Sampling Rate: How Much Speed is Necessary?
Signal speed is an obviously important criterion for data converters. For greater accuracy in rendering analog
signals, the data converter should support fast speeds in the sampling of incoming data. Following the guidance of
the Nyquist sampling theorem, the sampling rate should be higher than twice the highest frequency of the signal,
in order to generate an accurate reproduction of an analog signal in digital form. When evaluating analog IP, the
fastest signal that is targeted for sampling will determine how fast the sample rate must be. Under conventional
thinking, it would appear that faster is better. However, as we noted earlier, life is about trade-offs. So, more speed
also means higher power. The art is to strike a balance between just fast enough and just low enough power.
Data converters that support emerging high-speed communications protocols, such as WiGig (802.11ad), Long
Term Evolution (LTE), and LTE Advanced, require fast sampling rates. WiGig, for instance, runs on a 60GHz
spectrum, with a potential data throughput up to 7Gbps, or 2.16GHz bandwidth (a rate 10X higher than in
801.11n). As a result, the ADC needs to be 10X faster. Such capabilities would bring functions like wireless trans-
mission of uncompressed video to reality.
Resolution: a Look at Quality of Signal Sampling
How fast do you need to sample (as discussed above) and how many bits you select will determine how accurately
you represent your analog signal. This is resolution. The higher your resolution, the more accurate you will be
to represent your original analog signal. In other words, the combination of your sample rate and bit width will
determine the quality of the signal. When designing an SoC for an audio application, strive for as large a number
of sample rates (bits) as possible, to allow higher sensitivity to noise. For typical sensor applications, a sample rate
of six to seven bits can be sufficient.
Effective number of bits (ENOB) provides a holistic measurement of analog-to-digital (A to D) performance. This
metric assesses what the quality of the A to D conversion will be like in a real application, accounting for factors
including noise, temperature, and signals. An ideal ENOB is one that is as high as the number of bits in the ADC.
Architecture
There are a variety of ADC architecture types, each with varying sample rates, resolution, and power efficiency.
The successive approximation register (SAR) architecture is the most popular ADC architecture today, as it is very
fast and, with the reuse of each slice of ADC, it is also quite scalable. A single SAR data converter can have sample
rates from 100KSPS to 500MSPS (along with high power efficiency and medium-level resolution at 6 to 12 bits).
A parallel SAR converter (as shown in Figure 1) can boast sample rates from 500MSPS to 40GSPS, though with a
medium-level power efficiency and resolution (6 to 12 bits). By comparison, the Flash ADC architecture provides
the highest sample rates—100MSPS to 40GSPS—but it also comes with the lowest power efficiency and resolution
(three to six bits). Another popular ADC architecture, the sigma delta architecture, features high resolution (10 to
24 bits) but low sampling rate, at 10 SPS to 50 MSPS.
www.cadence.com
2
6 Key Considerations for Choosing the Right Analog IP