Fundamental Physical Layer Principles Underlying Uplink Power Control (PC) and Outer Loop Link Adaptation (OLLA)

Uplink power control (PC) is a widely recognized concept aimed at regulating UE power to attain a specified SINR. Simultaneously, uplink outer loop link adaptation (OLLA) involves determining the filtered operating SINR and its associated Modulation and Coding Scheme (MCS) to operate within that SINR range. It also considers errors occurring at the operating MCS, adjusting its decision to scale up or down the MCS for optimal data rate. While specific implementations may differ among vendor products, the basic functionality remains consistent across all products.

It's noteworthy to emphasize that during field trials, the UL PC/OLLA modules are the challenging components that demand a significant portion of developer time.

But the question is,

• what mandate the computation of target SINR in UL PC?
• why should we need to achieve a target SINR for a UE in UL PC?
• how to decide whether we need to increase or decrease the power for a UE in a UL PC?
• how to decide whether we need to use absolute TPC or accumulative TPC in UL PC?
• we should filter or use instantaneous Received-SINR in UL PC?
• why should we filter the instantaneous SINR in UL OLLA?
• why not operate on an MCS corresponding to an operating SINR always in UL OLLA?
• why should we need to scale up and down the MCS based on BLER?

Understanding the reasons and solutions to the aforementioned questions is rooted in certain fundamental physical layer (Layer-1) principles. This comprehension is crucial for the design, debugging, and analysis of the UL PC/OLLA algorithm, especially in dynamic channel conditions and unpredictable mobility within uncontrolled field environments.

The article is organized into the following sections,

Underlying Fundamental Principles

• Channel Inversion Strategy
• Water-Filling Strategy
• UL PC/OLLA Interaction

Conclusion

References

Underlying Fundamental Principles

Power Allocation Strategies:

There are two types of uplink power allocation strategies [3],

• Channel Inversion Strategy
• Water Filling Strategy

In an uncontrolled field environment, when a wireless signal is transmitted between a Base Station and a Mobile device, it experiences fading. Fading is a comprehensive term encompassing all wireless impairments caused by obstacles, both stationary and mobile, varying distances, and speeds.

Channel Inversion Strategy:

This is a strategy to handle slow fading; slow fading as the name suggests the fading happens slowly, pathloss and shadowing are the types of slow fading.

The strategy is to keep the received SNR constant irrespective of the channel gain. so that there is zero outage probability, and to control the transmit power such that the rate can be delivered no matter what the fading state is.

In this strategy, in the UL power control. for example, a TX is transmitting using 2dBm power per PRB and the receiver receives power of -98dBm per PRB, then the Channel Gain = ratio of the received signal power to the transmitted signal power, is -98dBm/2dBm = -100dB. This means the signal power is attenuated by 100dB due to fading. So the channel inversion strategy is to compensate the attenuated power to avoid outage and also to maintain a stable rate within the limit of the maximum transmit power fixed for that mobile.

In the 5G NR UL PC, these path-loss/shadowing compensations are done in two ways:

• Full compensation scheme
• Fractional Power control scheme or FPC

In a full compensation scheme, the computed attenuation due to path-loss/shadowing is fully compensated. The Full Compensation scheme aims to uphold a consistent Signal to Interference plus Noise Ratio (SINR) at the receiver. UEs elevate their transmit power to entirely offset any rise in path loss.

If you remember the PUSCH transmit power equation for the UE in the physical layer standard (38.213) is given as,

P_PUSCH = min { Pcmax , PO_PUSCH(i)+ 10×LOG(MPUSCH(i)) + [PL × ɑ(i)] + ?TF(i) + f(i)}

there is a below component in the equation,

α?PL

α = 1; for a full compensation scheme.

PL is the general term used to refer to path-loss/shadowing in 3GPP

Fractional Power Control (FPC) Scheme:

In the FPC scheme, where the computed attenuation due to path-loss/shadowing is compensated partially in the UE.

Whereas in,

α?PL

α = {0, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1}

PL is the general term used to refer to path-loss/shadowing in 3GPP.

for example, if Alpha = 0.8, 80% attenuation of path-loss/shadowing will be compensated in the UE and the BS will adjust the remaining transmission power in fractional steps; for example, if the signal quality is better than required, the power may be reduced by a certain percentage; if the quality is worse, it may be increased, this is called a Fractional Power Control scheme or FPC.

