Everyone nowadays buys ‘super-fast’ 802.11n or 802.11ac APs, but not everyone manages to get super-fast connection with them. In this article we’ll discuss the not-so-obvious or often overlooked nuances that can substantially improve (or impair) the Wi-Fi connection. Everything below is applicable both to small home networks with off-the-shelf consumer-grade routers using stock and advanced (DD-WRT, etc) firmware, as well as to enterprise-grade gear and networks. Thus, to keep it simple, I will be using the home WLAN as an example, since even the most advanced admins and most proficient engineers still live in residential areas with high neighbour AP density, and everyone wants fast and reliable Wi-Fi.
Before we begin, a few comments:
- The narrative is a little simplified, as you might want to explain some of these aspects to your neighbours, who probably have no clue of RF essentials, 802.11 standard or government regulatory policies.
- Everything below is a recommendation only. Some of the aspects might be inapplicable to your case. Every rule has an exception, and many of those exceptions are omitted for brevity and overall clarity. We can discuss any of those in the comments, though.
- Please note the word ‘unobvious’. Detailed proof for some of the theses might require deep-diving into standards, and I don’t want to do it without a serious reason (even though I had to a few times).
1. How to live and let live (you and your neighbours)
[1.1] It would seem trivial – crank the AP power to the max, get the max coverage, enjoy, right? Now, let’s think: it’s not just about AP signal reaching the client – the client’s signal has to get back to the AP as well. The AP transmit power is usually around 100mW (20dBm). Now, look into the spec sheet of your laptop/phone/tablet and find their transmit power. Found one? You’re extremely lucky – most vendors don’t even bother specifying it for the masses (the techies are using FCC ID search). Nevertheless, we can safely assume that the Tx power of most mobile clients is around 30-50 mW. Thus, if the AP broadcasts at 100mW, but the client – only at 50mW, there will be areas of coverage, where client would ‘hear’ the AP well, but the AP will hear the client poorly (or not hear at all) – asymmetry. This is addressed a bit by the AP receiver sensitivity, which is typically higher than the client’s one, but still, typically, some asymmetry remains (see the bottom of the post for some math on this). Note, that I’m emphasizing not range, but symmetry. There’s signal, but no link. Or signal is good, but link (especially uplink) is bad. This is important when Wi-Fi is used for online gaming or video conferencing (Skype, etc), not as much important for generic Internet access (unless you’re at the fringe of the coverage zone). This typically ends in blaming the ISP, blaming the router/tablet vendor, blaming the drivers, but not the inadequate network design.
Conclusion: it might happen, that in order to get better link quality, one might need to actually reduce the AP power. Which, agree, is not very obvious.
[1.2]. Another, not-so-obvious fact that adds to the asymmetry is that the majority of client devices must reduce Tx power on the ‘edge’ channels (1, 11/13/14 – depends where you’re at). Here’s an example for one of the iPhones (from FCC, sorry, source link no longer available – here’s a link to a post by Andrew vonNagy discussing the same ), listing the antenna port power:
- Channel 1: 13.50 dBm
- Channel 6: 17.00 dBm
- Channel 11: 13.50 dBm
As you can see, the Tx power on the ‘edge’ channels is ~2.3 times lower than in the middle. The reason is that Wi-Fi is boradband communications. As such, it is impossible to contain the whole spectrum of the signal in within the designated channel boundaries (part 2 will illustrate this). So manufacturers have to reduce the power on the ‘edges’ to ensure acceptable (defined by regulatory policies) level of signal bleed into frequencies neighbouring with the ISM range.
Conclusion: if your Wi-Fi does not work well in your bathroom – try migrating to channel 6 🙂
Ever wondered why faster rates have smaller effective distance? It was always attributed to signal fading – higher rates require more complex modulation, which, in turn, requires higher SNR (Signal-to-Noise Ratio) value. Turns out, in addition to that the transmitters may as well lower their power when transmitting at higher rates! This is done because at higher power values transmitters are more likely to introduce errors in the signal they create due to non-linearity in their behaviour (if you want more details – just go to the discussion). Here’s an example from a datasheet that Eduardo provided (the numbers are Tx Power, dBm):
The users of alternative consumer WLAN router firmwares (such as DD-WRT and Tomato) sometimes suffer from strange performance and link reliability issues, as these firmwares allow varying radio power levels (and other power settings) beyond what manufacturer has intended, resulting in poor Tx signal quality. This mainly is just a result of the shortcomings and compromises in hardware design (even if the APs are built on the same chipset), but certainly not something a non RF engineer would suspect first! 😉
That’s it for this time. Next time we’ll look into channels. Some extra math below, as promised. I wish WordPress had a ‘spoiler’ tag.
Our goal is to ensure that there’s no chance that STA will hear the AP, but AP will not hear the STA. This means SNR at Station is <= SNR at AP. Why SNR (signal-to-noise ratio), and not signal strength will be explained once I port 3.1].
- SNRSTA <= SNRAP . Let’s work it out:
- RxSTA – RxSensSTA <= RxAP – RxSensAP , or
- RxSensAP – RxSensSTA <= RxAP – RxSTA , where
- RxSTA/AP is the strength of a signal received at STA/AP from AP/STA.
- RxSensSTA/AP – receive sensitivity of STA/AP.
- We are simplifying here, by assuming there’s no noise impact on the link (which is not that bad – if the noise level is the same at AP and STA – it will not affect the symmetry).
Moving on, let’s expand what Rx is:
- RxSTA= TxAP + TxGainAP – PathLoss + RxGainSTA
- RxAP= TxSTA + TxGainSTA – PathLoss + RxGainAP , where
- TxAP/STA – AP/STA transmitted signal power, at antenna port
- TGainAP/STA– AP/STA antenna gain, including all the cable/wiring losses and directionality effects
- PathLoss – all the losses on the path between AP and STA.
- RxGainAP/STA – STA receive gain from antenna, including all the wiring losses and directionality effects.
Stay with me, we’re almost done… Note the following:
- PathLoss is the same in both directions.
- TxGain and RxGain for the same antenna is the same. .
- Thus TxGainAP == RxGainAP and TxGainSTA == RxGainSTA
- This is not strictly true for MIMO effects, but let’s hold this discussion off for later in the article.
- RxSensAP – RxSensSTA <= RxAP – RxSTA
- RxSensAP – RxSensSTA <= TxSTA + TxGainSTA – PathLoss + RxGainAP – [TxAP + TxGainAP – PathLoss + RxGainSTA]
- …opening the braces and applying RxGain == TxGain we see that most variables are eliminated…
- RxSensAP – RxSensSTA <= TxSTA
+ TxGainSTA – PathLoss + RxGainAP– TxAP – TxGainAP + PathLoss – RxGainSTA
- TxAP – TxSTA <= RxSensSTA – RxSensAP
Note how all the antenna gains and path losses were eliminated – the channel asymmetry depends only on Tx powers and Rx sensitivities of parties involved. Basically it says: “The AP power advantage should be compensated by AP sensitivity advantage“, which is exactly what we were trying to prove. Keep in mind that sensitivities are negative numbers when expressed in dBm, so better sensitivity is smaller number, thus it’s STA – AP, not AP – STA on the Rx side.
We can also quantify and characterize the asymmetry – is it uplink or downlink, and by how much?
- D = TxAP – TxSTA – (RxSensSTA – RxSensAP)
- When D = 0 we have perfectly symmetrical link. AP power advantage is compensated by Rx sensitivity advantage. SNR at both ends is the same, whenever client hears the AP – AP hears the client. We’re good. However, perfect balance is never reachable (see note on MIMO below), so let’s see what happens when D != 0.
- When D > 0 we have downlink advantage: it is possible for client to hear AP without AP hearing the client, or having asymmetrical up/down rates. Thus, client may experience various uplink problems. The solution in to reduce AP power to make D =0.
- When D < 0 AP sensitivity overcompensates for AP power advantage: uplink advantage.This may seem like a problem as well – it may result in AP responses not reaching the client, and all sorts of related issues. But it doesn’t. Homework: what could the potential problems be, and why we’re not seeing them? 🙂
Here’s an example of an enterprise-grade laptop with consumer-grade AP, 802.11g@54Mbps:
- HP 8440p laptop: Tx(STA) = 17dBm, RxSens(STA) = -76dBm@54Mbps
- D-Link DIR-615 AP: Tx(AP) = 20dBm, RxSens(AP) = -65dBm@54Mbps
- D = (20 – 17)- (-76 – -65) = 3 + 11 = 14 dBm > 0. A clear case of inferior AP, one will have to reduce AP power even beyond the client’s to balance the link!
Or get a better AP – same client with Motorola AP-7131N Enterprise-grade AP (Tx = 20dBm, RxSens = -82dBm@54Mbps):
- D = (20-17) – (-76 – -82) = 3 – 6 = -3 dBm < 0. Minor uplink advantage (with up to 4.7dB extra MRC gain depending on protocols and MCS used), no problems.
A note on MIMO effects.
All of the above is nice, but applies only to single-stream single-antenna scenario. Modern MIMO effects such as beamforming, MRC, STBC, etc (which do not always apply anyway) kind of break this rule in many ways.
For example, beamforming provides pure Tx gain, while MRC and STBC provide pure Rx gain. Some APs may use specialized antenna arrays for Tx, but totally different omni antenna for Rx. All this breaks the RxGain == TxGain assumption, since now we heed to add all these other gains into equation, besides the antenna gain (antenna RxGain always equals antenna TxGain – physics).
Furthermore, these gains may depend on the number of spatial streams, difference between number of Tx antennas and number of Rx antennas, client [im]mobility (MU-MIMO and beamforming) etc.
And this is why you MUST know the AP type/capabilities, client type/capabilities, the desired/allowed MCS AND use specialized software such as Ekahau or AirMagnet to predict WLAN performance and coverage more or less reliably these days (does not apply to “Telnet in the warehouse” or “1-room cafe” scenarios 🙂 )
Anyways, note that for many 1-antenna clients (which is still the majority of cheap/small devices out there) all these techs will only INCREASE the imbalance (great downlink, poor uplink)! Which, actually, can be leveraged by using, for example, protocols that do not require ACKs (i.e. multicast vs unicast, UDP vs TCP etc) – but that’s a whole different story!
P.S. Special thanks to Markus Burton, and T R who took care to read through all my math and found the hidden glitches. This section had been rewritten with fixes since then.