«2.1 Time Synchronization The actual need for time synchronization within an underwater acoustic network is not always present. It can be argued that ...»
2.2 Full-Duplex Links 11 Bandwidth-on-Demand (BoD) techniques. DAMA allocates channels to users when the users request an allocation. These channels are typically ﬁxed in size.
Alternately, BoD provides users a variable sized allocation depending on the request of the individual user. By allocating multiple channels to a user on demand, these channels may be used to inverse multiplex two or more message frames, thus providing a coarse version of BoD. It is this coarse BoD implemented via DAMA that is suggested.
2.2.2 Physical Layer
Time division duplexing is no option in an underwater network with the large propagation delays of the acoustic channel. Frequency division duplexing was demonstrated in  where data was transmitted between the shore and a ship.
Using two transducers on the ship, one for transmission and one for reception, with a distance of 33 m between them, they managed to transmit and receive simultaneously in adjacent frequency bands. The distance between the ship and shore was up to 4500 m. Reception was good on both the ship and the shore. The frequency bands are not known, but Chebyshev ﬁlters with 80 dB out-of-band rejection were used for side-lobe suppression.
Obtaining full-duplex can also be achieved by using Code Division Multiple Access (CDMA) based techniques, see Sect. 3.2. In  a test using CDMA based channelization schemes was performed in a bucket and a small lake with a distance up to 5 meter. Separate Tx and Rx transducers were used with a spacing of 30 cm.
Several channelization schemes were tested including frequency hop CDMA, time hop CDMA, direct sequence CDMA, and also hybrids thereof. Pulse position modulation was used for keying the data onto data symbols. In these tests, frequency hop CDMA performed best.
All demonstrations utilize a separate transducer for transmitting and receiving signals, and apparently the most successful  having a long distance of 33 m between them. On a node like an AUV this kind of distance is not available, and preferably it should be a single transducer both transmitting and receiving. No literature could be found on full-duplex transducers. Radio systems like maritime VHF and maritime ship radio stations have FDD channels using only a single antenna.
2.2.3 Concluding Notes
Fig. 2.3 The traditional MAC/PHY division (the OSI-model) available. But if a good full-duplex solution is found in the future, it could signiﬁcantly improve the performance of underwater acoustic network protocols.
2.3 Adaptive Data Rate By Adaptive Data Rate in this context it is meant that the communication system is able to utilize some knowledge about the present state of the communication channel so that both the coding and modulation methods can be adapted to this state. The goal is to maximize the system throughput under varying channel conditions.
One way to achieve this is to employ a technique which in the telecom industry is known as ‘‘Adaptive Coding and Modulation’’, or ACM. This is used today in both wired and wireless communication systems.
In order to use adaptive coding and modulation effectively, it is necessary to establish a close interaction between the operations of the physical layer (PHY) and the medium access control layer (MAC).
In a traditional telecom environment, where circuit-switched networks (ISDN and ATM) were the norm, there were clear distinctive lines between the responsibilities of the PHY and the MAC, as it is laid out in the OSI model and as shown in Fig. 2.3.
With the introduction of packet switched networks (Ethernet and others), these lines have been considerably blurred, and this has lead to a simpliﬁed model where some of the layers have reduced functionality and others are removed.
In a communication scenario that involves ACM, this model will have to be changed further in that some of the traditional MAC functionality, such as the selection of the modulation format and coding scheme, will have to be moved down to the PHY layer in order to be able to respond to the (quickly) changing channel conditions. This part of the MAC functionality is sometimes referred to as the ‘‘lower-level MAC functionality’’ (Fig. 2.4).
In this scenario the ‘‘higher level MAC functionality’’ is responsible for establishing the overall system parameters like quality-of-service (QoS) requirements for the individual links, and to organize the network for maximum system capacity, the latter being important in an ad-hoc/mesh network scenario.
2.3 Adaptive Data Rate 13
In this discussion of adaptive data rate, only the PHY and lower-level MAC functionality will be considered, and it is easier to take the bottom-up approach and identify the requirements of the PHY ﬁrst.
2.3.1 The Physical Layer The discussion of the PHY-layer will be based on a conceptual transmitter/receiver pair, as shown in Fig. 2.5, below. Note that more detailed discussions on physical layer technology for underwater acoustic communications are not part of the present study.
In a communication system, each terminal will at least have one such transmitter/receiver pair (or ‘‘transceiver’’). In addition, a transmit/receive switch circuitry is needed if time-division duplex (TDD) operation (not shown in the ﬁgure) is required.
In order to achieve maximum system capacity, this hypothetical system would have to be able to utilize all access techniques, such as frequency division multiple access (FDMA), time division multiple access (TDMA), code division multiple access (CDMA) and space division multiple access (SDMA). These access techniques are described elsewhere in this document, and will not be repeated here.
The various access techniques all have different requirements with respect to clock stability, frequency stability, linearity of up-/down-conversion chains and number and size of transducer elements, and will eventually be dictated by a cost/beneﬁt trade-off.
A short description of the various functions carried out in the building blocks of Fig. 2.5 is given below.
From the information source comes the user data to be transmitted over the link.
A forward error correction (FEC) encoder protects the data before transmission by adding special bits (parity) or bit-patterns that can later be utilized in the receiver to extract the original information bits. There are a number of different coding schemes that can be applied for this purpose, and they range from simple to advanced block-coding structures (from Hamming to Reed-Solomon) to convolutional and Turbo codes, and various concatenations of the above.
