# Increasing samples per symbol degrades correlation between signals

I am trying to implement a simple pseudo-noise (PN) ranging simulation. I have used this GNU Radio example, and I have adapted it a bit to suit my needs.

Here is a screenshot of my flowchart:

The most important change is that I have introduced a new variable called 'bps' (bits per second). I use it and the 'samp_rate' to calculate the samples-per-symbol by the following formula:

sps = int(samp_rate/bps)


Here is where the sps variable is used (see the red blocks):

Question:

Why increasing the bits per second (bps) and therefore decreasing the samples per symbol (sps) gives better correlation between the reference and "transmitted" signals?

Correlation with bps = 1.2k:

Correlation with bps = 4.8k: