Forum

Greetings,

My first topic on this forum will be a simple question:

Do other networks transmitting on the same channel affect my network's medium access? In other words: does my network's devices need to go through medium contention procedures because of other nearby network transmissions?

I believe the answer would be yes, because a collision is the incapability of correctly understanding a received packet at the PHY level due to multiple signals and packets arriving at the same time. I'm not completely sure, though.

thank you

Juan Carlos

You are correct, they will contend with each other. A normal Wi-Fi frame doesn't even indicate the SSID that it belongs to so there is no simple way for a STA or AP to know if the frame is from its own SSID or a different SSID.

GT

You would be interfered with at layer 1, but the MAC frame would be rejected due to an incorrect destination addr. Also the BSSID is included in the MAC frame. So the other networks frame would be discarded by the AP.

Juan Carlos and Ronald

One thing to note is that we have to be a bit careful with the term "interference". If we had two point to point links with continuous signals (carrying, say an E-1 or T-1) and they were set up on the same frequency (say 10 GHz) and the signals from one were being picked up by the other's antenna (due to a bad design or lack of knowledge of the existence of the other), then that interfering signal would be there as long as power was applied to both systems.

With Wi-Fi systems, there is a "sensing" mechanism which is like a "listen before speaking" type of system. If one radio detects the signal from another (on the same channel), it will "wait a bit" "until the air is clear", followed by some other calculated delays before transmitting. In this case there is no actual physical interference in the sense of one carrier being "sat on top of" the other carrier. In nearly all documentation you will see this listed as "co-channel interference". That is where the definition of "interference" becomes a bit hazy. If it is meant that the "other" signal has interfered with the first signal's ability to transmit, then that could be described as a form of interference, but only in the sense of system throughput and efficiency.

However, in the classical case of RF interference, if both transmitted at the same time, then that would truly be Radio Frequency Layer 1 interference.

David Coleman (one of the CWNA study guide authors) calls this "Co-Channel Co-operation" when one station detects the signal from another and "waits". The "Co-operation" expression is a good one, as it describes what the CSMA/CA mechanism was designed to do.

Even though many radios are designed to operate in "burst condition" mode, it is still possible for internal electronic circuitry faults to occur whereby a radio can transmit a "continuously on signal" (this is rare, but I have seen it happen). In that case, all other radios "listening in range" on that channel could possibly defer "permanently" depending upon signal levels etc.

These are small, but important points.

There is a lot more to this whole business, but this should hopefully help a little.

Dave

Thank you dave, ronald and GT for answering my question.

Here's another one I was thinking about, related to this topic: Is there any method / tool for calculating VoWIFI Capacity?

My first guess would be that, theorically speaking, an infinite number of calls could be supported by a single AP because of the medium contention mechanisms that should give a "fair shot" for everyone. However, as more VoWIFI clients transmit their own stream, there's a great chance that each frame will be sent even more delayed, thus decreasing throughput and, moreover, exceeding a threshold in which it is no longer feasible to support any new concurrent call.

There are a few tools out there that can help. This is a very complex area. There are many factors involved, such as whether the "network" is all voice, voice and data (and in what ratio ) etc. Call Admission Control (CAC, where used) can be involved to restrict the number of calls in progress when resources (bandwidth etc) are becoming low. A user attempting to connect in a CAC system will usually receive a form of "network busy tone". Vendors are pretty secretive about the exact calculation methods by which they figure out the max number of calls.

You wouldn't get an infinite number of calls due to a number of reasons....finite bandwidth to "be shared", AP resources (memory, processor "power" etc).

Over the last while, the IEEE and the Wi-Fi Alliance have developed functions which can help with voice QOS. There are special frames that can ask for the "reservation" of a certain amount of bandwidth requested by the "client".

Dave

Bye the way, if you can, try to work with your vendors on this one. There are a lot of vendor specific issues involved.

If you have a copy of the CWNA study guide (PWO-104), there is some useful info on page 506 e.g. "Polycom recommends a maximum of 12 calls per 11 Mbps cell". This is just a rough guide, but those guys work voice as their "bread and butter". You should drop them a note. I'm sure they would be able to help you.

When you are working with broadband transmission systems, you need to figure out when the "busy hour" is....i.e. the physical period of time during which there is maximum call activity. Telephone switch designers do not calculate the switch size directly as to the number of lines coming in to it, but to a subset of that number. They rely on the fact that statistically it is very unlikely that all lines will be "seized" simultaneously. On rare occasions, a scenario close to this has happened. After September 11, some exchanges crashed due to the massive seizure of lines.

Erlang calculations (traffic calculations are important):

http://www.erlang.com/

Dave

Phone people call this Tele-Traffic Engineering - or at least they used too.

The same kinds of calculations can be used for computing memory buffer requirements and interrupt queue sizes.

One thing to beware of though, is that people in the US and and the UK tend to look at the reciprocal of the values that the other one looks at. This can really set your head spinning, until you "get it" - and by then you'll understand the whole subject much better.

I once worked on a system with predefined queue sizes and my calculations showed that only a one element interrupt queue was necessary. I told myself this was insane - I wasn't that short of memory- so I made it two long.

You may say, why make them so small? Except for this one queue, I made the other 8 queues the size that theory said they needed to be. If they had ever overflowed, then some part of the system wasn't working the way it should have been. It was a good way of proving the design. None of them ever overflowed, in three years of 24/7.

You just have to understand the "traffic" patterns correctly.

I know this isn't exactly Wi-Fi related, but I had to add in my two cents worth regarding something that worked very well for me.

When you read manfactures recomedations of amount of phone calls on an AP, one must know where they are going. In a lab environment it is so easy to call another handset on the same AP and then you will have four voice streams on that AP instaed of two.

Martin

thank you very much for your replies.