# Convolutional Code: choosing a trellis and puncture pattern for code 2/3

I'm working on convolutional coding in which I'm using generator polynomials as {13, 15} with puncture pattern of {1 1 0 1} for code rate 2/3. I don't get expected BER trend. If I use puncture pattern as {1 1 1 0}, the BER curve behaves as expected. Is there a rule of choosing puncture patterns based on different trellis structures? what am I missing?

Any idea/ hint is appreciated.

• For your chosen example, there are four different puncture patterns $$(1,1,1,0), (1,1,0,1), (1,0,1,1), (0,1,1,1)$$ that will give a rate $\frac 23$ code. Why do you expect the BER vs SNR curve to be the same in all four cases? If anything, I would expect the $(1,1,1,0)$ and $(1,0,1,1)$ curves to be similar since both are deleting the second bit in alternate two-bit frames, and the curves for $(1,1,0,1)$ and $(0,1,1,1)$ to be similar since they are deleting the first bit in alternate two-bit frames. – Dilip Sarwate Aug 29 '19 at 18:57