In the last years, Orthogonal Frequency Division Multiplexing (OFDM) has become the dominant transmission
technique for both wired and wireless communication systems. Besides its advantages to combat multipath propagation
with simple receiver structures, it furthermore enables a flexible adaption of modulation schemes on a subcarrier
basis or group of subcarriers.
Fig. 1: Frequency-selective radio channel with adaptive modulation pattern
Fig. 1 shows the magnitude of the transfer function of a frequency-selective radio channel and the corresponding
optimal bandwidth efficiencies as a result of an adaptive modulation algorithm.
However, transmitter and receiver knowledge regarding the chosen bit allocation table (BAT) must be instantaneously
synchronized. In a static environment like the wired communication standard digital subscriber line (DSL), the information
about this bit loading pattern is explicitly signaled. Contrarily, accounting for time-varying propagation channels,
the BAT must be continuously updated in a mobile wireless setup. The required signaling overhead would significantly
reduce the effective data rate (typically by at least 4% in wireless local area networks, WLAN) even if sophisticated
source encoding techniques are applied.
In order to avoid such overhead losses, our approach is to automatically classify the modulation schemes on all
subcarriers at the receiver by utilizing side information available in time-division duplex (TDD) systems.
Furthermore, we introduce the new approach of a signaling-assisted modulation classification in which only a fraction
of the complete signaling information, called auxiliary information here, is transmitted. The main idea is to incorporate
this additional knowledge into the modulation classifier such that the reliability can be significantly increased on the
one hand. On the other hand, the amount of signaling overhead is greatly lowered compared to a pure signaling scheme.
The application of this concept to wireless OFDM systems with adaptive modulation puts a lot of new constraints on
optimal modulation classification algorithms. Especially exploiting the strong correlation of the uplink and downlink
channel in TDD systems helps to greatly enhance the classification reliability.
Fig. 2: Time-division duplex system model (TS: training sequence, CE: channel estimation, AM(C): adaptive modulation (and coding))
In order to illustrate the adaptive transmission concept, Fig. 2 shows the signal flow between the mobile station (MS)
and base station (BS):
- In receive mode of the i-th frame, a training-based channel estimation is carried out.
Based upon the estimated channel transfer functions Ĥ_{BS,n,ki} and Ĥ_{MS,n,ki} respectively,
the optimal coding scheme and BAT b_{BS,i} and b_{MS,i} are calculated.
- In transmit mode, a frame composed of a preamble, a signaling field
and information data modulated according to the BAT – calculated in 1) – is sent.
It is assumed that the coding scheme is kept constant during the depicted observation interval.
In order to be able to correctly decode the i-th received downlink frame (see current downlink frame in Fig. 2),
the BAT b_{BS,i} must be known at the receiver side and is automatically classified.
Fig. 3: Example constellation diagram of received 16QAM symbols at average SNR=20dB
As an example, the constellation diagram of received 16-QAM symbols at the average SNR of 20dB is shown in Fig. 3 for a
different number of symbols. Whereas the 16-QAM structure can be easily recognized based on 640 symbols in the left figure,
modulation classification from only 20 symbols seems rather challenging, but is still possible with high reliability
if side conditions of the transmission systems are utilized. Optimal algorithms based on a maximum-a-posteriori (MAP)
approach have been developed and their performance has been analyzed in terms of misclassification probability and
the influence on the effective bandwidth efficiency which takes the data rate reduction caused by signaling overhead
into account.
Fig. 4: Probability of BAT classification errors versus transmit SNR for 2×2- transmission with minimum
mean-squared error (MMSE), zero-forcing (ZF) and singular value decomposition (SVD) receiver,
b_{average} = 4 bit/s/Hz (left); effective bandwidth efficiency at target frame error ratio FER_{target} = 10^{-1} vs.
transmit SNR for 2×2 MMSE receiver (right)
The representative example in Fig. 4 verifies that the adaptive signaling-free transmission scheme
(green curve: adaptive, proposed) outperforms adaptive transmission with conventional signaling as well as
non-adaptive transmission in terms of the effective bandwidth efficiency in a typical indoor 2x2 MIMO scenario.