Abstract. This paper is a sequel of previous work, in which we proposed a model and computational technique to calculate the Erlang capacity of a single CDMA cell that supports elastic services. The present paper extends that base model by taking into account two important features of CDMA. First, we capture the impact of soft blocking by modeling the neighbor cell interference as a lognormally distributed random variable. Secondly, we model the impact of the outage by taking into account that in-progress sessions can be dropped with a probability that depends on the current load in the system. We then consider a system with elastic and rigid service classes and analyze the trade-off between the total (soft and hard) blocking probabilities on the one hand and the throughput and the session drop probabilities on the other.