This disclosure relates to control of reverse link rate and stability in a multi-user wireless communication system.
In a radio access network, such as a cellular network, it is often important to ensure that sectors in the network do not become overloaded. One way in which a sector can become overloaded is if many access terminals (e.g., cellular devices) in a sector transmit at high data rates, which increases the transmission power. In a first Evolution Data-Only (1xEV-DO) network, the 1xEV-DO protocol provides mechanisms for measuring sector load (referred to as a “rise-over thermal (ROT)” measurement) and reducing the transmission rates of ATs in the sector by transmitting a bit (referred to as the “reverse activity bit”) to ATs in a sector that is becoming overloaded. However, ROT measurement may not always be not available, and, in these cases, a radio access network may fail to detect and mitigate overloading of a sector.
In one aspect, the invention features a method for controlling the rate of transmission of an access terminal in a radio access network (e.g., a 1xEV-DO network) that includes, for one or more sectors in which the access terminal has a connection, determining a number of connections for the sector, and setting a rate limit for the access terminal based on one or more of the determined number of connections.
Implementations may include one or more of the following features. The method may include determining a number of connections for each sector in which the access terminal has a connection (e.g., sectors in which an AT is in soft or softer handoff). The method may also include determining a number of connections for a subset of sectors in which the access terminal has a connection (e.g., only those sectors in which the access terminal has a connection serviced by a common radio node).
Determination of a number of connections may include determining an effective number of connections for a sector. Determining an effective number of connections for a sector may include, for each connection in a sector, applying a weight to one or more characteristics of the connection to determine a weighted connection, and summing the weighed connections.
Setting a rate limit for the access terminal based on one or more of the determined number of connections may include using a rate-limit table to set a rate limit. A rate limit table may, for example, assign a first rate limit if a number of connections determined for the access terminal is within a first range of values, and a second rate limit if a number of connections determined for the access terminal is within a second range of values. If a number of connections is determined for multiple sectors in which an AT has a connection, the method may set a rate based on an determined number of connections for a sector having a value equal to or greater than any other effective number of connections determined for other sectors.
The method may including changing the way in which a number of connections is determined for a sector (e.g., applying different weights to connections in a sector) and/or a rate limit table used to set rate limitations based on the way in which the radio access network performs its RA bit estimation. If it uses ROT measurement to set or clear RA bits for ATs, the method may use one rate limit scheme, whereas if ROT measurement is not available (and the system uses, for example, post-automatic gain control-ROT estimation), then the method may use a different rate limit scheme. The method may include transmitting a rate limit message via a broadcast message or a unicast message to the access terminal. The method may also include transmitting to the access terminal a rate limit message immediately upon the access terminal establishing a connection in a sector.
In another aspect, the invention features a radio node configured to transmit radio signals to and receive radio signals from an access terminal having one or more connections in a group of one or more sectors of a radio access network. The radio node includes a processor and a medium bearing instructions to cause the processor to determine a number of connections for one or more sectors in which the access terminal has a connection and set a rate limit for the access terminal based on at least one of the determined number of connections.
Implementations may include one or more of the following features. The radio node may be configured to determine a number of connections for each of sector in the group of sectors in which the access terminal has a connection. If the radio node determines multiple numbers of connections for sectors in which the access terminal has a connection, the radio node may be configured to base a rate limit on the greatest determined number of connections.
The radio node may be configured to receive (e.g., from another radio node or from an radio network controller) an effective number of connections for other sectors in the radio access network in which the access terminal has a connection but is not in the group of sectors serviced by the radio node.
The radio node may be configured to determine an effective number of connections by, for example, weighting characteristics of each connection in a sector and summing the weighed connections. The radio node may be configured to set a rate limit based on a rate limit table that, for example, specifies a first rate limit if a number of effective connections determined for the access terminal is within a first range of values and specifies a second rate limit if a number of effective connections determined for the access terminal is within a second range of values. The radio node may also be configured to transmit a rate limit to the access terminal (e.g., via a broadcast or unicast message).
In another aspect, the invention features a system for controlling the rate of transmission of an access terminal in a radio access network that includes a processor and a medium bearing instructions to cause the processor to determine an effective number of connections for one or more sectors in which the access terminal has a connection and set a rate limit for the access terminal based on at least one determined effective number of connections.
