Why 802.11ax matters
802.11ax matters. This is not just another marketing trick to make you buy a new router (although it is true that you will need to buy one). Its goal is to increase the throughput by at least 4x. But mostly, it will help with multi-users to make it competitive with LTE-U and LAA (think crowded environments). 802.11ax finally might have a shot at solving Steve Job’s Moscone Wifi nightmare.
What's new in 802.11ax
The key features improvements in 802.11ax are the following:
- Better support multiple users (with OFDMA)
- More throughput by providing better OFDM symbol efficiency. Also more complex constellation (1024QAM) and more subcarriers.
- Uplink multi-user MIMO (UL MU-MIMO). Again this is for more throughput.
- Better outdoor support.
Oh yes, that’s a lot of acronyms. People like to throw those to make you look dumb. Well, let’s take a look at the meaning of each of them.
OFDMA
OFDMA: Stands for OFDM Access. This is the similar technology used by LTE to have multiple user share the OFDM bins. This is probably the technology that will make the most impact versus previous Wifi specs. Currently, all the subcarriers are only used for 1 client. With OFDMA, we can do Tx/RX to and from different subcarriers. We will dedicate an entire blog post for this.
UL MU-MIMO
UL MU-MIMO is very challenging. Much harder to implement than OFDMA. We would need to do beamforming to handle interference. But it can allow us to have multiple users.
More Throughtput
One simple way to increase the throughput is to have more subcarriers. To achieve this, we need smaller subcarrier spacing. Think of OFDM like a freeway. Now we improved the freeway by reducing the width of the freeway lanes to pack more cars. So instead of having a 4 lanes for 4 Cadillac Escalade, we have 16 lanes for 16 chevy sparks. The actual width of the freeway is the same, we just reorganized the lanes. For OFDM, the width could be 20MHz for example and the spacing is now 78.125kHz (instead of 312.5kHz). The actual reason behind this is to support outdoor transmission, since the multipath are longer. By reducing the spacing, we increase the symbol duration (remember frequency = c/wavelength). Since the symbol duration is longer, the multipath needs to be even longer for having intersymbol interference (ISI).
Let’s go in depth on the choices and tradeoffs:
The current subcarrier spacing is 312.5KHz. It was picked in 1999 and was mostly because of the PPM requirement on 5GHz oscillators (back in ‘99). The LTE subcarrier spacing is 15kHz and LTE was designed with mobility in mind. So 15kHz can support the long outdoor channel delay spread.
Let’s now talk about efficiency:
OFDM level : 3.2/ (3.2+0.8) = 32/40 = 80%. So it means 20% of all data and power spent is for header.
Now with 12.8us symbol time, our efficiency is now at 94.11% (12.8/ (12.8 + 0.8) ).
One of the tradeoff about efficiency is support for outdoor channels. Remember that OFDM has a cycle prefix and that prefix determines how much ISI we can support until it starts affecting our subcarriers. The longer the prefix, the more ISI we can support.
Therefore, it was decided in 802.11ax to support 3 options. 0.8us, 1.6us,3.2us.
Spectral efficiency formula:
CP = 0.8us. SYM = 12.8us (symbol time)
\(= \left ( 1 - PER \right ) \left (\frac{1}{1+\frac{CP}{CP+SYM}} \right )\left ( R \right ) \left ( bps \right ) \left ( \frac{tone_{used}}{tone_{total} } \right )\)
\(= \left ( 1 - PER \right ) \left (\frac{1}{1+\frac{0.8}{0.8+12.8}} \right )\left ( \frac{5}{6} \right ) \left ( 8bps \right ) \left ( \frac{234}{256 } \right )\)
We can see that 802.11ax with 20Mhz, 256FFT with R=⅚ and 256QAM, we can achieve almost up to 6bps/Hz. Where 802.11ac could only achieve 5bps/Hz at most (80Mhz, 256QAM, R=⅚).
So we can say that 802.11ax (6bps) is 20% more spectrally efficient than 802.11ac (5bps) and it's still not comparing apple to apple, since we are comparing 80MHz bandwidth for 802.11ac vs 20Mhz for 802.11ax.
In another post, we will go in details on OFDMA and it brings to 802.11ax.
Note the information in this are based on the draft version on January 2016. The final version is likely to change (but not by much hopefully)