One of the less visible parts of developing a new generation of mobile technology is deciding what assumptions should be used when evaluating it. Before different radio techniques can be compared in 3GPP, they need to be tested using agreed Link-Level Simulation (LLS) and System-Level Simulation (SLS) models. Those models inevitably simplify reality, but if the simplifications become too far removed from real devices, they can also distort the conclusions.
UE antenna modelling is a good example. The Nokia white paper, New 3GPP UE Antenna and Near-field Blocking Models for Handheld Devices, looks at changes made in 3GPP Release 19 to make the antenna model used for handheld devices more representative of modern smartphones. Although Release 19 is still part of 5G-Advanced, Nokia describes the work as preparation for 6G, with the revised model expected to be applicable to 6G devices from Release 20 onwards.
The interesting part is not simply that 3GPP has created another antenna model. The changes address some fairly fundamental assumptions about antenna radiation patterns, physical antenna placement, polarisation, antenna combining and what happens when a person actually holds the phone.
The starting point is surprisingly old. Nokia points out that the UE antenna model used for 5G in Release 15 is essentially the same model defined during the early LTE work in Release 8 in 2008. For an individual antenna port, simulations could use an isotropic radiator or a dipole-type radiation pattern. That abstraction was useful because it made simulations manageable, but smartphones have changed considerably.
Modern handsets contain multiple antennas integrated around the chassis. Their radiation patterns are not isotropic, their maximum gain does not necessarily point in the same direction, and their efficiency and polarisation depend on their physical position in the phone. The antennas also interact with the chassis, neighbouring antennas and, importantly, the person holding the handset.
The old model also treats multiple antenna ports rather like a conventional Uniform Linear Array, or ULA. In this type of model, antenna elements can be separated by a convenient electrical spacing such as half a wavelength, λ/2. That may be a reasonable approximation for some base-station arrays or other regularly arranged antenna systems. It is much less representative of a smartphone.
The physical positions of antennas in a handset are largely dictated by the dimensions and industrial design of the device. Their spacing therefore remains physically fixed while the electrical distance between them changes with frequency. Smartphone antennas also do not have identical radiation characteristics or perfectly aligned phase centres. This becomes particularly important for MIMO and antenna combining.
Nokia investigated the problem using a detailed 6G reference smartphone. The Mechanical Computer-Aided Design, or MCAD, model contains 22 potential antenna locations covering different frequency groups. The study then uses electromagnetic simulations to characterise antenna behaviour under free-space conditions as well as different user interactions, including CTIA-defined right-hand browsing and head-and-hand configurations together with a Nokia-defined dual-hand browsing grip.
The simulations were carried out at 700, 2000, 2600, 3800, 4600 and 7800 MHz. Even without the user holding the device, the resulting radiation patterns demonstrate how different real handset antennas are from ideal isotropic radiators. At different frequencies, maximum gain occurs in different angular directions. There can also be large differences between the best and worst antenna ports for the same direction.
Nokia's simulations show that antenna gain imbalances of more than 10 dB are not unusual even before user interaction is considered. Adding the different hand and head configurations can increase that imbalance to more than 30 dB. At the 50th percentile, Nokia observes antenna gain imbalances of around 10 to 20 dB when the user is holding the phone. That can have a direct effect on simulated MIMO performance, coverage, interference and mobility. This is the problem that the Release 19 model tries to address.
Instead of modelling the UE antenna as an isotropic radiator, Release 19 introduces a directive reference antenna radiation gain pattern for handheld devices. Importantly, 3GPP did not create a completely new mathematical framework for this. The model reuses the antenna element equations already defined in TS 38.901, but changes the parameters to better represent UE antennas. The resulting reference antenna has:
- maximum directional gain of 5.3 dBi
- 3 dB beamwidth of 125° in both θ and φ
- front-to-back ratio of 22.5 dB
Nokia's comparison with its electromagnetic handset simulations shows that this directive pattern provides a considerably better approximation than an isotropic radiator. There is a deliberate compromise here. The electromagnetic radiation pattern of a real antenna changes with frequency, but using a complete frequency-dependent handset antenna model would make large 3GPP simulation campaigns significantly more complicated.
The Release 19 model therefore uses the same reference antenna radiation pattern across frequencies. Frequency-dependent behaviour is introduced in other ways, particularly through the physical placement and subsequent combining of the antennas. This is an important distinction. The objective is not to reproduce the exact antenna pattern of a particular smartphone. It is to retain enough of the characteristics of real devices that the system-level behaviour becomes more representative.
