Understand receiver dynamic range and AGC tradeoffs in WiMAX femtocells and picocells (Part 2 of 2)

Optimal MAC scheduling and resource allocation to improve receiver dynamic range
The calculations shown in the Part 1 of this article assumed a –45 dBm maximum input signal. As with the noise figure (NF) requirement in the standard, it is quite possible that the maximum signal level requirement is inadequate for a performance-differentiated solution. Further, while the range calculations from Part 1 predict acceptable performance, the increase in NF due to the large input signal does require that mobiles far from the base station transmit at power that could interfere with other networks. The MAC can provide margin for this issue.

Uplink subframe
The uplink subframe can be divided into a maximum of three zones. The uplink zones that need to be supported (WiMAX Forum Mobile Radio Conformance Tests–Wave 2 Amendment v1.0.1) are the uplink partial usage of subcarriers (UL PUSC) and band adaptive modulation and coding (AMC). In the UL subframe, each zone can consist of signals from multiple customer premises equipment sites (CPE), each with its own burst profile. The burst profile used by each mobile subscriber is specified by the base station MAC based on channel quality measurements. The UL-MAP transmitted in the datalink subframe specifies the burst profile to be used by each mobile subscriber in the cell.

Uplink permutation zones
At the PHY level, the UL PUSC symbol structure has a tile structure, where each tile contains four subcarriers over three OFDM symbols (8 data and 4 pilot subcarriers). A subchannel consists of 6 tiles, which translates to 24 subcarriers over 3 symbols. The subcarriers that form the subchannel are randomly allocated across the subset of used subcarriers and are nonadjacent.

At the MAC layer, the smallest time/frequency allocation allowed is a slot of 48 data subcarriers over three OFDM symbols for UL PUSC. This restricts the smallest allocation in time to 3 symbols. The allocation and mapping of MAC slots in the UL PUSC zone is based on the algorithm specified in 8.4.3.4 of the 802.16e specification (IEEE 802.16e-2005 Amendment and Corrigendum to IEEE Standard 802.16-2004). The mapping of slots to the subchannels occurs such that slots are allocated to increasing symbol index. Once the end of the zone is reached, the mapping continues in the next available subchannel in the lowest numbered symbol.

The band AMC 2×3 symbol structure requires adjacent subcarrier permutation for the allocations. Here, the basic allocation unit is a set of 9 contiguous subcarriers referred to as a bin. A subchannel is defined as 2 bins over 3 symbols for AMC 2×3. The MAC slots are allocated in band AMC with the help of a subchannel offset index in the UL MAP. The slots can be allocated with increasing subcarrier index over the 3 symbol duration for AMC 2×3. Once the data has been allocated over the first 3 symbols in the zone, the allocation moves to the next set of 3 symbols.

An UL subframe can contain up to a maximum of three zones (either UL PUSC or band AMC zones).

MAC Allocations based on mobile subscriber distance from base station
Since femtocells do not need to support a very large number of users, it is possible for the MAC to schedule access to near and far users on the uplink such that the receiver's gain is optimally set for each group of users. Thus, the MAC scheduling could be modified to improve the dynamic range of the base station receiver, which will consequently result in a wider coverage area for the base station.

Consider a scenario with 2 users (user A and user B) close to the base station and two users (C and D) at the maximum distance from the base station (Figure 1).

Figure 1: Figure 1: User distances from base station
(Click on image to enlarge)

The base station MAC allocates resources and chooses a burst profile based on channel quality measurements and bandwidth requirements of an mobile subscriber. The MAC resource allocation and scheduling algorithm is chosen based on system performance requirements and the designer is not restricted by the standard to any specific implementation.

To improve the dynamic range on the receiver base station, and hence the coverage area of the base station, it is proposed that the MAC allocate zones such that the near users are allocated slots in one zone and the far users in a different zone.

Figure 2 shows a possible technique that could be employed for resource allocations by the MAC.

