How to Implement Flexible RF Sampling Architectures Using High-Speed Data Converters
[Copy link]
How high-speed data converters enable flexible RF sampling architecturesMarc Stackler, Product Applications Engineer, Asia Pacific, Teledyne e2v Semiconductor, Asia Pacific For many years, digital transceivers have been used in many types of applications, including terrestrial cellular networks, satellite communications and radar-based surveillance, earth observation and monitoring. Their performance directly affects the efficiency and system cost of new 5G mobile networks. In the past, transceiver system engineers used IF architectures for these applications. Now, recent advances in high-speed data converters have made possible new architectures based on RF sampling. Compared with previous solutions, this has many advantages at the system level. These advantages include not only "SWAP-C" - reducing size, weight, power and cost, but also shortening time to market and the flexibility brought by software-defined radio (SDR) and software-defined microwave (SDM). This allows engineers to use the same hardware system that supports multiple configurations and requirements in different applications. Before discussing how the latest generation of high speed data converters achieves these advantages, let’s take a look at two different transceiver system architectures. IF Architectures IF architectures require the addition of specific RF hardware to generate the RF frequency through one or more IF stages. These stages are called upconverters on the transmit (TX) side (which converts low frequencies to high frequencies) and downconverters on the receive (RX) side (which converts high frequencies to low frequencies). Figure 1 shows the two-stage frequency conversion architecture. Figure1. The upconverter stage consists of a mixer, whose frequency is provided by the local oscillator (LO), to accomplish the frequency conversion. This is followed by a filter stage to remove images introduced by mixing or amplification. Figure 2 shows the sequential conversion stages of this two-stage frequency conversion example. Frequency images must be carefully handled to avoid performance degradation due to aliasing and distortion. We will not discuss this in detail here. Figure2. The link of two-stage frequency conversion On the transmitter side, the first upconverter converts the baseband or first Nyquist domain signal to an intermediate frequency (IF1), and the second upconverter converts the IF1 signal to an RF signal. On the receiver side, the process is exactly the same (RF signal is converted to IF1, and then the IF1 signal is converted to baseband or first Nyquist domain signal), and the ADC then converts the analog signal to a digital signal for digital processing for demodulation. The number of frequency converters and the value of the IF are different for different applications, and the implementation on the transmitter and receiver side may not be the same. The IF architecture was invented during World War I and has been widely used since then, mainly because it was the only solution for digitally processing RF signals. Today, the main advantage of this solution is that it can provide an interface between high-frequency RF signals and data converters that only support baseband frequency conversion. Since data converters have been limited to low frequencies and digital domain data conversion for many years, people have been forced to use specific analog schemes to process the RF spectrum and digitize it. The main drawback of this approach is the increase in the amount of RF hardware, which reduces SWAP-C and performance. Another drawback is the lack of flexibility because the IF is already determined by the LO frequency and the input/output frequency of the data converter. This drawback is difficult to detect without comparing it to an alternative architecture. RF Sampling Architecture The RF sampling architecture converts the RF frequency directly to the digital domain without any upconverter or downconverter. It uses a wideband data converter to recover the RF signal (as shown in Figure 3). 393667 BlinkMacSystemFont, "]Figure 3. RF sampling architecture system’s TX and RX sides RF sampling architectures require analog-to-digital converters (ADCs) and digital-to-analog converters (DACs) that can directly convert signals in high Nyquist zones. A Nyquist zone is a frequency band defined by the Shannon-Nyquist sampling theorem with a width equal to half the data converter’s sampling rate and can be expressed as [/2;(+1)/2], where k is an integer. Figure 4 shows how a signal can be sampled directly in the fourth Nyquist zone. Figure4. RF Sampling in the Fourth Nyquist Zone RF sampling is key to minimizing the amount of processing in the analog domain. Moving more processing into the digital domain increases system flexibility—enabling hardware reuse across multiple platforms, reducing time to market and certification costs, and lowering risk. In addition to increased flexibility, RF sampling architectures reduce cost and power by removing analog frequency converters. After decades of development, system engineers now have good reasons to adopt RF sampling architectures. To date, the bottleneck for applications