A complete list of the VHOL, C, and C++ source code used in this thesis is available upon request. This dissertation examines two rlsmg technologies in the world of wireless cellular communications - COMA and Program Radio.
Progression
This thesis is a presentation of two key concepts in the current trends in wireless communication. Many of the advantages of COMA, which explain its emergence at the height of GSM, are described in this thesis.
Chapter Outline
In the case of the project, this external data source is a PC with a DSP-based PCI communication card. This usually OSP-based operation has been studied in detail in the context of program radio.
Introduction to Spread Spectrum
The Nature of Spread Spectrum
- Direct Sequence Spread Spectrum System (OSSSS)
- Freque ncy Hopping Spread Spectrum System (FHSSS)
- Hybrid Direct Seq uencelFrequency Hopping Systems
- The Orthogonal Condition
- Maximal Length Codes
- Walsh-Hadamard C odes
The foundation of successful signal transmission in a spread spectrum system is the selection of pseudo-random sequences. However, the main advantage of Gold codes over maximum length sequences can be seen when examining the cross-correlation.
Asynch ronous CDMA
- The M USIC Algorithm
If the source of the previously described interference is due to multipath propagation, it is preferable that the received amplitudes are uncorrelated. They conclude that "the behavior of the MUSIC algorithm is very similar when rectangular pulses are used and when square root-increased cosine pulses are used".
Conclusioll
Ottersten, in [30] discusses the application of the MUSIC algorithm to systems with band-limited pulse shapes. However, due to the computational complexity of the MUSIC algorithm and other delay estimation algorithms for OS-COMA systems.
The System Model
Next, the importance of Walsh codes is examined in the context of the multiuser detector, since the use of these codes produces a particularly elegant solution.
System Mathematics
As already planned, this chapter will examine the MMSE variation of the multiuser detector. In the case of the MMSE multi-user detector, a filter matrix H is derived which, when multiplied by the received data vector, will produce a stream of all the current user's data.
Iterative Solutions
Since both Gold codes and Walsh codes converge to the correct solution within one iteration, it is of value to observe the performance of the iterative solution in the presence of AWGN. This was performed for 100,000 bits for each signal-to-noise ratio, both with Gold codes and Walsh codes.
Fixed Coefficiellt Solul;olls - The Walsh Code Case
Once again, these desirable properties of the Walsh codes derive from their very low in-phase cross-correlation values. In the earlier case of the MMSE detector, several parameters of the system were required to be known at the receiver, including signatures and timing of all relevant users.
Conclusion
The disadvantage of the tile adaptive filter is that it requires an initial training set. This is a symbol string known to the recipient and sent by a specific user with a specific code.
The Software Radio Platform
- Establishing the Need for a Flexible Platform
- Meeting tire Need - Defining tire Software Radio E lements
- Resource Partitioning
- The Alcafel Altech Telecol1ls Flexible Radio Platform
- Conclusion
The flexibility of the software radio extends to the area of field troubleshooting and improvement. The re-configurable nature of software radio enables the concept to find a natural home in international mobile radio. After reflecting on the issues already discussed, this reconfigurable element should form as large a part of the software radio as possible.
This simulation process adds another dynamic to the usability of the software radio - that of high-speed channel simulation. The Alcatel Altech Telecoms Oexible radio platform offers many of the features and flexibility inherent in the software radio concept.
Design and Synthesis
- Design Tools
- VHDL
- Vendor-Specific Design Tools
- Synthesis Tools
- Designing for Synthesis
- Data Manipulation Techniques
- Bit-Level Manipulation Approach I
- Bit-Level Manipulation Approach 2
- Comparison of Approach I and Approach 2
- Data Storage and Data Types
- Conclusioll
This language has been the focal point in relation to the development of techniques for implementing the COMA system. In fact, 90% of deployment time is spent compiling, optimizing, synthesizing, directing, and mapping the design. For this purpose, data storage and data types must be examined, and this can only be achieved in light of the data manipulation techniques that have been used in the project.
One of the fundamental techniques in this project is mastering the manipulation of high-speed bit streams. However, the designer must design within the context of currently available synthesis tools.
System Overview
Th e System Block Diagram
The bidirectional interface connected to the RS-485 conversion module is responsible for generating a series of clock signals for the communication bus that exists between the software radio and the DSP subsystem. This interface also serves to properly buffer and package the data from the MMSE bit estimator and pass it on to the DSP. The bidirectional interface is also responsible for unpacking the information received from the DSP subsystem and presenting it to the Walsh encoder in an appropriate font.
In Chapter 1I, this transmitted signal will be distorted with internally generated AWGN and Rayleigh fading before joining the receiver filter. The pulse detector is capable of being shifted in timc to present a different set of sampled chips to the MMSE bit evaluator.
Interfacing
- The Se-BUS Standard
FSYNCN SCLK
DATA
Processing the DSP to Software Radio Stream
- The Reorderin g Process
- Serial to Parallel Co nversion
Conversely, the process must be reversed when data is sent back to the DSP system. Also note that from the timing analysis, each odd falling edge of the double speed serial clock corresponds to the beginning of a bit period. Once all 32 time slots have been processed in this way, the send buffer is shifted to the right (line 13) and the byte counter is reset.
