Generic selectors
Exact matches only
Search in title
Search in content
Post Type Selectors

Initial MCS tuning

  • This topic has 11 replies, 1 voice, and was last updated 16 years ago by Pan.
Viewing 12 posts - 1 through 12 (of 12 total)
  • Author
    Posts
  • #51295
    Leila
    Guest

    Hello every one,
    I am working with Alcatel BSS (release B9).
    I am trying to increase Ul/DL initial MCS. Throughput has increased but drop has also increased. The tested area has a good radio conditions
    Is there any way to lead with drop increasing ?
    Thanks in advance

    #51296
    Pix
    Guest

    Which MCS INIT have you chosen in UL and in DL ?

    A higher MCS is weaker, and any interference or low coverage can lead to a call drop.

    INIT MCS should be limited to 6 in DL, and in UL you can choose 6 or even a lower value.

    There is no point from using higher MCS if it leads to TBF drops 😉 The MCS adaptation (from lower MCS to higher MCS) is a quite reactive mechanism, that will increase efficiently the MCS if the radio conditions are good enough.

    #51297
    Leila
    Guest

    Thank you Pix,
    We have choose MCS6 as initial MCS in UL and MCS9 as initial MCS in DL.
    We notice a little increase in drop.
    Throughput per PDCH and per TBF increase in UL which is expected.
    But Throughput per PDCH and per TBF decrease significantly in DL and remain stable.
    Also, retransmission rate decrease in DL.
    Please Pix, have you any explanation to throughput per PDCH and per TBF in DL direction ?
    Thanks a lot !

    #51298
    Pix
    Guest

    Hi Leila,

    Well, MCS9 is a very very bad idea. This coding scheme offers zero protection, and it requires a C/I above 30dB and a RXLEV DL > 85dBm in order to provide acceptable throughput.

    I strongly recommend you move back to MCS-6, or even MCS-5. Beware, MCS3 & MCS4 are not resistant at all.

    FYI, this is the list of coding scheme from most resistant to less resistant :
    MCS1
    MCS2
    MCS5
    MCS6
    MCS3
    MCS7
    MCS8
    MCS4
    MCS9

    this is approximate.

    regards,
    pix

    #51299
    Bilal
    Guest

    Hi Leila,

    Throughput per PDCH and per TBF are different because PDCH is one data channel and TBF is one data session which can use multiple PDCHs and one PDCH can also be shared by multiple TBFs.

    #51300
    Leila
    Guest

    Hi Pix,
    Just a thing normally there is no protection (MCS9 usage) retransmissions normally increase
    But retransmission decreases in our case
    What do you think ?

    #51301
    Pix
    Guest

    The TBF drops before it could make a retransmission ?
    Maybe the TBF establishment failure is much higher ? Or the overall downlink LLC traffic is much lower ?

    What about MCS usage ? CAn you check whether the TBF remain in the MCS9 or do they dynamically adapt to lower MCS ?

    You need to give me more details about your QoS 🙂

    #51302
    Pan
    Guest

    Hello, Leila!
    Which initial DL-MCS and MCS Families you had before MCS change? And now (MCS and Families)?
    Is Dynamic Link Adaptation Mechanism implemented in your network?
    If it is, then what BLER (Block Erasure Ratio or retransmission rate) threshold values you have for downgrade switching points between MCS9 and other MCS (MCS9 -> MCSy). These thresholds are varied for different switching points.
    Have you a BLER threshold that is lower than it was before change?

    #51303
    Pix
    Guest

    Pan,

    I’m surprise, I thought the thresholds to swith from MCSi to MCSj (EGPRS only) are fixed and defined by the 3GPP, they are based on MEAN BEP and CV BEP measurements. The tables can be found in the standards.

    Is it different in your equipment ?

    The thresholds are modifiable only in GPRS (CS1/2/3/4).

    #51304
    Pan
    Guest

    Hello, Pix!
    You meaning the table in 45.008 (AnnexD)? I think, it is only for example. In our RAN the triggering of the switch does not use separate measurements of channel quality, but
    it is executed by analyzing the number of blocks to be repeated (not acknowledged
    blocks) versus the number of transmitted blocks in total. Link adaptation is then
    based on BLER measurements (indirect measures of the radio quality). They are really fixed in BSC database and stored in pre-calculated matrix tables, one for each possible RF environment, since switching points depend on the actual RF scenario (lowDiversity or highDiversity). These switching points are varied for different transitions MCSi->MCSj, different RF scenario, and whether the incremental redundancy (IR)used or not.

    #51305
    Pix
    Guest

    So your BLER thresholds are fixed, and not changeable, right ?
    When I read your previous post, I thought you asked Leila to check the BLER settings in her network. Since it can’t be changed, I don’t think there is much to do on this side.

    I don’t know if alcatel is using 3GPP tables or not, but they are using the same measurements methods : mean bep, cv bep.
    Then there are different tables depending on APD, on IR, on GMSK/8PSK, etc, as recommended by 3gpp.

    #51306
    Pan
    Guest

    Yes, Pix. Our thresholds are not changeable by O&M action. By O&M, it is possible to select the suitable matrix table, containing all of the ideal switching points
    (downgrade/upgrade switching points from/to all coding schemes) for the particular RF
    scenario, by selecting the right radio environment.
    When I write my previous post, I imlied following.. For example:
    We have EDGE with Incremental Redundancy and RF scenario with HIGHDIVercity// Then,
    switching point for MCS9 to MCS7 transition is BLER > 35%,
    but switching point for MCS5 to MCS2 transition is BLER > 70%.
    If there is only FamilyB (MCS7,5,2) was implemented with init MCS5, then the retransmission rate will be greater, than in case when both families were implemented – FamilyB ((MCS7,5,2)) and FamilyA (MCS9,6,3).
    This is assumption only 🙂 Laila should give us more details.

Viewing 12 posts - 1 through 12 (of 12 total)
  • The forum ‘Telecom Design’ is closed to new topics and replies.