using RF sampling architectures has been the capabilities of the data converters. Today’s high-speed data converters are fully capable of meeting the requirements of RF sampling architectures up to C-band and even X-band, but performance is not adequate at higher bands, such as K-band, E-band, and V-band, which will be used by future 5G backhaul systems. Table 1 gives a high-level comparison of IF sampling and RF sampling. RF sampling has many advantages, but it also brings new challenges and requirements, especially for data converters. Table1. High-Level Comparison of IF andRF Sampling The Role of Data ConvertersThe performance of today’s transceivers is often limited by the data converter, which plays a key role at the interface between the digital and analog domains. Regardless of the architecture chosen, the performance and specifications of the ADC and DAC must be considered. However, RF sampling architectures require additional attention to certain key specifications, including the analog bandwidth of the data converter and the output mode of the DAC. Before explaining why these parameters are so important for this architecture and how they affect system performance, it is useful to review how the Shannon-Nyquist sampling theorem theoretically defines this implementation. The Shannon-Nyquist sampling theorem states that the sampled signal can be reconstructed when the sampling rate is at least twice the total bandwidth of the sampled signal (i.e., ≥2, where =()). This theorem is often stated as a version for the first Nyquist zone, i.e., ≥2. This is sufficient for baseband systems, but in order to achieve high Nyquist zone RF sampling, the full version of the theorem must be understood. The aliasing effect causes each odd-numbered Nyquist zone to produce a signal with the same information, and each even-numbered Nyquist zone to produce a signal with opposite information, as shown in Figure 5. By performing anti-aliasing filtering in the Nyquist zone of interest, RF sampling can focus on a single signal that contains useful information. [ font=-apple-system-font, BlinkMacSystemFont, "]Picture5. Aliasing principle The first additional specification to consider for RF sampling is the analog bandwidth of the ADC and DAC. The bandwidth of the device will have an effect similar to that of a low-pass filter, limiting the signal frequencies that can be effectively converted with a specified accuracy. Figure 6 shows a simple example where after the second Nyquist zone, the signal is significantly attenuated due to bandwidth limitations.
Figure6. Example of the effect of the analog bandwidth of the ADC/DAC on RF sampling capability Additionally, the total bandwidth of the analog front end must be considered. The amplifiers and filters in the ADC and DAC analog front ends affect the total bandwidth that can be restored by the transceiver. Bandwidth is not the only parameter that affects direct RF sampling performance. Performance can vary greatly depending on the process and architecture of the data converter. For example, some ADCs and DACs use CMOS processes with bandwidths exceeding 6 GHz, but performance degrades significantly starting at 3-5 GHz. This is one of the main reasons for using bipolar and BiCMOS processes to manufacture ADCs and DACs, as these processes guarantee good performance even when the conversion frequency is higher than the converter bandwidth. Of course, at high output frequencies, bandwidth has a significant impact on the output power generated, thus limiting the performance of the system. However, for applications that do not require very high output power, devices that can generate high-frequency signals above the nominal analog bandwidth can be used for RF sampling. When selecting a data converter for RF sampling, both bandwidth and high-frequency performance must be considered."]Another DAC parameter that affects the performance of RF sampling is the output mode, or more precisely, the output power it produces. DACs can generate output signals in different output modes. For example, the latest generation of DACs offer four different output modes: Non-Return to Zero (NRZ) – The most common output mode, outputting the sampled value at full sampling period. Return to Zero (RTZ) – This output mode is also very common, outputting sample values in half of the sampling period and outputting 0 in the other half of the period. Narrow Return to Zero Mode (NRTZ) – This output mode provides a flexible solution between NRZ mode and RTZ mode, outputting sample values in a certain section (X) of the sampling period and outputting 0 in the part before and after (1-X)/2. RF Mode (RF) – This output mode is mainly oriented towards RF sampling, outputting the sampled value in half of the sampling period and the opposite value in the other half of the period. The effect of the output mode on the output power can be more easily understood from the frequency response of these modes. The appropriate mode should be selected based on the Nyquist zone of interest to maximize the output power. That is, the mode should be selected based on the sampling rate and the generated RF frequency. For example, if the sampling rate is 6Gsps and the output frequency is in the C-band (4-8GHz), the