Voice Quality
CDMA versus GSM 

• I can't find a single real-life study or survey that proclaims that GSM voice quality is better than CDMA voice quality. Every study points to the opposite conclusion, or is too close to call. I am talking about surveys and studies by independent entities. I can easily find studies at both the GSM and CDMA trade organizations, each of which claim superior voice quality; in my view these studies are too biased to be taken seriously.
   
• The quality (or lack thereof) of the GSM networks in the U.S. masks the fact that GSM is actually capable of providing better voice quality than CDMA. In other words, all things being equal, GSM will sound better than CDMA, but all things are not equal.
   
• In a perfect world, with 800 Mhz CDMA on an uncongested network versus 800 Mhz GSM on an uncongested network, GSM will indeed sound better than CDMA. If the GSM advocates are saying that GSM in Europe and Asia sounds better than CDMA in the U.S., I'll agree completely, as I've experienced it personally. When the U.S. GSM networks sound better than the U.S. CDMA networks, and I believe that this will eventually be the case, I will certainly change the voice quality ratings for GSM in the technology comparison table from "Good" to "Excellent."
   
• The comparisons put forward by GSM advocates invariably compare a congested CDMA network to a GSM network;  indeed, under this scenario there will be a very big difference in voice quality.
   
• There are not many places where CDMA congestion is actually a problem. At special events, where network congestion is a real possibility, both the CDMA and the TDMA/GSM carriers roll in mobile cells to handle it. These mobile cells prevent deterioration of the CDMA voice quality and prevent "system busy" problems on TDMA and GSM networks.
   
• Both CDMA and GSM are "improving" their vocoders in order to increase capacity. Both camps claim that the vocoders with lower bit rates sound the same or better than the higher rate vocoders they are replacing. CDMA is further along in reducing the bit rates than GSM. CDMA networks are switching from the  QCELP-13 vocoder (13kb/s) to EVRC vocoders with an 8kb/s coding rate. GSM networks are switching from a 13kb/s EFR vocoder to an AMR (adaptive multi-rate) vocoder. AMR vocoders vary their bit rate based on network quality. There was an attempt by Vodafone to use half-rate GSM vocoders but the poor voice quality resulted in so many complaints that they dropped it. Next, the CDMA networks are going to  deploy SMV (selectable mode vocoder) vocoders. Naturally, the CDMA lobbying group claims that the SMV vocoder is better than the AMR vocoder (and has posted the results of listening tests). It is safe to say that the CDMA EVRC vocoder is worse than the GSM AMR vocoder, but it is also likely that the SMV vocoder will improve CDMA voice quality to equal or better than the GSM AMR vocoder. SMV vocoders still give the CDMA carriers the ability to trade capacity for quality. Note that a handset with an older, higher bit-rate vocoder will continue to work just fine on networks with the "improved" vocoders, and it may be a bad decision to "upgrade" your handset to one that has an "improved" vocoder.
   
• People will put up with different things based on their being mobile or stationary, and based on whether they are making calls or receiving calls. The result of Vodafone's failed attempt to use half-rate GSM vocoders in Europe showed that users valued voice quality over the ability to make a call at all. If a user is stationary they would probably be willing to hit re-dial a few times (or a few dozen  times) to get a dedicated GSM slot that is of discernibly higher voice quality than CDMA. If they are mobile, they would likely prefer not having the call drop as they move from cell to cell, even at the expense of lower voice quality. The big difference is for incoming calls. Subscribers don't want to miss calls, especially emergency calls (knock on wood), because there are no free GSM slots; yet this is exactly what I've experienced with GSM in its current state in the U.S..
    
• This debate reminds me of the Ethernet/Token Ring debate of a decade ago. In a very heavily loaded network it is possible for Ethernet to be slower than Token Ring. But the reality was that this did not happen because if the network became too congested it was simply split into segments. Similarly, if a CDMA cell is consistently congested then another cell will be added; it is even possible to add more cells to one cell site via directional antennas. Wow, CDMA has a lot in common with CSMA/CD.