In one specific implementation, the processor of the system may be part of a radio network controller. In cases where the processor is part of a radio network controller, the instructions may be written to cause the processor to determine a number of connections for each of sector in which the access terminal has a connection. If the access terminal has connections in multiple sectors, the radio network controller may set a rate limit based on the sector with the greatest number of connections.
In another particular implementation, the processor of the system may be part of a radio node that services a group of sectors. The instructions may be written to determine a number of connections for each of the sectors in the group of sectors in which the AT has a connection. The radio node may be configured to set a rate limit based only on this determination. The radio node may also be configured to receive information (e.g., from an RNC or from other RNs) on numbers of connections in sectors other than the group of sectors with which the AT has a connection and may be configured to set a rate limit based also on these numbers of connections. The system may also include a radio network controller that is configured to determine a number of connections for sectors outside the RN's group of sectors with which the AT has a connection. The radio network controller may be configured to also send the AT a rate limit message based on these numbers.
Referring to
The radio access network 20, which may cover a large service area, includes one or more Access Sub-Networks (ASN's), e.g., ASN 22, each anchored by a Radio Network Controller (RNC) 34 communicating with several Radio Nodes (RN's) 10, 12, 14 using a private or public IP backhaul network 32. Each RN may support multiple sectors, such as the three sectors shown in
ASN 22 is connected over a public or private IP network 36 to one or more Packet Data Serving Node's (PDSN's), e.g., PDSN 38. The PDSN, in turn, receives and transmits data packets (e.g., voice over IP packets) to a server 42 via the Internet 40. In some implementations, the functions of a PDSN and an RNC are combined into a single device.
Each AT is in communication with a radio node, e.g., RN 10, via an air link 28a, 28b. An air link comprises a forward link, which carries data transmitted from an RN to an AT, and a reverse link, which carries data transmitted from the AT to the RN. As an AT moves from one sector to another sector serviced by the same RN, it undergoes a “softer handoff” between the sectors. Similarly, when an AT moves from one sector to another sector serviced by different RNs, it undergoes a “soft handoff” between the RNs. When an AT is in soft or softer handoff, it will have connections in multiple sectors.
As shown in
As shown in
The pole capacity is the theoretical capacity of access terminals supported by a radio node. The pole capacity of an RN is a function of the signal-to-noise ratio (SNR) necessary to achieve a certain aggregate data rate with a certain aggregate error rate within a cell. Pole capacity is a theoretical maximum capacity of a cell and it is often advisable to design radio access networks in which the number of active ATs within a cell is limited to some percentage of the pole capacity, e.g., 50% to 75% of the pole capacity, which corresponds to an Rise-Over-Thermal (ROT) of 3 or 6 dB respectively.
A radio node, such as RN 10 shown in
When a radio node is able to measure sector load using ROT, it can control load on the sector by setting (or clearing) the reverse activity (RA) bit in the reverse activity channel on the forward link. More particularly, the radio node compares the ROT value with a predetermined threshold, which is referred to as the RA bit threshold. The RA bit threshold corresponds to some percentage of nominal sector loading (typical values are about 50 to 60% of nominal sector loading). If the ROT value is above the RA bit threshold, then the radio node sets the RA bit, otherwise the radio node clears the bit.
The RN transmits an RA bit every RAB length slots over the RA medium access control (MAC) channel (which is a component channel of the reverse activity channel). When an AT receives data on a MAC channel with the RA bit set, the AT executes a “coin-flip” algorithm to determine whether to freeze or reduce its transmit rate. If the coin-flip has a first outcome, the AT freezes its transmit rate, if the coin-flip has a second outcome, the AT decreases its rate from its current rate to the next lowest rate defined by IS-856. By reducing the rate at which ATs transmit on the reverse link, ATs transmit at less power and cause less interference in the sector, which decreases the ATs usage of the sector's capacity.
In some cases, the RN's radio interface is not available, and, as a result, sector load cannot be measured using ROT. In this situation, sector load is measured by the RN's modem using post-Automatic Gain Control (AGC)-ROT estimation. A post-AGC-ROT estimate is not as accurate as a ROT measure because it typically does not account for interference from ATs that do not have active connections on that sector (interference generated by other sectors). Because of post-AGC-ROT estimation may be inaccurate, the RA bit may not get set when a sector is becoming overloaded. If the RA bit is not properly set, ATs in the sector are free to transmit at higher rates (and thus also higher powers), which may eventually lead to an overpowering of the reverse link. As the number of active ATs in the sector grow, the problem is exacerbated because as the number of active connections grows, the effective SNR of all ATs denigrates until eventually the ATs in the sector lose the reverse link.