RF mode has the maximum output power in the second or third Nyquist zone, with only a 5dB impact on the output power. The frequency response of the output mode depends on the sampling rate. A DAC operating at a 12Gsps sampling rate and an output frequency of C-band is at most in the second Nyquist zone, where the NRZ mode is most suitable. For data converters used in RF sampling, in addition to the conventional requirements, they must be able to achieve sufficient output power and performance indicators at the target RF frequency, including analog bandwidth, dynamic performance and output mode (applicable to DAC only). Once the appropriate data converter is selected, the RF sampling architecture can be applied to the system. Architecture Comparison Using the LatestADCsTo highlight the advantages of RF sampling over IF architectures, Teledyne e2v’s latest generation of ADCs is available in three different configurations: a two-stage IF architecture, a single-stage IF architecture, and an RF sampling architecture. Table 2 shows the power consumption and noise of the three different configurations. As can be seen, removing each downconverter stage significantly reduces the system power consumption, from dual downconverters to RF sampling architectures by about 25%. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2 shows the power consumption and noise of the three different configurations. As can be seen, removing each downconverter stage significantly reduces the system power consumption, from dual downconverters to RF sampling architectures by about 25%. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Performance Comparison of IF and RF Sampling Architectures Using Latest GenerationADCs
While IF architectures have a long history in various forms, today’s advances in data converters have made them less and less suitable for RF applications. RF sampling architectures have a clear SWAP-C advantage and are evolving with the development of new generations of high speed, high bandwidth data converters. Advances in data converter technology will result in higher performance and support for direct RF sampling at higher frequencies. 51)]· Non-Return-to-Zero (NRZ) – The most common output mode, outputting sample values in the full sampling period· Return-to-Zero (RTZ) – This output mode is also very common, outputting sample values in half a sampling period and 0 in the other half period · Narrow Return-to-Zero (NRTZ) – This output mode provides a flexible solution between NRZ and RTZ modes, outputting sample values in a certain part (X) of the sampling period and outputting 0 in the part before and after (1-X)/2 · RF Mode (RF) – This output mode is mainly for RF sampling, outputting sample values in half of the sampling period and outputting the opposite value in the other half of the period "]The effect of the output mode on output power can be more easily understood from the frequency responses of these modes. The appropriate mode should be selected based on the Nyquist zone of interest to maximize output power. That is, the mode should be selected based on the sampling rate and the resulting RF frequency. For example, at a sampling rate of 6Gsps and an output frequency in the C-band (4-8GHz), the output power of the RF mode is maximized in the second or third Nyquist zone, with only a 5dB effect on the output power. The frequency response of the output mode depends on the sampling rate. A DAC operating at a sampling rate of 12Gsps and an output frequency in the C-band would be at most in the second Nyquist zone, where the NRZ mode would be most suitable. For data converters used in RF sampling, in addition to the normal requirements, they must be able to achieve sufficient output power and performance indicators at the target RF frequency, including analog bandwidth, dynamic performance, and output mode (applicable to DAC only). Once the appropriate data converter is selected, the RF sampling architecture can be applied in the system. Architecture comparison using the latestADC[color=rgb(51, 51, To highlight the advantages of RF sampling over IF architectures, Teledyne e2v’s latest generation of ADCs is available in three different configurations: a two-stage IF architecture, a single-stage IF architecture, and an RF sampling architecture. Table 2 shows the power consumption and noise of the three different configurations. It can be seen that removing each downconverter stage significantly reduces the power consumption of the system, by about 25% from dual downconverters to RF sampling architectures. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2. Performance Comparison of IF and RF Sampling Architectures Using Latest Generation ADCs 393672 "]While IF architectures in various forms have a long history, today’s advances in data converters have made this architecture less and less suitable for RF applications. RF sampling architectures have clear SWAP-C advantages and are evolving with the development of new generations of high-speed, high-bandwidth data converters. Advances in data converter technology will result in higher performance and support for higher frequency direct RF sampling. 51)]· Non-Return-to-Zero (NRZ) – The most common output mode, outputting sample values in the full sampling period· Return-to-Zero (RTZ) – This output mode is also very common, outputting sample values in half a sampling period and 0 in the other half period · Narrow Return-to-Zero (NRTZ) – This output mode provides a flexible solution between NRZ and RTZ modes, outputting sample values in a certain part (X) of the sampling period and outputting 0 in the part before and after (1-X)/2 · RF Mode (RF) – This output mode is mainly for RF sampling, outputting sample values in half of the sampling period and outputting the opposite value in the other half of the period "]The effect of the output mode on output power can be more easily understood from the frequency responses of these modes. The appropriate mode should be selected based on the Nyquist zone of interest to maximize output power. That is, the mode should be selected based on the sampling rate and the resulting RF frequency. For example, at a sampling rate of 6Gsps and an output frequency in the C-band (4-8GHz), the output power of the RF mode is maximized in the second or third Nyquist zone, with only a 5dB effect on the output power. The frequency response of the output mode depends on the sampling rate. A DAC operating at a sampling rate of 12Gsps and an output frequency in the C-band would be at most in the second Nyquist zone, where the NRZ mode would be most suitable. For data converters used in RF sampling, in addition to the normal requirements, they must be able to achieve sufficient output power and performance indicators at the target RF frequency, including analog bandwidth, dynamic performance, and output mode (applicable to DAC only). Once the appropriate data converter is selected, the RF sampling architecture can be applied in the system. Architecture comparison using the latestADC[color=rgb(51, 51, To highlight the advantages of RF sampling over IF architectures, Teledyne e2v’s latest generation of ADCs is available in three different configurations: a two-stage IF architecture, a single-stage IF architecture, and an RF sampling architecture. Table 2 shows the power consumption and noise of the three different configurations. It can be seen that removing each downconverter stage significantly reduces the power consumption of the system, by about 25% from dual downconverters to RF sampling architectures. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2. Performance Comparison of IF and RF Sampling Architectures Using Latest Generation ADCs 393672 "]While IF architectures in various forms have a long history, today’s advances in data converters have made this architecture less and less suitable for RF applications. RF sampling architectures have clear SWAP-C advantages and are evolving with the development of new generations of high-speed, high-bandwidth data converters. Advances in data converter technology will result in higher performance and support for higher frequency direct RF sampling. "]· Return to Zero (RTZ) – This output mode is also very common. It outputs sample values in half of the sampling period and outputs 0 in the other half of the period.· Narrow Return to Zero (NRTZ) – This output mode provides a flexible solution between NRZ and RTZ modes. It outputs sample values in a certain section (X) of the sampling period and outputs 0 in the part before and after (1-X)/2. · RF Mode (RF) – This output mode is mainly oriented towards RF sampling, outputting the sampled value in half of the sampling period and the opposite value in the other half of the period. The effect of the output mode on the output power can be more easily understood from the frequency response of these modes. The appropriate mode should be selected based on the Nyquist zone of interest to maximize the output power. That is, the mode should be selected based on the sampling rate and the generated RF frequency. For example, if the sampling rate is 6Gsps and the output frequency is in the C-band (4-8GHz), the RF mode has the maximum output power in the second or third Nyquist zone, with only a 5dB impact on the output power. The frequency response of the output mode depends on the sampling rate. A DAC operating at a 12Gsps sampling rate and an output frequency of C-band is at most in the second Nyquist zone, where the NRZ mode is most suitable. For data converters used in RF sampling, in addition to the conventional requirements, they must be able to achieve sufficient output power and performance indicators at the target RF frequency, including analog bandwidth, dynamic performance and output mode (applicable to DAC only). Once the appropriate data converter is selected, the RF sampling architecture can be applied to the system. Architecture Comparison Using the LatestADCsTo highlight the advantages of RF sampling over IF architectures, Teledyne e2v’s latest generation of ADCs is available in three different configurations: a two-stage IF architecture, a single-stage IF architecture, and an RF sampling architecture. Table 2 shows the power consumption and noise of the three different configurations. As can be seen, removing each downconverter stage significantly reduces the system power consumption, from dual downconverters to RF sampling architectures by about 25%. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2 shows the power consumption and noise of the three different configurations. As can be seen, removing each downconverter stage significantly reduces the system power consumption, from dual downconverters to RF sampling architectures by about 25%. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Performance Comparison of IF and RF Sampling Architectures Using Latest GenerationADCs
