You need to have good Idea of SACCH Multiframe and speech coding to understand fully these concpets of BER, RXQUAL, FER, DTX and SUB and FULL values.
SACCH Multiframe consists of 4 26 frames containf 24 speech/data bursts and 1 SACCH and 1 Idle timeslots.
so there are total 96 timeslots for speech, 4 timeslots for each SACCH and IDLE.
Full measurements are made over the complete SACCH Multiframe while SUB Measurements are made over the mandatory timeslots that is SACCH and SID (Silence Discriptor).
When DL DTX is being used no need to worry about Full values and just check the SUB values for FER and RXQUAL.
each 20msec speech is coded into 240bits and is divided into three categories accordint to their importance
Calss 1a = 50bits (Most impotant)
Class 1b=132 bits (Less important)
and class2=78bits (no impotance and are not used in BER Calculation)
3 CRC bits are attached with Class 1a 50 bits and then these 53 bits along with 132 bits of class 1b and 4 tail bits a total of 189 are 1/2 convolutionally coded.
This coding doubles the number of bits to 378 and these 378 bits are used for BER calculation out of total 456 bits excluding class 2 bits 78 bits.