Analytics Drives Signal Processing

This post sponsored by Texas Instruments.

Mixed signal designs span a wide realm of applications, from a meager few MHz 8-bit processor with A/Ds and D/As bolted on, to high-speed parallel multi-channel sampling and multi-core processors running in the GHz ranges. And each processor’s architecture serves the designers’ needs. You would not want to use a sophisticated DSP processor in a waffle iron; the iron’s needs are just not that critical. Then again, you also probably wouldn’t want a little 4-bit processor making decisions about keeping your heart beating.

But in most every case, mixed signal designs involve reading real-world data, using these values in a rudimentary algorithm, and outputting some control or feedback level also in the analog realm. Mixed signal technology effectively replaces analog components with digital control systems. And let’s not underestimate the effectiveness of processing analog signals in the digital. It has given us many of the devices in use today.

But processing is getting more sophisticated and integrated. A mere peripheral function today would overextend the entire resource of yesterday’s processor solutions. Look at all the processors and simultaneous functions occurring in a modern-day automobile, for example. Each modern car uses multiple processors on a network performing “sense-and-send” as well as “process-and-act” functions.

Processors are being asked to do more, and to make higher-level decisions. Consider all the processing that occurs every day in the palm of your hand with a typical smartphone or tablet. It is clear that higher-functioning integration is needed, and Texas Instruments’ MP3 encoder/decoder with capacitive-touch interface, OLED display & SD card mass storage meets that need. This device combines several highly integrated functions that designers may typically need in a present-day handheld device and includes fully implemented mixed-signal systems that use internal analytics—saving design and debug time.

In addition to audio-in and -out streaming, the capacitive touch sensor interface shows how to detect and analyze functions for scroll-wheel proximity sensors, and multi-touch and gesture recognition. These functions require more than just detect-and-act personalities. Higher-level analytic processing is fundamental as well.

Vision is another area where mere sense and stream is not enough. Gesture detection, recognition, as well as image clarification and filtering are a part of many next-generation drawing boards. In addition, DSP cores as well as RISC and CISC processors all play a critical role.

Scalability is a factor as well. Once algorithms and functions are refined, the ability to integrate that functionality into next-generation designs is key. Texas Instruments KeyStone Architecture allows a chip’s infrastructure to share processing and software development resources. It also scales well (Figure 1).

Figure 1: The KeyStone Architecture allows shared-chip infrastructure resources for multi-core designs. (Source: Texas Instruments)

This can be demonstrated today using the Texas Instruments Ultra-Low Power Vision Analytics Module. Partnering with the SNAP sensor module, it uses the Texas Instruments 200 MHz low-power C5517 single-core DSP processor with dual MACs, FFT accelerator, and on-chip 320 KB of image analysis memory for video and image acquisition. This allows the on-chip processing of many functions to reduce cost and size. Schematics, design guides, design files, and BOMs are also available online.

A common subset of these and many other emerging applications involves the use of higher-level analytics. Both the highly integrated functional blocks and techniques may be new to many designers and fortunately, TI supplies a multitude of reference designs and demo kits to help speed this along. Another helpful resource is the Embedded Analytics Paper, which highlights examples such as advanced automotive systems, facial recognition systems, gestures, and more.

Another good resource is the Demystifying DSP Programming paper that shows how intelligent architectures woven together with good fairly high-level mixed-signal performance can result in reliable systems that make more advanced decisions. Analytics is the next layer above high-speed processors and mixed-signal and DSP functionality.

TMS320C5534
For more information about this product, click here.
To request a sample, click here.
To buy now at Mouser, click here.

TMS320C6671
For more information about this product, click here.
To request a sample, click here.
To buy now at Mouser, click here.

TMS320C6474
For more information about this product, click here.
To request a sample, click here.
To buy now at Mouser, click here.

OMAP-L138
For more information about this product, click here.
To request a sample, click here.
To buy now at Mouser, click here.

TMS320C6748
For more information about this product, click here.
To request a sample, click here.
To buy now at Mouser, click here.

TMS320VC5416
For more information about this product, click here.
To request a sample, click here.
To buy now at Mouser, click here.

TMS320C6727B
For more information about this product, click here.
To request a sample, click here.
To buy now at Mouser, click here.

Author: Jon Gabay