A three terminal non-volatile memory (NVM) cell for a CMOS IC is formed by either a standard CMOS process flow or a slightly modified CMOS process flow. The NVM cell includes read and injection transistors that share a common floating gate. The floating gate includes a portion disposed over the channel region of the read transistor, a portion disposed over the channel region of the injection transistor, and a portion extending into an enlarged drain diffusion area away from the channel regions, whereby the gate-to-drain capacitance is higher than the gate-to-source capacitances. The source/drain of the injection transistor are formed using different LDD implants to achieve faster program/erase. Alternatively, an optional CHE enhancing implant is added to the source/drain of the injection transistor to enhance CHE programming. Both HV LDD and LV LDD implants are introduced together enabling LDD implant merging under the floating gate extension.