Your browser is not supported. Please update it.

11 December 2014

Avoiding WiFi becoming 4K’s bottleneck

Even before 4K, WiFi had become the new performance bottleneck for video streaming. Unlike the fixed broadband access network WiFi is inherently un-deterministic and therefore unpredictable, so that in some homes it works fine for premium HD delivery, while in others receiving the same headline broadband bit rate, QoS is intermittent and poor.

This was reflected in a recent survey by Arris finding that 53% of North American users reported streaming issues over WiFi, while 8% had persistent and serious problems. The story becomes more interesting when we drill down, since Arris seemed to identify different underlying causes for poor performance than other studies, for example from Turkish WiFi streaming specialist AirTies.

Arris identified congestion as the principle issue, at present mostly confined to the 2.4 GHz band. The current IEEE 802.11ac standard, which was finally ratified in January 2014, introduced the parallel 5 GHz band largely for this reason and the main point of the recently published Arris paper was to argue that operators should prepare for that band to become congested as well.

Arris noted that the 2.4 GHz spectrum was particularly full in US Multiple Dwelling Units (MDUs), where the incidence of problems is higher than for single dwelling homes. In MDUs there is a higher incidence of contention between neighbors’ WiFi networks, which has become an issue as video streaming has increased.

There are also plenty of MDUs in some of Europe’s leading markets, especially Germany, but a key difference is that building walls tend to be thicker, containing more water, which soaks up the microwaves and also metal, which can reflect or divert signals. This may actually help reduce contention between neighboring WiFi networks, but increases attenuation, so that a device separated from the nearest WiFi Access Point (AP) by say a floor and two walls can easily obtain only 10% to 20% of the bit rate of one in the same room.

The AirTies study found that average throughput between a device and an AP over a full 802.11ac WiFi link with 3X3 MIMO with beam forming on both ends was 802 Mbps when both are in the same room, 593 Mbps when there are two typical walls separating them and just 96 Mbps when they are separated by a floor as well. AirTies pointed out that the situation is even worse in many real deployments where there are multiple mobile devices with varying capability.

AirTies also found that contention between multiple devices became worse in rooms further away from the nearest AP. AirTies uses this study as ammunition to sell its mesh technology approach, which certainly does tackle the issue of attenuation quite effectively, especially in multi-floor buildings containing a lot of reinforced concrete. It replaces the typical star topology where central APs communicate with clients in a one-to-many arrangement, with a flat mesh where the end devices themselves participate as APs.

This reduces distances and also the re-transmission overhead of central APs that have to receive signals from the source devices and then send to the destination. But we would like to see a definitive study backing up AirTies’ claims that most homes can achieve effectively deterministic WiFi performance with just two dedicated APs and almost all with three, without any need for a wired backbone. One indisputable point made by AirTies is that although the 5 GHz band does deliver more bandwidth having more bandwidth in it, for the same reason it suffers from greater attenuation, increasing the appeal of some form of mesh technology.

The Arris argument is that whatever technology is adopted, enabling 4K streaming over WiFi is going to be a challenge. This is backed up by wireless semiconductor company Qualcomm, which pointed out that mobile device makers are being the first to adopt 4K.

Leading 802.11 ac chip maker Quantenna has also presented findings from several studies, showing that contrary to earlier expectations, 4K makes a discernible difference to quality even for screens as small as 10 inch tablets. The key measure of perceived display clarity and sharpness is not the headline resolution, such as 1920 x 1080 in 1080p, but the pixel density per inch (PPI), according to Quantenna. This makes sense because pixel size varies with screen dimensions and so from a given distance a tablet will give a sharper display than a 60 inch TV. Tablets tend to be viewed from close range and at 1080p resolution their PPI is around 220, while the average human eye can discern 344 and somebody with 20/20 vision 573.

A study by Sharp Devices Europe found that in practice humans could discern even higher PPIs by a factor of up to 10 because the brain is capable of resolving details that are slightly offset within the picture, such as the relative alignment of two line segments. The technical explanation is that a line in an image stimulates many photoreceptors in the retina along its length, enabling neural processing to fine-tune object locations based on these multiple inputs.

This may all be bad news for WiFi since the logic of Quantenna’s argument is that on a tablet viewed from 10 inches, most humans will also be able to discern a noticeable difference between 4K at 3840 x 2160 and 8K at 7690 x 4329, doubling the PPI again. Arris argues that MSOs at any rate will need a radically new architecture to enable 4K over WiFi, never mind 8K if that is ever deployed. This will combine measures within the physical wireless domain with higher level techniques aimed at delivering sustained QoS to make WiFi acceptably deterministic.

Arris indicates, somewhat contradicting the AirTies line, that wired assistance may be required. Arris suggests that future cable gateways based on DOCSIS 3.1 will have options for range extension through high speed Gigabit Ethernet, Dual Band MoCA or G.Hn, to other locations in the house. On the other hand AirTies has suggested to us that it is working on just the same type of architecture, but is perhaps not ready to talk about it quite yet, using powerline as an alternative backhaul for WiFi instead of just more WiFI.

The most interesting aspects of the Arris white paper, entitled The Quest to send 4K Video over WiFi Networks, concerns the integration between the service level and wireless domain in order to reconcile the requirements of premium 4K streams with “ordinary web traffic”.  Currently WiFi APs tend to operate independently of the service level, but in future will more likely need to communicate with higher level video management agents. For example the management system might want to ask an AP whether a new 4K session can be supported. The AP will then apply emerging tools to determine whether or not a new 4K session can be supported without impacting other services. If it turns out that a new 4K session can only be supported by restricting the bandwidth available to other best effort services, then the system can decide what to do, perhaps downgrading the 4K stream to 1080p for the time being.

More complex situations can arise, but Arris’ point is that it is time for operators to start considering these now, which is why it has published its white paper.