Transmission In Mimo With Minimum Power Consumption Computer Science Essay

Published Date: 02 Nov 2017

Abstrac- In this paper, we proposed a novel scheme for transmission of image over MIMO channel using cognitive radio. Compressive sampling is used for spectrum sensing with minimum system power consumption. The dynamic spectrum is sensed using the Modulated WideBand Converter (MWC). The OFDM is used to modulate the compressed image. The optimal and sub optimal algorithms are developed under the rate constraint to minimize the power consumption of the system. According to the target rate of transmission, the number of antenna is activated for data transmission. The simulation results demonstrate that sub-nyquist sampling system used to sense the spectrum ensures with faster and low rate processing, significant power savings up to 80% are achieved when compared with the conventional methods.

Keywords- Cognitive radio, Compressive sampling, MIMO, OFDM modulation, Power Minimization

Introduction

Multimedia data transmission over wireless channels has become more popular today due to the tremendous development in wireless cellular systems. The spatial dimension in MIMO channels provides an extra dimension to separate users, allowing more aggressive reuse of time and frequency resources, thereby increasing the network capacity1. The increasing demand for wireless services and the advent of new wireless technologies imposes an impending limit on the amount of spectrum that will be available for these technologies to be deployed. According to studies mentioned in A, allocated and licensed spectrum is sparsely used. This information has yielded a solution to the limited available spectrum, which involves sharing of the allocated spectrum between primary users who own the licensing for a particular band, and secondary users. The latter would own licensing to also transmit on those bands, but on a series of conditions imposed by regulation. Cognitive radios (CR) implement this solution on the side of secondary users, by requiring the radio to collect information on the state of the spectrum before making a transmission decision B. In this manner, CRs can adaptively maximize the utilization of the entire spectrum. Conventional signal processing approaches for spectrum sensing to assume detection of a narrowband signal in noise. Most commonly used techniques include matched filtering, energy detection, and cyclostationary detection. The energy detection method is not robust against noise uncertainty and fails to differentiate between signals. Other methods, such as matched filtering and preamble detection, require strong prior knowledge about the primary user. A new sampling paradigm based upon sparse sampling named Compressive Sensing (CS) provides a sampling mechanism at rates lower than the Nyquist rates3. The theory of Compressed Sensing (CS) can lead to recover certain signals from far fewer samples than traditional methods 6. In this system, various power consumption models have been proposed and used for the power optimization by adapting modulation, coding, antenna configuration and radiated power to channel conditions and QoS requirements10.

System Model

The system in fig. 1 includes cognitive radio environment, compressive spectrum sensing, and channel estimation, adjusting parameters like modulation, image compression and antenna selection to reduce the power consumption.

Fig. 1 System Architecture

The process starts with cognitive radio environment simulation which performs the spectrum sensing and channel capacity estimation using compressive sampling. According to different channel capacity estimation and modulation scheme parameters are adjusted and transmitted when the target rate is achieved with minimum number of antenna. In the receiver, the signal is demodulated, and then image is reconstructed.

Transmission and Reception

A MIMO system with 4X4 antenna is used for transmission reception. The random process x (n) is wide-sense stationary is taken for processing. The sequence observed is

(1)

The L X L auto covariance ( ) matrix for y (n) in (2) is represented with the addition of auto covariance matrix for the signal x (n) and the auto covariance matrix for the noise multiplied with the identity matrix (I)

(2)

The Singular Value Decomposition (SVD) is used for finding the eigenvalues. It is a reliable and accurate method for measuring eigen values. Let the covariance matrix be

(3)

Using the SVD, Q in (3) can be factorized as, where U, V are the left and right singular vector and ∑ = diag (λ1, λ2, λ3….. λm) is a diagonal matrix whose non negative entries are the square roots of the positive Eigen values of QQ*. * denotes the conjugate transpose. These non negative values are called singular values of Q and they are arranged in a decreasing order with the largest on in the upper left – hand corner. It is noted the QQ* = R so this method allows to calculate the Eigen values without to build the covariance matrix.

Recently, there is a need for faster sampling and less power consumption of the system. In the proposed system, this can be achieved by the compressed sampling and active branch selection of the system.

