"As the" Real World "" "" "" Digital World "and 1 and 0, the data converter is one of the key elements in modern signal processing. In the past 30 years, a large number of innovative technologies have emerged in the field of data conversion. These technologies not only boost from medical imaging to cellular communications, and then to consumer audio and video, performance improvement and architectural progress, but also play an important role in achieving new applications.
Continuous expansion of broadband communication and high performance imaging applications highlights special importance of high-speed data conversion: the converter can process signals with bandwidth ranges above 10 MHz to 1 GHz. People have achieved these high rates through a variety of converter architectures, each with its advantages. Switching on the simulation domain and digital domain in the high speed also puts some special challenges for signal integrity - not only the analog signal, but also the clock and data signals. Understanding these issues is not only important for component selection, but also affects the choice of overall system architecture.
In many technical fields, we are accustomed to associating technological progress and higher rates: from Ethernet to wireless LAN to cellular mobile network, the essence of data communication is constantly increasing data transmission rate. Through the advancement of clock rates, microprocessors, digital signal processors, and FPGA have been developed very quickly. These devices are mainly due to the narrowing etching processes, resulting in transistors with faster switching, smaller volume (and lower power consumption). These advances have created a environment where processing power and data bands are exponentially grow. These powerful digital engines bring a signal and data processing requirements that are also exponentially growing: from a static image to the video, to the bandwidth spectrum, whether it is wired or wireless, is it. A processor running a running clock rate may effectively process a signal with a bandwidth of 1 MHz to 10 MHz: The processor of the run clock rate Dumer GHz can process a signal with a bandwidth of hundreds of MHz.
Naturally, stronger processing capabilities, higher processing rates can result in faster data conversion: broadband signals expand their bandwidth (often achieved physical or regulatory spectrum limit), the imaging system seeks to improve pixel processing capacity per second To deal with higher resolution images more quickly. The system architecture is new to use to utilize the extremely high processing performance, which also have a trend of parallel processing, which may mean the need for multi-channel data converters.
Another important change on the architecture is to go to multi-carrier / multi-channel, and even the trend of software definition systems. Traditional analog intensive systems Complete many signal conditioning operations (filtering, amplification, frequency conversion) in the analog domain; digitally process the signal is digitally processed after sufficient preparation. An example is FM broadcast: the channel width of the given station is usually 200 kHz, and the FM band ranges from 88 MHz to 108 MHz. The conventional receiver converts the frequency of the target station to the intermediate frequency of 10.7 MHz, filter out all other channels and enlarge the signal to demodulation. The multi-carrier architecture digitizes the entire 20 MHz FM band and uses digital processing techniques to select and restore the target radio. Although multi-carrier schemes need to be complicated with more circuits, it has a large system advantage: the system can restore multiple radio stations at the same time, including the Edge Radio. If the design is properly designed, the multi-carrier system can even be reconfigured to support new standards (eg, a new HD station assigned to the radio sideband). The final goal of this method is to adopt a broadband digitizer that can accept all bands and a powerful processor that can restore any signals: this is the so-called software definition radio. Equivalent architectures in other fields - software definition meter, software definition cameras, etc. We can treat these as virtualized signal processing equivalents. Such a flexible architecture is possible to be a powerful digital processing technology and high speed, high performance data conversion technology.
Figure 2. Multi-carrier example
Bandwidth and dynamic range
Whether it is analog or digital signal processing, its basic dimensions are bandwidth and dynamic range - the two factors determine the amount of information that can be handled by the system. In the field of communications, Craward's theory uses these two dimensions to describe the basic theoretical limit of the amount of information that the information that can be carried by a communication channel, but its principle is suitable for multiple fields. For imaging systems, bandwidth determines the amount of pixels that can be processed at a given time, and the dynamic range determines the intensity or color range between the flexible detectable light source and the pixel saturation point.
The available bandwidth of the data converter has a basic theoretical limit set by the Nyquist sampling theory - in order to represent or process a signal with a bandwidth F, we need to use a data converter that runs at least 2 f in the running sampling rate (please note This Law applies to any sampling data system - analog or numbers apply). For actual systems, a certain amount of overcapacity can greatly simplify the design of the system, and therefore, a more typical value is 2.5 to 3 times the signal bandwidth. As mentioned earlier, increasing processing capabilities can improve the ability of the system to handle higher bandwidth, while cellular phones, cable systems, wired and wireless local area networks, image processing, and instrumentation and other systems are growing with higher bandwidth. This continuous improvement of bandwidth demand requires a higher sampling rate.
