Frequency-modulation is usually produced with an individual carrier. For some applications it is interesting to produce several frequency-modulated carriers using the same modulation signal. Such a system could be used to provide, for example, the FM frequencies in a motorway tunnel or as jammers.

In our work we simulated, developed and built such a multiple-frequency modulator. Our concept is based as far as possible on digital building blocks. A baseband signal modulates a sinusoidal carrier. Either a mono- or a FCC-stereosignal can be used as modulation signal. The modulated carrier is transferred onto several carrier frequencies with a single-sideband mixer. The parameters of these carrier frequencies are loaded via PC into the hardware simultaneously. The modulated carrier frequencies do not yet have the desired final values, however the correct distances to each other. Finally two additional analog mixers transfer the carrier package to the required frequency range.

We were careful to follow the international broadcast standards, in order that any customary FM-radio can receive the generated signals.

The diploma project consists of two sections:

Part A: Development of a test system

Our task was to develop a test system which delays radiograms directly in real time, attenuates them by different amounts and adds noise to them. For signal handling we used a Digital Signal Processor (DSP), which can be controlled by serial connection from a PC. We implemented a graphical user interface on the PC, which controls the DSP via the serial RS-232 interface. On the DSP, individual modules were implemented for the delay, for attenuating the signals and for the generation of noise, which were linked together depending upon mode. An electronic adapter circuit was implemented that interfaces inputs and outputs to the real radio environment. So at the end of our project we have a functioning test system, which can be used directly in the real environment.

Part B: Analysis of radio signals

This part consisted of testing the suitability of criteria for the quality evaluation of radiograms, which are used by our partner company in their system. Furthermore, a basis for the detection of feedbacks in radiograms should be found. We developed a Matlab program which reads the radiograms directly from a wave file, analyzes them and outputs the results of the analysis graphically. During a critical review of the criteria we found that manipulations of individual spectral lines by band limitations and in-band modulations impair the function. The investigation of feedback sequences showed that a detection cannot be determined easily from the quality factors. A promising approach was found in the continuity of the quality factors of the signal level.

SPEARS stands for speaker recognition system. SPEARS is a, in Matlab realised, simulation for a speaker verification system based on LPC-Model of human speech production. The speech signal is analysed by a LPC-Processor which extracts the main features of the signal in order to find the acoustic characteristics of the speaker. This acoustic representation, the speech pattern of a specific speaker, is then compared with other speech patterns in the systems storage in order to find the best match. An artificial neural network classifies the speech patterns according to the speakers it is trained to recognise. Finally a decision logic decides which speaker has been recognised. The pattern comparison is based on a dynamic time warping algorithm. The algorithm uses two embedded for loops and is based on a symmetrical local continuity constraint. For a speaker dependent word recognition the system reached a total goal of 95% in a silent environment and a total of 90% in adverse environment. For a word dependent speaker recognition the system reached a total goal of 100% for both environmental conditions. Finally we proved the sensitivity of the linear predictive coding (LPC) spectral analysis model to noise.

Why is a two-wire-system desirable for several subscribers? For usual intercom-systems are several connectionwires asked, which brings a big expense of time in installation. A system based on a two-wire-conduction makes it much easier.

A special feature of the two-wire-system is to connect tonesignals together with supply voltage on a two-wire-conduction at once. We had the possibility to develop the system from the beginning. On a first step we defined the timing of the system. After that we worked out every single component.

The microphon-signal is brought by a automatic gain controled amplifier, what it brings on a certain level important for transmission. The transmission is realised by frequency modulation, which guarantees a high quality and shoots malfunctions of 50Hz netfrequency. The appartments are connected by dual tone signals. The loudspeaker is also used as a microphon, which brings some more swichings with. The system succeeds except the switch of the changeover.