While IF architectures have a long history in various forms, today’s advances in data converters have made them less and less suitable for RF applications. RF sampling architectures have a clear SWAP-C advantage and are evolving with the development of new generations of high speed, high bandwidth data converters. Advances in data converter technology will result in higher performance and support for direct RF sampling at higher frequencies. "]· Return to Zero (RTZ) – This output mode is also very common. It outputs sample values in half of the sampling period and outputs 0 in the other half of the period.· Narrow Return to Zero (NRTZ) – This output mode provides a flexible solution between NRZ and RTZ modes. It outputs sample values in a certain section (X) of the sampling period and outputs 0 in the part before and after (1-X)/2. · RF Mode (RF) – This output mode is mainly oriented towards RF sampling, outputting the sampled value in half of the sampling period and the opposite value in the other half of the period. The effect of the output mode on the output power can be more easily understood from the frequency response of these modes. The appropriate mode should be selected based on the Nyquist zone of interest to maximize the output power. That is, the mode should be selected based on the sampling rate and the generated RF frequency. For example, if the sampling rate is 6Gsps and the output frequency is in the C-band (4-8GHz), the RF mode has the maximum output power in the second or third Nyquist zone, with only a 5dB impact on the output power. The frequency response of the output mode depends on the sampling rate. A DAC operating at a 12Gsps sampling rate and an output frequency of C-band is at most in the second Nyquist zone, where the NRZ mode is most suitable. For data converters used in RF sampling, in addition to the conventional requirements, they must be able to achieve sufficient output power and performance indicators at the target RF frequency, including analog bandwidth, dynamic performance and output mode (applicable to DAC only). Once the appropriate data converter is selected, the RF sampling architecture can be applied to the system. Architecture Comparison Using the LatestADCsTo highlight the advantages of RF sampling over IF architectures, Teledyne e2v’s latest generation of ADCs is available in three different configurations: a two-stage IF architecture, a single-stage IF architecture, and an RF sampling architecture. Table 2 shows the power consumption and noise of the three different configurations. As can be seen, removing each downconverter stage significantly reduces the system power consumption, from dual downconverters to RF sampling architectures by about 25%. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2 shows the power consumption and noise of the three different configurations. As can be seen, removing each downconverter stage significantly reduces the system power consumption, from dual downconverters to RF sampling architectures by about 25%. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Performance Comparison of IF and RF Sampling Architectures Using Latest GenerationADCs
While IF architectures have a long history in various forms, today’s advances in data converters have made them less and less suitable for RF applications. RF sampling architectures have a clear SWAP-C advantage and are evolving with the development of new generations of high speed, high bandwidth data converters. Advances in data converter technology will result in higher performance and support for direct RF sampling at higher frequencies. "]The effect of the output mode on output power can be more easily understood from the frequency responses of these modes. The appropriate mode should be selected based on the Nyquist zone of interest to maximize output power. That is, the mode should be selected based on the sampling rate and the resulting RF frequency. For example, at a sampling rate of 6Gsps and an output frequency in the C-band (4-8GHz), the output power of the RF mode is maximized in the second or third Nyquist zone, with only a 5dB effect on the output power.The frequency response of the output mode depends on the sampling rate. A DAC operating at a sampling rate of 12Gsps and an output frequency in the C-band would be at most in the second Nyquist zone, where the NRZ mode would be most suitable. For data converters used in RF sampling, in addition to the normal requirements, they must be able to achieve sufficient output power and performance indicators at the target RF frequency, including analog bandwidth, dynamic performance, and output mode (applicable to DAC only). Once the appropriate data converter is selected, the RF sampling architecture can be applied in the system. Architecture comparison using the latestADC[color=rgb(51, 51, To highlight the advantages of RF sampling over IF architectures, Teledyne e2v’s latest generation of ADCs is available in three different configurations: a two-stage IF architecture, a single-stage IF architecture, and an RF sampling architecture. Table 2 shows the power consumption and noise of the three different configurations. It can be seen that removing each downconverter stage significantly reduces the power consumption of the system, by about 25% from dual downconverters to RF sampling architectures. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2. Performance Comparison of IF and RF Sampling Architectures Using Latest Generation ADCs 393672 "]While IF architectures in various forms have a long history, today’s advances in data converters have made this architecture less and less suitable for RF applications. RF sampling architectures have clear SWAP-C advantages and are evolving with the development of new generations of high-speed, high-bandwidth data converters. Advances in data converter technology will result in higher performance and support for higher frequency direct RF sampling. "]The effect of the output mode on output power can be more easily understood from the frequency responses of these modes. The appropriate mode should be selected based on the Nyquist zone of interest to maximize output power. That is, the mode should be selected based on the sampling rate and the resulting RF frequency. For example, at a sampling rate of 6Gsps and an output frequency in the C-band (4-8GHz), the output power of the RF mode is maximized in the second or third Nyquist zone, with only a 5dB effect on the output power.The frequency response of the output mode depends on the sampling rate. A DAC operating at a sampling rate of 12Gsps and an output frequency in the C-band would be at most in the second Nyquist zone, where the NRZ mode would be most suitable. For data converters used in RF sampling, in addition to the normal requirements, they must be able to achieve sufficient output power and performance indicators at the target RF frequency, including analog bandwidth, dynamic performance, and output mode (applicable to DAC only). Once the appropriate data converter is selected, the RF sampling architecture can be applied in the system. Architecture comparison using the latestADC[color=rgb(51, 51, To highlight the advantages of RF sampling over IF architectures, Teledyne e2v’s latest generation of ADCs is available in three different configurations: a two-stage IF architecture, a single-stage IF architecture, and an RF sampling architecture. Table 2 shows the power consumption and noise of the three different configurations. It can be seen that removing each downconverter stage significantly reduces the power consumption of the system, by about 25% from dual downconverters to RF sampling architectures. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2. Performance Comparison of IF and RF Sampling Architectures Using Latest Generation ADCs 393672 "]While IF architectures in various forms have a long history, today’s advances in data converters have made this architecture less and less suitable for RF applications. RF sampling architectures have clear SWAP-C advantages and are evolving with the development of new generations of high-speed, high-bandwidth data converters. Advances in data converter technology will result in higher performance and support for higher frequency direct RF sampling. "]Table 2 shows the power consumption and noise of the three different configurations. As can be seen, removing each downconverter stage significantly reduces the system power consumption, from dual downconverters to RF sampling architectures, by about 25%. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2. Performance comparison of IF and RF sampling architectures using the latest generationADC "]
While IF architectures in various forms have a long history, today’s advances in data converters have made them less and less suitable for RF applications. RF sampling architectures have a clear SWAP-C advantage and are evolving with the development of a new generation of high-speed, high-bandwidth data converters. Advances in data converter technology will result in higher performance, enabling direct RF sampling at higher frequencies. "]Table 2 shows the power consumption and noise of the three different configurations. As can be seen, removing each downconverter stage significantly reduces the system power consumption, from dual downconverters to RF sampling architectures, by about 25%. At the same time, the noise performance does not change. This is because the noise performance is usually determined by the first-stage amplifier if it can provide sufficient gain. Table 2. Performance comparison of IF and RF sampling architectures using the latest generationADC "]
While IF architectures in various forms have a long history, today’s advances in data converters have made them less and less suitable for RF applications. RF sampling architectures have a clear SWAP-C advantage and are evolving with the development of a new generation of high-speed, high-bandwidth data converters. Advances in data converter technology will result in higher performance, enabling direct RF sampling at higher frequencies.
|
提升卡
变色卡
千斤顶