Abstract for group VoIP Embedded VoIP Solution With Techniques to Reduce Quantization Noise The project is a re-attempt / continuation of VoIP phone project from the previous spring in which a group tried to implement but failed due to reasons beyond our knowledge. We are attempting to develop a VoIP client (VoIP phones) which would send out packetized (RTP/UDP) packets to a peer VoIP phone . The voice data would be captured through the Audio interface on the board. Thereby , it would undergo ADC and using well known algorithms (codecs) we would convert the quantized data to packets which would be sent in the form of RTP / UDP packets (designed especially for such a purpose). It would also involve either implementing or downloading an open source IP stack for the Xilinx Spartan boardand would also require us to write our own MAC protocol. TCP would be used for connection setup while UDP would be used for sending the packets. We might need to implement our own APIs for the transport layer protocols that we would use and would certainly implement our own RTP stack. The other half would involve reducing the noise in the signal introduced by the quantization process. A brief description of this technique (known as companding) is given below. In a fixed-point digital processor, small signals (signals with envelopes much less than the overflow tolerance) are more distorted by quantization noise than large signals (signals with envelopes near the overflow tolerance). Roughly speaking, small signals do not take full advantage of the available bits. Normally, signals in the processor are sometimes small and sometimes large (they have a large dynamic range); as such, when they are small, they become very distorted. Companding is a technique in which the dynamic range of an input signal is compressed before quantizing (before the analog-to-digital converter, or ADC), yielding a signal which is always large at the input to the ADC. At the output of the digital-to-analog converter (DAC), the signal is expanded back to its full dynamic range. The digital signal therefore takes full advantage of all the bits available in the fixed-point processor.