Page 16 - EngineerIT September 2021

AUDIO PROCESSING

Planning for success in real-time

acoustic processing

By David Katz, Infotainment systems architecture director

ow latency, real-time acoustic processing is a key factor priority SoC tasks are serviced. These spikes may occur, for
in many embedded processing applications, among them example, due to SoC-centric activities such as media rendering,
Lvoice pre-processing, speech recognition and active noise browsing or app execution on the system. Whenever the spikes
cancellation (ANC). As real-time performance requirements surpass 100% CPU loading, the SoC is no longer operating in
steadily increase within these application domains, developers real time, and this will result in audio dropouts.
need to adopt a strategic mindset to properly accommodate Audio DSPs, on the other hand, are architected for low
these needs. latency throughout the signal processing path, from sampled
Given the substantial performance offered by many larger audio input to composite (for example, audio + anti-noise) speaker
systems on chips, it may be tempting to simply load these devices output. L1 instruction and data SRAM, the single-cycle memory
with any additional tasks that arise, but it’s important to understand closest to the processor core, is ample enough to support many
that latency and determinism are critical elements that can easily processing algorithms without offloading intermediate data to
lead to major real-time system problems if not considered carefully. off-chip memory. Additionally, on-chip L2 memory (further from
This article will explore issues designers should consider when the core but still much faster access than off-chip DRAM) helps
choosing between an SoC and dedicated audio DSP in order to provide a buffer for intermediate data operations when L1 SRAM
avoid unpleasant surprises in their real-time acoustic systems. storage is exceeded. Finally, audio DSPs commonly run a real-
Low latency acoustic systems cover a wide range of time operating system (RTOS) that ensures incoming data can be
applications. For instance, in the automotive space alone, processed and sent out to its target destination before new input
low latency is critical for personal audio zones, road noise data arrives, thus ensuring that data buffers don’t overflow during
cancellation and in-car communication systems, to name a few. real-time operation.
With the emerging trend of vehicle electrification, RNC The actual latency in system boot — often measured by time-
becomes even more important because there’s no combustion to-audio availability — can also be an important metric, especially
engine generating noticeable noise. Therefore, the sounds in automotive systems where audible warnings must be broadcast
associated with the car-to-road interface become much more within a certain window from start-up. In the SoC world, where
noticeable and offensive. Reducing this noise not only creates it’s typical to have a lengthy boot sequence that involves bringing
a more comfortable riding experience, but it also reduces up the operating system for the entire device, it can be difficult or
driver fatigue. There are numerous challenges associated with impossible to meet this start-up requirement. On the other hand,
implementing a low latency acoustic system on an SoC as a stand-alone audio DSP running its own RTOs, unaffected by
opposed to on a dedicated audio DSP. These include issues other extraneous system priorities, can be optimised for fast boot
of latency, scalability, upgradeability, algorithm considerations, that comfortably satisfies the time-to-audio requirement.
hardware acceleration and customer support. Let’s examine each
of these in turn.
Latency
The issue of latency in real-time acoustic processing systems is
an important one. If the processor cannot keep up with the real-
time data movement and computational demands of the system,
unacceptable audio drops can result.
Typically, SoCs have small on-chip SRAMs and, accordingly,
must rely on cache for most local memory accesses. This
introduces non-deterministic availability of code and data, and
it also increases processing latency. For a real-time application
such as ANC, this alone can be a deal breaker. However, there’s
also the fact that SoCs run non-real-time operating systems
that manage heavy multi-tasking loads. This amplifies the non-
deterministic operating characteristic of the system, making it very
difficult to support relatively complex acoustic processing in a
multi-tasking environment.
Figure 1 shows a concrete example of an SoC running a real- Figure 1: Instantaneous CPU loads for a representative SoC
time audio processing load, where CPU loading spikes as higher running high audio memory processing in addition to other tasks. 1

EngineerIT | September 2021 | 14