I wanted to ping the DSP hivemind for general thoughts on what would be the best modulation type to use for low frequency underwater communications. I have chosen this project since I can learn a lot from it.

Some context:

  • Low frequency as in < 500 Hz (for both carrier and modulated data obviously)
  • BPS of say, 200 Hz would be good.
  • Will certainly have multipath.
  • Frequencies can be smeared due to doppler by a maximum factor of about 0.3% of the original frequency.

What I have found thus far:

  • I was thinking about OFDM, but I have learnt that while channel estimation is much easier, it is much more sensitive to doppler effects.
  • I was also thinking about chirp modulation, has anyone ever done something like that?

What are your thoughts?

Edit: I have attached some of what I believe are 'worst case' scenarios of the (multipath channel, for a bps = 200 Hz). The channel is in terms of number of bits in the time domain so you can more easily see how many bits pass before the next reflection arrives.

Case 1: enter image description here Case 2: enter image description here Case 3: enter image description here Case 4: enter image description here


  • As we can see, I almost always have a second path of nearly equal magnitude but opposite phase ready to stick it to my main path.
  • For a packet of 1000 bits @ 200 bps, (5 seconds), I suppose the channel could change significantly... but at the same time, we have total control over the packet length and content.
  • We may assume that frequency offsets due to doppler is relatively 'well behaved', ie, no sudden 'jerks'. Frequency offsets due to carrier mismatches can also be thought of as well behaved.
  • 1
    $\begingroup$ Is the frequency offset constant or time-varying? If it changes, how quickly? Over the course of a symbol? A packet? Many packets? Do you have a model for what the channel will look like? $\endgroup$
    – Jason R
    Oct 31, 2011 at 17:32
  • $\begingroup$ This is ELF electromagnetic waves, right? Not sound waves? $\endgroup$
    – endolith
    Oct 31, 2011 at 20:37
  • 2
    $\begingroup$ @endolith No this is definitely sound waves... $\endgroup$
    – Spacey
    Oct 31, 2011 at 20:38
  • $\begingroup$ @JasonR I have edited the question to try and answer your feedback, please let me know if you need any other inputs, etc. Thanks. Also, nothing is set in stone at the moment as far as packet durations, etc. BPS is also not set in stone but 200 would be nice. $\endgroup$
    – Spacey
    Nov 1, 2011 at 1:49

1 Answer 1


That is a nasty signal environment. I would do a little DSP judo and make the multi-path work for you by using a rake receiver, which should make the multi-path signals increase your SNR rather than diminish it.

Rake receivers have only been used (as far as I know, anyway) in CDMA systems, but that is not because they can only be used in CDMA systems. The issue is the symbol period. The following quote is from a paper on rake receivers:

RAKE receiver attempts to collect the time-shifted versions of the original signal by providing a separate correlation receiver for each of the multipath signals. This can be done due to multipath components are practically uncorrelated from another when their relative propagation delay exceeds a chip period.

Thus, the reason that they were used in CDMA systems and not, say, GSM, was because the chip speed was so much faster, and thus the multipath signals would not arrive in the same chip/symbol. That was not true for GSM and other "narrowband" signals. Even though your bit rate is very low, the multipath signals still seem to have delays greater than one symbol, so this constraint should not be a problem.

The other problem is detecting the multipath. This could, I would think, be accomplished with known data sequences, such as a preamble, but I confess that I am not an expert on rake receivers.

If the rake receiver does not work out, you could still use a long equalizer to handle the multipath.

Regarding the other elements of the system, if there is no feedback I would do a QPSK signal with FEC (probably turbo codes). If you do have feedback I would do the same but also dynamically change the modulation type from QPSK up to 16-QAM or something, depending on how well the signal is coming through.

EDIT: After thinking about it a bit I realized why rake receivers and CDMA systems go hand-in-hand. The problem is that even if you detect a multi-path signal it doesn't do you a lot of good unless it has a positive SINR. By definition at most one multi-path signal can have a positive SINR, because for all the other multi-path signals the strongest overwhelms them.

That is where despreading comes in. Once the multi-path signal is despread it will have a positive SINR, assuming the spreading factor is large enough to overcome the initial negative SINR. Given this I think the right solution is to relax the 500 Hz limit, use a fairly large spreading factor and a rake receiver to combine the various multi-path signals.

  • $\begingroup$ Thanks - Yeah, I looked into the RAKE, however, isnt this only doable for CDNA spread spectrum systems? I dont see how one could lock on to the various signals from the various multi-paths otherwise... $\endgroup$
    – Spacey
    Apr 9, 2012 at 15:55
  • $\begingroup$ @Mohammad I edited my answer to respond to your question. $\endgroup$
    – Jim Clay
    Apr 9, 2012 at 17:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.