| ??? 02/03/03 14:44 Read: times |
#38169 - RE: Sound sampling for PWM? Responding to: ???'s previous message |
Hi Andy,
Ten years ago I build a (Philips 87c552 based) full duplex voice scrambler machine. The processing was in the time domain (packet scrambler) and was done entirely with the 87c552. I don’t exactly remember but I think I forced a little the clock (about 16MHz). The 87c552 has a multiplexed 50uS 10-bits ADC and two 8-bits PWM based DAC. I used 2 of the ADC’s channels, both the PWM DACs and an external 8KB RAM memory for 0.5-second voice acquisition in both directions. The program was in the internal 8K eprom. Also I used 4 identical 2-poles Chebyshev LPF in a Sallen-Key configuration. The voice band was limited 3.4kHz and the sampling frequency was 8 ksps. The PWM frequency was ~24KHz (at the upper limit of the uP). The scrambling process was over 0.5 s in 16ms packets in full duplex mode. The full processing chain was as follows: (Microphone) ->(LPF)->(8bit ADC) ->(uP scrambler)->(8bit PWM)->(LPF)-> (phone-line)-> (LPF)->(8bit ADC) ->(uP descrambler)->(8bit PWM)->(LPF)->(speaker). Conclusions: 1. The S/N ratio for 8- bits PWM DAC was identical as for an 8-bits parallel DAC. 2. The quality of the voice was near to an 8-bit PCM system (without scrambling) 3. The quality is more dependent from the input antialiasing LPF then from the output LPF witch may be an also a 1-pole RC LPF. 4. 8-bits 8Ksps PWM is good enough for a telephone voice player. Regards, Cristian Ceausescu |