Compressed sampling specifically yields a sub-Nyquist sampling criterion. Sub-Nyquist sampling, also known as compressive sampling or compressed sensing, refers to the problem of recovering signals by samples much fewer than suggested by Nyquist rate. MWC is used to convert the analog to digital signal. The system does not require knowledge of the frequency support in either the sampling or the recovery stage. The goal of the modulator is to alias the spectrum into baseband. The modulated output is then low pass filtered and sampled at a low rate. The rate can be as low as the expected width of an individual transmission. The power spectral density (PSD) estimation of x(n) is invoked in order to locate the energy concentration within each spectrum slice. The hamming window size is W ≥ 2B/ fres. The frequency resolution is given by

(4)

The threshold (µ) is determined by

(5)

The signal detector does this giving a binary output, it is 0 when an unused frequency band is found, and it is 1 when a signal is detected. So, when an unused frequency band is detected it is allocated to the secondary user. The threshold is calculated using (5), if the power is larger than the threshold, the channel is busy and represented as 1 otherwise represented as 0. Then the image is JPEG compressed and converted to bit streams. As shown in the fig.2 the bit streams are then modulated with the OFDM modulation technique. JPEG compression process starts with 8x8 block separation then DCT is coefficients are calculated for each block. The coefficients are then quantized and run length encoded.

The time discrete equivalent is the inverse discrete Fourier transform (IDFT). So, the IDFT and the discrete Fourier transform (DFT) are used for modulating and demodulating the data constellation on the orthogonal subcarriers. The discrete samples of the OFDM symbol is given by

Where (6)

Following the parallel-to-serial (P/S) conversion, the baseband OFDM signal s(n) is upsampled and passed through the digital-to-analog (D/A) converter to convert the digital signal into an analog signal. The baseband OFDM signal is then low pass filtered, upconverted to the desired centering frequency using a mixer and a local oscillator (LO), and amplified for transmission by the power amplifier (PA). The receiver performs the reverse operation of the transmitter, mixing the RF signal to baseband for processing. Then, the signal is low pass filtered, converted to digital signal using an analog-to-digital (A/D) converter, and downsampled. The serial stream of sampled time signal is converted into parallel streams with a S/P converter and the cyclic prefix is discarded from the received composite signal, rm,n. Then, the DFT in (7) is used to transform the time domain data into frequency domain

Where (7)

These parallel streams are then demodulated to yield digital data and are multiplexed together using the P/S converter to yield the serial bit stream, and delivered to the data sink. At the receiver, frame detection is an important task. Moreover, frequency and timing synchronization is required before the OFDM symbol can be correctly demodulated. A known preamble is sent before each OFDM frame to allow receiver synchronization and channel estimation, as well as an initial acquisition of the frequency offset.

The system is provided with four antennas, these antennas are connected with power amplifiers. This PA may be class A amplifier or class B amplifier is taken into account. When the target rate is achieved then only the transmission is allowed otherwise the transmission is not allowed. The more transmit branches the MIMO system uses, the more power is consumed by the system. Therefore, the minimum number of branches (PA) as the system power consumption is linearly proportional to the number of active branches as long as the target rate can be achieved.

JPEG ENCODER

FEC ENCODER

SERIAL

TO

PARALLEL

INTERLEAVER

PILOT

INSERTION

QAM MAPPING

ANTENNA

SELECTION

IQ

MODULATOR

MODULATED

WIDE BAND

CONVERTER

(MWC)

PARALLEL

TO

SERIAL

JPEG

DECODER

FEC DECODER

PARALLEL TO

SERIAL

DEINTERLEA-

VING

CHANNEL

CORRELATION

QAM

DEMAPPER

IQ

DEMODULATOR

MODULATED

WIDE BAND

CONVERTER

(MWC)

SERIAL

TO

PARALLEL

IFFT (TX)

FFT(RX)

ANTENNA

SELECTION

LNA

HPA

AFC CLOCK

RECOVERY

AGC AMP

INPUT IMAGE

RECONSTRUCTED

IMAGE

TIME AND

FREQUENCY

SYNCRONIZA-

TION

Fig. 2. The System Architecture

The power of the system is calculated by considering radiated power as in (8) of PA. The number of branches is increased only when the current number of active branches cannot satisfy the rate requirement. The total radiated power (9) is calculated for the system. The power amplifiers considered are class A and class B and the power radiation for these amplifiers are calculated by (10). The radiated power from the transmit branches is

(8)

The total radiated power becomes

(9)

where is the nth element of and the radiated power from transmit branch n. The system power consumption from branch can be expressed as

(10)

The system power consumption reduction is

(11)

where Pcon is the system power consumption of MIMO system with a conventional power allocation scheme (i.e., the water-filling algorithm and the equal power allocation), and Pcog the system power consumption of a MIMO system with the proposed CR framework. The MIMO systems employing Class A PAs and Class B PAs are simulated for each algorithm using (12).

