Thanks. You are right, 4096QAM is 64x64 constallation, 2^6 x 2^6 = 12bit/symbol. I also wrote "Friis equasion" instaed of Shannon... Maybe because I often refer to both equasions to verify "revolutionaly communication method" on news :-)
For comparison, the US digital TV standard requires about 15.1 dB of SNR to achieve reception, in a gaussian channel. US DTV uses channel bandwidth of 5.3 MHz (and guard bands that bring this up to the 6 MHz channel width), and a net capacity of 19.29 Mb/s. The Shannon limit for a 5.3 MHz channel carrying 19.3 Mb/s is 10.6 dB of SNR minimum, required.
So this now-20-year-old standard is already only 4.5 dB from the Shannon limit. Unless Shannon's limit can be proved to have been violated, there ain't any 10 dB gains to be had here.
DVB-T2, the new European DTV standard, gets even closer. Last time I checked, it was ~ 3 dB from the Shannon limit.
So, all of this tells me that we're not looking at any "breakthrough in modulation." We're looking at refinements, much like DVB-T2 refined DVB-T1. Marginally better FEC codes, clever tricks on twisting the constellation, better interleaving, and so on. Small improvements that provide a small but measurable improvement.
Also, a significant point here. The purpose of OFDM is NOT to improve spectral efficiency. It is to improve resistance to multipath distortions. There's no such thing as a free lunch. What you pay, with OFDM, is moving away from the Shannon limit. So if a new modulation standard goes back to a single carrier approach, with improved equalizers, no one should be surprised. Equalizers benefit from Moore's law, after all. They are bound to improve over the decades.
I guess the catch is that QAM 4096 may be near impossible to implement taking into account the practical implementation issues (EVM etc etc). I think the benefit must be that Magnacom has a scheme theoretically almost equivalent to coded QAM4096. Thats a big deal.
Today's codes achieve very close to this (I have heard). So coding+QAM4096~=capacity. So would be interesting to see which way this company goes especially if you decide to toss the popular OFDM-QAM combination. Who would want them for 1-2 dB improvement ?
Frii's equasion is C=Blog2(1+S/N), while C=data rate (bit/sec), B is bandwidth (Hz), S/N is simple signal-noise ratio (not in dB).
While is is easy to define B, but assuming S/N is trickey part. Typical WiFi usecase is signal level around -70dBm, while background noise floor will be around -90dBm, so SNR is about 20dB.
Based on 802.11ac standard, 256QAM MCS9 datarate is 200Mbps per stream with 40MHz channel, short GI, coding rate 5/6. Raw (pre-FEC) rate is 240Mbps.
So it proves 802.11ac MCS9 is within typical usecase, even though 2dB margin is pretty low. Of course "typical" usecase could be varied - you'll get -50 - -60dBm signal if your PC is close to AP (within 10ft) so MCS9 will be much more practical.
Theoritically, 4096QAM (64x64 constallation) is 16bit/symbol so it should have x2 datarate than 256QAM (16x16 constallaton, 8bit/symbol). Thus, we can assume 802.11ac 4096QAM must have x2 datarate than 256QAM MCS9.
It shows "4096QAM WiFi" will be inpractical (so I don't think WiFi will adopt more-than 256QAM modulation / stream). Even if their claim of +10dB advantage is correct, 26dB SNR will be still tough to find in public wireless networ (WiFi or LTE).
However, it will make sense to backhaul, where dedicated frequency band is used with much higher TX power and highly tuned directional antenna.