Figure 2: MAC Scheduling of users allocated based on channel quality and distance of users from base station. User A and user B are close to the base station and users C and D are at the edge of the cell area.

(Click on image to enlarge)

The MAC does its scheduling and allocates resources based on channel quality measurements it initiates. The channel quality measurement, which includes RSSI and SINR measurements, will correlate to the relative distances of the individual users from the base station.

For example, if users A and B have an average SINR of 20 dB and 18 dB, respectively, and users C and D have SINR of 10 dB and 8 dB, respectively, this would indicate that A and B are close to the base station and C and D are at some distance. The SINR could also depend on other factors (e.g. presence of walls and floors), but these factors can be treated as an increased separation between mobile subscriber and base station.

As shown in Figure 2, users A and B are allocated subchannels or slots in the first 6 symbols and in the last 6 symbols (symbols 13 through 18), since these two users are closer to the BS. Users C and D, which are at the edge of the cell, are allocated slots or subchannels in symbols 7 through 12. The allocation described above is possible in the UL subframe for both UL PUSC and for band AMC zones.

For band AMC zones, since the allocations are based on adjacent subcarrier permutations, user allocation as described above would be a relatively straightforward operation. For UL PUSC, since the 802.16e standard requires MAC slot allocations with increasing symbol index, user separation would only be possible if near and far users can be separated into different UL zones (i.e. one UL PUSC zone could be allocated to the near users and another PUSC zone to the users that are further from the base station). Increased demand or bandwidth from a certain set of users could be accommodated by increasing the duration of the PUSC zone servicing those users.

At this point, it is important to consider the ranging channel. The ranging channel is used for initial ranging, periodic ranging, bandwidth request and handover ranging and could be active over a whole UL subframe. The ranging channel consists of 6 subchannels out of a total of 35 subchannels for UL PUSC (1024 FFT). For optional UL PUSC, the ranging subchannel consists of 8 subchannels out of a total of 48 subchannels. A mobile subscriber transmits the 144 length code for 2 symbols or two unique codes over 4 symbols. The ranging channel, especially for a femtocell or picocell application should be active infrequently in the uplink.

It is also important to distinguish between the total received power and the power density (or power/subcarrier) that a single mobile subscriber can transmit. A mobile subscriber that is very close to the base station may be transmitting at a high power density but not on every subcarrier. For example, with 841 used subcarriers, an average power of –45 dBm for the WiMAX signal equates to a power density of 37.6 pW/subcarrier. With the power level at 37.6 pW/subcarrier, an mobile subscriber that's transmitting on only 1 subchannel (24 subcarriers) would produce –60 dBm at the base station input.

A MS transmitting such that the average received power is –95 dBm (minimum sensitivity for QPSK) equates to a power density of 0.38 fW/subcarrier. If 29 of the subchannels are received at a power density of 0.38 fW/subcarrier and 6 subchannels are received at 37.6 pW/subcarrier, the average input power is –53 dBm at the receiver input for those 2 to 4 symbols that include ranging signals from a near MS. Since the base station has a priori knowledge of the start of the ranging channel, the BBIC should necessarily allow the AGC to lock during the CP time of the first symbol of the ranging channel. Figure 3 shows that a gain of 63 dB results in sufficient SNR at minimum sensitivity (–95 dBm for QPSK and –85 dBm for 16QAM-3/4) and sufficient EVM for a signal at –53 dBm input level.

Figure 3: Receiver EVM vs input power and receiver gain (Click on image to enlarge)

The following section shows how the AGC can be used to lock to the changing signal power levels at the input to the receiver of a transceiver, such as Analog Devices' AD935x, multiple times within a UL subframe.

