GPRS is designed for transmitting packet data and supposed to take its radio resource from the pool of channels unused by GSM voice services originally. In real operating network

the number of channels allocated to GPRS is not only a random variable depending on the voice traffic

but also comprised of fixed PDCH to guarantee the quality of data service. In this paper

an empirical method is proposed

named Campbell method. It is validated in CMCC network. two other methods are also introduced for performance compare.