DSP Application On One Dimention Signal

IEEE Paper Review: Implementation of a modem based on Quadrature Amplitude Modulation using TMS320C6201 DSP
Authors: N Dahnoun and D R Clewer
Publisher: IEEE
Published in: IEE Colloquium on DSP Chips in Real-Time Instrumentation and Display Systems 
Location: Leicester, UK, 
Date of Conference: 24 September 1997

Summary: 
Modems have traditionally been used to enable data communications via PSTN. This is severely band limited channel, requiring use of highly bandwidth efficient modulation schemes such as Quadrature Amplitude Modulation (QAM) combined with forward error correction and data compression in order to achieve bit rates up to 56 Kbps. The advances in DSP technology have recently enabled these bits rate to be achieved in commercial equipment at a reasonable cost.TMS320C6201 device boasts a maximum of 1600 MIPS at a clock speed of 200 MHz which is said to be more than a magnitude faster than any other fixed-point DSP. Cable modems allowing 36 Mbps transmission rates could be implemented using the same. Each module in transmitter as well as receiver was tested individually by writing a test program to read data in from a file, run the module and write the results on an output file. The data output were compared with expected output to judge failure and success of the DSP processor. To conclude, it was found that programming large functions in C on the DSP processor is both quick and easier than programming in assembler 

Link: http://ieeexplore.ieee.org/document/659698/ 


Patent Review: Method for generating modern transmission signals with quadrature amplitude modulation.
Application Number: 06/136733 
Application Date: 04/27/1982 
Filling Date: 04/02/1980 
Inventors: Gockler, Heinz and Hofmeister, Hagen

Summary: 

The present invention relates to a method for generating modem transmission signals with quadrature amplitude modulation (QAM) in which scanned values of elementary bandpass signals are formed in digitally coded form, the elementary bandpass signals are weighted with coefficients depending on the information to be transmitted and the modem transmission signals are formed by superposition of the thus weighted elementary bandpass signals. It is therefore the object of the present invention to provide a method for generating modem transmission signals which operates with less calculating operations and thus enables the circuit to operate at a greater speed and with greater accuracy. The above object is accomplished according to one embodiment of the invention in that in a method for generating modem transmission signals with quadrature amplitude modulation (QAM). Another advantage of the invention is that in the case of m-ASK where m=2, 3, 4, . . .  the elementary bandpass signals are additionally weighted with factors corresponding to the amplitude stages characteristic for the information to be transmitted.

 Link: http://www.freepatentsonline.com/4327439.html

Basic operations using DSP processor

In this experiment,a new project was created and a simple code was written to demonstrate various arithmetic and logical instructions in DSP processor.The DSP kit used was TMS320F28335.The output of the instructions was stored in the registers and these values were obtained in real-time using the debugging functionality of the software.

FIR Filter Using Frequency Sampling Method

This was the last experiment which was to be performed in the lab. 

hence everyone was happy that we successfully completed rest of the experiments.

We took two case: Low Pass FIR Filter and High Pass FIR Filter using Scilab. We specified the stop band and pass band attenuation, stop band and pass band frequencies and sampling frequency in each case.

 we observed Phase response for both LPF and HPF and concluded that they  remains same as long as the order is kept same in both cases.

calulated values of passband and stopband gain

we got the output.

FIR Filter Design using window Method

In FIR filter, poles are always located at origin. For causal & stable filter all the poles must lie inside a unit circle therefore FIR filter is always stable. FIR filter can have a linear phase response only when h(n) is symmetric & anti-symmetric.

For this particular experiment we used window method for designing purpose. In this method, the output h(n) is calculated by multiplying hd(n) with particular window function which can be chosen depending upon As value.

We designed LPF & BPF for specified As,Ap,Fp,Fs & Fsamp values. After doing calculations on values derived from magnitude response we obtained As & Ap values. (calculated).

The phase response linearly varies with frequency.

DESIGN OF CHEBYSHEV IIR FILTER

The analog chebyshev filter was converted into digital chebyshev filter using BLT method. The magnitude response for both LPF and HPF were plotted. Ripples in the pass band were observed and monotonic in stop band. The number of the peaks in the ripple defined the order of the filter.

For same input specifications for butterworth and chebyshev filter, order of chebyshev filter was smaller. Ultimately, this leads to lesser cost for the hardware required. It is better for the real time application.

DESIGN OF BUTTERWORTH FILTER

In this experiment we were supposed to find the output of digital Butterworth Filter. First the order of the filter and cut off frequency is calculated. After this normalized transfer function is calculated from which denormalised transfer function is calculated by replacing 's'. Then z transform is obtained by BLT and the output graph of magnitude v/s frequency is observed.

(Filtering of long data sequence) OVERLAP ADD METHOD

In this experiment of the course was based on Linear FIR Filtering Methods - Overlap Add Method and Overlap Save Method. We performed filtering of long data signal using Overlap Add Method using Scilab. A long data signal was taken as the input with length equal to 24. Impulse response was taken as {1,2,3}. The input signal was decomposed into smaller subsequences with length equal to 6. In order to use the radix 2 FFT algorithm we had to use zero padding to make the length of each subsquence equal to 8. Linear Convolution by Circular Convolution, was performed using zero padded h(n) and the individual subsequences. Output y(n) was computed and verified manually. Thus, we can say that OAM is better for real time filtering of long signals.

FAST FOURIER TRANSFORM

After learning DFT,our next experiment was Fast Fourier Transform(FFT). We were asked to take a N input sequence. I took 8 input sequence on which FFT was performed. The code was written in c++ programming language.FFT performs 12 complex multiplications, 24 complex additions, 48 real multiplications and 72 real additions. After calculating FFT we were asked to calculate inverse FFT(IFFT) to check whether we get the same input sequence again. The speed of operation in FFT increases as compared to that of DFT. The Operations are carried out parallely in FFT.

DISCRETE FOURIER TRANSFORM

In this second experiment of DSPP we were suposed to take a N point signal x(n) as the input and find Discrete Fourier Tramsform of the signal. Also we had to observe the effect of zero padding on the magnitude spectrum of the signal. We used C to write the code for the experiment. After obtaing the output with and without zero padding we learnt that with zero padding, the frequemcy spacing decreases, error in representation decreases and also there is an increase in the resolution and quality of the spectrum. 

DISCRETE CONVOLUTION AND CORRELATION

With this experiment we came back to our basics, the C language. In this experiment we performed Linear convolution, Auto-correlation and Cross-correlation of two signals. while performing correlation we observed the following statuses of the signal.

              As we performed auto-correlation, we found that the signal is symmetric with respect to its center point and the value of the center point is the greatest.The cross-correlation of the time delayed  signal causes the output signal to be delayed by the same amount as that of the time shift applied to the original input signal.               The cross-correlation of a time delayed and a zero-padded signal also gives a similar result as that in the case above, but the output signal will be delayed by the total amount as the time shift and the zero padding given.