The DSP can conveniently encode these parameters as bytes and words and then serialize them to the program radio. When the bits are shifted, the rising edge of the clock is used to map the accumulated byte to the output port of the entity.
Conclusion
Note how line 5 once again achieves sync by resetting the master counter while frame sync is low. Note how, since the single-rate serial clock is used, the frame sync is low only once during a clock cycle, removing the need to avoid a second frame sync reset. Fundamental to any COMA system is an effective means of sharding baseband data, thereby spreading information over a higher bandwidth.
This requires careful attention to timing issues, especially when considering the coupling between the spread data source and the baseband filter. This chapter will then look at the particular approach used in implementing a Walsh encoder for 16 simultaneous users, each with a baseband data rate of 64 kBps.
Th e Entity Structure
Additionally, this particular implementation is intended to emulate n downlink and will therefore need to chip and overlay information from many users. To this end, a review of some VHDL code fragments will also introduce the reader to some of the more difficult implementation and timing issues. The nature of the pilot signal and code will be discussed in 10.4, Choice of Pilot Code and Bit Sequence.
Hence, each chip must be transferred with at most 15 others to generate the entity output. This relates to 33 possible output levels per bit, and this is reflected in the "stack out" output, accompanied by the relevant clock.
The Code Dalabase
The Slacking Process
This clock is used to accurately provide the transminer filter with the time required for its filtering operation. Once this is complete, the database is rotated 16 bits to the left, loading bits 255 to 240 with the appropriate code for the next user.
Conclusion
Base-Band Filtering
- Bandwidth
- The Rectangular Filter
- The Raised Cosine Pulse Filter
- Matched Filtering
- Characterising the Matched Fi lter
- Implementillg tlte Trallsmitter alld R eceiver Filters
- The Digital FIR Filter
- Impl eme nt a tion of the Tra nsmitter Fi lter
- Implementation of the Receiver Filter
- Conclusion
In this way the receiver can estimate the nature of the received signal quite accurately. Line 2 of the table shows that So is multiplied by ho. In the case of the transminer filter, the convolution was done with IIN times the sampling rate.
The width of the input samples was a global parameter with default value 12 (the bit width of the ADC). For example, in the example given, the corresponding bits for bit 0 of each of the multipliers form the number (0111) when added together (the enlarged digits in the multiplier row form this number).
MMSE Detection
- The MMSE System Model
- Timing Analysis
- Implementation Architecture
- The Bit Est imator Entity
- The Pilot Tracker
The pulse detector keeps track of the number of samples clocked in from the receiver. The implementation issues surrounding some of the entities outlined in Figure 10.1 will be explained in the following sections. As a result, the system will attempt to scan to the right so that the zero count occurs at the actual position of the next peak.
The bit estimator unit is responsible for convolving the pulses from the pulse detector with the rows of the fixed MMSE filter matrix. This value corresponds to the maximum available value for the correlation of the expected samples with the fixed MMSE filter values.
Simulating Channel Distortion in the Software Radio
- System Overview
- Specifying Ullcorrelated Rayleigh Fading
- Generating Vncorrelated Rayleigh Random Variables
- Implementing Ullcorrelated Rayleigh Fadillg
- Rayleigh Fading Results
- Implementing an Additive White Gallssian Noise Generator
- Specifying a Fast Norma l Distribution Generator
- Uniform Number Ge nerati on
- Implementation of Specific SNR
- Additive White Gaussian Noise Results
- Persollal Computer Simulatioll
- Simulation Interface
- The DSP Layer
- The Simulation Processing
- Results
- Results of System Subjected to A WGN
- Results of System Subjected to Uncorrelated Rayleigh Fading
- Co nclusion
Here it is possible to use::tht: software radio to generate noise. X is the average of the set of numbers labeled x, n is the number of elements in the set. This project used C++ routines and Figure 11.15 reflects the inherent encapsulation associated with C++ objects.
The software radio uses the RS-485 connection to report count values to the DSP layer. For the current generation of Rayleigh fading, the variance was found (due to round-off errors) to be 0.2788.
Conclusion and Recommendations
Baseband shaping of the COMA signal was undertaken in a set of matched finite impulse response filters. Despite the success of the project, there are many recommendations that will enable the flexible radio platform to become an even more useful tool. This may include using the inter-FPGA communication bus for transferring signals to an off-board DSP chip.
Manual For The 1/0 Processing Board Of The Dsp·Based Digital Transceiver System", Alcatel Altech Telecomms, 2000, Saarnisaari, "Robustness of the MUSIC algorithm to subspace size estimation errors: delay estimation in DS/SS in the presence of interference", University of Oulu, Telecommunications Laboratory, 1999.
Apl}Cndix
- l Kron eck er Products
- Iterative Matrix Inversion
From this, it is possible to derive the number of mathematical operations performed in a single iteration step for each of the iteration techniques. The ratio of the values in equation 14.4 to those calculated for the iteration procedures will give an indication of the efficiency of the iteration technique. From equation 14.8, it is possible to deduce the number of multiplications and additions needed.
In Equation 14.5, the inverse of S is also a lower and diagonal (nxn) matrix, so it is interesting to see how many additions and multiplications will be performed per iteration. The effective number of iterations is the point at which it is more efficient to use the Gauss-Jordan reduction technique instead of the Gauss-Seidel iterative scheme.
Chapter 14 Appendix
- Gold Code Sequence Generation