If the bandwidth is intuitive and easy to understand, then the dimension of dynamic range may be slightly smoked. In signal processing, the dynamic range represents the distribution range between the system can handle and the signal that does not have saturated or clipped, and the system can effectively capture the distribution range. We can consider two types of dynamic range: configurable dynamic ranges can be implemented by placing a programmable gain amplifier (PGA) before the low resolution analog to digital converter (ADC) (assuming to 12-bit configurable dynamic range, in one Place a 4-bit PGA before the 8-bit converter): When the gain is set to a low value, this configuration can capture a large signal without exceeding the converter. When the signal is in an hour, the PGA can be set to a high gain to enlarge the signal to the noise of the converter. The signal may be a signal of strong signal or signal, or it may be a bright or dimming pixel in the imaging system. This configurable dynamic range may be very effective for only a traditional signal processing architecture that only tries to restore a signal.
The instantaneous dynamic range is more powerful: In this configuration, the system has sufficient dynamic range, which can capture large signals without generating a clipping, while also recovering small signals - now, we may need a 14-bit converter . This principle is suitable for a variety of applications - restoring strong stations or weak radio signals, recovering mobile signals, or restoring the super brightness and overdage of images. While the system tends to use a more complex signal processing algorithm, the demand for dynamic range is also the tall to the boat. In this case, the system can handle more signals - if all signals have the same intensity, and need to handle twice the signal, it is necessary to increase the dynamic range of 3 dB (in the case where all other conditions are equal). It may be more importantly, as mentioned earlier, if the system needs to handle strong signals and weak signals at the same time, the incremental requirements of the dynamic range may be much larger.
Dynamic range different metrics
In digital signal processing, the key parameters of the dynamic range are the number of bits in the signal representation, or the call length: one 32-bit processor has more dynamic ranges than one 16-bit processor. Excessive signals will clip-bearing - this is a highly nonlinear operation that will destroy the integrity of most signals. Excessive signal - smaller than 1 LSB - will become unable to detect and lose. This limited resolution is often referred to as quantization errors, or quantizes noise, which may be an important factor under detectability.
Quantization noise is also a factor in a hybrid signal system, but there are several factors to determine the available dynamic range of the data converter, and each factor has its own dynamic range
The signal ratio (SNR)--converter is the ratio of the total noise of the frequency band. This noise may come from quantization noise (as described above), thermal noise (all in all reality systems) or other error items (such as jitter).
Static Nonlinearity - Differential Nonlinearity (DNL) and Point Nonlinearity (INL) - Measure Non-ideality of DC Transfer Function from Directance Enter the output end of the output (DNL usually determines the dynamics of the imaging system) scope).
Total harmonic distortion-static and dynamic nonlinearity generates homophones, which may effectively block other signals. THD usually limits the effective dynamic range of the audio system.
No spurious dynamic range (SFDR) - Considering the spectrum spur, whether it is second order or third-order harmonic clock, even 60 Hz "" 嗡 "" noise. Since the spectral sound or spur, the SFDR is used to represent a good indicator for the available dynamic range in many communication systems.
There are other technical specifications - in fact, each application may have its own effective dynamic range description. At the beginning, the resolution of the data converter is a good alternative indicator of its dynamic range, but it is very important to choose the correct technical specifications when truly decisive. The key principle is, the more you better. Although many systems can immediately realize that higher signal processing bandwidth is required, the demand for dynamic range may not be so intuitive, even if it is more demanding.
It is worth noting that although bandwidth and dynamic range are two main dimensions of signal processing, it is necessary to consider the third dimension, that is, efficiency: This helps us answer such a question: "" In order to achieve additional performance, I How much cost does it take? "We can see the cost from the purchase price, but for data converters and other electronic signal processing applications, a technique for pure, measure costs is power consumption. The higher the performance of the system - larger bandwidth or dynamic range - often consumes more electrical energy. With the advancement of technology, we have tried to reduce power consumption while increasing bandwidth and dynamic range.
main application
As mentioned earlier, each application has different requirements in the basic signal dimension, and in a given application, there may be a variety of performance. For example, a 1 million-pixel camera with a 10 megapixel camera. Figure 4 shows the bandwidth and dynamic ranges of some different applications. The upper half of the figure is generally referred to as a converter having a high speed-sampling rate of 25 MHz and or more, which can effectively process the bandwidth of 10 MHz or more.
It should be noted that this application map is not still constant. Existing applications may utilize new, higher performance techniques to enhance their functions - such as a HD camera or a higher resolution of 3D ultrasound equipment. In addition, a new application will emerge each year - a large part of the new application will be outside the outer edge of the performance boundary: benefited from a new combination of high-speed and high resolution. As a result, the edge of the converter continues to expand, just like the ripple in the pond.