• The partial compensation is done to adapt to changes in the inter-cell interference situation or to correct the path-loss measurements and power ampli?er errors, aperiodic close-loop adjustments (Transmit Power Control (TPC) command) can also be applied.
• It is proven FPC is an efficient scheme and it is used widely in LTE and 5G NR Power Control algorithms. The FPC scheme partially compensates for the path-loss, which means that not all the users have the same SINR requirements, for example, a cell center user will require a 30dB SINR, and can add all available PRBs in BW, with peak TX power of 23dBm; whereas a cell edge UE will require a 5dB SINR, and can add only a few PRBs in BW, with peak TX power of 23dBm. The FPC scheme improves air-interface spectral efficiency and increases average cell throughput by reducing intercell interference.

How the P0, α, and IoT Parameters impact the FPC Scheme:

The parameter, P0 (represents the Base-Station received power per PRB assuming a path loss of 0 dB; it can be called Target Power Spectral Density), is also important in the FPC scheme.

The P0 is mainly defined in two ways. In the first method of P0 selection, α is fixed as one selected optimum value, such as 0.8, and P0 is searched for the optimum value in each simulation scenario. For example. we can simulate different P0 values say, -84dBm to -65dBm for a fixed Alpha, say 0.8, and we can plot the simulation results of these scenarios in Sector SE (average throughput) vs Cell-Edge SE (average cell edge throughput), the P0 which gives the highest Sector SE and Cell-Edge SE should be selected. If the P0 is not well selected, the sector SE and cell-edge SE degrades at the same time.

Some of the ideas for selecting P0 and α: To plot the average cell/sector spectral efficiency and the fifth percentile of the UE throughput, respectively, versus α for different values of P0. For example, α will be x-axis of the plot taking all the possible values defined in 3GPP from 0 to 1, and y-axis is the average cell SE for one plot and the average 5th percentile of the UE throughput for another plot; for different P0 values like P0 =-120dBm/-100dBm/-80dBm/-60dBm/-40dBm/-20dBm, different simulations can be done and plotted. Using these two plots Avg. cell SE Vs. Alpha and Avg. 5th percentile UE throughput Vs. Alpha, the P0/Alpha combo should be selected from these plots based on our requirements like, P0/Alpha which is good for Avg. cell SE (Or) P0/Alpha which is good for Avg. 5th percentile UE throughput (Or) P0/Alpha which will give balanced best SE and cell edge throughput.

In the second method of P0 selection, P0 is transformed from α value and cell-edge target SNR and given by [1],

P0 = α x (SNRTarget?+ PSDn) + (1-α) x PSDMax;

where,

• SNRTarget? is the cell edge target SNR,
• PSDn is the thermal noise PSD,
• and PSDMax is the maximum transmission power PSD for the assigned resource size.

- FPC allows for smooth and gradual adjustments to the transmission power, helping to maintain a balance between conserving power and ensuring reliable communication. It adjusts the power in small fractions or steps, typically expressed as a percentage of the total power.

- The FPC permits the decline of received SINR with increasing path loss, meaning that the received SINR decreases as the UE moves toward the cell edge. The UE adjusts its transmit power at a diminished rate in response to increasing path loss (path loss compensation), i.e., increases in path loss are only partially compensated. For example, if a cell center user is seeing 100dB path-loss, for alpha=0.8, it will compensate 80dB, and the remaining 20dB is dynamically handled, so the received SINR = P0 - (1-Alpha)PL - I-N, -79dBm - 20dB - I-N = 99dBm - I-N. Where a cell edge user is seeing a 130dB path-loss, for alpha=0.8, it will compensate 104dB, and the remaining 26dB is dynamically handled, so the received SINR = P0 - (1-Alpha)PL - I-N, -79dBm - 26dB - I-N = -105dBm - I-N. So, the FPC permits the decline of received SINR with increasing path-loss. In the FPC, the received power per PRB decreases as the path loss increases. Whereas in a full compensation scheme, the received power per PRB remains constant as path loss increases.

- The FPC enables UEs at the cell center to transmit at reduced power per PRB compared to UEs at the cell edge. For example, a cell center user with received SINR = -99dBm - I-N per PRB from the above example, can choose better MCS, and the I+N will be low, so it can transmit reduced power per PRB, but at the same time add more number of PRBs, so the cell center user is bandwidth limited user. Whereas a cell edge user will have high I+N and 26dB (from the above example) remaining PL to compensate, so it has to send each PRB with high power, so for lower SINR it has to use low MCS and use high transmit power per PRB, it will not be able to add max PRBs, as the user is power limited.

A higher P0 (with fixed α) means an overall received SINR per user increase. On the other hand, a lower α (with fixed P0) not only decreases the SINR but also spreads the curve which leads to a higher differentiation in terms of experienced SINR between cell-edge and cell-center users. In a real system, however, the interference level is also a function of P0 and α which is expected to have a certain impact on the SINR distribution. The below plot illustrates it [2].

Let’s de?ne the experienced received SINR per user S as [1]:

S=P?L?IoT ?N [dB]

P = P0

L = (1-α) x PL

IoT = Interference over Thermal, the IoT level in the linear domain is calculated as the ratio of interference plus thermal noise over thermal noise.

N = thermal noise

• An increase in P0 is mapped into a considerably lower rise in the experienced SINR since such power boosting also means increasing the power of all users, and hence, the level of interference. This behavior will be evident in interference-limited cells.
• A similar effect would occur when changing the path-loss compensation factor for a given P0, i.e. a decrease in α would power down all UEs, which would lead to a lower SINR, a decrease than expected for a constant IoT level.
• There is a strong dependency between P0 and α on the expected SINR and IoT distributions. To keep the same IoT operating point, both P0 and α parameters require to be tuned at the same time, e.g. increasing P0 when α decreases.

The FPC Scheme Simulation Results and Analysis:

• Let us see the results of two simulation scenarios using open-loop power control [2]:

1. The CDF of the per-UE experienced average SINR is shown in the below plot, for different path-loss compensation factor (α) values where P0 is tuned so that the interference is kept constant. - In the first cell (Macro 1 plot), cell-edge users can meet the SINR requirements imposed by the FPC algorithm since there is no limitation in power as they are in an interference-limited scenario. In the same cell, cell-center users experience a higher SINR, which leads to higher MCS selection, and hence, higher throughput. The cell-center users can do a more pro?table use of power boosting than cell-edge users. If the interference level is lower, cell-edge users can easily ?t the SINR requirements set by the FPC scheme, as evident from the plot. However, a higher IoT means an overall power increase, which makes cell-edge users transmit at maximum power in any case. So users start to be limited by power, diminishing the effects of the FPC algorithm. For example, if the P0 is fixed to -79dBm, received SINR = P0 -L - IoT - N, N is fixed more or less, L is variable but as P0 is fixed, transmit power will be boosted proportionately. Assuming a constant level of interference and noise, the per-UE experienced SINR distribution is directly obtained from the path-loss distribution scaled according to the OLPC parameters like P0/Alpha/IoT.- In the second cell (Macro 3 plot), where noise becomes the main constraint, cell-edge users are in any case transmitting at maximum power so they tend to experience a similar SINR regardless of the path-loss compensation factor. This behavior predicts a less noticeable effect of the FPC algorithm on the outage performance. For example, if the P0 is fixed to -79dBm, received SINR = P0 -L - IoT - N, L is variable but as P0 is fixed, transmit power will be boosted proportionately. Compared to the first cell, which can meet the required SINR, though interference is limited, all cell edge UEs perceived different received SINR because there is some TX power headroom left out for boosting throughput. But in the case of the second cell, the users have to compensate for high noise, so all users perceived similar received SINR, with no TX power headroom left out for boosting throughput. The below plot illustrates it [2].

2. The performance of the FPC algorithm is analyzed, where the cell capacity (a) and coverage (b) are plotted as a function of the average IoT for different α values in the first cell scenario. It can be seen that an FPC approach translates into a gain in cell throughput but penalizes cell-edge performance for a given IoT operating point. This is simulated in a cell where there is no limitation in power as they are in an interference-limited scenario. The below plot illustrates it [2].