One of the biggest improvements is therefore not just the antenna pattern but where those patterns are placed. Release 19 defines eight candidate antenna locations around the edge of a reference 150 mm × 70 mm handheld device. The reference antenna pattern is rotated at each position so that the direction of maximum gain is aligned with the X-axis, Y-axis or one of the defined diagonal directions, including orientations tilted by 25° relative to the Y-axis. This is what Figure 18 (top) shows particularly well.
Each antenna consequently has its peak radiation in a different angular direction. More importantly, the antenna positions are defined by physical distances on the handset rather than by an artificial fixed electrical spacing such as λ/2. This means the electrical separation between two antenna ports automatically varies with carrier frequency. For a fixed physical separation d, d/λ increases as the wavelength becomes shorter. In other words, the same two antennas are electrically much further apart at 7.8 GHz than at 2 GHz.
That changes the array factor and therefore the combined radiation pattern. The new model also improves the treatment of polarisation. A real smartphone antenna does not necessarily radiate with a simple, perfectly linear polarisation in every direction. The total radiation field contains θ and φ components, and their relative magnitudes depend on antenna position and observation direction.
The Release 19 reference gain pattern is initially defined in an Antenna Coordinate System, or ACS. It is rotated into the handset's Local Coordinate System, LCS, according to the position of each antenna, and from there into the Global Coordinate System, GCS, according to the orientation of the handset. The existing rotation framework in TS 38.901 can then be used to derive individual θ and φ field components for the antenna locations.
This is significant because the antenna ports around the handset no longer behave like perfectly orthogonal identical elements. Their effective polarisation characteristics depend on their location and the orientation of the device.
Perhaps the most interesting results in the Nokia paper concern what happens when several UE antenna ports are combined. For an ideal ULA, the result is relatively predictable. With identical antenna elements, controlled spacing and suitable phase relationships, increasing the number of coherently combined elements can produce the familiar array gain. A handset is more complicated.
Its individual antennas have different radiation patterns, and their high-gain regions occur in different directions. Their θ and φ field components can also have different relative phases. Because typical handset antennas are single-feed structures, one precoding phase is applied to an antenna port. It cannot independently optimise the phase relationship of its θ and φ components. Consequently, adding more antenna ports does not automatically provide the ideal theoretical combining gain.
Nokia demonstrates this using its electromagnetic 6G smartphone model. With a full precoding codebook using 90° phase increments, the increase in maximum combined gain ranges from 0.9 to 3.8 dB, depending on frequency, with an average increase of 2.4 dB over best-antenna selection. That is quite different from simply assuming that adding antennas to an ideal array always gives the expected textbook increase in array gain.
The new 3GPP reference patterns do not contain the complete complex phase information available from Nokia's electromagnetic simulations. The Release 19 model therefore combines antennas using an Array Factor (AF) calculated from the physical distances between antenna ports, the operating frequency and the applied precoding vector. The result is frequency-dependent combined radiation patterns with ripples, nulls and different directions of maximum gain. This is exactly what one would expect from fixed antenna locations on a handset as their electrical separation changes with frequency.
Using the same 90° precoding codebook, the new 3GPP model produces an average improvement in maximum gain of about 2.6 dB, compared with the 2.4 dB average obtained with the much more detailed Nokia electromagnetic handset model. Nokia notes that the 3GPP model shows less frequency variation because it deliberately uses one reference antenna pattern across frequencies.
Another interesting result concerns how many antenna ports should actually be combined. For the new model, averaged across the frequencies studied, the maximum gain for a particular angular direction is obtained using:
- 1 AP: 2.1%
- 2 APs: 66.1%
- 3 APs: 12.2%
- 4 APs: 19.6%
This is very different from the legacy ULA model, where using all four antenna ports consistently produces the maximum gain across the angular domain. The Nokia reference smartphone does not produce exactly the same AP distribution as the new 3GPP model, but it shows the same fundamental behaviour: more antenna ports are not automatically better in every direction.
This may appear to be a relatively low-level antenna modelling issue, but it can influence conclusions about MIMO, precoding, coverage and interference when new radio techniques are evaluated. There is another problem that the old UE model handles even less realistically: the user.
