LTE Paging Mechanism

Since I got a lot of requests to explain the LTE paging mechanism and paging capacity, so I decided to write a couple of short articles on these topics.

Paging is a mechanism to initiate services for UEs that are in idle mode. Idle mode transition is important to conserve battery of the UEs. If a UE is in connected mode and it has no data to send or receive, then the eNB waits for a specific timer (UE Inactivity Timer) and once that timer expires, the eNB sends the UE to idle mode. This is done by sending a RRC Release message to the UE.

However, a procedure has to be defined to access the users in idle mode if there is downlink data for them. This process is the paging mechanism. The main parameters that determine the paging capacity at the eNB are given below:

Default Paging Cycle: This is known as “T” and it is the value after which the UE wakes up from idle mode to read the paging messages. Usually it is set to 1280 ms or 128 frames. This means that if a UE woke up in frame # 3 then it will wake up again in frame # 131 (after 128 frames) and then at frame # 259 and so on. It will read the PDCCH and if the PDCCH directs it to a paging message, it will read the paging message. However, this only indicates that there is a paging message in this subframe but it might not be addressed to that UE as there can be multiple UEs using the same paging cycle. “T” is transmitted in SIB-2 but can also be shared by the MME and in that case the smaller value of the two is used. For instance, if T in SIB-2 is 1280ms and T shared by MME is 640ms then the system will use 640ms as the T’s value.

NB: This parameter depicts the number of paging subframes within a Default Paging Cycle. If NB is set equal to T, then every frame contains one Paging subframe. If NB is set to 2T, then every frame will contain 2 paging subframes. If NB is set to 1/2T then every second frame will have a paging subframe. So, in short, this parameter can impact the paging capacity but at the same time it can increase or decrease the paging overhead as well. This is also shared in SIB-2.

LTE Paging Mechanism

Paging Mechanism: These parameters are contained in the SIB-2 so the UE knows it’s paging subframe or paging occasion from the parameters in SIB-2 and from the function of it’s IMSI value. The UE will wake-up at the subframe where it expects it’s PO and will read the paging messages. The paging messages contain the TMSI value so if the UE does not find it’s TMSI inside the paging messages, it will assume that it is not paged, and it will go back to idle mode. However, if the UE finds it’s TMSI in one of the paging messages, it will understand that the page is addressed to it and it will initiate a RRC Connection.

From the core’s perspective, if there is downlink data, the SGW will inform the MME and the MME will send a paging message to the eNB over the S1 interface. The S1 paging message contains a UE ID value which is obtained from a function of the IMSI. Based on this number and the value of N (N=min(NB,T)), the eNB calculates the PO of the UE and it buffers the paging message until the PO. Since the UE already knows it’s IMSI and the eNB derives it from the UE ID in the S1 Paging message, both the UE and eNB are synchronized and they use the same PO subframe. Once, the UE receives the paging message addressed to itself, it initiates RACH process by sending a RACH preamble to the eNB. The eNB responds with a Random Access Response. The UE then sends the RRC Connection Request and the cause value of this request is set to mt-access which indicate that the UE is initiating the connection due to a paging message. The eNB responds with RRC Connection Setup and UE finally sends the RRC Connection Setup Complete to the eNB resulting in completion of the RRC Connection. The eNB sends a S1 Initial UE Message to the MME with the Service Request and this message is considered by the MME as the paging success.

LTE Paging Mechanism

Timer T3413 : The MME maintains a timer T3413 for paging messages. Once it sends the paging message to the UE, it starts the T3413 and it stops it after receiving the Service Request in S1 Initial UE Message. In case, the Service Request is not received and T3413 expires then the MME will resend the paging message to the UE.

However, this timer only works for PS pages and not for CS pages. In case of CS page, the page is retransmitted from the CS core.

Impact of Default Paging Cycle : The paging cycle or T as explained above tells the periodicity with which the UE wakes up to check the paging messages. The smaller the value of T, the UE will need to wake up more frequently. This means that a smaller value of T will lead to higher battery consumption or utilization and UEs will have to be charged more frequently. However, a larger value means that the latency for the UE will increase. For instance, if we use 1280ms for T, the incoming pages at the eNB will be buffered longer on average and that will increase delay. This is not significant for data services but it can be important for CS pages (CSFB) or for VoLTE and can increase call setup time. So, a compromise needs to be maintained between battery consumption and expected paging delay.

Paging Failures : Another important aspect is to ensure that paging messages are successfully reaching the destined users. The paging failures can be caused by decoding issues if the paging modulation scheme is aggressive or if the PDCCH allocation for paging is not using a robust aggregation layer. So, it is a good idea to keep the Paging MCS on the PDSCH much lower in order to ensure successful decoding and also to ensure that the PDCCH is using a conservative aggregation level like Aggregation Level 4 or 8.