In the above plot for Average IoT vs cell capacity, alpha=0.6 gives the better capacity at 14dB IoT, one of the reasons is the P0 is chosen very high as -58dBm, but at the same, it gives less coverage at 14dB IoT compared to alpha=0.8. If we want to choose the best capacity-giving parameters then alpha=0.6 is good, but if we want to give both capacity and coverage, alpha=0.8 is good. Whereas the alpha=0.8 gives balanced capacity and coverage at 12dB IoT.

Why should alpha=0.6 should perform better in capacity and poor in coverage at 14dB IoT? The main reason is P0 =-58dBm, which is too high power, so the received SINR is increased for all users. But alpha is low 0.6, so it splits SINR differentiation between cell center/middle and edge users. As the cell edge users' SINR is low, it directly impacts the coverage. At the same time the increased SINR for cell center/middle users due to high P0 and low alpha, is translated to high capacity at the cost of cell edge users PRB share.

Why should alpha=0.8 should fetch balanced good capacity and coverage at 12dB IoT? All the users received SINR increased because of high P0=-58dBm, and at the same time alpha=0.8 which enabled the cell edge user to obtain good receive SINR so they can achieve comparatively good throughput than alpha=0.6, but at the cost of cell center/middle users. As the PRB share of cell center/middle users is taken by cell edge users, the capacity does not match the alpha=0.6, because the cell edge users operate at lower MCS than cell center/middle users. At the same time due to good SINR for cell edge users in alpha=0.8, there is a significantly better coverage than alpha-0.6.

Why does the change in the alpha parameter impact the coverage? If a cell edge user is seeing a 130dB path-loss, for alpha=0.8, it will compensate 104dB, and the remaining 26dB is dynamically handled, so the received SINR = P0 - (1-Alpha)PL - I-N, -58dBm - 26dB - I-N = -84dBm - I-N. This cell edge user will have high I+N and 26dB (from the above example) remaining PL to compensate. If a cell edge user is seeing a 130dB path-loss, for alpha=0.6, it will compensate 78dB, and the remaining 52dB is dynamically handled, so the received SINR = P0 - (1-Alpha)PL - I-N, -58dBm - 52dB - I-N = -110dBm - I-N. This cell edge user will have high I+N and 52dB (from the above example) remaining PL to compensate. So, there is a difference in received SINR for alpha 0.6 vs 0.8, the alpha=0.6 got low SINR. Because of the low SINR in the alpha=0.6, the coverage is significantly impacted. Though these kinds of difference in SINR is evident in the above open loop power control-based simulation results only; whereas in the closed loop power control TPC commands will be issued to the cell edge user to compensate for I+N and path-loss, in that case, the difference will not be seen.

For example, alpha=0.6 achieved good capacity compared to any other configured alphas in the simulation result, but at the IoT=14dB, at the same time, 14dB IoT impacted the coverage significantly, refer to the below table [2]. Whereas the alpha=0.8 achieves balanced capacity at 12dB IoT and for 12dB IoT it can give better coverage.

Water-Filling Strategy:

This is the strategy to handle fast-fading in the wireless channel. In the slow fading scenario, we are interested in achieving a target data rate within a coherence-time period of the channel. In the fast-fading case, one is now concerned with the rate averaged over many coherence-time periods.

In the fast fading, the wireless channel varies quickly with the frequency. Fast fading originates due to the effects of constructive and destructive interference patterns which is caused due to multipath. Doppler spread leads to frequency dispersion and time-selective fading. The multipath and Doppler shift are the types of fast-fading.

The coherence time is the time duration over which the channel is constant and is considered to be not varying. Such channel variation is much more significant in wireless communications systems, due to Doppler effects.

