I'm working on FSK demodulator with the TMS320C6711 DSK. I've done this demodulation with the method where the signal and its delayed version are multiplied together and then low-pass filtered. But this demodulation technic is only working if the sample rate is an integer multiple of the baud rate. Does someone know a technic or an algorithm which allow a demodulation of a 300 baud rate signal with a 8000hz sample rate? Is there an other technique that's not as sensitive to this? thanks, Benoit

# Sample rate and baud rate issue

B=E9nou wrote:> I'm working on FSK demodulator with the TMS320C6711 DSK. I've done this > demodulation with the method where the signal and its delayed version are > multiplied together and then low-pass filtered. But this demodulation > technic is only working if the sample rate is an integer multiple of the > baud rate. Does someone know a technic or an algorithm which allow a > demodulation of a 300 baud rate signal with a 8000hz sample rate? Is there > an other technique that's not as sensitive to this? > > thanks, > > BenoitI would suggest that you resample the 8000 sps data by a factor of 3/8 before you do the delay-multiply operation. Then you'll have 10 samples/bit and be able to realize the desired delay exactly. John

B=E9nou wrote:> I'm working on FSK demodulator with the TMS320C6711 DSK. I've done this > demodulation with the method where the signal and its delayed version are > multiplied together and then low-pass filtered. But this demodulation > technic is only working if the sample rate is an integer multiple of the > baud rate. Does someone know a technic or an algorithm which allow a > demodulation of a 300 baud rate signal with a 8000hz sample rate? Is there > an other technique that's not as sensitive to this? > > thanks, > > BenoitI do not think there is a practical technique that will allow non-integer samples per symbol for a number of reasons. 1) Timing recovery techniques usually work with integer number of samples per symbol 2) you will have to "select" the symbol from the signal at some point in the processing and having a non integer sample per symbol will probably complicate things. So, you will probably need to resample as a first step. Suggest a higher rate (eg. 9600Hz) to avoid potential aliasing problems.=20 -V

> >B=E9nou wrote: >> I'm working on FSK demodulator with the TMS320C6711 DSK. I've donethis>> demodulation with the method where the signal and its delayed versionare>> multiplied together and then low-pass filtered. But this demodulation >> technic is only working if the sample rate is an integer multiple ofthe>> baud rate. Does someone know a technic or an algorithm which allow a >> demodulation of a 300 baud rate signal with a 8000hz sample rate? Isthere>> an other technique that's not as sensitive to this? >> >> thanks, >> >> Benoit > >I would suggest that you resample the 8000 sps data by a factor of 3/8 >before you do the delay-multiply operation. Then you'll have 10 >samples/bit and be able to realize the desired delay exactly. > >John > >Hi John, Thanks for your informative response. Regards Benoit