Design Article
Optimize SDR performance
Joel Kirshman AWR corporation
12/15/2011 5:58 PM EST
The concept of software defined radio (SDR) has existed for many years. Consequently, you can find many descriptions of an SDR. A concise definition of an SDR is a radio in which some or all of the physical layer functions are software-defined. The physical layer function is the layer within the wireless protocol in which processing of RF, IF, or baseband signals (including channel coding) occurs. Many of today’s SDRs have part of the signal processing implemented in software.
The ability to implement in software numerous waveforms and receiver functions is appealing from the perspective of cost, complexity, programmability, and more. A goal of SDR is to produce an entire radio in software, downconverting the received analog signal to baseband at the antenna port and eliminating much of the hardware in between. Today, SDR is increasingly implemented in both military and commercial networks in more modest ways. The challenge with any SDR design is that it is a complex system that requires analysis from the component through system levels in order to achieve the desired performance. This article explains the attributes of SDR, the design challenges, and it includes hands-on examples of how AWR’s Visual System Simulator (VSS) software for system simulation of RF end-to-end architecture design can be used to address these challenges.
SDR in brief
There are many specific processes that define a radio as an SDR. A modern day cell phone, for example, can be considered an SDR because most (if not all) of the baseband processing is performed by a digital signal processing (DSP) chip. SDRs enable a single transceiver to accommodate many different waveforms, which makes it an essential ingredient in creating, for instance, a universally-compatible nationwide public safety network. SDR has long been attractive to manufacturers, commercial wireless carriers, and the military because it enables a single hardware platform to accommodate a wide array of modulated waveforms by implementing them in software. As a result, it eliminates physical components such as mixers, filters, amplifiers, modulators and demodulators, and detectors. In addition, it creates a multi-mode, multi-band, multi-functional communications platform that can be dynamically programmed locally or remotely to accommodate new waveforms and enhancements to existing ones.
Simulating SDRs
In order to understand and predict the behavior of a particular SDR design, it is necessary to construct the signals per the baseband physical layer specifications, such as WCDMA, WiMAX, or LTE. This requires designers to implement channel coding, interleaving, and pulse shaping. Once these signals are generated, the simulation can be used to determine the impact of impairments on the RF/IF portion of the SDR, making measurements such as error vector magnitude (EVM) or bit error rate (BER) for example. With this information, designers can make trade offs between different components or circuit implementations while injecting interfering signals and adjacent channels into the RF path to simulate a real-world environment.
SDR systems must inherently accommodate a variety of modulation schemes that vary in their complexity within the same system architecture. For instance, higher-order schemes such as 64QAM are used when signal conditions permit in order to deliver higher data rates. So, the impact of an RF link on the BER performance of a 16QAM versus that of a 64QAM signal is important to understand. Unfortunately, understanding the impact that RF impairments have on today’s complex modulated signals is not a straightforward procedure, nor can it be accurately predicted with ‘gut calls’ or back-of-the-envelope calculations.
Critical SDR parameters required in a simulation tool include:
In the following example, AWR’s VSS software is used to virtually modify the values of components to achieve optimal results for an SDR receiver designed to capture 16QAM and 64QAM signals. The goal is to use VSS to determine a suitable P1dB for the front-end amplifier as well as a phase noise mask for the local oscillator (LO). Determining a suitable P1dB prevents a device from being driven into compression as well as getting reasonable power added efficiency (PAE). Achieving a tolerable phase-noise level will enable the device to meet BER requirements.
Figure 1 shows the initial 16QAM system. To complete a system analysis, first, the signal characteristics are defined and a swept simulation of signal-to-noise ratio (SNR) versus BER is performed. In this example, the results of the BER simulation shown in Figure 2 (red squares) agree with theory (the black reference curve). Next, a behavioral amplifier model is added to the receive link just after the additive white Gaussian noise (AWGN) channel. This amplifier is used to determine a suitable P1dB to maintain a BER of 10-6 at an SNR of about 20dB (the optimal performance characteristics for this SDR).