• In the water-filling strategy in general, the transmitter allocates more power when the channel is good, taking advantage of the better channel condition, and less or even no power when the channel is poor. This is precisely the opposite of the channel inversion strategy.
• In the fast fading case, the channel over N such coherence periods can be modeled as a parallel channel with N sub-channels which fade independently. In the context of the frequency-selective channel, water-filling is done over the OFDM sub-carriers or PRBs; in this discussion, we discuss the water-filling which is done over time. In both cases, the basic problem is that of power allocation over a parallel N sub-channel. The optimal power P` allocated to the Nth coherence period depends on the channel gain h in that coherence period; hence, the optimal power at any time depends only on the channel gain h at that time.
• The water-filling capacity requires a variable-rate coding scheme; this scheme consists of a set of codes of different rates, one for each channel state or coherence period. When the channel is in state h, the code for that state is used. This can be done since both the transmitter and the receiver can track the channel. A transmit power of P is used when the channel gain is h. The rate of that code is therefore dependent on power, channel gain, and others. No coding across channel states is necessary. Thus, knowledge of the channel state at the transmitter not only allows dynamic power allocation (DPC) but simplifies the code design problem as one can now use codes designed for the AWGN channel.

Dynamic Power Allocation (DPC) Scheme:

• DPC is designed for more rapid adaptation to changing conditions, which can be beneficial in scenarios where the channel conditions vary quickly, such as in a mobile environment. DPC typically adjusts the transmission power in larger steps or dynamically adapts the power based on the specific conditions of the communication channel.
• The underlying assumption is that the CSI or UL channel, over the entire system bandwidth, is available at the base station every TTI, for all active UEs in the corresponding cell, with a given resolution in the frequency domain and a certain Gaussian-distributed error. So, in the 5G UL OLLA, an operating SINR will be maintained from this measured CSI, and it will be converted to MCS for the current PUSCH transmission scheduled, the Operating SINR/MCS combo can be used throughout the coherence period. Due to SINR measurement errors and interference variability, the Block Error Rate (BLER) tends to deviate from the target. To compensate for such bias, an UL OLLA algorithm can be implemented to give corrections to the operating SINR value used for choosing the proper MCS, for example, SINR can scaled up and down in a granular scale in response to BLER, to compensate for the bias.
• Generally, the channel side information at the receiver (CSIR) under Rayleigh fading is considered for the following analysis. At high SNR, however, the capacity is insensitive to the received power per degree of freedom, and varying the amount of transmit power as a function of the channel state yields a minimal gain, refer to the below figure. At low SNR, the capacity is quite sensitive to the received power (linear, in fact) and so the boost in received power from optimal transmit power allocation provides significant gain. At low SNR, the water-filling strategy transmits information only when the channel is very good, when there is very little amount of water, the water ends up at the bottom of the vessel (refer to the below figure). Thus, dynamic power allocation is more important in the power-limited (low SNR) regime than in the bandwidth-limited (high SNR) regime.

• Since the AWGN capacity is the same as the CSIR capacity at low SNR, this leads to the interesting conclusion that with full CSI, the capacity of the fading channel can be much larger than when there is no fading. This is in contrast to the CSIR case where the fading channel capacity is always less than the capacity of the AWGN channel with the same average SNR. The gain comes from the fact that in a fading channel, channel fluctuations create peaks and deep nulls, but when the energy per degree of freedom is small, the sender opportunistically transmits only when the channel is near its peak. In a non-fading AWGN channel, the channel stays constant at the average level and there are no peaks to take advantage of. For models like Rayleigh fading, the channel gain is unbounded. Hence, theoretically, the gain of the fading channel water-filling capacity over the AWGN channel capacity is also unbounded. However, to get very large relative gains, one has to operate at very low SNR. Overall, the performance gain from full CSI is not that large compared to CSIR, unless the SNR is very low.
• While the water-filling strategy enhances the long-term throughput by transmitting during favorable channel conditions, it introduces a significant delay. In this context, it is valuable to compare the water-filling power allocation strategy with the channel inversion strategy. Channel inversion, although less power-efficient than water-filling, expends substantial power to invert the channel during unfavorable conditions. However, the information flow rate remains consistent across all fading states, resulting in a delay that is independent of channel variations. Consequently, the channel inversion strategy can be seen as a power allocation strategy limited by delay. Under an average power constraint, the maximum achievable rate using this strategy can be considered a delay-limited capacity. For applications with stringent delay requirements (like voice traffic, which has a stringent delay requirement and requires a consistent throughput), this delay-limited capacity may offer a more relevant performance metric than the water-filling capacity.
• Lack of channel diversity typically leads to a limited delay-limited capacity. Conversely, with an increase in channel diversity, the chances of encountering a weak channel decrease, resulting in reduced average power consumption required to maintain a specific delay-limited rate. To put it simply, greater channel diversity under a constant average power constraint enhances the delay-limited capacity.

