Check this process to understand the VoIP function in the background

1. When you speak into the regular telephone, your voice is passed through an ATA (Analog Telephone Adapter) which converts your analog voice into digital sequence of packets using ADC (Analog to Digital Converter). Every VoIP provider would give an ATA to every subscriber when you sign-up for the service. Remember ATA is only needed if you have an Analog phone, if you're using currently available digital phone in the market, you don't need an ATA.
2. The digitally converted analog voice is now arranged into bit and gets compressed into a standard format which can rush over the Ethernet cable or any form of digital media very quickly and efficiently. The DSP (Digital Signal Processors) will be performing this compression using appropriate Codec's which are there to further segment the voice signals into multiples frames and split them as voice packets. Some of the compression standards with there associated bandwidth are listed below:
   1. PCM, Pulse Code Modulation, Standard ITU-T G.711, 64Kbps
   2. CS-ACELP, Standard ITU-T G.729 and G.729a, 8Kbps
   3. ADPCM, Adaptive differential PCM, Standard ITU-T G.726, up to 40Kbps
   4. LD-CELP, Standard ITU-T G.728, 16KbpsMP-MLQ, Standard ITU-T G.723.1, 6.3Kbps, Truespeech
   5. ACELP, Standard ITU-T G.723.1, 5.3Kbps, Truespeech
   6. LPC-10, able to reach 2.5 Kbps

   Some of the phone also uses the G.711 Codec and G.723 Codec which are becoming very popular choice for any IP telephone applications as well as they are smaller in size and performs higher compression giving an ease of transportation over the internet.

3. Once the data is compressed it will now be encapsulated within IP packets done at the layer 3 of the ISO model. VoIP is a layer 3 protocol which uses different layer 2 point to point protocols as its transport and rest is all left to RTP (Real-time Transport Protocol) which runs over the UDP (User Datagram Protocol). UDP packets are generally connection less packets which means for some reason if they were lost somewhere on the internet, the gone forever. Majority of RTP based application do not use the connection oriented TCP packets as it slows down the transmission rate. The connection becomes reliable but slow. Internet layer pickup the source and destination IP address and formulated a path. After which is instruct the data link line layer to prepare control protocols and send the packet in bits over the physical medium.
4. The packets uses the H323 (which is the standard call control) protocol to get transmitted over to the MGCP (Media Gateway Control Protocol) or SIP (Session Initiation Protocol) which sends them over to the destination.

Once the data packet has arrived at the destination (or the receiver) it goes through steps 4 to 1 (in reverse direction) where the de-encapsulation of the packets happens and the compressed audio is decompresses using same coded and then comes out the original digital packet. Hear it and the process continues all the time for either party in the conversation.