection rate is also observed
when the number of coherent accumulations is
decreased.
Simulation results are given in Fig. 7 with a
frequency error of 1 kHz. Here, the proposed preamble
format with a shorter coherence time (e.g. 3c N ) has
better performance than the preamble format with one
long sequence.
V. DISCUSSION
In a 5G system with a large number of receiver
antennas, the computational complexity of FFT
processing in the receiver is also large. With dedicated
antenna-signal processing used for random-access
preamble detection, a lot of special hardware must be
included. This negatively impacts complexity, design
effort, and energy consumption. With the proposed
preamble format as described in section IIA, the
special large FFT is avoided. Instead, the same FFT as
used for data reception can advantageously be applied
to the random-access receiver, as illustrated in Fig. 2.
Fig. 5. Switching between analog beamforming candidates
Fig. 6. Miss detection rate without frequency errors

Some ICI (Inter-Carrier Interference) may result
from the proposed preamble format in the first and last
OFDM symbols. This is due to that the preamble will
not cover the beginning of the first FFT window, when
the preamble is received slightly delayed compared to
Fig. 2, as in Fig. 3, causing ICI to data transmissions.
In the same manner, the preamble will only cover
some fraction of the last FFT window. However, the
other OFDM symbols are not subject to ICI from the
proposed preamble. This is in contrast to the preamble
format with one long sequence where the preamble
generates ICI for all OFDM symbols, see Fig. 1. Also,
the preamble detector will suffer from ICI due to
concurrent data transmissions, which is not the case
for the proposed 5G format with a preamble based on
short sequences and a detector as described in section
IIB.
The proposed receiver as in section IIB can only
estimate a delay of the preamble up to the length of the
short sequence shortT . If the preamble is delayed more
than this length shortT , then the delay will be wrapped
to a biased estimate into the interval ),0 shortT . With
numerology according to LTE release 8, the maximum
delay corresponds to the length of the uplink DFTS-
OFDM symbol, i.e. sT 6,66short  . This corresponds
to a maximum distance from UE to the base station of
km102shortmax  cTR . Compare with the
preamble with one long sequence, where a length of
s800 T corresponds to a maximum distance of
km120 . The proposed preamble is thus designed for
significantly smaller cells, as envisioned for future 5G
systems, as compared to the preamble format with one
long sequence. For 5G systems with higher carrier
frequencies, larger subcarrier spacing is likely
attractive for OFDM based systems to reduce the
impact from phase noise. For example, with a
subcarrier spacing of kHz75f the length of the
short symbol equals μs3,13short T and a maximum
distance from UE to the base station equals km2 .
The coherent addition of signals should be reduced
when the coherence time is low. This coherence time
is depending on the rate of time variation of all
distortions between baseband transmitter and receiver.
For example, high Doppler spread, large Doppler shift,
frequency errors or large phase noise leads to a
decreased coherence time such that the time should be
reduced for which the coherent additions are done.
Large phase noise is believed to be an issue for 5G
systems with many antennas in which the cost for each
receiver must be low.
The number of available preamble sequences is
reduced when reducing the length of the sequence. On
the other hand, the use of narrow beamforming in a 5G
system reduces the impact of interference from other
UEs. This is an area for further research.
VI. CONLUSIONS
A new preamble format is proposed suitable for
5G systems with many antennas. Here, a preamble is
constructed based on repeating a short sequence of the
same length as the length of the OFDM symbol. In this
way, a 5G preamble detector can be designed which
reuse the same FFT, for both user data and random-
access preamble detection. By removing FFTs
dedicated to random-access only a lot of special
processing and hardware can be avoided which comes
at a reduced cost in material and design effort, as well
as reduced energy consumption which are considered
as important features of a 5G system. With time
domain beamforming, the proposed format provides a
possibility to evaluate several beamforming candidates
for one individual preamble transmission occasion.
This beamforming is improving the link budget for 5G
systems which typically operates on millimeter wave
frequencies. Simulation results indicate a performance
of the proposed preamble format which is robust
against transmitter and receiver imperfections such as
frequency errors. The proposed preamble format is
thus suitable for low cost hardware as envisioned for
future 5G systems.
ACKNOWLEDGMENT
The authors would like to thank Zhao Zhenshan
for support with the simulation results.
REFERENCES
[1] 3GPP TS 36.211, V12.1.0 (2014-03), 3rd Generation
Partnership Project; Technical Specification Group Radio
Access Network; Evolved Universal Terrestrial Radio Access
(E-UTRA); Physical Channels and Modulation, Release 12
[2] 3GPP TS 36.321, V12.1.0 (2014-03), 3rd Generation
Partnership Project; Technical Specification Group Radio
Access Network; Evolved Universal Terrestrial Radio Access
(E-UTRA); Medium Access Control (MAC) protocol
specification, Release 12
[3] E. Dahlman, S. Parkvall, J. Sköld “4G LTE/LTE-Advanced
for Mobile Broadband,” Academic Press, Second Edition
2014
[4] S. Sesia. I. Toufik. M Baker “LTE, The UMTS Long Term
Evolution, From Theory to Practice”, Second Edition, John
Wiley & Sons Ltd., 2011
[5] S. Kay “Fundamentals of Statistical Signal Processing 2:
Detection theory,” Prentice Hall, New Jersey, 1998
Fig. 7. Miss detection rate with frequency errors

Conference Paper: Random Access Preamble Format for Systems with Many Antennas

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Conference Paper: Random Access Preamble Format for Systems with Many Antennas (20)

More from Ericsson (20)

Recently uploaded (20)

Conference Paper: Random Access Preamble Format for Systems with Many Antennas