The third-order intercept point (IP3) is set to about 10dB above the P1dB point, and amplifier gain is maintained at 10dB. The results of a BER simulation shown in Figure 2A show that with P1dB power of 10dB, a received SNR of about 22dB is required to achieve 10-6 BER. However, if P1dB power is set to 12dB, the desired results are achieved (Figure 2B).
After adding a filter data file to the system just after the amplifier block, it is possible to evaluate BER performance. The 16QAM signal uses a root-raised cosine (RRC), pulse-shaping filter with an alpha of 0.35, and a bandwidth of about 50kHz. The center frequency of the signal is 10MHz. The selected filter has 0dB of attenuation in the passband and a bandwidth of about 52kHz. Attenuation rolls off to about -50dB at 9.9 and 10.1MHz from the center frequency. The 16QAM signal can be detected with the desired BER results. Next, single-sideband (SSB) phase noise of the downconverter is added to the received signal, and it is applied directly to the downconverter’s local oscillator (LO). Performing the simulation in this manner allows direct analysis of the impact of phase noise on BER. The system diagram and the resulting performance with phase noise added are shown in Figure 3. After adding phase noise, the minimum SNR required by the receiver to achieve a BER of 10-6 is about 22dB.
The ability to implement in software numerous waveforms and receiver functions is appealing from the perspective of cost, complexity, programmability, and more. A goal of SDR is to produce an entire radio in software, downconverting the received analog signal to baseband at the antenna port and eliminating much of the hardware in between. Today, SDR is increasingly implemented in both military and commercial networks in more modest ways. The challenge with any SDR design is that it is a complex system that requires analysis from the component through system levels in order to achieve the desired performance. This article explains the attributes of SDR, the design challenges, and it includes hands-on examples of how AWR’s Visual System Simulator (VSS) software for system simulation of RF end-to-end architecture design can be used to address these challenges.
SDR in brief
There are many specific processes that define a radio as an SDR. A modern day cell phone, for example, can be considered an SDR because most (if not all) of the baseband processing is performed by a digital signal processing (DSP) chip. SDRs enable a single transceiver to accommodate many different waveforms, which makes it an essential ingredient in creating, for instance, a universally-compatible nationwide public safety network. SDR has long been attractive to manufacturers, commercial wireless carriers, and the military because it enables a single hardware platform to accommodate a wide array of modulated waveforms by implementing them in software. As a result, it eliminates physical components such as mixers, filters, amplifiers, modulators and demodulators, and detectors. In addition, it creates a multi-mode, multi-band, multi-functional communications platform that can be dynamically programmed locally or remotely to accommodate new waveforms and enhancements to existing ones.
Simulating SDRs
In order to understand and predict the behavior of a particular SDR design, it is necessary to construct the signals per the baseband physical layer specifications, such as WCDMA, WiMAX, or LTE. This requires designers to implement channel coding, interleaving, and pulse shaping. Once these signals are generated, the simulation can be used to determine the impact of impairments on the RF/IF portion of the SDR, making measurements such as error vector magnitude (EVM) or bit error rate (BER) for example. With this information, designers can make trade offs between different components or circuit implementations while injecting interfering signals and adjacent channels into the RF path to simulate a real-world environment.
SDR systems must inherently accommodate a variety of modulation schemes that vary in their complexity within the same system architecture. For instance, higher-order schemes such as 64QAM are used when signal conditions permit in order to deliver higher data rates. So, the impact of an RF link on the BER performance of a 16QAM versus that of a 64QAM signal is important to understand. Unfortunately, understanding the impact that RF impairments have on today’s complex modulated signals is not a straightforward procedure, nor can it be accurately predicted with ‘gut calls’ or back-of-the-envelope calculations.
