Millimeter-Wave Beamforming Overview & Design Choices
Current cellular 4G networks face a multitude of challenges. Soaring demand for mobile high-resolution multimedia applications brings these networks ever closer to their practical limits. 5G networks are envisioned to ease the burden on the current infrastructure by offering significantly higher data rates through increased channel bandwidths. Considering the shortage of available frequencies traditionally used for mobile communications, mm-wave bands became a suitable alternative. The large bandwidth available at these frequencies helps to offer data rates that satisfy 5G demands.
However, the mobile environment at these mm-wave bands is far more complex than at the currently used frequencies. Higher propagation losses that vary greatly depending on the environment require an updated network infrastructure and new hardware concepts.
Beamforming antenna arrays will play an important role in 5G implementations since even handsets can accommodate a larger number of antenna elements at mm-wave frequencies. Aside from a higher directive gain, these antenna types offer complex beamforming capabilities. This enables an increase in the capacity of cellular networks by improving the signal to interference ratio (SIR) through direct targeting of user groups. The narrow transmit beams simultaneously lower the amount of interference in the radio environment and make it possible to maintain sufficient signal power at the receiver terminal at larger distances in rural areas.
This paper gives an overview of the beamforming technology including signals, antennas and current transceiver architectures. All equations presented in this paper apply to linear antenna arrays, which for the purpose of this paper are defined as an array of equally spaced, individually excitable n radiating elements placed along one axis in a coordinate system.
Beamforming in general works with simple CW-signals as well as with complex modulated waveforms. Candidate waveforms for 5G are a current research topic, since many of today’s implementations suffer great disadvantages at millimeter wave bands.
Phase Coherent Signal Generation
An important prerequisite for every beamforming architecture is a phase coherent signal. This term means that there is a defined and stable phase relationship between all RF carriers. A fixed delta phase between the carriers, as shown in Figure 1, can be used to steer the main lobe in a desired direction.
Phase coherence can be achieved by coupling multiple signal generators via a common reference (i.e. 10 MHz). A closer inspection of the instantaneous differential phase (“delta phase”) of these RF signals shows instability due to:
- Phase noise of the two synthesizers
- “Weak” coupling at 10 MHz and a long synthesis chain up to the RF output
- Temperature differences that cause a change in the effective electrical length of some synthesizer components
Because of the dominance of the second factor, the only way to stabilize the phase between two signal generators is to use a common synthesizer/LO source. This measure simultaneously eliminates the first factor.
All signals radiated from any kind of antenna share the same basic characteristics. Multipath fading and delay spread significantly reduce the capacity of a cellular network. Congestion of the available channels and co-channel interference further reduce the practical network capacity.
Free Field Attenuation: Electromagnetic waves are attenuated while travelling from the transmitter to the receiver. The free field attenuation describes the attenuation that the signal will suffer due to the distance between the two stations.
The Friis formula determines the free field attenuation:
Where 𝑃𝑟,𝑑𝐵 is the received power level in dB, 𝑃𝑡,𝑑𝐵 the transmitted power and 𝐺𝑟,𝑑𝐵 and 𝐺𝑡,𝑑𝐵 the receive and transmit antenna gain in dBi. Figure 2 (left) illustrates the free field attenuation over a large frequency band.
Even in case of a perfect line of sight (LoS) transmission, there are many different factors that additionally affect the magnitude of the received signal. As shown in Figure 2 (right), the resulting overall attenuation varies greatly depending on the frequency and radiation environment.
Fading: The phase shift in multipath signals is non-constant due to the time variant nature of the channel. The following expression shows the time-dependent received multipath signal, where the complex values 𝑎𝑛 (𝑡) and 𝑒−𝑗𝜃𝑛(𝑡) describe the change in amplitude and phase for the transmit path n:
The signals add up constructively or destructively depending on the current phase shift. The received signal consists of a multitude of scattered components, making it a random process. Based on a sufficient amount of scattered components, this can be seen as a complex Gaussian process. This results in the creation of small fade zones in the coverage area, which is called Rayleigh-Fading.
A special case of fading is the phase cancellation, which occurs when multipath signals are 180° out of phase from each other. The cancellation, and thus the attenuation of the signal, depends largely on the amplitude and phase balance. A 30 dB difference, for example, corresponds approximately to a 0.1 dB and 1.0 degree matching error.
Delay Spread: This effect is also due to the multipath nature of signal propagation. It describes the difference between the time of arrival of the earliest and latest significant multipath component. Typically the earliest component is the LoS transmission. In case of large delay spreads the signal will be impaired by inter-symbol interferences, which dramatically increase the bit error rate (BER).
Modern beamforming antenna architectures can help to mitigate these problems by adapting to the channel. This way, delayed multipath components can be ignored or significantly reduced through beam steering. Antennas that are designed to adapt and change their radiation pattern in order to adjust to the RF environment are called active phased array antennas.
Millimeter-wave bands potentially enable high bandwidths. To date, the limited use of these high frequencies is a result of adverse propagation effects in particular due to obstacles in the LoS. Several transceiver architectures have been developed to compensate for these issues by focusing the received or transmitted beams in a desired direction. All these solutions make use of smaller antenna element sizes due to higher carrier frequencies that enable the construction of larger antenna arrays.