Coherence time and prediction uncertainty:

Effective rate adaptation (OLLA) hinges on precise tracking and forecasting of the channel at the transmitter. This is feasible only when the coherence time of the channel significantly exceeds the interval between when the channel is assessed at the mobile/base station and when the packet is transmitted at the base station or mobile.

• When walking at a speed of 3 km/h with a carrier frequency fc=1.9 GHz, the coherence time is approximately 25 ms, allowing for accurate channel prediction. However, at a slow driving speed of 30 km/h, the coherence time decreases to 2.5 ms, making precise channel tracking challenging. At a faster speed of 120 km/h, the coherence time drops below 1 ms, rendering channel tracking impossible, and transmitter CSI unavailable.
• In the low SNR regime, where h follows the stationary distribution of the fading. Thus, to determine an appropriate transmission rate across this fast-fading channel, it suffices for the mobile/base station to predict the average SINR over the transmission time of the packet, and this average is quite easy to predict. This is the reason UL OLLA filters the operating SINR before mapping the SINR to MCS.
• The challenging scenario lies in the intermediate zone between very slow and very fast-fading scenarios. In this range, there is substantial uncertainty in channel prediction, yet the time diversity over the packet transmission time is not extensive. This channel uncertainty has to be taken into account by being more conservative in predicting the SINR and in requesting a rate. This resembles an outage scenario, with the distinction that channel randomness is conditional on the predicted value. The requested rate should be configured to achieve a specified target outage probability. This is the reason the OLLA's IIR filter co-efficient is very important and needs to be conservative. For example, the IIR filter coefficient of 10% or 0.1, is more conservative than using the filter coefficient of 20% or 0.2. The different roles of coding in the predicted SINR: when the predicted SINR is accurate, the main role of coding is to combat the additive Gaussian noise; and, if the predicted SINR is not accurate, the coding combats the residual randomness in the channel by exploiting the available time diversity.
• To reduce the loss in performance due to the conservativeness of the channel prediction, the hybrid-ARQ mechanism for the repetition-coded multiple slot packets; this way, a rate higher than the requested rate can be achieved if the actual SINR is higher than the predicted SINR. The HARQ Incremental Redundancy allows for dynamic adaptation of transmission parameters to the actual channel conditions, and if the observed SINR is better than predicted, the sender can increase the transmission rate in subsequent retransmissions. If the actual SINR is better than the previous transmission's predicted SINR, the sender can use a higher MCS, which may involve higher-order modulation or more aggressive channel coding. If the actual SINR is worse than the previous transmission's predicted SINR, the sender might choose a more robust MCS with a lower data rate but better error correction capabilities. This flexibility is crucial for achieving higher throughput while maintaining reliable communication in varying and dynamic wireless environments.

UL PC/OLLA Interaction

It is important to understand how the UL PC and OLLA interact. There are two types of UL PC and OLLA interaction:

• Unsynchronized?UL PC/OLLA,
• Synchronized UL PC/OLLA.

The following explanation aims to provide a typical implementation perspective; however, it's important to note that the design aspects listed below are not obligatory. One can always apply intelligent alternatives that may circumvent the mentioned design aspects.

Unsynchronized?UL PC/OLLA:

UL PC will independently compute the power up and down, without interacting with UL OLLA. After UL PC is executed, UL OLLA will take the power up/down input and compute whether UL OLLA has to increase the number of PRBs by reducing the MCS (OR) or decreasing the number of PRBs by increasing the MCS. Both operations are to achieve better throughput than the current Max PRB allocation possible with UL OLLA computed MCS.

In unsynchronized UL PC/OLLA interaction, the UL PC uses the fractional power control (FPC) and channel inversion strategy; where the alpha will be configured to compensate for the path-loss partially and the FPC will permit the cell center UE to transmit in a low power per PRB, whereas the cell edge UE will be allowed to transmit in high power per PRB. This FPC strategy will be achieved by computing a Received-SINR and Target-SINR for a PRB for each UE, these two SINR differences will be compensated in small fractions or steps by UL PC independently without interacting with UL OLLA.

