Generic selectors
Exact matches only
Search in title
Search in content

Reply To: Erlang Formula


Hi osprey,
I see you still have doubts about RxQual Sub vs Full and RxQual vs SQI.

Sub vs Full values (this applies not only to RxQual, but to RxLev and FER):

GSM can use discontinuous transmission (DTX) – when there is no sound to transmit, the phone and/or the BTS may not transmit all frames on the timeslot. This means that when you have DTX the RF signal is not present on the whole timeslot. So, when DTX is active and was used (it may be enabled, but not used in specific instants if you “speak too much ;-)) the RxLev and RxQual measured across the whole timeslot period will be meaningless. For this situation, there are measurements of RxLev and RxQual which are taken only on the obligatory frames (not on the actual tranmitted frames, which would complicate matters). These are the “Sub” values.

So “Sub” values are always valid, but may not be as precise as Full values, whereas Full values are more precise, but may be wrong when you have DTX. The network and the phone “know” when to consider the Sub or the Full values for handover decisions because there’s a “DTX was used” flag transmitted.

SQI vs RxQual:

RxQual is part of the GSM standard (05.08), and it is used basically for handover/power control. It is directly related to Bit Error Rate (there is a table on the standard). For each BER range, am RxQual value from 0 to 7 is used.

SQI, as Erik has pointed out, is more true to the actual user experience in terms of voice quality. RxQual alone is not a direct estimative of voice quality because:
– it does not take into account FER and CRC (you may have bad bits spread on different frames, which are corrected by the redundancy protection with no impact on the voice, or all on the same frame, which will cause the frame to be discarded, impacting the voice quality)
– it does not take into account the vocoder in use (even with RxQual and BER = 0, a half rate vocoder won’t give you the voice quality of an enhanced full rate vocoder, and for AMR it will depend on the percentage of bit rate used for each instant)
– it does not take into account the handovers (if you have lots of handovers, voice quality will drop even if RxQual is good – Ok, generally with good RxQual, there will be no handovers, but…)

SQI takes all these factors into account and also their statistical distribution (isolated FER occurences may have little or no impact on voice, for example), and has a very high correlation with actual human scoring of voice quality (better than 90%).

Summarizing, RxQual is more an RF quality paramenter, and SQI is a voice quality parameter.

Check this:
from question 12 to 25. It has good information about SQI.


Joao S Veiga