Another issue that causes paging failures is frequent TAUs (Tracking Area Update). When the UE moves from one tracking area to another tracking area, it initiates a TAU and if during this process a paging message is received then the UE might not be able to read it as it is undergoing a TAU. The problem here is that when the UE is in idle mode, the MME knows it’s location at a TA or TAL (Tracking Area List) level. So, when the page comes for the UE, MME sends this paging message to all the eNBs in that TA or TAL. If the UE moves from one TA/TAL to another TA/TAL then the UE needs to do a TAU to tell the MME that the UE has moved to another TA/TAL and thus the MME will send the next paging messages for that UE to the new TA/TAL. If the UE is undergoing TAU and the TAU is not completed yet and a paging message comes for this UE, the MME will send it to the old TA/TAL and the UE will not be able to receive it. Thus, it is very important that the TAC planning is properly done to mitigate this issue. I will address the TAC dimensioning and paging capacity in the next short article.

In case of any queries or feedback, please drop a comment below and I would love to respond and help.

The following two tabs change content below.

Ali Khalid

5G SME & Solution Architect | NB-IoT | VoLTE | Massive MIMO at Ooredoo Oman
Ali Khalid is a Senior LTE/VoLTE RNPO Expert and 5G Solution Architect who has successfully delivered a number of LTE RNPO Projects in different regions across the globe including Pakistan, Bahrain, UAE, Qatar, Nigeria, Turkey and Oman. In case of any questions or feedback, please feel free to drop a comment below or connect with him on LinkedIn.

42 thoughts on “LTE Paging Mechanism”

  1. Thank you for this topic.
    Could you please explain the relation and the restriction between paging cycle and DRX in LTE ?
    Thanks again for your effort.

    1. The Paging cycle and the DRX cycle are usually used for the same thing which is described as “T” in the article. However, Connected mode DRX is a different thing that is also introduced to conserve battery for the UEs. The Users which are in RRC Connected mode can be sent to sleep mode using the connected mode DRX. This sleep mode is usually a small value for instance 320 ms is commonly used.
      In short, the paging cycle or drx cycle is used by UEs in Idle mode while the Connected Mode DRX is used in RRC Connected mode as the name implies.

      1. Many Thanks.
        (1) I saw that non optimized value of DRX Cycle could lead to a degradation in paging Success Rate in Core side.
        could you please explain ?
        And (2) does inactivity Timer of UE (to go to idle mode) has a relation with the DRX sleep/active ?

        1. For the first one, kindly explain more as I am unable to get the question? Secondly, inactivity timer does not have a direct impact with DRX sleep/active time under normal conditions but it will interact with it in case of dynamic drx which is not usually used.

          1. for (1) I mean, what happen if Network whant to page the UE and that UE is in sleep pahse ?
            Should NW wait for active phase of UE ? if yes, How NW will know when UE will get up ?
            So in case UE is in sleep, i think the paging will be re send many time.

          2. The active and sleep phase of DRX happens in RRC Connected mode and paging is supposed to come in Idle mode (under normal conditions) so this issue should not happen. Moreover, the eNB also maintains a table for each UE containing the UE’s sleep and active time details.

  2. “1st Query”
    In 2nd snapshot

    There are 3 RRC Paging Message , first 2 are not for UE A as MME S1AP Paging Message was sent after the 2nd RRC Paging Message and since it was sent before 3rd RRC Paging Message thus UE A recognized that there is a paging message meant for it (UE A)
    Although UE will wake up during all 3 (as per the snapshot example) RRC Paging Message (Paging Occasion during which UE is suppose to wake up)
    Right ?

    2nd “Query”

    Consider nB as 2T here
    As per the Paging Occasion lookup table
    Ns => It sort of determines the number of subframes (during a Radio Frame) in which UEs are suppose to wake up
    i_s=> it determines the specific Paging Occasion (subframe 0 or 4 or 5 or 9) for a UE

    correct me if i am wrong here
    After all the calculation ,
    If Ns comes out to be 2 and i_s comes to be 0
    then according to the table one of the specific UE will need to wake up during “Subframe 4” but having said that there won’t be a single UE which needs to wake up or is going to be paged so another UE whose calculated Ns=2 as well but i_s=1 then it will need to wake up during “Subframe 9”

    In short what i mean to ask is this
    i_s=0 is basically telling that specific UE that you need to wake up in subframe 4 but Ns will inform that these subframes (4 and 9) are to be checked regardless of the value of i_s=0 & a specific UE wakes up ONLY during one subframe within a Paging Frame

  3. Thank you for the insightful article the topic

    Can you kindly shed some light on Aggregation Level? What does it means and how it relates to robustness?

  4. Hi,

    I have several questions, if you allow me 🙂
    1. How long a paging message is buffered? If at first PO paging cannot be sent due to congestion, then paging will be discarded?
    2. Under huge traffic(concert/stadium area) with T=64rf and NB=1/T it is supposed that paging channel is congested. Why avg Ue paged in one PO is ~7 if, according to configuration, 16 Ue can be paged in one PO? I was expecting channel to be full. I know Ue paging location is computed based on TMSI, which is a random value not decided in RAN, but under huge traffic shouldn’t all PO be full?
    3. If PDCCH is congested then paging is discarded immediately or eNB will buffer the message until resources will be available?