The existing 3GPP stochastic self-blockage model defines a blockage region in the angular domain. For the defined region, an attenuation of 30 dB is applied. The problem is that this angular attenuation is effectively common to the antennas on the device. That does not resemble what happens when somebody holds a smartphone. A finger may be immediately adjacent to one antenna. Another antenna a few centimetres away may be partly affected, while an antenna at the opposite end of the device may be almost untouched. The human hand can change not only the received power but the radiation pattern, antenna efficiency and directivity. Nokia's electromagnetic simulations show this clearly.
For example, with the CTIA right-hand browsing grip, the bottom and side antennas can be heavily affected while the top antennas remain comparatively close to their free-space behaviour. At 3800 MHz, the simulated reduction in efficiency relative to free space ranges from less than 1 dB for some antenna positions to 18.6 dB for one heavily affected antenna. The old common angular blockage region cannot reproduce those large antenna-to-antenna differences. Nokia's comparison shows cases where the legacy blockage approach underestimates the reduction in maximum antenna gain and efficiency by more than 10 dB for some antennas.
Release 19 therefore introduces a different approach based on Spatial Non-Stationarity, or SNS, at the UE side. Instead of assuming the same blockage behaviour for every antenna, the near-field blockage model assigns attenuation according to the physical antenna location. The attenuation factors are derived from electromagnetic simulations using hand and head phantoms. Different antenna ports can therefore experience very different losses for the same user scenario.
Figure 30 is particularly useful because it compares three cases: the free-space antenna, the concept behind the new near-field blockage model and the detailed CTIA right-hand electromagnetic simulation. For Ant#1 in the example, the change is relatively modest. For Ant#3, the loss is much greater. This produces the antenna-port power imbalance that would be expected when the user's hand is physically close to only some parts of the handset.
The final Release 19 near-field attenuation values are derived from simulations for one-hand browsing, dual-hand browsing and head-plus-hand operation. For the 1 to 8.4 GHz range, Nokia uses averages from simulations at 2000, 2600, 3800, 4600 and 7800 MHz. The resulting attenuation values vary significantly by antenna location. For the one-hand case, for example, the 1 to 8.4 GHz values range from around 0.6 to 0.7 dB at the least affected positions to 10.8 dB at one of the most heavily affected positions. With the dual-hand grip, different antenna locations become the strongly attenuated ones.
It is worth noting that this near-field model is not intended to eliminate the existing stochastic user-body blockage model completely. Nokia notes that the existing blockage model can still be used to represent body blockage, although the angular region may need to be reconsidered for that purpose. The new mechanism is trying to capture something different: the local interaction between the hand or head and individual antennas on the handset. This distinction becomes increasingly important as more antennas and more advanced MIMO techniques are incorporated into mobile devices.
The overall Release 19 change can therefore be thought of as moving from an idealised UE consisting of several uniform antenna elements towards an abstract handset that retains some of the essential properties of a real smartphone. The model now includes a directive rather than isotropic reference pattern, physically meaningful antenna positions, frequency-dependent electrical separation, location-dependent polarisation characteristics and individual near-field attenuation associated with how the user holds the device. It is still deliberately much simpler than a full electromagnetic handset model. That is necessary if it is to be used practically in large-scale 3GPP simulations.
But Nokia's results suggest that the new abstraction behaves much more like a real handset than the model inherited from the early LTE era. This is particularly relevant for 6G. Before 3GPP can decide whether a new waveform, MIMO technique, precoding method, mobility enhancement or other radio feature offers a meaningful improvement, the assumptions underneath the simulations need to be sufficiently realistic. Otherwise there is a risk of optimising the radio interface for an idealised UE that does not behave like the device in somebody's hand.
In that sense, the new Release 19 UE antenna and near-field blockage models are not simply another 5G-Advanced enhancement. They are part of the groundwork needed to evaluate 6G more realistically.
The Nokia white paper is highly recommended for anyone interested in 3GPP simulations, antenna modelling or the assumptions that will underpin early 6G evaluations. It contains considerably more detail than covered here, including the full electromagnetic reference handset, antenna efficiencies and directivities, θ and φ field components, antenna combining equations, array factors, CCDF results and the near-field attenuation tables.
Related Posts:
- Free 6G Training: Nokia's Four Blueprints for Building 6G from Day One
- Free 6G Training: MediaTek’s Device View of the Road to 6G
- Free 6G Training: Qualcomm Keynote on 'Advancing Wireless Innovations towards 6G — from a Device’s Viewpoint' from B6GS 2023
- Free 6G Training: The 6G devices of 2030
Comments
Post a Comment