LTE uses OFDMA, which has nothing to do with CDMA (no superposition and coding separation). OFDMA is a combination of frequency and time division: time wise, th

LTE uses OFDMA, which has nothing to do with CDMA (no superposition and coding separation). OFDMA is a combination of frequency and time division: time wise, the allocation unit is a subframe of 1 ms (to be compared with a 2G GSM burst duration of 0.577 ms, so not that far) and frequency wise a subframe is made of "resource blocks" (RBs) of 12 subcarriers of 15 kHz each (so 1 RB = 180 kHz total). What's allocated to a device is a set of RB in a subframe, and this is repeated as needed accross subframes in time. 3G used a 10 ms allocation unit, but with HSPA it's been reduced to 2 ms to become more "TDM-like" (it turns out that in real life code superposition is not so good, better split in time and do less superposition: less interference, less power consumption, better results). I honestly don't know the source of the 2G noise, but from the above it's not as simple as TDM vs. no TDM, as current 3G and LTE both in practice have a time-based allocation component on roughly similar granularity as 2G. One difference could be the transmission power: 2G can go up to 2W, while LTE is 200 mW maximum so 10 times less. I don't have 3G peak Tx power in mind, but in practice I guess it should be closer to LTE than 2G (interference minimization in a same frequency / reuse 1 network: both LTE and CDMA are reuse 1 deployment, while 2G is not). That will certainly cause a 2G device to have higher power spikes. And I guess the details of the transmit pattern are likely to play a role too.