Automatic gain control
The automatic gain control (AGC) algorithm on the receiver incorporates some efficient methodologies to settle at the optimal gain in the smallest time duration possible. Some of the salient features of the gain control algorithm are:

• AGC can be used to lock onto a new gain every symbol if required.
• Time to lock the AGC gain will depend on input signal variation from symbol-to-symbol, but the gain lock time should be less than the duration of the cyclic prefix.
• A maximum gain index word can be programmed to allow the user to customize the range of the receiver for fast gain locking.
• The lock level or range of the AGC can be specified such that the final settling gain of the AGC is based on the lock level word.

The AGC also allows real-time control of the gain with the help of the EN_AGC pin. This signal, when low, freezes the gain of the receiver. When the signal goes high, the gain freeze is released and the AGC attempts to reacquire gain lock (based on channel conditions and input signal levels). The time to acquire gain lock depends on changes in input power levels. If the change in input power levels is small from when the AGC last locked the receiver gain, the gain lock time could be small. Figure 4 shows a proposed method of operating the AGC for maximizing dynamic range on the receiver.

Figure 4: AGC relocking every 3 symbols. CP time is 11.2 μs for 10 MHz mode. Time required to lock the gain is considerably less than CP time (≈3-6 μs)
(Click on image to enlarge)

If the MAC allocates uplink resources through the UL-MAP based on the scheme described in Figure 2, then the EN_AGC pin controlled by the BBIC can be used to create an extremely fast symbol locking AGC. From Figure 4, the EN_AGC pin goes high at the start of the UL subframe. The AGC settles to the optimal gain index (depending on average input signal level and the programmable lock level), locks the gain before the EN_AGC pin goes low, effectively freezing the receiver gain.

After freezing the gain for three consecutive symbols, the EN_AGC signal from the BBIC is driven high during the CP time (11.2 μs for a 10 MHz BW signal) of the fourth symbol. The AGC state machine resets, the AGC calculates a new gain index depending on the input power level and locks the gain (as shown by the gain lock signal from the AGC state machine going high). All this occurs in a time period less than the CP time. At the end of the CP time, the BBIC takes the EN_AGC signal low effectively freezing the gain for the next three symbols.

The three symbol window used here as an example could be changed to a six symbol window or larger. The BBIC has absolute control of how often the AGC needs to relock within a UL subframe.

The MAC allocation scheme and the 'fast AGC locking operation' described above, if implemented, could result in an improved dynamic range for the receiver. This range would exceed the RCT specification and allow for increased cell coverage.

Conclusion
As noted above, the femtocell must meet the necessary technical requirements while remaining affordable to consumers. This paper shows that by using an integrated WiMAX transceiver; it is possible to meet the RCT specification. If increased margin (over and above the RCT) is required, the modified MAC allocation algorithm in combination with the fast locking AGC operation described above can be used to minimize dynamic range requirements and extend usable range. The above analysis can also be applied to the picocell case. The main consideration is that the nearest MS would be at a greater distance than what was assumed for the femtocell analysis thus affecting the dynamic range calculations; however, the proposed MAC algorithm should remain valid for the picocell case.

About the authors
Tony Montalvo is the Director of Analog Devices' Raleigh Design Center in Raleigh, North Carolina. Before joining Analog Devices, Tony. led the RF IC group at Ericsson Inc and was involved with the design of Flash memories at Advanced Micro Devices.. He received a B.S. in Physics from Loyola University in 1985, an M.S.E.E. from Columbia University in 1987 and a Ph.D. from North Carolina State University in 1995 where he is also an Adjunct Professor.

Manish Manglani is a systems design engineer focusing on wireless infrastructure products at Analog Devices Inc. He graduated with a Master's degree in Electrical Engineering from Virginia Tech in 2001.

Chris Cloninger is the marketing manager for the WiMAX transceiver products at Analog Devices Inc. Previously, Chris worked as the Marketing and Application Manager for Analog Devices' High Speed Analog-Digital Converters group. He holds a Bachelor's degree in Computer Engineering from Clemson University.

For questions or comments, please send an email to wimaxtransceivers@analog.com.