To prevent overloading when ROT measurement is not available, RAN 20 (shown in
The generalized rate-weighting algorithm assigns weights to connections in a sector to determine an effective number of connections in the sector. If an AT is in softer handoff and has connections in two or more sectors serviced by the same RN, the generalized rate-weighting algorithm is applied to each of the sectors with which the AT has a connection. In some implementations, if the AT is in soft handoff and has connection in two or more sectors serviced by different RNs, the generalized rate-weighting algorithm is applied to each of the sectors with which the AT has a connection. As mentioned above, if an AT has connections in multiple sectors, the RAN may base the ATs rate limit on the largest effective number of connections of the sectors with which the AT has a connection.
The generalized rate-weighting algorithm provides systems engineers the ability to weight the following characteristics of connections in a sector:
In addition to variably weighting characteristics of connections in a sector, the generalized rate-weighting algorithm can be programmed to weight connections differently based on their transmit rate. Thus, for example, the soft/softer handoff factor of a connection transmitting at one rate can be weighted differently than the soft/softer handoff factor of another connection transmitting at a different rate.
After weighting the connections in a sector, the weighted connections in the sector are summed to obtain the effective number of connections in the sector. For example, in some implementations the generalized weight-rating algorithm is programmed according to Table I.
As shown, the generalized rate-weighting algorithm has been programmed only to consider whether a connection has data to send and ignores other characteristics of a connection. In addition, the generalized rate-weighting algorithm ignores connections that are transmitting at the lowest 1xEV-DO rate of 9.6 Kbps. Thus, if a sector has 10 connections and 4 of the 10 connections have data to send and of these 4 connections, one is transmitting at 9.6 Kbps, the generalized rate-weighting algorithm (as programmed according to Table I) will determine that there are 3 effective number of connections in the sector.
The RAN then sets a maximum rate limit for ATs in the sector based on the effective number of connections. If an AT has multiple connections for which an effective number of connections has been determined, the RAN may set the rate limit based on the largest effective number of connections of the sector with which the AT has a connection. In some implementations, the RAN uses the maximum rate limit table shown in Table II to set a maximum rate limit for the sector.
Thus, if the effective number of connections determined by the generalized rate-weighting algorithm is between 1 and 7, the RAN permits the ATs to transmit at the maximum 1xEV-DO rate of 153.6 Kbps. If the effective number of connections is between 8 and 59, the RAN limits the ATs in the sector to transmit at the second-highest 1xEV-DO rate of 76.8 Kbps. The current iteration of 1xEV-DO limits the number of connection in a sector to 59, which is why the table does not account for situations where the effective number of connections is over 59. However, future iterations of 1xEV-DO may increase the maximum number of connections in a sector, in which case, the rate limit table would be extended. Additionally, some implementations may use other rate limit tables to limit transmit rates of ATs based on characteristics of a sector (e.g., the hardware capabilities of a sector, the presence of objects like buildings that cause interference in a sector, etc.).
Some implementations may use any one or a combination of the connection characteristics described above to determine an effective number of connections. In addition, weighting of connection characteristics can vary depending on the transmit rate of the connection.
The RAN can be configured to transmit rate limit messages by way of unicast or broadcast messaging. For broadcast messaging, the RAN transmits a rate limit message to all ATs in a sector over the control channel on the forward link. For unicast messaging, the RAN transmits a rate limit message sent to individual ATs over the forward traffic channel on the forward link.
When an AT establishes a connection with a sector, it does so at the lowest rate, i.e., 9.6 Kbps, and will remain at this rate until it receives a rate limit message. Because it may take some time before the RAN transmits a rate limit message (e.g., a RAN may be configured to periodically broadcast rate limit messages in a sector), in some implementations the RAN is configured to send a rate limit message at the maximum rate of 153.6 Kbps immediately after a new connection is established to prevent an AT from needlessly remaining at a low transmit rate while waiting for a first rate limit message.
