I understand the advantages of SC-FDMA over OFDM and I also understand the disadvantage of additional complexity (FFT block), are there another disadvantages?

Also, if $N$ has to be bigger than $M$ ($N>M$) then I don't understand why I don't lose rate as a result of padding with zeros instead of transmitting another data subcarriers.

• Channel estimation using pilots is harder, because you do not have orthogonal data on each frequency bin, compared to OFDM. That's why e.g. LTE uplink resorts to a midamble for channel estimation in uplink
• Application of non-linear MIMO detection algorithms is harder (i.e. more complex), because you essentially have coupled MxM elements from each transmit antenna, compared to 1x1 elements when using OFDM.
• In high SNR regions I think SC-FDMA does not perform as good as OFDM when considering a coded transmission with linear detectors.

Regarding the data loss: Even if $M<N$, there is no real loss, because the used bandwidth of your signal is only M frequency bins (ignoring any guard carriers). So, any other system can use the remaining N-M frequency bins.

• Thanks! I'm wondering if converting the ofdm structure in DVB-T2 downlink into SC-FDM will make things better, what do you think? I guess you won't need the rotated constellation mechanism as you have inherent diversity over the frequency bins and hence you have greater immunity to fading and 0dB echo channels. – Gabizon Feb 22 '17 at 20:17
• @Gabe Well, yes you have frequency diversity over the bins, but at the same time there is ISI in a multipath channel (without proper equalization). To really exploit this frequency diversity you would need 1) a very high frequency-selective channel 2) at least LMMSE detector and high SNR 3) high-complexity non-linear detector like sphere decoder. It's up to you to decide if this litte gain in diversity is worth the effort. You can look for "interleaved SC-FDMA" for more frequency-selectivity, or even vector-OFDM, which spreads all data over a very wide frequency range. – Maximilian Matthé Feb 22 '17 at 20:40
• I'm discussing mainly that scenario of 0 dB echo channel (one and only one another path as powerful as the LOS path). Why does 2 true?and what do you need sphere decoder for? There is no MIMO involved in DVB-T2. Also,that's why VOFDM may be not relevant here. Is there another improvement to the DVB-T2 standard that you think might help in some scenarios? – Gabizon Feb 22 '17 at 21:16
• 0dB echo path or any other multipath, you essentially have the following signal model for your single-carrier: y=Hd+n, where n is noise, H is the MxM (circulant) channel matrix and $d$ is the transmitted data. It's equivalent to a MxM MIMO system. You can now do linear equalization by "dividing by H". Here, only the LMMSE equalizer can gain diversity, but only for higher SNR. The ZF-equalizer will not gain diversity. Or you can use a non-linear technique, like a sphere-decoder to get to the d-values with even better performance. VOFDM gives you more frequency diversity by wider data spreading. – Maximilian Matthé Feb 22 '17 at 21:28
• I tried reading Xia's paper about VOFDM but I haven't understood it entriely. How does it work exactly, how can it help me with frequency diversity and what is the cost? – Gabizon Feb 24 '17 at 10:00