Critical SDR parameters required in a simulation tool include:
- Modulation accuracy or EVM
- Fixed-point implementation
- ADC/DAC quantization level
- BER performance
- Carrier to noise ratio (CNR)
- Spurious-free dynamic range (SFDR)
- Adjacent channel-power ratio (ACPR)
- Carrier to third-order-intermodulation ratio (C/IM3dBc)
- Spectral mask requirement
In the following example, AWR’s VSS software is used to virtually modify the values of components to achieve optimal results for an SDR receiver designed to capture 16QAM and 64QAM signals. The goal is to use VSS to determine a suitable P1dB for the front-end amplifier as well as a phase noise mask for the local oscillator (LO). Determining a suitable P1dB prevents a device from being driven into compression as well as getting reasonable power added efficiency (PAE). Achieving a tolerable phase-noise level will enable the device to meet BER requirements.
Figure 1.
The 16QAM system is shown schematically and the BER simulation results. The red squares closely match those of theory (the black reference curve).
Figure 1 shows the initial 16QAM system. To complete a system analysis, first, the signal characteristics are defined and a swept simulation of signal-to-noise ratio (SNR) versus BER is performed. In this example, the results of the BER simulation shown in Figure 2 (red squares) agree with theory (the black reference curve). Next, a behavioral amplifier model is added to the receive link just after the additive white Gaussian noise (AWGN) channel. This amplifier is used to determine a suitable P1dB to maintain a BER of 10-6 at an SNR of about 20dB (the optimal performance characteristics for this SDR).
Figure 2A. A BER simulation of the 16QAM system with power at 1-dB gain compression of 10dB shows that a received SNR of about 22dB is required to achieve a BER of 10-6.
Figure 2B.
If RF power is set to 12dB, the desired results (BER of 10-6, SNR of 20dB) are achieved.
The third-order intercept point (IP3) is set to about 10dB above the P1dB point, and amplifier gain is maintained at 10dB. The results of a BER simulation shown in Figure 2A show that with P1dB power of 10dB, a received SNR of about 22dB is required to achieve 10-6 BER. However, if P1dB power is set to 12dB, the desired results are achieved (Figure 2B).
After adding a filter data file to the system just after the amplifier block, it is possible to evaluate BER performance. The 16QAM signal uses a root-raised cosine (RRC), pulse-shaping filter with an alpha of 0.35, and a bandwidth of about 50kHz. The center frequency of the signal is 10MHz. The selected filter has 0dB of attenuation in the passband and a bandwidth of about 52kHz. Attenuation rolls off to about -50dB at 9.9 and 10.1MHz from the center frequency. The 16QAM signal can be detected with the desired BER results. Next, single-sideband (SSB) phase noise of the downconverter is added to the received signal, and it is applied directly to the downconverter’s local oscillator (LO). Performing the simulation in this manner allows direct analysis of the impact of phase noise on BER. The system diagram and the resulting performance with phase noise added are shown in Figure 3. After adding phase noise, the minimum SNR required by the receiver to achieve a BER of 10-6 is about 22dB.
Figure 3.
The modified 16QAM receive link and BER performance including phase noise.
|
|
|
|
|
Navigate to related information
Paul Stoaks
12/22/2011 2:42 PM EST
This is a good article and it addresses a thorny problem for SDR design. Getting the RF, modulation, and EC behaviors correct is critical to waveform success. The recent drone capture in Iran actually highlights the importance of this, in my opinion. Noise and jam resistance are critical features of a good waveform.
Another area of great interest to SDR designers is the optimization of the waveform at the baseband and networking level. Modern waveforms face daunting requirements in these areas and engineers need tools to assist in ensuring that designs "fit" into the target hardware platfroms. Foresight Systems M&S (www.foresight-mands.com) helps engineers optimize the mating of waveform software and hardware. Using our Resource Aware Modeling and Simulation (RAMS) approach, we can predict memory and processor utilization, jitter, latency, etc. enabling engineers to effectively optimize the waveform design.
Sign in to Reply
Dr DSP
12/31/2011 1:12 PM EST
It is great to see some simulation results in detail. Too often articles don't give enough background and specifics like this one does. Thanx!
Sign in to Reply