In some implementations determination of the effective number of connections and rate control is determined by the RNC. For example, as shown in
RNC 52 includes a reverse rate and stability control algorithm 63, which, as described above, uses the generalized rate-weighting algorithm to determine an effective number of connections for an AT over ALL sectors that the AT is in handoff with and then uses a rate limit table (e.g., Table II above) to set rate limits for the AT.
Rate limit messages 64 can be transmitted to ATs in the sector either via unicast or broadcast messaging. Unicast messages are reliable, but an individual message needs to be sent to each AT. This may increase the processing load on the RNC and signaling traffic between the RAN and the AT. If unicast messaging is used, the rate limit message can be transmitted either periodically or whenever there is a change in the rate limit.
Broadcast messages are best effort (and thus not as reliable as unicast messages). However, a broadcast rate limit message can be a single message per sector transmitted over the control channel of the sector. Thus, broadcasting conserves RNC processing resources and signaling bandwidth between RN and AT. As with unicast messaging, a broadcast rate limit message can be transmitted either periodically or whenever there is a change in the rate limit. It should be noted that broadcast rate limit messages normally need to be transmitted to the AT through its serving sector, and there can be ambiguity in choosing the serving sector when the AT is switching sectors. However, periodic broadcasted rate limit messages tend to diminish this problem. Furthermore, due to the format of a broadcasted rate limit message, non-serving sectors sometimes have to transmit rate limit information for the AT, and, therefore, in some implementations, rate limit information for an AT is broadcasted from all sectors in handoff with an AT targeted for a rate limit.
In some implementations, a reverse rate and stability control algorithm is implemented at the radio node. For example, as shown in
In the implementations illustrated in FIGS. 4 and 5A-5B a reverse rate and stability control algorithm may be executed for a particular AT in all of the sectors in which the AT has a connection. This can be considered an optimal approach. In some implementations, however, determination of the effective number of connections is performed only on a subset of the sectors with which an AT has a connection. For example, an RN may determine an effective number of connections for the sectors with which an AT has a connection AND are serviced by the RN. Thus, the RAN would not determine an effective number of connections for sectors in which the AT is in soft handoff. An approach that considers a subset of the sectors with which an AT has a connection can be considered a sub-optimal approach, but nonetheless can prevent sector overload without having to implement inter-RN messaging or additional RNC-RN messaging across the backhaul network. For example, as shown in
In some implementations, a hybrid approach between the optimal and sub-optimal approaches is taken. For example, as shown in
As described above, rate limit messages can be transmitted to affected ATs either view unicast or broadcast messaging. In some implementations, ATs that are not in soft or softer handoff receive rate limit messages via broadcast messaging from the radio node in which they have a connection, whereas ATs that are in soft or soft-softer handoff receive unicast or broadcast rate limit messages from the RNC. For ATs in soft handoff, the RN transmits a broadcast rate limit message corresponding to no handoff that will be overridden by the unicast message transmitted from the RNC. For connections in soft-softer handoff, the RN transmits a broadcast rate limit message corresponding to softer handoff that will be overridden by the unicast message transmitted from the RNC.
A simulation performed on an implementation using the sub-optimal approach with a reverse rate and stability control algorithm using a generalized rate-weighting algorithm programmed according to Table I above and a maximum rate limit table programmed according to Table II above. The simulation showed an improved reverse link sector throughput while still ensuring stability of the reverse link versus an implementation in which all ATs were limited to a transmit rate of 38.4 Kbps. Embodiments on existing systems select a conservative rate limit (such as 38.4 Kbps) to maintain reverse link stability. The simulations were based on the Strawman models provided the 3GPP2 standard bodies (3GPP2-1xEV-DO Evaluation Methodology (V1.3)) for wireless network simulation. Simulation results for a heavily loaded sector having 18 connections are shown in Table III.
Note that the simulation of a RAN using a reverse rate and stability control algorithm achieved better sector throughput while operating at acceptable ROT than that of a RAN using which limited the transmit rate to 38.4 Kbps for all ATs in a sector.
