Underwater Acoustic Communication


Jim has long encouraged us to focus efforts on underwater acoustic communication. And now he's starting to prototype ideas based on research from UCSD.

regarding underwater communications - I have been looking into scaling down an underwater modem project that my daughter worked on (her masters thesis). FPGAs were used and I have been playing with some low power/cost ones that work with the BeagleBone.

With a modem, we could transmit and receive the types of messages described above in a GPS like way and thus figure out the geometry.

Link to research: http://cseweb.ucsd.edu/~kastner/index.php?q=research/underwater_com...

First project would be to adapt to ROV size form factor and power constraints. Further work could miniaturize even further so that this could be worn by divers, etc. In the swarm scenario we would at least have relative position across the swarm.

Let me know who might be interested in further work.

Anyone want to help?


Hi David:

Albeit my field of work is Fluid dynamics, I've been profesionally involved on GMDSS and AIS coding/TX-RX projects (Office requirements, you know).

Would be glad helping If I can.

I think the AIS data packaging system could be useful. It allows for a very robust and compact data transmission.



Interesting throught, although I would assume a compact, self correcting packet format (such as PACTOR http://en.wikipedia.org/wiki/PACTOR ) might be a better direction. Not an expert in this area either, but have some limited experience with u/w coms (in the Navy).


Absolutely. I've been floundering on some projects and would love to collaborate on a directed effort to get traction again. As for PACTOR, has it been used for subsurface communications? Seems bandwidth limited as well.


If I recall my ham stuff, PACTOR uses FSK (frequency-shit keying) as it's modulation scheme. FSK itself is used in u/w comms, as is PSK (phase-shift keying). The WHOI micro-modem uses both, and can switch between modulation schemes depending on bandwidth needs/acoustic conditions (more: Mirco-Modem)

I think the idea of building atop ham radio communication protocols is a great one, since there's already a lot of code out there (simple example: the AX.25 protocol has been implemented in the linux kernel for years), and probably a group of hams/devs who might be interested in a new application.


Thanks all for the interest. I am planning to use Jennifer's thesis work as a starting point. http://cseweb.ucsd.edu/~kastner/papers/ms-thesis-trezzo.pdf

The thesis includes background and a survey of currently available uw/modem products/projects with their characteristics. There are quite a few useful references at the end.

One of the challenges is to scale this down so it can fit inside the ROV. Current work used a larger FPGA development board (ZedBoard for the modems section) and another board for the Analog Transceiver and RF section (Toyon Chilipepper FMC). Power amps and some form of transducer are also needed. The UCSD project fabricated their own potted piezo-ceramic ring fro this.

Please take a look and let me know your interest.




Dear David:

I've been making a little research and asking some fellows at my University Dpt, regarding acoustic possitioning and precission concerns.

The carrier frequency reaches 100 Khz in high quality equipments, but most common is 45 Khz.

Taking 1500 m/s as an averaged sound speed, this frequency, returns a theoretical maximum accuracy of 0.033 m. But it would suppose a 100% signal profit and a 100% band with, wich is not possible.

Hence, if 5 cm is taken as the required possition accuracy, with a 45 Khz carrier frequency, the data rate would be, 30000 bauds, wich looks enough for granting transmission quality.

Best regards



which and not wich

By the way,

signal profit and a 100% band width


I would love to help with this! Have been thinking of this since I first saw the prptotype!


Hi Marius:

Im thinking about something similar to ......

Three beacons lying on the bottom. Each one transmits a particular signal, lets say, a song (binary sequence), that starts at a given UTC time. (GPS like, but acoustic).

One of the beacons is the Master one.

Lets say, master beacon is number 1, and the other two are 2 and 3.

Vectors 1-2, and 1-3, define a 2D vectorial space, which base matrix is design by them.

Vector 2-3, defines the error between theoretical and measured lectures.

Hence, two beacons are Tx/Rx, while the third could only be Tx.

Master beacon, from theoretical(geometric) datas and error, modifies the vectorial base, sending the correction data beside its own signal.

The ROV, receives at least two beacons (master + another). From each particuar songs Rov software obtains:

Beacon identification-Time from transmission start, and hence, by means of each message phase lag, the distance to each one.

Time stamp can be substituted by a Known message transmission time.

Beside its own message, master beacon sends the correction to the vectors forming the vectorial reference base.

ROV software, solves for the base change matrix, returning the desired kind of possition coordinates.

Which coordinates ? It will all depend on the software as its just a simple matter of maths.

As an information:

Ais systems, send the whole navigation and safety information of a ship, packed in a 168 bits chain.

By the way, I've alredy developed that coding software for another project, and would be pleased to share it with the ROV project if it can be useful.

AIS parsing example:

' type 1, 2, 3
(first bit is 1 )
1) Message ID, 6 bits 1-6
2) Repeat Indicator, 2 bits 7-8
3) User ID, 30 bits 9-38
4) Navigation status, 4 bits 39-42
5) Rate of turn, 8 bits 43-50
6) SOG, 10 bits 51-60
7) Position accuracy, 1 bit 61
8) Longitude, 28 bits 62-89
9) Latitude, 27 bits 90-116
10) COG, 12 bits 117-128
11) True Heading, 9 bits 129-137
12) UTC second when report generated, 6 bits 138-143
13) Regional Application, 4 bits 144-147
14) Spare, 1 bit 148
15) RAIM Flag, 1 bit 149
16) Communications State, 19 bits 150-168

Of course thats the IMO/ITU standard, but we could use the same coding but using message blocks for our particular needs.



Hi Jim:

I've read the thesis you've recomended.

Quite interesting albeit most of it lays out of my field. Unfortunatelly electronics hardware is not my thing.

Anyway hope to be useful in other fields.



I should get a pair of piezoceramic transducers in late November to experiment with. I will try to recreate some of the work in the research papers with scaled down electronics as a starting point.



Hi Jim:

Albeit electronics is not my thing, "What to send and what do once signals are received, certainly is, or better said, has been".

Let me explain an idea about possition error reduction. Perhaps it could help.

As far as I know, the ROV is fitted with accelerometers (6 degrees of freedom).

A heading magnetic compass is subjected to electromagnetic interference from the ROV electronics and to geographic and local magnetic variations, hence its not a reliable heading sensor.

Would be a good thing to provide the ROV with a gyroscopic heading sensor.

Well, with accelerometers and a reliable heading sensor a very accurate Dead Reckoning navigation can be done.

Lets now suppose the ROV receives two signals, from two beacons, that once proccesed and corrected by all factors described at Jennifer's thesis, will still result in a possition error, may be bigger than the ROV size.

Hence it can be imagined that the ROV possition is located inside a romboidal area, composed by the intersection of four circles, which radius are computed from distance to each beacon and estimated distance error.

Most probable ROV possition, can be supposed to be at the romboidal area diagonals intersection.

After a given time lapse, from accelerometers and heading, another possition can be estimated (Second Dead Rekconing location).

At the same time, a second romboidal area and its diagonals are obtained from the acoustic sensors.

Maths say that uncertainty is carried along the path and that increases along the studied function (in our case, the ROV path as a function of time).

Hence first diagonals and romboid can be translated to the second location taking as translation point, the diagonals intersection, obtaining a romboids intersection, and four diagonals (With a short time lapse, uncertainty growth can be neglected).

From the new geometrical structure, a reduced uncertainty can be computed and Dead Reckoning and signal processing corrections are corrected.

By a combination of external references and internal possition guessing, location will be improved along time, instead of being blurred by errors accumulation.

As you can see, all starts from the assumption that signals are correctly sent and received.

I can refine and work on the mathematical development if you think it coud fit our needs.