14 2 Topics Bordering the Physical Layer
Fig. 2.5 Conceptual transmit/receive functionality. See text for explanation In the interleaver, the encoded data are repositioned in the data-stream according to a predeﬁned structure. This is to avoid loss of data caused by impulsive noise which would be detrimental to a convolutional code decoding process. (This is not so much of a problem with block-codes.) In the frame-assembly block, the information is inserted into a frame where preambles like unique words (for frame alignment) or various pilot-assisted modulation (PSAM) bits and post-ambles like CRC’s and/or end-of-frame (EOF) delimiters are placed.
In the constellation mapper, the individual bits of the frame are mapped onto a suitable alphabet of symbols, later to be modulated onto two orthogonal waveforms (for a quadrature modulated signal).
For a coherently modulated signal, the alphabets can be a set of symbols belonging to a modulation format like e.g. binary phase shift keying (BPSK), quaternary phase shift keying (QPSK) and higher order like quadrature amplitude modulation (QAM).
For a code division multiple access system based on a direct sequence spread spectrum (DSSS) technique the alphabet is a set of orthogonal spreading codes.
For a non-coherently modulated signal, the alphabet is e.g. a set of frequencies in a multiple frequency shift keying (MFSK) modulation format.
A code division multiple access system can also be constructed by using a set of frequencies in a frequency hopping pattern orthogonal for each code.
The symbol set is then modulated onto two orthogonal carrier waveforms (cosine and sine), where the carrier frequency is generated through a numerical controlled oscillator (NCO). In Fig. 2.5, a direct-to-carrier type of system is shown. The transmitted signal, s(t), is a real band-pass signal.
At the receive side, the signals are converted directly from carrier (real bandpass signal) to complex base-band, the demodulation frequency (and phase, in case of a coherent demodulation scheme) are again controlled by a NCO. The exact frequency and phase are controlled by the carrier-frequency- and phase-recovery sub-system.
2.3 Adaptive Data Rate 15 Not shown in the ﬁgure are the necessary signal/burst acquisition sub-systems.
The information bearing signals are then extracted from the signal constellation and fed to an equalizer to remove inter-symbol-interference (ISI) induced by the channel.
In the frame disassembly operation, the information bearing signal is extracted and fed to the de-interleaver before FEC decoding and the result is then fed to the end-user.
In , it is claimed that a coherent modulation scheme based on Phase Shift Keying (BPSK/QPSK), in conjunction with an adaptive decision feedback equalizer (DFE) and a spatial diversity receiver is an effective way of combating the effect of multipath fading in a shallow water environment. It is, however, admitted in the article that the excessive delay spread, often several hundred symbols, make it too computational complex for real-time operation for such a system.
2.3.2 Medium Access Control, Lower Level
In order to be able to utilize the communication channel effectively, an optimally conﬁgured system will have to be able to change both the modulation format and the coding scheme in order to adapt to the current channel conditions.
To make this work, it would be necessary to collect information about the current channel conditions, convey this information back to the transmitter side and use this to select the modulation and coding schemes for the next outgoing burst.
The information to be used in this process could be e.g. an estimate of the signal-to-noise ratio and/or the delay and Doppler spread of the channel, and this information could be extracted from known symbols in the frame structure like the pre-amble (e.g. a unique-word) or the PSAM symbols (if they are used).
However, any addition of extra symbols to aid in these processes would incur an overhead and thus reduce the information data rate.
Other, more indirect means of extracting the same information could be to monitor the states of an equalizer or the decoding depth of the trellis in an (convolutional code decoder) FEC decoding process.
A different way of obtaining information about the channel state could be to use a special burst sent in advance as a channel probe, as proposed in . In their system, such a probe is sent just before the actual information burst is transmitted, as shown in Fig. 2.6.
Here station 1 is transmitting a probe before the actual data is transmitted, and this is used at station 2 to extract some key properties of the channel state at the moment of reception. This information is then afﬁxed to the data in the next outgoing burst from the station 2. If the turn-around time is short this is a more or less correct description of the channel, and it is used at station 1 for the next outgoing burst from this station.
16 2 Topics Bordering the Physical Layer
Fig. 2.6 Using a channel probe to extract information about the channel In addition, if the channel can be treated as reciprocal, station 2 can itself use this information to set the parameters for the next outgoing burst, and/or to construct a better probe for more detailed channel measurements.
The approach described in the paper is a very simple adaptation of this, but it lacks a conclusive statement about the system improvements, if any, that can be achieved using this technique. It is not clear why a special probe is required to do this, as the same result can be achieved by simply embedding the same information into the pre-amble of the burst itself.
2.3.3 Adaptive Data Rate in ARQ Systems
Adaptive Coding and Modulation is something that is normally closely connected to the physical layer processes of a communication system, as described in the previous sub-sections. This is due to the fact that very low latency is required in order for the system to respond to the changing channel conditions.
If ACM methods based on direct measurements of the current channel conditions are not possible, e.g. where the time-delays in the system are so severe that the channel measurements are obsolete by the time the system can respond to these, an ACM approach based on the use of information extracted from the ARQ system (see Chap. 4) can be foreseen.
One way to achieve this could be that instead of using an approach where the previous packet is blindly repeated when an ACK time-out occurs or a NAK is received, one would use a scheme where the ﬁrst re-transmission is just a re-transmission of the packet. If this transmission is also unsuccessful, the packet is reformatted into a longer burst where the modulation and coding schemes are strengthened (more energy per bit and/or more protection bits added).
If the transmission still fails, this back-off procedure is repeated until all possibilities are exhausted and the system ﬁnally breaks down.
When the ACK packets start to arrive, the modulation and coding overhead is gradually reduced until the maximum possible information bandwidth is re-established on the channel. The transmit side will always be aware of the current maximum channel capacity based on the reception of the ACK-packets, since this will conﬁrm that the receiver is able to decode the packets.
2.3 Adaptive Data Rate 17
2.3.4 Summary and Conclusions