In a system where more than one method for ROT measurement and/or post-AGC-ROT estimation is available, the system may use different rate weightings in the generalized rate-weighting algorithm may differ depending on the ROT/post-AGC-ROT estimation method used. Similarly, the system may use different rate limit tables (e.g., Table II above) depending on the ROT/post-AGC-ROT estimation method used. For example, if the system uses highly accurate RA bit estimation method (e.g., ROT measurement), the system may employ one weight limit table, whereas if the system uses a less accurate RA bit estimation method (e.g., a post-AGC-estimation ROT method), it may switch to a more conservative rate limit table.
A number of embodiments of the invention have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the invention, and, accordingly, other embodiments are within the scope of the following claims.
Number | Name | Date | Kind |
---|---|---|---|
4031469 | Johnson | Jun 1977 | A |
4072900 | Ray | Feb 1978 | A |
5631604 | Dent et al. | May 1997 | A |
5828677 | Sayeed et al. | Oct 1998 | A |
5842140 | Dent et al. | Nov 1998 | A |
5857147 | Gardner et al. | Jan 1999 | A |
5884187 | Ziv et al. | Mar 1999 | A |
5923650 | Chen et al. | Jul 1999 | A |
5982760 | Chen | Nov 1999 | A |
6226525 | Boch et al. | May 2001 | B1 |
6507744 | Han et al. | Jan 2003 | B1 |
6567420 | Tiedemann, Jr. et al. | May 2003 | B1 |
6633552 | Ling et al. | Oct 2003 | B1 |
6697378 | Patel | Feb 2004 | B1 |
6711144 | Kim et al. | Mar 2004 | B1 |
6731618 | Chung et al. | May 2004 | B1 |
6741862 | Chung et al. | May 2004 | B2 |
6781999 | Eyuboglu et al. | Aug 2004 | B2 |
6904286 | Dantu | Jun 2005 | B1 |
6965564 | Coffman | Nov 2005 | B2 |
6983153 | Jain et al. | Jan 2006 | B2 |
6999425 | Cheng et al. | Feb 2006 | B2 |
7016686 | Spaling et al. | Mar 2006 | B2 |
7031141 | Kuriyama | Apr 2006 | B2 |
7082317 | Yano et al. | Jul 2006 | B2 |
7085570 | Tigerstedt et al. | Aug 2006 | B2 |
7120134 | Tiedemann et al. | Oct 2006 | B2 |
7120447 | Chheda et al. | Oct 2006 | B1 |
7136353 | Ha et al. | Nov 2006 | B2 |
7136666 | Charriere et al. | Nov 2006 | B2 |
7142548 | Fong et al. | Nov 2006 | B2 |
7170871 | Eyuboglu et al. | Jan 2007 | B2 |
7200391 | Chung et al. | Apr 2007 | B2 |
7206291 | Soldani et al. | Apr 2007 | B2 |
7242958 | Chung et al. | Jul 2007 | B2 |
7248875 | Schreuder et al. | Jul 2007 | B2 |
7248889 | Schwarz et al. | Jul 2007 | B2 |
7268674 | Bohler et al. | Sep 2007 | B2 |
7277446 | Abi-Nassif et al. | Oct 2007 | B1 |
7299278 | Ch'ng | Nov 2007 | B2 |
7305241 | Hirvonen et al. | Dec 2007 | B2 |
7366230 | Jonsson | Apr 2008 | B2 |
7466669 | Hosein | Dec 2008 | B2 |
20020021687 | Toki et al. | Feb 2002 | A1 |
20020072385 | Salvarani et al. | Jun 2002 | A1 |
20020111183 | Lundby | Aug 2002 | A1 |
20020193118 | Jain et al. | Dec 2002 | A1 |
20020196749 | Eyuboglu et al. | Dec 2002 | A1 |
20030072294 | Wei et al. | Apr 2003 | A1 |
20030083092 | Kim et al. | May 2003 | A1 |
20030092463 | Charriere et al. | May 2003 | A1 |
20030100311 | Chung et al. | May 2003 | A1 |
20040038697 | Attar et al. | Feb 2004 | A1 |
20040047305 | Ulupinar | Mar 2004 | A1 |
