Problem Statement
I’m trying to get a handle on the Correlation Estimator block, so keep things pretty simple after having studied the example example_corr_est.grc
In the attached CorrelatorExeriments2.grc, I’m creating a steady flow of random packets. The original Access Code I was using is 0xFE6B2840, which has decent autocorrelation properties. Move to a stream that swings from -1 to +1, no transmit
pulse shaping filter at this time. Feed this into the Correlation Estimator Block with a fairly high ThreshHold (.995).
RandomPDU à Tagged Stream
à Corr Estimator (dynamic)
à Sink
What I end up with is many “correlation hits” across a single packet. Tightening up the threshold to .999 yields no correlation hits?
Switched to a simple 32 bit ‘1010’ preamble. That preamble is producing ~22 “correlation hits”. Would expect to get multiple peaks, but they all have equal |Corr|^2 outputs (see pic). Might indicate that too short a sequence was being
correlated against, but the DataVector being fed in is (supposed to be) 32 bits
Questions:
- With a 32-bit ‘1010’ preamble, why are all the |Corr|^2 values the same? Would have expected the proper alignment to produce a larger peak…
- Is this block running a traditional correlation calculation?
- Is this block running a traditional correlation calculation?
- In a perfect simulation environment with no impairments, why does a threshold of .999 produce no matches on 0xFE6B2840? Again, seems to say that maybe the reference sequence doesn’t
match, but I don’t see the configuration error…
- How does one avoid timing hits in the downstream Symbol Sync block on false positives (not yet in the sim, but I’ve read it re-centers on the tags)?
- Why does setting the “max bytes” in the Random PDU generator below 1307 cause the flowgraph to freeze up (even if you adjust Min Bytes in tandem)?
Notes about the .grc
- Tag “packet_len” marks the beginnings of each packet and serves as the Time Sink trigger
- The 0xFE6B2840 AC is in the sim but disabled, switching to it requires enabling/disable the relevant Default Header Format Obj and variable AC_Hex
- Don’t yet feel completely comfortable with the Modulate Vector, an extra set of eyes on the setup here would be appreciated
Other notes / suggestions
- the Example_corr_est.grc included with GR seems to have some problems in 3.8.1.0. I’ve adjusted it to where it passes casual inspection, example_corr_est2.grc attached.
- would be useful for the Documentation block / wiki, to mention the role that tags are intended to play (with perhaps a mention/pointer to the example flowgraph)
