KPI: 3G2G_HO/Utran_CS_calls

[s52d]

Hi!

How to tell if a UTRAN cell is on the coverage edge?

Nice KPI where it is clearly seen how well is coverage is ratio between
number of handovers to GSM and all CS calls started in the cell.
Purists can compensate for handovers: but they are mostly simmetrical.

Numbers are a bit suprising: more then 10% of all voice calls end with handover to the GSM.
While cells with good coverage has this ratio over 20, edge cells are around 3, and isolated rural cells
can be even below one (handovers...).
Quite important: numbers are stable (simmilar numbers every day), so we can use them for cell
qualification and different parameter settings.

It is also good pointer if there is something wrong with the cell: if city center cell behaves like
coverage edge one, it is quite likely something is wrong beside coverage holes.

BR
s52d

[danikd]

I think there is nothing special in description that you've put here.
I'll try to elaborate:
1. For real 3G cells on the coverage edge the number of 3G-2G HO should not be too high once the large hysteresis is maintained between GSM and UMTS networks for idle and active mode i.e. 3G network selected only when EcNo and RSCP are above -12dB and -100dBm respectively, while event 2D for active connection for IRAT HO defined for EcNo and RSCP -16dBm and -108dBm for instance. Within the 3G network typical setting for 3G cell reselection is defined for EcNo and RSCP -18dB and -115dBm respectively. The rural mobile behavior is more affected by RSCP and less by EcNo once no overlapping between surrounding cells.
2. For cells in the urban network that are affected by heavy load and significant overlapping between surrounding cells even large hysteresis can not compensate large variation in EcNo behavior. The unstable EcNo behavior can be easily explained HS transmissions or even HS with 64QAM, where Total Transmit Power is close to maximal for specific cell and overall EcNo has spiky behavior. It is very easy to estimate the possible degradation...or let's say possible damage while operating HS over R99 carrier. Most significant damage (too many causes for IRAT HO) can be observed on cells with low CPICH power (27-30dBm) and rather high Maximal available power for carrier on this cell (46dBm) for instance. very simple calculation for CPICH EcNo estimation while HS Transmission exist....30-46......-16dBm already!!!!!!! This is a trigger for IRAT HO....ok ok ...there is code ortogonality within the cell....but it is urban area with cells overlapping where surrounding cell affecting significantly and making EcNo even worse.

Please, remind, do not be surprised when for cells with high power amplifiers and low CPICH power you'll get to many IRAT HO.

[boring]

hi guys
i fully agree with danikd.
HSDPA load (own or neighbouring cells) reduces significantly measured EcNo which results in pushing traffic to 2G for in-coverage cells.
we are in the processes to change EcNo 2d thresholds in order to keep traffic for longer on 3G.
i will let you know what are our findings

[jayk]

Better will be to distinguish the 3G-2G handovers based on EcIo and RSCP. The kind of analysis you are doing, should consider handovers based on RSCP only. The stats you provided seems to driven more by EcIo in city, considering the parameters for re-selection and handovers are consistent in network.

[s52d]

Hello friends!

Yes, I fully understand why this KPI works, this is not an question.
Thanks for explanation, it might motivate some collegues to actually try it.
It measures effective voice coverage gaps, including cell breathing, compared to GSM 900.


On KPI story: we can not have 100 KPIs. We can have 100 performance indicators, but KEY PI are only few.
Those we regulary check, and those who tell us something. Like dropped call ratio.
On this one (gsm_ho/cs_call) I like simplicity and clear understanding of meaning.
It correlates well with other, more complex information (like coverage, type of use etc), and it might
point to some problem. Also, it is not easy to differentiate cells based on simple KPIs.

For me, this is something to look at. I can change numbers with some other parameters (like handover settings),
and actually we use it for automatic parametrization: those few parameters we tune per cell, not per RNC/network.
(This is called SON these days, but for me real SON is far, far away...)


Best regards,
s52d

[danikd]

Very good idea to do classification based on EcNo or RSCP for IRAT HO causes. You can do it easily in E******* network based on GPEH data, where entire IRAT HO process fully presented with all RF information correlated.

As for s52d statement on limited KPIs number and SON, I would like to say that as for number of KPIs you are 100% correct that it should a limited number of Key PI for level one performance monitoring, but for solving issues like IRAT....KPIs is not sufficient and you need more info about RF environment and causes. IRAT HO problem is pretty serious you can split at least for two parts:
1. Dedicated mode IRAT HOs that may significantly affect overall network Drop Rate especially for CS connections, since it is a HHO with RAT change.
2. Idle mode 3G coverage lost and ping-pong 3G selection/reselection where excessive signalling for LU exists in UTRAN and GSM networks, this may cause serious problem of lost incoming calls in case where LACs between GSM and UTMS are not synchronized (i.e. not the same). During RAT selection and LU your mobile is not available for incoming calls for 3-5 seconds. Once RAT Reselection occurs too often, then probability to get Mobile Is Off (MIO) too high.

As for SON....to make dynamic changes for IRAT HO parameters per cell basis....I think this is not efficient at all. The real SON taking care for most significant issues that may affect the root cause for unnecessary IRAT HO....bad RSCP and bad EcNo....this is can be treated efficiently in SON via antenna etilt optimization...such SON already exists and performing.....if you would like to know more on this issue I can share some info with you.