Mixed-signal solutions are an inevitable trend. We want our electrical devices to interact with us. Our car doors should unlock as we approach. Our computer should turn itself on as we sit down in front of it—or at least enhance its screensaver to show how excited it is that we are returning. The screens on our phones should dim to save power when we are in low-light conditions. And, thermostats should keep our homes in a comfortable range of temperatures.
Humans are analog, meaning they are infinitely variable, while electronics communicate with their own language of voltages and currents. In the electronic realm, the initial readings from sensors (see the figure) are typically conditioned by input circuits, converted to digital signals, processed, converted back to analog signals, and properly conditioned to interact with us again.
The basic block diagram has remained relatively constant, absorbing changes in operating frequency, channel bandwidth, processing power, and technology. Before starting a discussion of system-level issues, let’s look at each of the blocks (from left to right) in a little more detail.
Steps In The Mixed-Signal Chain
Sensors bridge the gap between the physical world and the electrical one. All of the technology examples given above need sensors. They can be RF, proximity, ambient light, or temperature sensors. Temperature and pressure sensors are two of the most popular.
In the industrial market, there is growing interest in using accurate flow sensors. Proximity sensors bounce ultraviolet light to signal when something comes close. Ambient light sensors are a vital sentry in portable devices with any kind of screen, saving power and prolonging battery life.
The next step in the mixed-signal chain is the input amplifier, which must properly accept the signal from the sensor without loading or distorting the signal. The wide variety of sensors suggests that a broad range of amplifiers may be needed to pair correctly with the sensor. Buzzwords include the instrumentation amplifier, chopper-stabilized amplifier, low-noise amplifier, and input bias cancellation amplifier. Each system solution has different requirements and should be matched with the amplifier optimized for its needs.
If filtering is needed (in most cases, it is needed), it may be wrapped around this amplifier or added in series to the system. Filtering is an art all by itself. A handful of programs is available online to help you design the circuit your system requires. One new tool that combines practical circuit techniques with the expertise of years in filter design is Intersil’s iSim Active Filter Designer (http://web.transim.com/iSim/). It’s powerful and it’s free.
The analog-to-digital converter (ADC) is the most crucial selection in the signal chain and is often one of the first blocks chosen. Your choice will determine the number of bits in the system, the speed of the system, and one of the main power-consuming blocks. Choices of topology create different tradeoffs, depending on what’s required.
The mere mention of a few topology names is enough to send an ordinary person running in another direction: delta-sigma, pipelined, successive approximation, flash, and integrating. The most daunting topology to understand seems to be the delta-sigma ADC. This oversampling device typically operates at low frequencies, although a few break the megahertz barrier. Delta-sigmas offer the highest resolution of the pack, like 24 bits, in applications for weighing, temperature control, and instrumentation.
Continue to next page
Successive approximation ADCs, called SARs because they use a successive approximation register, offer a tradeoff of resolution and speed. They operate in a range from 1 kHz to a couple of megahertz and provide mid-range precision. An integrating converter is slow, taking time to average the input signal. That makes the integrating converter great for dc measurements since it can filter power-supply noise (50 or 60 Hz).
The opposite side of the tradeoff is the flash converter, which can operate over 1 GHz but is limited to about 10 bits of accuracy. To achieve this speed, the flash converter burns lots of power because it calculates the conversion in one step.
If you choose to back off on the power and use a two-stage solution, it is called a multi-stage converter. If you back off even more to three or more stages, then you’d typically call it a pipelined converter. Pipelined ADCs run from around 500 kHz to 500 MHz and can provide resolutions up to 16 bits.
A microcontroller or an FPGA commonly provides the digital signal processing. These blocks can be chosen for their resolution, speed, footprint (size), and power consumption. Most of the time, designers choose these blocks because they have used them before and are familiar with them.
Designers also look for compliance with the signals generated by the ADC or required by the digital-to-analog converter (DAC). In fact, many of these parts are available with built-in ADCs and DACs. These built-in converters are sufficient for simple solutions, but may not provide the performance available in discrete ADC and DAC packages.
If you do decide to use a discrete DAC package, you again have an array of speed, resolution, power, and performance issues, just like the ADCs. One delta-sigma topology for DACs oversamples similarly to its high-quality ADC counterpart. The R-2R and the resistor string are two simpler DACs. While the R-2R configuration relies on matching, the resistor string can guarantee monotonicity. (Each increase in input voltage provides a correlated increase in output voltage.)
The output amplifier buffers the DAC from whatever load needs to be driven. In some cases, this amplifier must also perform current-to-voltage conversion, depending on the output signal of the DAC. Filtering may be required in this stage, like the requirements of the input half.
To keep the discussion relatively simple, we haven’t addressed the other parts of the system. Remember that each block needs clean power rails, and many need voltages either from a voltage reference or digital potentiometer. Also, good power-supply bypassing and excellent layout techniques will increase the chances of success.
Mixed-Signal System Issues
Now that we’ve investigated each block, let’s see how we bring it together. Cost motivates many design choices. If cost is your primary concern, then using a microcontroller with a built-in ADC and DAC may be your best option. The next option is for the designers who prefer to choose ADCs and DACs in operating pairs—with similar characteristics and usually from the same vendor. Demonstration boards are typically available to reduce debugging time.
Many designers are pushing for the integration of both sides of the signal chain. Technology has advanced enough to allow the op amp and some filtering to be co-packaged, if not included in the same die as the converter. This has an obvious advantage in footprint and ease of use. It seems destined in commodity products. If system designers could save space by selecting an op amp-converter combination, why wouldn’t they want to integrate the entire mixed-signal path in one path? Nothing. They would! That’s why application-specific standard products (ASSPs) are widely used.
But why might you choose to de-integrate? There are many reasons. The product or application might be new and not defined to a level where it makes sense to fund an ASSP design. Second, there is no flexibility. What if you wanted to upgrade the filter to a higher order to compensate for a new, powerful interferer? What if you wanted to try a new configuration of converter? What if you had to build a prototype quickly? What if a small design change would allow your system design to be more flexible and accommodate more applications and more customers?
Finally, what if you wanted lower power? Many converters need 1.8-V power supplies while those op amps may need 3.3 V or 5 V to achieve the dynamic range/common-mode rejection ratio (CMRR) needed by the system. The discrete solution offers more options and optimizations. Many experienced system designers who have mastered board layout and supply bypassing opt for the discrete solution to retain their options.
Integration makes our lives easier with system designs, as long as it doesn’t limit our capability at the same time.