Sonar processing: Back to basics
Like the rest of the world, the oceans and the vast spaces beneath them are growing more dangerous. International adversaries are projecting power more aggressively with fighting ships and submarines. Smaller, quieter vessels are being employed, and reverberation-rich littoral waters are now key to protecting shorelines. Sonars and sonar processing need to keep up with the threat.
Oddly, however, the reverse may be happening in some cases. Cost pressures as well as knowledge gaps are thought to be driving some countries to adopt simpler, single-sensor sonar systems in lieu of more capable towed or hull-mounted arrays with hundreds or thousands of individual sensors. Single-sensor sonars lack the resolution, beam-steering agility, and target-detection capability of multiple-sensor systems.
Single-sensor systems have a cost advantage over more complex, multiple-channel systems. The processing hardware and software are likewise more affordable, but are they cost-effective? If signal detection is the goal, the simpler, cheaper systems clearly leave something to be desired.
A single-sensor sonar is relatively less capable than its multiple-sensor counterpart of picking a signal out of the noise environment. Real signals are additive, whereas noise is not. That means that the more sensors you have, the better probability there will be of detecting a relevant signal.
If the decision is made to adopt a multiple-sensor solution, the next step is to choose a processing architecture. Submarines boast very limited real estate, so users ideally would want to squeeze as many processing channels as possible into the smallest hardware footprint at the lowest cost per channel.
Users can choose backplane solutions such as VME or VPX, in which data acquisition (DAQ) and data processing cards are housed in the same chassis, interconnected by a bus or fabric topology. Alternatively, users can opt for more distributed systems in which data-acquisition and data-processing functions are more physically distinct, but where front-end acquisition modules offer greater channel density and analog-to-digital (A/D) or digital-to-analog (D/A) throughput. Both architectures are supported by subsystem vendors.
There are merits to both approaches. The first consolidates front- and back-end processing in a single, convenient package, using standardized commercial off-the-shelf (COTS) embedded technologies that are well supported and understood. DAQ cards can be combined with single-board computers with Intel or PowerPC architectures, together with multiprocessor digital signal processing cards equipped with field-programmable gate arrays (FPGAs) and general-purpose processors. Newer cards and chassis can be added as applications grow.
The distributed approach, on the other hand, enables front-end data acquisition, alignment, and digitization tasks to be maximized by dedicated, high-density resources yet to be sectioned off from the processing tasks as a “black box” capability. These front-end DAQ boxes – which can be ganged together and synchronized for greater channel count – can then “serve” digitized data via high-speed networks to inexpensive, general-purpose commercial desktop computers rather than embedded cards.
Unlike bus boards, standalone DAQ elements require no processors, operating systems, drivers, and programming. They are simply network-attached devices that are controlled over Ethernet via TCP/IP or UDP commands. This approach allows easier growth of front-end capability while reducing the buildup of components such as power supplies, integrated circuits, and chassis, associated with multiples of VME or VPX systems.
An example of the second approach is the GE Intelligent Platforms daqNet, a 1U-format acoustic front-end system that features 192 analog channels, digital control channels, and redundant Gigabit Ethernet ports, with analog-to-digital sampling/conversion frequencies of 625 kHz/channel at 24-bit resolution (maximum).
Sensor processing is a two-way street, however. If the sonar system is active as well as passive, transmit as well as receive processing chains are required. This reality means implementing D/A channels as embedded cards or dedicated resources. In the interest of stealth, target detection sonars may not transmit, but applications such as navigation will commonly use both transmit and receive functions. These systems also benefit from a multisensor approach that enhances the agility of the desired acoustic signal.
Whatever the choice of processing architecture, there’s an argument for going back to basics. If the objective of a sonar acquisition is to maximize the resolution, agility, speed, reliability, and overall target detection capability of the military platform – and to get the most complete picture possible of the operating environment – a multiple-sensor system is something to be considered.