At the same time, it should also be remembered that most applications require attention to power consumption: For portable / battery power supply applications, power consumption may be major technical restrictions, even the line power supply system, we also start discovering, signal processing components (simulation) Also, the number of power consumption will restrict the performance of the system in a given physical area.
Technical development trend and innovation - how to achieve ...
In view of these applications, the industry responded to this by continuous technological progress in continuous improvement of high-speed data converter performance. Technical Supply of High Speed Data Converters From the following factors:
Process technology: Moore Law and Data Converter - Semiconductor Industry has been obvious to the achievements of continuous promotion of digital processing performance, and its main drive factors are the huge progress made by the wafer processing process in the process of taking more space. The switch rate of the deep submicron CMOS transistors far exceeds its predecessors, allowing the controller, digital processor, and the RMS rate of the FPGA, with a number of steps of GHz. The mixed signal circuit like a data converter can also utilize these advances in the field of etching process, which borrows "" Moore Law "to achieve a higher rate - but for the mixed signal circuit, this is cost: More advanced etching processes have a continuous reduced trend. This means that the signal swing of the analog circuit is reduced, which increases the difficulty of maintaining analog signals over thermal noise flooring or more: a higher rate is obtained at a rate of consumer.
Architecture (this is not a data converter in the original era) - At the same time, in the past 20 years, the high-speed data converter architecture has also seen several wave innovations, which is higher than the amazing effect. Bandwidth, greater dynamic range made great contributions. Traditionally, there are a variety of architectures for high-speed analog-to-digital converters, including full parallel architecture (FOLDING), interleaved and pipeline architecture, which are still very popular today. Later, the architecture traditionally used for low-speed applications also joined high-speed application camps, including successive approximation registers (SAR) and - these architectures made original changes for high-speed applications. Each architecture has its own advantages and disadvantages: some applications generally determine the architecture based on these compromise. For high speed DACs, architectures are generally a switching current mode structure. However, such structures have many variants; the rate of switching capacitance structures is steadily increasing, and it is still very popular in some embedded high-speed applications.
Digital auxiliary method - For many years, high-speed data converter circuit technology has also achieved brilliant innovation achievements in the process and architecture. The calibration method has been a decades of history, playing a crucial role in compensating integrated circuit components and improving the dynamic range of circuits. Calibration has surpassed the category of static error correction, more and more for compensating for dynamic nonlinearity, including establishing errors and harmonic distortion.
In short, innovation in these areas greatly promotes the development of high-speed data conversion.
accomplish
Implementing a broadband mixed signal system not only to select the correct data converter - these systems may have strict requirements for other parts of the signal chain. Similarly, the challenge is a dynamic range implemented within a wide bandwidth range - making more signals into the digital domain, and take advantage of the processing power of the digital domain.
Broadband and Signal Conditioning - In a conventional single carrier system, signal conditioning is to eliminate useless signals as soon as possible, then enlarge the target signal. This often involves selective filtering and a narrowband system for the target signal fine-tuning. These fine-tuning circuits may be very available in achieving gain.Effect, and in some cases, frequency planning technology can be utilized to ensure that harmonics or other stranders are excluded. Broadband systems cannot use these narrowband technology, and broadband amplification may face huge challenges in these systems.
Data Interface - Traditional CMOS interface does not support significantly more than 100 MHz data rate - and low voltage differential swing (LVDS) data interface running speed up to 800 MHz to 1 GHz. For larger data rates, we can use multiple bus interfaces or use the SerDES interface. The modern data converter uses a speed up to 12.5 GSPS (specifications JESD204B standard) - a plurality of data channels can be used to support different combinations of resolution and rate in the converter interface. These interfaces themselves may be very complicated.
Clock Interface - The processing of high-speed signals may also be very difficult to process the quality of the clock used in the system. The jitter / error in the time domain converts the noise or error in the signal, as shown in FIG. When the processing rate is greater than 100 MHz, the clock jitter or phase noise may become a limiting factor for the converter available dynamic range. Digital clocks may not be able to take such systems and may need to use high-performance clocks.
Figure 5. Ways of clock error becomes signal error
in conclusion
The pace of going to widen wide signal and software definition system is constantly accelerating, the industry is constantly pushing new, emerges in building better, faster data converters, pushing the bandwidth, dynamic range, and efficacy three dimensions push new steps.
Reprinted from Verid Electronic Market Network. "
Our other product:
