Maximizing the Benefits of the Latest Generation of Intelligent Sensors

By Matthew Hoare, Field Applications Engineer, Future Electronics (UK)

Introduction The requirement to sense environmental and other parameters from the outside world is as old as electronics itself. This has always been achieved in the past by using materials the characteristics of which change in some measurable way in response to changes in a physical parameter such as heat or movement.

These materials often react in ways we would like to control, but cannot. This is because they react in a ‘dumb’ way to natural phenomena: for example, devices used to measure light also often change in response to heat. Further, dumb sensors’ physical responses to environmental phenomena are generally not linear.

Two major improvements in the design and manufacture of sensors have made them easier and more convenient to use. First, new materials and signal pre-processing provide a more usable output.

Second, modern fabrication technology allows devices to be placed in the same package as the sensor. The resulting product is better described as a highly integrated sensing solution rather than a sensor. With on-board signal conditioning, the output, in either an analog or digital state, is available for direct interfacing to the system.

Crucially, this eliminates the need for additional external conditioning circuitry in most cases. Traditionally, the responsibility for designing an effective sensor conditioning circuit fell on the user, not the sensor manufacturer – albeit that the manufacturer supplied guidelines and application notes to help the design engineer. So the performance of the sensor on criteria such as sensitivity, noise immunity and linearity would ultimately become more a function of the quality of the design surrounding the sensor than of the sensor itself.

In intelligent sensors, signal conditioning is done by the device manufacturer. Since the datasheet specifications promise a standard, linear output, complicated calibration routines can now be avoided; some devices even offer self-test functions to establish sensor integrity before and during operation.

 

The Benefits of Integration and Intelligence

So what difference does the use of an intelligent sensor make to the design engineer? Most obviously, an intelligent sensor can save space, because of the vastly decreased requirement for external signal conditioning circuitry.

Using intelligent sensors can also reduce design time. In many cases, the output from the sensor can be directly interfaced to a microcontroller without the need for any signal conditioning. Using dumb sensors, by contrast, requires the integration of sensitive parts such as expensive instrumentation amplifiers.

 

Intelligent Sensors in Practice

As things stand, there are many more ‘dumb’ sensors and sensor types than there are ‘intelligent’ replacements. But where an intelligent sensor can be used, the choices the design engineer can make alter dramatically. One sensor category well populated with ‘intelligent’ devices is accelerometers. Conventional analog-output accelerometers are generally used to provide a constant output to a system controller. A processing overhead is permanently placed on the controller, so it has to be specified appropriately to provide enough processing bandwidth.

But what if all the application requires is an occasional interrupt, alerting the system to special conditions? In this case, it would be wasteful to be constantly polling the controller with outputs from the accelerometer. The latest digitaloutput accelerometers from companies such as Freescale Semiconductor can be programmed to send an output only when certain conditions are met. Since they connect to the system controller over a serial link, it is easy to program them by writing configuration data to their internal registers.

This allows the design engineer to segregate the processing of the acceleration signal outside the microcontroller, leaving the microcontroller free to manage the appropriate response to this signal. This makes the system’s architecture simpler, reduces processing overhead on the microcontroller, allows for a lower-specification microcontroller.

The same effect can be seen in any sensing application in which intelligent sensors are able to be used. Touch sensors are a good example. Manufacturers of consumer goods such as kitchen equipment and media players are eager to use touch sensing technology to replace traditional keys and buttons in their user interfaces.

But the use of a traditional dumb touch sensor places a huge signal conditioning and signal processing overhead on the system. It is also bulky and expensive to manufacture because of the need for multiple signal conditioning components such as operating amplifiers, filters and comparators.

But a device such as the CapSense PSoC (Programmable System-on-Chip) from Cypress Semiconductor operates as an intelligent touch sensing solution. In fact, the PSoC device is not a ‘sensor’ in the normal meaning of the word at all – it is a programmable mixed-signal array with an embedded microcontroller core. This device is able to absorb the analog and digital circuitry in touch sensing systems, and provide the appropriate outputs for processing by the on-board controller core.

Another example can be found in the arena of light sensors. Silicon devices generally respond very differently to light than a human eye does. An intelligent light sensor from Avago Technologies, however – the APDS-9002 – uses a carefully designed transistor to mirror very closely the response of the human eye.

This means that its output is immediately usable in lighting applications in which the product must respond to light in the same way as a user does. Using any conventional light sensor, the design engineer would have to design signal conditioning and processing circuitry in order to filter out the ‘useless’ wavelengths that the human eye does not respond to.

Figure 1 shows this detector interfaced to an analog-to-digital converter (ADC) input on a microcontroller that controls a light source.

 

 

Conclusion

Before the advent of intelligent sensors, the design task started with the sensor itself - the thermocouple, the strain gauge, etc. and then started the long, laborious task of calibrating, conditioning and processing the output and interfacing it to a system controller. Overheads placed on the controller by the sensor often forced the adoption of a high specification for the controller that was not required by any other part of the system. It was a highly fragmented, difficult and expensive approach.

Now, in many instances, the engineer can use a single for the signal conditioning and processing circuitry, and that provides a serial interface to a controller. So now the design engineer’s choice of MCU can be driven by the real requirements of the design – as it should be – rather than by the overhead created by the sensor(s). Indeed, sensing can begin to be treated as a ‘black box’ function dropped on to the board, and scarce analog expertise redirected to where it is most useful.

Further, the customization of such devices over a serial data link opens a whole new world of flexibility. Alarm thresholds and other preset parameters are no longer static; they can be dynamically modified. Power requirements can be reduced by instructing the device to “sleep” until a particular condition is encountered or the device is “woken” by command.

The only thing that should now delay the adoption of intelligent sensors is lack of availability of devices. Future Electronics expects that problem to be solved in the coming years as silicon and MEMS device manufacturers implement their product development roadmaps.