20040109424 | Chheda | Jun 2004 | A1 |
20040110534 | Chung et al. | Jun 2004 | A1 |
20040179494 | Attar et al. | Sep 2004 | A1 |
20040179525 | Balasubramanian et al. | Sep 2004 | A1 |
20040185868 | Jain et al. | Sep 2004 | A1 |
20040202136 | Attar et al. | Oct 2004 | A1 |
20040213182 | Huh et al. | Oct 2004 | A1 |
20040229604 | Fong et al. | Nov 2004 | A1 |
20050047375 | Kwon et al. | Mar 2005 | A1 |
20050107090 | Hosein | May 2005 | A1 |
20050107091 | Vannithamby et al. | May 2005 | A1 |
20050124369 | Attar et al. | Jun 2005 | A1 |
20050169301 | Jain et al. | Aug 2005 | A1 |
20050213555 | Eyuboglu et al. | Sep 2005 | A1 |
20050243749 | Mehrabanzad et al. | Nov 2005 | A1 |
20050245279 | Mehrabanzad et al. | Nov 2005 | A1 |
20050250511 | Xiao et al. | Nov 2005 | A1 |
20060067422 | Chung | Mar 2006 | A1 |
20060067451 | Pollman et al. | Mar 2006 | A1 |
20060126509 | Abi-Nassif | Jun 2006 | A1 |
20060135173 | Vannithamby | Jun 2006 | A1 |
20060135189 | Nagaraj et al. | Jun 2006 | A1 |
20060159045 | Ananthaiyer et al. | Jul 2006 | A1 |
20060176187 | Bohler et al. | Aug 2006 | A1 |
20060215608 | Lee et al. | Sep 2006 | A1 |
20060240782 | Pollman et al. | Oct 2006 | A1 |
20060252429 | Chen et al. | Nov 2006 | A1 |
20060268798 | Kim et al. | Nov 2006 | A1 |
20060291420 | Ng | Dec 2006 | A1 |
20060294241 | Cherian et al. | Dec 2006 | A1 |
20070026884 | Rao | Feb 2007 | A1 |
20070058628 | Rao et al. | Mar 2007 | A1 |
20070077948 | Sharma et al. | Apr 2007 | A1 |
20070097916 | Eyuboglu et al. | May 2007 | A1 |
20070101015 | Larsson et al. | May 2007 | A1 |
20070115896 | To et al. | May 2007 | A1 |
20070140172 | Garg et al. | Jun 2007 | A1 |
20070140184 | Garg et al. | Jun 2007 | A1 |
20070140185 | Garg et al. | Jun 2007 | A1 |
20070140218 | Nair et al. | Jun 2007 | A1 |
20070155329 | Mehrabanzad et al. | Jul 2007 | A1 |
20070202826 | Dean | Aug 2007 | A1 |
20070220573 | Chiussi et al. | Sep 2007 | A1 |
20070230419 | Raman et al. | Oct 2007 | A1 |
20070238442 | Mate et al. | Oct 2007 | A1 |
20070238476 | Raman et al. | Oct 2007 | A1 |
20070242648 | Garg et al. | Oct 2007 | A1 |
20070248035 | Sang et al. | Oct 2007 | A1 |
20070248042 | Harikumar et al. | Oct 2007 | A1 |
20080003988 | Richardson | Jan 2008 | A1 |
20080013488 | Garg et al. | Jan 2008 | A1 |
20080062925 | Mate et al. | Mar 2008 | A1 |
20080065752 | Ch'ng et al. | Mar 2008 | A1 |
20080069020 | Richardson | Mar 2008 | A1 |
20080069028 | Richardson | Mar 2008 | A1 |
20080076398 | Mate et al. | Mar 2008 | A1 |
20080117842 | Rao | May 2008 | A1 |
20080119172 | Rao et al. | May 2008 | A1 |
20080120417 | Harikumar et al. | May 2008 | A1 |
20080139203 | Ng et al. | Jun 2008 | A1 |
20080146232 | Knisely | Jun 2008 | A1 |
20080151843 | Valmikam et al. | Jun 2008 | A1 |
20080159236 | Ch'ng et al. | Jul 2008 | A1 |
20080162924 | Chinitz et al. | Jul 2008 | A1 |
20080162926 | Xiong et al. | Jul 2008 | A1 |
20080253550 | Ch'ng et al. | Oct 2008 | A1 |
20080254792 | Ch'ng | Oct 2008 | A1 |
20090034440 | Samar et al. | Feb 2009 | A1 |
20090170547 | Raghothaman et al. | Jul 2009 | A1 |
Number | Date | Country | |
---|---|---|---|
20060159045 A1 | Jul 2006 | US |