Transmission Algorithm

Step 1: Signal x(t) is sensed and sampled at sub- Nyquist rate x(n).

Step 2: Channel is estimated using SVD.

Step 3: Power Spectral Density is estimated for the sampled signal x(n).

Step 4: Frequency resolution (fres), window size (W) and threshold (µ) is calculated.

Step 5: The regions which are smaller than µ are united and those regions which are smaller width than bmin are isolated.

Step 6: If there is any isolated regions

Allocate the isolated region to the user.

Source Coding is performed using JPEG Compression technique.

Channel Coding is performed using OFDM with the channel parameter calculated by Step 2.

Else go to step 1.

Step 7: Initialize the number of active antenna to one.

Step 8: Calculate and store the power allocation, target rate and achieved rate.

Step 9: Calculate the radiated power, consumed power and total consumed power.

Step 10: If power allocation achieves the target rate of the user

The signal is transmitted with that antenna configuration and achieved rate with minimum power consumption.

Else increase the number of active antenna and goto step 8.

Reception Algorithm

Step 1: Signal x(t) is sensed and sampled at sub- Nyquist rate x(n).

Step 2: Channel is estimated using SVD.

Step 3: Power Spectral Density estimated for the signal x(n).

Step 4: Frequency resolution (fres), window size (W) and threshold (µ) is calculated.

Step 5: The region of interest is isolated.

Step 6: If there is any isolated region, the channel coding is

performed using OFDM with the channel parameter

calculated by Step 2. Else go to step 1.

Step 7: JPEG decompression is done and image is reconstructed.

RESULTS AND DISCUSSION

The proposed framework is evaluated under the correlated Rayleigh fading channel and the comparison of the quasi optimal algorithm, sub optimal 1 algorithm and sub optimal 2 algorithm for power allocation scheme in terms of radiated power minimization is simulated.

For sub nyquist frequency of 10 GHz and window size is [0, 0.5] MHz, the spectrum can be divided into 195 subchannels. The threshold is calculated, if the power is larger than the threshold, the channel is busy and represented as 1 otherwise represented as 0. As shown in fig 3, fig 4, fig 5 the signal is received, sampled at the sub nyquist rate and reconstructed. The sampling duration is 0.02µs.

Fig. 3 Original signal transmitted by the receiver to estimate the MIMO channel.

Fig. 4 Original signal added with noise in the MIMO channel and sensed by the transmitter.

Fig. 5. Sub Nyquist sampled signal and the spectrum holes are detected.

The reconstructed image after transmission is shown in fig.6. The presence of blocking artifact in JPEG is unavoidable due to 8 x 8 block selector in JPEG encoder. The PSNR value obtained after reconstruction of the image is 33.96. The bit stream of the image is modulated with the OFDM modulation and the bandpass signal is shown in fig. 7. The orthogonality helps to transmits more data simultaneously. Comparing the power savings between the quasi optimal, sub optimal 1 and sub optimal 2 in Figs. 8 and 9 it is clear that at some target rates, using more antennas can achieve further power saving.

Fig.6. Input image and reconstructed image at the receiver.

Fig.7. Bandpass signal of the Orthogonal Frequency Division Multiplexing (OFDM) in cognitive radio systems.

Fig. 8. Power saving of a MIMO systems for class A PA.

Fig. 9. Power saving of a MIMO systems for class B PA.

V. CONCLUSION

Thus the image is transmitted over a MIMO using cognitive radio approach with minimal power consumption of the system. The sub- Nyquist sampling system is used for the ADC & DAC of the OFDM which ensures the faster with low rate processing and this gives the ability to process the transmitted bands without requiring the prior knowledge of it. This helps the CR to understand the channel, transmitter, receiver, bandwidth response etc. The systems power consumption is reduced up to 80% than the conventional power allocation method. This framework and algorithms achieve significant power reduction for power Amplifier.