    1. This can be implementation specific as some vendors have a timer to discard a paging message and in that case the message will be buffered until that timer expires while some vendors discard it if it could not be sent in the first PO. For the paging channel, what was your bandwidth configuration for the event? As far as PDCCH is concerned, I don’t think that is a common issue as the paging message has higher priority in the PO subframe so PDCCH should be assigned to it first.

        1. Then the subframe should not be full with 7 paging messages. On average a UE paging message takes around 6 to 8 bytes over the air and the default MCS index is around 0 in terms of code rate. This means that around 3 PRBs should be used per UE and therefore, with 20 MHz and 100 PRBs, 7 UEs cannot congest the resources. In order to confirm this, take a trace over S1 and Uu and verify the count of incoming paging messages and compare with the outgoing RRC pages in a small interval. It seems that the incoming paging count was not high.

          1. this means that around 3 PRBs should be used per UE and therefore,– How you get to this info from paging message size and no of paging message

  5. I could see some OEM going with maxpagingrecords as 16 with 5Mhz+20mhz carrier whereas some going with 7(considering lower bandwidth carrier while deciding maxpagingrecords).. Any hard limits on setting it to 7(maxpagingrecords) in case of 5+10+20 scenario?

    1. As explained in another comment, on average a UE paging message takes around 6 to 8 bytes over the air and the default MCS index is around 0 in terms of code rate. This means that around 3 PRBs should be used per UE and in order to ensure that the subframe is not congested with paging, usually a lower value is used for maximum paging records for 10MHz compared to 20MHz.

  6. Dear Ali.. i have one doubt in paging procedure in LTE.. how paging during csfb and ps data call happens . If i m in connected mode.. if i rcv any email or skype cll then paging will happen or not.. in csfb procedure how many pagings happen.. is it paging to lte once and paging to umts as well or paging to lte and reaponse to umts..

    1. In connected mode, there is no need for PS paging as the data can be forwarded to UE without any paging requirement since the UE has a valid RRC connection and it is reading the PDCCH. In CSFB scenario, the page comes from the CS domain so the MME will still need to send the page to the UE and in that case it is a NAS message and once the UE gets that NAS message in connected mode, the UE will initiate ESR to begin the CSFB procedure. Once the UE moves to CS RAT which is typically UMTS, it will send the paging response to the MSC.

  7. If i have a paging for csfb call. That i rcvd in 4g and ue released to legacy nw…and in the legacy nw paging response didnt happen.. So msc will repage again to where.. 3g or 4g.. ? How this will happen.. or call wll b failed . ?

  8. helpful as always. Thank you so much Ali.
    Btw I have a case relate to paging csfb in our network. Basically when UE move from 3G to 4G, UE will initiated the RRC connection (mo-signalling) to perform TAU procedure. As soon as TAU complete message appeared, this RRC will be release right away. which means this RRC only last about 0.5s.
    However, if I purposely call this UE, in order to interupt this TAU procedure, which means when TAU complete message was sent out, the paging will be received by the UE (cause RRC is still there, it will be CS service notification), the RRC will last about more 10s. The strange thing (to me) is: when receive the CS notification, UE didnt send out the Extended service request (ESR) to do the CSFB. the UE will wait this RRC to be release, which means 10s after and initiates new RRC and send ESR (this will cause timers waiting the paging response in MSC expire and csfb fail, I extended this timer to temporary solved this issue).
    So it looks like the first RRC (mo-signalling for TAU, only SRB1, no SRB2 or DRB is included) is not suitable for ESR, but I cannot offcially explained this and why UE delayed ESR and have to wait to make a new RRC to do it
    Do you have any suggestion?

  9. HI,
    I have a question here. Assume the timer T3413 value in MME is 10sec and default paging cycle is 640ms, how many times can eNodeB send Paging messages to UE ? Does it try to send paging message for every DRX cycle ? does eNodeB aware of time out value for paging in MME ?

    1. From the eNB’s perspective, under normal conditions, it should send the paging message only once. The UE can be anywhere in the tracking area and since it is a broadcast, the UE is not supposed to send an acknowledgement to the eNB that it has received the paging message. So, the timer on the MME is not an issue for the eNB. However, if the MME does not get a service request for the paging and the timer expires, it will resend the page and once again, the eNB will forward it to the UEs.

  10. Hi Khalid
    Thanks for wonderful info.
    Please can u explain how different bandwidth affect the paging handling capacity per sec. And what is the calculation behind it.

  11. HI , I have another Query, I could see that in the calculation of PF, there is an equation PF = SFN mod T = (T div N) x (UE_ID mod N), and where N is N = min(T, nB), and T can come from SIB2( assuming 128), or can be defined by upper layer as well. Can u further put some insight on if MME or UE can define their own DRX cycles. I could see that there is considerable difference in PSR of VoLTE and that of PS call in LTE in our network.

  12. Hi Ali,
    can you please explain how page is repeated in case of CS paging. which timer is used on cs core side ?

Leave a Reply

Your email address will not be published. Required fields are marked *