Why FSK demodulator output different bitstreams in case of the same inputs of different lengths?

[0711]

Hi everyone,

   Recently I was trying to do FSK demodulation with Labview and my USRP2.

   I did the following two tests, but found a strange phenomenon:

   In the first test, I use the demodulator to demodulate a waveform of 18399 samples, where the signal occupies about the first 12000 samples, and the FSK demodulator outputs a bitsream of 1116 bits.

   The input waveform is shown in the following figure.

  

   In the second test, I use the demodulator to just demodulate the first 12000 samples of the waveform,  and I found that the demodulator outputs a bitstream of 705 bits.

  The input waveform is shown in the following figure:

  

   According to my understanding, as the input of the 2nd test  is the same as the first part of the first 1st test, the output bitstream of the 2st test must be equal to the output of the 1st test. However, they looks very different. In the first test, even the sync  sequence of bits cannot be found. 

    Can any one  help me explain why?  Thanks in advance!  

   In the attachment, I post the program and my test data.

[Anthony_F]

Hello 0711,

Are you using the Demodulate FSK function provided with modulation toolkit? I did not see the demodulation VI in the zip file and the main VI is not executable.

If I understand correctly the second signal is just a subset (first 12,000 samples) of the first signal.  Do the first 705 bits match in both signals? 

For 16 bits/sample and 18399 samples you should get back about 1149 bits. For 16 bits/sample and 12000 samples you should get back about 750 bits. Both of these seem to loosly correspond with the number of samples you are giving the demodulate VI and the output bitstream. 

Have you zoomed into the acquired waveform to see if it is not distorted? Are you able to spot the frequency shifts by inspection?

[0711]

Hello Anthony F,

   Yes, I use the modulation toolkit.But I encapulate it into a subvi. Sorry for that I forget to attach it.

[Anthony_F]

Another place to check for inconsistencies is the symbol mappings array in the FSK System Parameters cluster as well as continuous/discontinuous phase setting.

[0711]

Hello Anthony,

   Yes, you are correct. The second input  signal is just a subset of the first signal, and it is the same as the first 12000 samples of the second signal, But the demodulation of the second signal is totally different than the first 705 bits of the second signal.

    I get the second signal by extracting and dumping the first 12000 samples of the first signal. So both the two signals have the same start time, the same sample rate, but diffferent sample length.

[0711]

Dear Annthony,

   The symbol mappings array in the FSK System Parameters cluster and the  continuous/discontinuous phase setting are same in both tests.

[0711]

Sorry Everyone,

   I checked the program last night. Now I package all the subvis in the attachment. Thanks for your help