Image
image
image
image


Design Articles

New System Partitioning

Using new generation low-power audio converters with signal processing capabilities to partition your system can reduce time-to-market

By Luca Cacioli, Texas Instruments

The product lifecycle of the consumer electronics industry is directly related to consumers’ tastes in electronic products, which evolves quickly. What was a must-have item a few years ago easily can be a product overstocked on retailers’ shelves or in manufacturers’ or distributors’ warehouses. In the current economic environment where companies are carefully scrutinizing the return on investment (ROI) of research and development (R&D) and marketing, the successful companies are those that match or even forecast this swift change.

One possible strategy to combat this ever-shrinking window of opportunity is to build products on flexible platforms. If a product based on such a platform fails and the platform is not the issue, companies can create new products by reusing the same platform, thereby reducing developments time and cost.

Cell phones, portable navigation devices and portable media players can greatly benefit from this idea because they are subject to the mood of consumers who like to buy the newest electronic device of the year. Today, these devices are built around one main host processor that acts as the brain of the device and runs all the software needed to make the product work, such as audio and video streaming or Internet connections. Implementing a novel system partitioning, in such a way that not everything is controlled by these processors, can help reduce the development time.

The recent introduction of very low-power audio converters with embedded miniDSPs and powerful graphical development environments is a step in that direction. The converters can run all the audio processing algorithms the product needs and the graphical nature of their software make them easy to program. Moreover, companies that have audio converters with miniDSPs often offer audio algorithms with these devices. These algorithms are fully tested and usually royalty-free, eliminating the need for customers to write software or pay royalties to third-party developers.

It’s logical to offload the audio processing onto these converters. Let’s consider the case of a failing product whose audio is fine as is. In this situation, if an audio converter with a miniDSP was used, audio algorithms would not need to be rewritten and the audio development effort would be virtually nonexistent. New products would then be able to reuse the same audio chain.

Now let’s analyze one of the most time consuming tasks in developing a new product: regression tests. New consumer electronic products are loaded with software and run a variety of applications. These complicated devices are hard to test, and even harder to fix once testing uncovers a flaw. This new method of partitioning the system can reduce the regression complexity and decrease the time to complete it. However, this is not just about speed, but also about quality. With less new code to write engineers can focus on other tasks, thus potentially yielding better results.

figure 1
Figure 1: Power consumption per MIPS of two devices,
a low-power general purpose DSP, and the low-power
audio miniDSP integrated in audio converters.

Not every audio converter with miniDSP is necessarily used in battery-powered applications, but those that are make very efficient use of their resources. While they may not be the lowest power DSPs in the industry on a mW per million instructions per second (mW/MIPS) basis, they may use less power than DSPs to run audio algorithms. This is because they are designed to run only audio algorithms and very efficiently, and at the expense of less MIPS. Figure 1 depicts a hypothetical example.

Following is a hypothetical example:
Power consumption per MIPS:

  • DSP                 0.15mW
  • miniDSP           0.2mW

From this example, we can see that the general purpose DSP uses less power on a MIPS basis. However, the outcome changes when we look at the actual power consumed to run an algorithm. Let’s consider algorithm A (it could be anything from bass boost, to dynamic range compression, to noise reduction):

MIPS needed to run algorithm A:

  • DSP                 20
  • miniDSP           10

By looking at Figure 1, we can conclude that the total power consumed to run algorithm ‘A’ is:

  • DSP                 3mW
  • miniDSP           2mW

Audio codecs with miniDSP are audio processors that also integrate a data conversion section. Some of them can operate at supply voltages as low as 1.26V, thus, reducing the power consumed by the device.

The combination of low-power signal processing and ultra-low power data conversion in one chip offers significant power-saving opportunities in conventional system architectures that include a host processor and codec. The total power consumption to run algorithm A and play it through headphones can be less than 10mW. Figure 2 shows where an audio CODEC fits into a system.

figure 2
Figure 2: High-level system block diagram

Now, let’s discuss support. In the current system partitioning model, what happens when someone designing a product encounters difficulty with audio processing? Who will come to the rescue when audio processing is just a small part of the system software (i.e., run on the host processor)? On the other hand, let’s consider a scenario where the audio converter is tasked to run audio algorithms. If the need arises, the designer can access engineers who live and breathe audio and will be able to thoroughly address any audio-related query.

Some of the audio converters from Texas Instruments integrate a miniDSP. These devices are capable of running advanced audio processing algorithms. They enable engineers to rethink their systems’ partitioning and, therefore, to take advantage of the benefits that come with it. Texas Instruments has a complete family of devices with miniDSPs that are supported by a graphical development environment called PurePathTM Studio. They differ mainly by the analog interfaces (ADCs and DACs) between the miniDSP and the real world. For example, one such device is the TLV320AIC3254, a low-power efficient stereo audio codec.
 
In summary, running audio processing algorithms on audio codecs allows manufacturers to customize their products very quickly and with lower risks. These devices are designed to run audio algorithms and do so very efficiently. Algorithms are available and, in most cases, are royalty-free. Finally, by interacting with audio experts, engineers and managers can more easily create products with unique audio features. Being able to quickly satisfy end-users’ needs with differentiated products increases the likelihood of commanding price premiums and increasing revenues and profits.

References

For more information on the TLV320AIC3254 and the PurePathTM Studio, visit: www.ti.com/tlv320aic3254-ca.

About the Author

Luca Cacioli

Luca Cacioli is the Portable Audio Marketing Manager at Texas Instruments. He received his BSEE/MSEE from Universita’ di Firenze, Florence, Italy, and his MBA from Southern Methodist University (SMU), Dallas, Texas. Luca enjoys soccer and traveling (…and he does a lot of both). He can be reached at ti_lucacacioli@list.ti.com.

Texas Instruments Inc.
Dallas, TX
(800) 336-5236
www.ti.com



Bookmark and Share


Insert your comment

Author Name(required):

Author Web Site:

Author email address(required):

Comment(required):

Please Introduce Secure Code:


image
image