Usually two variables are used for beamforming: amplitude and phase. The combination of these two factors is used to improve side lobe suppression or steering nulls. Phase and amplitude for each antenna element n are combined in a complex weight wn. The complex weight is then applied to the signal that is fed to the corresponding antenna.
Figure 3 shows a basic implementation of an analog beamforming transmitter architecture. This architecture consists of only one RF chain and multiple phase shifters that feed an antenna array.
The first practical analog beamforming antennas date back to 1961. The steering was carried out with a selective RF switch and fixed phase shifters. The basics of this method are still used to date, albeit with advanced hardware and improved precoding algorithms. These enhancements enable separate control of the phase of each element. Unlike early passive architectures, the beam can be steered not only to discrete but virtually any angle using active beamforming antennas. True to its name, this type of beamforming is achieved in the analog domain at RF frequencies or an intermediate frequency.
This architecture is used today in high-end millimeter-wave systems as diverse as radar and short-range communication systems like IEEE 802.11ad. Analog beamforming architectures are not as expensive and complex as the other approaches described in this paper. On the other hand, implementing a multi-stream transmission with analog beamforming is a highly complex task.
In order to calculate the phase weightings, a uniformly spaced linear array with element spacing d is assumed. Considering the receive scenario shown in Figure 4, the antenna array must be in the far field of the incoming signal so that the arriving wave front is approximately planar. If the signal arrives at an angle 𝜃 off the antenna boresight, the wave must travel an additional distance 𝑑∗𝑠𝑖𝑛𝜃 to arrive at each successive element, as illustrated in Figure 4. This translates to an element specific delay, which can be converted to a frequency dependent phase shift of the signal:
The frequency dependency translates into an effect called beam squint. The main lobe of an antenna array at a defined frequency can be steered to a certain angle using phase offsets calculated with the above expression. If the antenna elements are now fed with a signal of a different frequency, the main lobe will veer off by a certain angle. Since the phase relations were calculated with a certain carrier frequency in mind, the actual angle of the main lobe shifts according to the current frequency. Radar applications with large bandwidths especially suffer inaccuracies due to this effect.
Figure 5 shows the impact of beam squint as a function of the frequency for a linear array of four elements. The main lobe was steered to 15° at a frequency of 30 GHz. Using the above expression this was achieved by a phase offset Δ𝜑 of 141° per element. Due to the large bandwidth used, the beam squint effect is clearly visible at the lower frequencies, where the main lobe is located at 25°.
The above expression can be converted to a frequency independent term by using time delays instead of frequency offsets:
This means that the frequency dependency is eliminated if the setup is fitted with delay lines instead of phase shifters. The corresponding receiver setup is shown in Figure 6. The delay lines 𝑡0 to 𝑡2 compensate for the time delay Δ𝑡, which is an effect of the angle of the incident wave. As a result, the received signals should be perfectly aligned and will thus add constructively when summed up.
The performance of the analog architecture can be further improved by additionally changing the magnitude of the signals incident to the radiators.
While analog beamforming is generally restricted to one RF chain even when using large-number antenna arrays, digital beamforming in theory supports as many RF chains as there are antenna elements. If suitable precoding is done in the digital baseband, this yields higher flexibility regarding the transmission and reception. The additional degree of freedom can be leveraged to perform advanced techniques like multi-beam MIMO. These advantages result in the highest theoretical performance possible compared to other beamforming architectures.
Figure 7 illustrates the general digital beamforming transmitter architecture with multiple RF chains.
Beam squint is a well-known problem for analog beamforming architectures using phase offsets. This is a serious drawback considering current 5G plans to make use of large bandwidths in the mm-wave band. Digital control of the RF chain enables optimization of the phases according to the frequency over a large band.
Nonetheless, digital beamforming may not always be ideally suited for practical implementations regarding 5G applications. The very high complexity and requirements regarding the hardware may significantly increase cost, energy consumption and complicate integration in mobile devices. Digital beamforming is better suited for use in base stations, since performance outweighs mobility in this case.
Digital beamforming can accommodate multi-stream transmission and serve multiple users simultaneously, which is a key driver of the technology.
Hybrid beamforming has been proposed as a possible solution that is able to combine the advantages of both analog and digital beamforming architectures. First results from implementations featuring this architecture have been presented in prototype level.
A significant cost reduction can be achieved by reducing the number of complete RF chains. This does also lead to lower overall power consumption. Since the number of converters is significantly lower than the number of antennas, there are less degrees of freedom for digital baseband processing. Thus the number of simultaneously supported streams is reduced compared to full blown digital beamforming. The resulting performance gap is expected to be relatively low due to the specific channel characteristics in millimeter-wave bands.
The schematic architecture of a hybrid beamforming transmitter is shown in Figure 8. The precoding is divided between the analog and digital domains. In theory, it is possible that every amplifier is interconnected to every radiating element.
Results and Outlook
It is already widely accepted that beamforming will play an important role in the implementation of next generation networks. Many 5G topics are still subjects of ongoing research, but the general direction taken by the industry includes small as well as large beamforming arrays, the latter in part only made feasible by the shorter wavelengths encountered in millimeter-wave bands.
For a more in depth look at this, view the full, continually updated, whitepaper here: