Co-channel interference

Co-channel interference

By CWNP On 01/15/2008 - 16 Comments

While 802.11a/b/g stations connect to APs at data rates anywhere from 54 Mbps down to 1 Mbps, when the network is properly designed, data rates are kept as high as possible.  It's important to consider, however, that just because the slowest client might be connected at 12 Mbps in a cell, that doesn't mean that the RF signal just stops right there at that client.  On the contrary, the signal keeps going well past where a user might connect at 1 Mbps.  Even at this great distance from the AP, the RF energy is strong enough to cause clients to defer transmissions due to "busy" clear channel assessments.  This distance might be hundreds of feet indoors, depending on the environment.  In addition to the energy emitted by AP transmissions, one must consider the energy emitted by client transmissions as well.  Clients move away from APs while transmitting, and thus cause co-channel interference at a much greater range even than the AP can cause. 

Consider a scenario where a client is connected to an AP at 36 Mbps at a distance of 50 feet.  The AP has its own interference range that may extend to 200 feet (4 times the current connection distance), and the station's inteference range may be 250 feet (200 + 50) from the AP if transmitting at the same output power as the AP.  If the client moves to 100 feet from the AP and connects at 6 Mbps, its interference range may be 200 feet around itself, but that will mean 300 feet from the AP with which it's connected.  

Given this scenario, it's obvious that with recommended AP spacing of 50-70 feet and 15-20% cell overlap that co-channel interference will be a tremendous throughput thief across an entire WLAN.  In addition to a significant loss of throughput, increased jitter and delay will be evident in VoWLAN phones.  Although WLAN QoS prioritizes WLAN traffic, this occurs after the CCA and therefore prioritization does not overcome the jitter and delay introduced by CCA.  While follows most manufacturer's guidelines on AP spacing and cell overlap, an AP on channel 1 (for example) will almost always still interfere with many other APs on channel 1.  The same will hold true for APs on channel 6 and 11.  

If you're thinking that deployment in the 5 GHz UNII bands will let you escape this problem, think again.  While 5 GHz signals have less effective range than 2.4 GHz signals, APs are spaced closer together to meet the recommended cell overlap.  This causes exactly the same problem in 5 GHz deployments.  By now you must be wondering, "What's the answer to this problem?"  I will let Cisco Systems give you a hint as to where the answer does and does not lie:

"It is not an effective strategy to reduce the overlap in order to reduce co-channel interference.  As users satisfaction can be greatly affected by poor roaming performance.  In contrast, call capacity can be addressed in planning and design."
                                Cisco Systems
                                VoWLAN Design Guide v4.1
                                December 28, 2007

My opinion of how to address this problem starts with choosing the right system architecture for the application being deployed.  It doesn't lie in proper cell planning.

16 Responses to Co-channel interference

Subscribe by Email
play game Says:
06/14/2018 at 03:01am
Hello guys you have to visit here and save the all function in windows for find battery icon this is the way for show battery icon taskbar so must be thanks full to visit here and get the update

Ionso Ionso Says:
05/05/2018 at 02:37am
We all should get the co channel shared interference operations and updates on review blog. Then everyone will come forward to make better features and essay online features. So, we can manage their needs and essay channel features.

07/01/2010 at 06:58am
I clearly don't understand how 2.3-2.4 GHz WiMAX can interfere with 2.4-2.5GHz WiFi, they are completely non overlapping bands. If I m wrong.... can somebody give me a strong proof or experimental results????
I need exact proof not a bare theoretical hypothesis.
Thanks in advance :)

03/11/2008 at 13:01pm

What were you trying to say in the final sentence? It seems to be laced with innuendo... but you didn't finish by giving any recommendation. I'd really like to know what solution you think works for co-channel interference.

02/04/2008 at 11:59am
Is this discussion ignoring any design that uses directional antennas, which will help reduce the time slot sharing capacity issue on a single channel? And also ignoring the more complicated design issues connected with beam-forming smart antennas? Would an AP with a smart antenna:
reduce its CCA holdoffs from most directions;
simultaneously increase gain relative to a particular client;
allow the client to reduce EIRP, causing fewer collisions and CCA holdoffs;
simultaneously reduce required CCA holdoffs at other APs in random directions?

01/28/2008 at 04:32am
(having to break this into multiple messages due to posting size limits)

How about this to visualize my point:
Let's assume that in our 31 AP example, the 2.8 APs/channel are so hyperbolically inefficient that they waste 85% of the useable bandwidth while SCA is so amazingly perfect in it's operation that it wastes 0% of the bandwidth. So...

SCA has a total usable bandwidth of 100% of one channel.

MCA has a total usable bandwidth of 165% of one channel (15% of each of 11 channels). If you'd prefer 9 channels, it's still 135%.

For even more amusement, you can compare 20MHz MCA vs. 40MHz SCA and still come out significantly ahead:
SCA = 200% of one 20MHz channel
MCA = 330% of one 20MHz channel (assuming 22 channels to match the 11 40MHz ones, so 1.4APs/channel)

Which shipping enterprise AP can use more than 9 non-overlapping 40MHz channels simultaneously? Meru is shipping an older chipset which can't do DFS2 radar detection, so they are stuck with 9 usable 20MHz channels, or best-case 4 40MHz in US. Last I checked no other enterprise APs were shipping yet.

01/28/2008 at 03:59am
I'm getting flashbacks. This thread reminds me of token-ring vs. ethernet debates over a decade ago.

Can you cite some independant sources for your first point? You keep repeating the same assertions without citing any proof. I'd love to read something from somebody who wasn't paid by an SCA vendor to run a specific test.

On your second point: Why in the world would I put all the MCA APs on the same channel for any kind of comparison? All that does is provide a perfectly skewed set of numbers by creating absolute channel scarcity (i.e. only 1 channel).

SCA is most valuable (and MCA least) when there is channel scarcity - as the number of available channels increases, the value of MCA increases while SCA maintains the same level (absent blanket overlays). If you keep this in mind, a lot of the Meru marketing message make sense - they have a vested interest in making people believe there is still channel scarcity on 5GHz after 11n mitigates propogation issues and DFS2 adds more channels. The only debate would be at what number of channels MCA becomes more valuable than SCA. I think you could have an interesting debate about it in the 2.4GHz band, but not the 5GHz band.

01/25/2008 at 11:22am
Consider the following two points:

- I fully agree that a system that requires more APs to provide the same throughput for a given area would therefore be more expensive. But, as it turns out, intelligent SCAs can be capable of either, for a given area, providing the same throughput with less APs, or providing more throughput with equal numbers of APs. Thus, the cost advantage actually seems to fall to SCAs. By the way, this is not considering that MCAs scale out far quicker than SCAs do, and so MCAs might not be able to compete at the higher end of capacity.

- 2.8 APs/channel are more damaging to an MCA than 30 APs/channel are to an SCA, within the same neighborhood. This must be the most surprising to MCA designers, but it is repeatable. I mentioned a few of the reasons for this earlier. Fundamentally, it comes down to the difference in behaviors of SCA and MCA APs. One good way to understand that is to imagine placing a number of MCA APs on the same channel, and comparing to an equal number of SCA APs in the same situation. The difference in performance and system behavior will be revealing.

In any event, it is true that Cisco only provides 9 40MHz channels, rather than the 11 that others provide. However, I figured that was a temporary limitation, and not worth considering here.

01/23/2008 at 04:00am
If there were 11 40MZ channels and 30 other APs "in range", any given AP will interfere with at most 2 other APs when using MCA (31 APs divided by 11 channels = 2.8APs/channel - round it up to three). In a SCA that single AP would interfere with all 30 of it's 'neighbors'. Incidently, the highest number of 40MHz channels I've seen supported by a shipping enterprise AP is 9. In the future you should probably use that number, it puts the results slightly more in your favor: 3.4 APs/channel in MCA vs. 31 AP/channel in SCA.

If you add in the 11n wrinkle of longer range at similar densities you'll end up hurting SCA deployments even more - every new AP that is within range ends up sharing bandwidth with the first AP as opposed to every 9th with a MCA in this scenario.

01/23/2008 at 03:24am
I didn't mention channel blanketing because I don't consider it a very practical solution. Yes, the blanketing can greatly increase the scalability of an SCA system, but it is at a prohibitively high cost. Most enterprises would have a hard time purchasing a second complete AP deployment (APs, POE, cable runs, edge ports, controller capacity, etc...) on top of the first. That's a really expensive proposition to justify.

01/22/2008 at 20:51pm
11n adds an interesting wrinkle to the problem. With the extra receive gain, from both processing gain and added antennas, 11n APs and clients pick up interference at higher ranges than legacy devices do. This makes mixed-mode deployments very tough: APs cannot be spaced further out to accommodate the higher ranges without jeopardizing coverage for legacy clients. Thus, co-channel interference increases under 11n. Note that this is independent of MIMO: 11n devices detect and defer transmission using the same rules as 11g and 11a devices, just at longer range.

Frank, it is true that there are many more channels in 5GHz now. However, let's continue to use the 200 ft. interference range and the conservative 70 ft. AP spacing from the above discussion. Using typical hexagonal spacings, a little math reveals that an AP will interfere with its nearest 30 neighbors, thus requiring 31 channels to avoid overlap. Now, even if those numbers shock you, it is difficult to imagine how anyone can deploy the 11 40MHz channels and avoid the problem. In the end, it's the same: it is very hard to minimize overlap, and the usual game of musical chairs for setting channel and power levels is generally ineffective.

For SCAs, I would think of this differently. If you treat overlap as a problem to be solved upfront, rather than as something to be avoided later, you could design a system that largely solves these problems. An intelligent SCA can recover the capacity lost in an MCA due to collisions, density, aggressively roaming or sticky clients, and hidden or unexpected co-channel interference. One-channel SCA deployments can, in many cases, greatly outperform multi-channel MCAs. Therefore, the opposite conclusion applies: SCAs scale horizontally (space) and vertically (channels) better than MCAs.

I'd suggest the following metric for comparing MCAs and SCAs: find the peak throughput using the same number of APs in the same area. That gives you the apples-to-apples comparison.

01/22/2008 at 06:41am
I'm vendor-neutral and technology neutral. :)

01/21/2008 at 21:27pm

Reading between the lines, it sounds like you're suggesting that SCA is the better solution. 'Slipshod' points out how SCA has problems scaling, but omits that overlaying with multiple channels can alleviate some of those concerns (Extricom calls them 'blankets').

With the many more channels available at 5 GHz range, plus their reduced propagation, and the enhanced reliability gained from MIMO, I think a 5 GHz-based 802.11n system with minimal co-channel interference and impact can be designed. Much harder to do at 2.4 GHz, though.


01/18/2008 at 14:01pm
This is the primary reason Single-Channel Architectures (SCA) have issues scaling. Even if you could have perfect coordination of transmissions, you are still slicing up the same pie and sharing amongst every device over a large area.
If you assume the APs are uniformly distributed, you will sharing bandwidth and introducing latency across 3x as many APs in 2.4GHz vs. MCA. The story is even worse in 5GHz where the number of available channels makes it anywhere from 9x to 21x as many, depending on the number of 5GHz channels supported by the radio and regulatory domain.

Nasser- Under normal 802.11 operation they will not cause dropped connections. If some transmitters are using QoS (WMM or other) and others are not, the bandwidth and latency will not be shared evenly. Many (most?) enterprise WLAN vendors also have "containment" features which will attack other APs which have been identified as "rogue APs". The containment/countermeasures will cause severe disruptions and disconnections for the rogue AP... Once 802.11w (protected management frames) is in use existing containment/countermeasure methods should no longer work, but that is a LONG way off.

01/18/2008 at 11:50am
This is more of a question then a comment I guess!

Is the only down side from having more than two access points on the same channel that the bandwidth/throughput is shared? or are there other problems, like dropping connections that occur?

01/18/2008 at 08:16am
very interesting article, thanks Devin.
It had never occurred to me that this could happen!


<< prev - comments page 1 of 1 - next >>

Leave a Reply

Please login or sign-up to add your comment.
Success Stories

I literally just came out of the testing centre having taken the CWDP exam. The certification process opened my mind to different techniques and solutions. This knowledge can only broaden your perspective. Great job, CWNP, you have a great thing going on here.

Read More

Working through the CWNP coursework and certifications helped not only to deepen my technical knowledge and understanding, but also it boosted my confidence. The hard work it took to earn my CWNE has been rewarding in so many ways.

Read More

I want to commend you and all at CWNP for having a great organization. You really 'raise the bar' on knowing Wi-Fi well. I have learned a ton of information that is helping my job experience and personal career goals, because of my CWAP/CWDP/CWSP studies. Kudos to all at CWNP.

Read More