In FPC, the crucial distinction in UL PC lies in the requirement for the Received-SINR to be an instantaneous SINR. This is vital for tracking the instantaneous channel gain and compensating for the two SINR differences. This approach is effective because UL PC does not interact with UL OLLA information, such as Modulation and Coding Scheme (MCS) and the current number of Physical Resource Blocks (PRBs) for Physical Uplink Shared Channel (PUSCH). Whereas, this information is maintained using an Infinite Impulse Response (IIR)-filtered SINR in UL OLLA.

FPC is characterized by small, incremental adjustments to power, providing a smoother and more gradual response to changing conditions; for this reason, FPC uses "Accumulated TPC" commands in the UL PC algorithm.

These unsynchronized UL PC/OLLA interacting FPC power control algorithms are better suited for delay-limited application traffic for example, Conversational Voice/Video, Real-time gaming, Mission Critical Voice/Video, and Live uplink streaming, demanding Guaranteed Bit Rate (GBR).

Synchronized UL PC/LA:

UL PC and LA work in a synchronized way. The power up/down computation is coupled with increasing/decreasing the number of PRBs and MCS. For example, if 1 dB power is increased, whether we have to reduce the MCS and increase PRBs OR keep the same PRBs and increase the MCS computed. And vice versa, if 1 dB power is decreased, whether we have to decrease the PRBs and increase the MCS OR whether to keep the same PRBs and decrease the MCS computed.

In synchronized UL PC/OLLA interaction, the UL PC uses the dynamic power control (DPC) and channel inversion strategy; where the alpha will be configured to compensate for the path-loss partially and the DPC will permit the cell center UE to compensate for all the channel gain power per PRB, whereas the cell edge UE will be allowed to transmit in high power per PRB when the channel state h is very good and transmit with low power per PRB when the channel state h is not very good. This DPC strategy will be achieved by computing a Received-SINR and Target-SINR for a PRB for each UE, these two SINR differences will be compensated by UL PC after interacting with UL OLLA based on a condition. For example, for a low SINR user, if the two SINR difference is a positive value, the UL PC has two option whether to boost the power or not, this decision will be made on the condition whether the power boost fetches significant data rate, means the low SINR user is in a good channel state h. For a high SINR user, the positive difference value will always fetch a significant data rate, where the user's chances of being in a good channel state are high; so the power boost is always the case.

In DPC, the primary distinction in UL PC lies in the requirement for the Received-SINR to be an IIR-filtered SINR. This is because UL PC interacts with UL OLLA information, such as MCS and the current number of PRBs for the PUSCH to decide on the power up and down. As previously discussed, UL OLLA SINR is filtered, and MCS is adjusted based on the Block Error Rate (BLER). Maintaining two Received SINR values, one for UL PC and one for UL OLLA for power adjustment decisions, is neither feasible nor efficient.

DPC involves more dynamic and potentially larger adjustments to power, offering a faster response to rapid changes in the radio environment; for this reason, DPC uses "Absolute TPC" commands in the UL PC algorithm.

These synchronized UL PC/OLLA interacting DPC power control algorithms are better suited for non-delay limited application traffic, for example, Voice/Video Buffered streaming, TCP-Based data, and Mission Critical data, demanding Non-Guaranteed Bit Rate (NGBR).

Conclusion

We have explored the fundamental physical layer principles underlying each design aspect employed in typical Uplink Power Control (UL PC) and Outer Loop Link Adaptation (OLLA) algorithms implemented in commercial products.

The challenge lies in determining the most effective UL PC/OLLA strategy/power control type, and interaction, for both slow-fading and fast-fading channel conditions, considering various UE geometries such as Cell Center, Middle, and Edge. Making informed decisions on power, Modulation and Coding Scheme (MCS), and Physical Resource Blocks (PRBs) adjustments is the secret sauce for any carrier-grade UL PC and OLLA in a product.

References

[1] Uplink Power Control for MIMO-OFDMA Cellular Systems (zte.com.cn)

[2] (PDF) Performance of uplink fractional power control in UTRAN LTE (researchgate.net)

[3] web.stanford.edu/~dntse/papers/book121004.pdf