Janus general discussion forum Help

Forums » Janus general discussion forum » Replacing the HFM sweep with a 32-chip FH sequence for detection & estimation

Replacing the HFM sweep with a 32-chip FH sequence for detection & estimation

posts: 46 user offline
Both Ivor Nissen and Dale have independently urged us to implement the 2009 workshop decision to replace the HFM sweep with a FH sequence.

Ivor's interest in this, if I understand correctly, is that it simplifies the analogue electronics and processing. Since one needs the FH capability in any case for the main message, why not use it also for the detection preamble?

Dale's motivation is from the signal processing perspective, that an orthogonal set of FH chips can be used without a guard time with superior detection and estimation performance, if properly handled.

Our original concern was that the Doppler performance of the FH sequence seemed inferior to that of the HFM, but this now seems like an artifact of our coherent detector processing, rather than a fundamental signal processing constraint. Indeed, from an information theory point of view, there should be nothing that can be done with a 4 or 6-chip HFM that cannot be done better with a 32-chip FH sequence.

Paul vanWalree is working with us to make comparative transmissions on our semi-permanent modem farm in which we will be using two different channel probe signals, bracketing both ver 0.3 and 1.0 JANUS signals and two types of DSSS signals with similar bandwidth and baud rates for comparison.

It is our intention to provide detector implementation examples in the ver 1.0 release that will transparently handle either the ver 1.0 standard or the older ver 0.3 for legacy purposes.


posts: 1 user offline
I would be interested to learn about the detection algorithm for a 32-chip FH sequence, and its robustness to channel variations within the sequence duration. With chirp and m-sequence detection preambles, and matched-filter processing, the optimal pulse length for detection is of the order of the coherence time of the acoustic channel. I wonder whether this is also a consideration for a frequency-hop detection preamble.

Med vennlig hilsen,


posts: 46 user offline
On another aspect of implementing the 32-chip FH preamble, we have a choice of the bit sequence to use. That is, we could transmit all '0's, all '1''s or some (presumably pseudo-random) sequence in between. If we go for only '1's or '0's, this can simplify the computational task for an incoherent detector (we only have to search for energy in 13 frequency slots). Does anyone see an advantage to using a specific sequence?


Show posts: