Spread Spectrum and Modulation/Encoding Techniques used in 802.11 Wireless

Written By Eric Jackowski

The goal of this paper is to educate those curious on how information is sent in wireless LANs. This paper is about the physical way the signal is sent, that is to say we will not be preforming a dissection of the 802.11 protocols but rather the manner in which the ones and zeros make their way through the air. We will be covering the various spread spectrum technologies used such as FHSS, DSSS and OFDM. Beyond the spread spectrum technologies we will also be discussing Barker Codes, CCK, DB/QFSM and other fun stuff, so sit back and enjoy the ride.

Narrowband:

In order to know what spread spectrum is, it helps to know what it is not, and most importantly ? why it is not. Narrowband transmissions are just the opposite of spread spectrum. They use high power over a small frequency. All the information is transmitted over that single frequency (well, a very narrow range of frequencies at least) and the high power is utilized in order to make the signal detectable over the background.

The main problem with narrowband transmissions is their susceptibility to corruption and jamming. The signal occupies a very small realm of the spectrum, therefor any interference will cause significant degradation of the signal. Since the signal is operating at the same frequency all the time, someone only has but to set up a transmitter on the same frequency at high power to drown out the signal.

This is not to say that narrow band is inherently evil. Narrowband transmitters/receivers have their place in the world, the radio in your car used narrowband transmissions for example. You will also notice your radio does not have the best quality. When quality is of paramount importance narrowband just can not offer what is needed, what we need is spread spectrum

FHSS:

The first spread spectrum technique we are going to discuss is frequency hopping spread spectrum (FHSS). This was the pioneer of spread spectrum and definitely has advantages, however, those advantages are also quickly superseded by its faults. The idea behind FHSS is a simple one; get out of the way of noise.

The concept of spread spectrum is directly in conflict with the goals of the bodies whom govern the use of the airwaves. The purpose of such organizations is to restrict the use of the airwaves so that they do not become congested. The idea of spread spectrum is to spread a single signal out over many frequencies.

The original 802.11 standard defines the use of FHSS and gives it a range of 79Mhz from 2.402Ghz to 2.48. In order for FHSS to work a couple of details need to be hashed out between the AP and client. One thing that they need to decide upon is the dwell time. The dwell time dictates how long the IR will transmit on a particular frequency. In addition to the dwell time a hop time needs to be calculated. The hop time defines how long it takes to hop from one frequency to another. Lastly in order to communicate a hop pattern needs to be determined. This is a list in essence, it lists the order in which the signal will hop, once it reaches the end of the list it starts all over again. All three of these values need to be synchronized between the AP and client or communication will not take place.

If one were to consider the information provided in the previous paragraph one would correctly assume that the fastest transmission would be achieved through a dwell time of infinite. The signal would loose no time hoping from one frequency to another. This is absolutely correct, and also absolutely defeats the purpose of spread spectrum.

The IEEE has therefore enforced a rule which states that a FHSS IR must have at least 6Mhz of separation between hops and a maximum dwell time of 400ms for any 30 second time span. How you divide that 400ms up is completely up to the designer of the system. This was originally in accordance with the FCC, however the rules were later (31-08-00) changed but the IEEE did not bother to update its requirements since FHSS had pretty much been phased out already.

FHSS can achieve data rates of 1 and 2 Mbps (well, with the FCC change that gets a boast to 10Mbps but that does not effect 802.11). Due to the extremely slow nature of FHSS its use has faded from this plane of existence. The information is provided here for educational purposes only, the only place you will probably find equipment utilizing FHSS would be at a garage sale.

DSSS:

After FHSS came Direct Sequence Spread Spectrum (DSSS). Unlike FHSS DSSS does not have a bunch of dedicated channels. What DSSS does is define a center frequency for a channel and kinda leaves the rest to the wind. It does stipulate that at plus or minus 11Mhz the signal’s power must be decreased by 30dB and at plus or minus 22Mhz, 50dB. Channels are considered “non over lapping” if they are separated by 25Mhz

In the United States the FCC has provided 11 channels for 802.11b/g transmissions. Again, these channels are dictated by their center frequencies and start at 2.412Ghz for channel one and go to 2.462Ghz for channel 11. Interestingly enough 14 channels are written into the standard yet no country at all makes use of the 14th, it used to be that Japan *only* used channel 14 but their laws changed and their channels now mirror Europe (ETSI).

DSSS is utilized in 802.11b wireless LANs and is sometimes referred to as HR-DSSS, the HR standing for “Hight Rate” quickly being phased out for faster and better technologies. DSSS is a step up from FHSS offering data rates of 1, 2, 5.5 and 11 Mbps which is perfectly adequate if all you plan to do is surf the web with your broad band Internet connection, but not good for much more than that.

The main problem with DSSS is interference from over lapping channels. Channels need to be 25Mhz apart to be considered “non overlapping” (however the reality of this is not so much the case as we will read about later). The channels themselves however, are only separated by 5Mhz, using quick math we can see that 5 * 5 = 25, ergo we need five channels of separation to get a non overlapping channel. This leaves us with channel 1, 6, and 11 or 2 and 7, etc.

OFDM:

Finally we come to the most interesting of the spread spectrum techniques. I suppose the first thing that we need to discuss about Orthogonal Frequency Division Multiplexing (OFDM) is that it technically is not even spread spectrum. Another interesting aspect about OFDM is that it is used in both 802.11a and 802.11g which are not interoperable. Meanwhile, 802.11b and 802.11g are interoperable, but their spread spectrum technologies (DSSS and OFDM respectively) are not.

OFDM then is this weird foray into spread spectrum. It masquerades as spread spectrum when it is not, does not play well with itself, and plays fine with strangers. Just what is this strange form of “spread spectrum” then?

First let us explore the strangeness of 802.11a and 802.11g. These two standards are exactly the same, yet they do not work together. What separates these technologies is the RF spectrum. 802.11g works in the 2.4Ghz spectrum while 802.11a is all alone in the 5Ghz spectrum. We will discuss why this is later. Ok, so that is one mystery solved and one left.

The other oddity of OFDM that we discovered was that it is used in 802.11g which is backwards compatible with 802.11b which uses DSSS. How can these two different technologies talk to each other? They can not! Solves that problem right there. What happens is that 802.11g devices will enable something called protected mode which encodes the headers with DSSS and then the payload is delivered via OFDM (More on this later).

Oh wait, that was only two of the three issues we had with OFDM, the other being that it is not even a spread spectrum technology. Well, if it is not spread spectrum, then what is it? Basically, OFDM is a bunch of narrowband transmissions sent in parallel. Interesting concept no? Here is how it works: frequencies have harmonics, musicians are very familiar with this concept, but harmonics are not limited just to the audible frequencies, they apply to any frequency. Harmonics are great in music, but in RF they have the nasty habit of corrupting signals. What OFDM does is use some math to create waves whose harmonics perfectly overlap causing them to cancel out. Now, the hardware to actually do this is quite expensive so what consumer chips use is something called Digital Signal Processing (DSP) which takes care of the dirty work a whole lot cheaper. OFDM takes a single channel and splits it into 52 sub-carriers (300Khz wide) which all transmit simultaneously with their harmonics canceling each other out leaving a clean signal.

Modulation Techniques:

There are all kinds of ways to modulate a wave to send a signal. Your car radio for example probably can be tuned to FM or AM radio. FM uses Frequency Modulation which alters the frequency of the wave in order to represent the data. On the other hand AM radio uses Amplitude Modulation which, you guessed it, alters the amplitude of the wave. As it happens, the property of a wave most likely to change as it propagates though space is amplitude which is why AM never really caught on. FM is better but still not resilient enough which is why all spread spectrum technologies utilize phase modulation.

Phase modulation changes the phase of the wave to represent information. The different types of spread spectrum utilize different kinds of phase modulation but they all are essentially the same process. DSSS for example can use DBPSK or DQPSK. These two may look different, but you will notice only one letter changed and it simply denotes the number of phases which can be detected.

Differential Binary Phase Shift Keying (DBPSK) is the most basic form of phase shift technology. As the name would lead one to assume it is capable of recognizing two distinct shifts, one of 0 degrees, and one of 180 degrees. We could therefore label a phase shit of 0 degrees as a “0” and a shift of 180 degrees at a “1”.

There is no reason we have to limit ourselves to simple binary shifting though, Differential Quaternary Phase Shift Keying (DQPSK) takes it to the next level by providing 4 separate shifts, 0 degrees, 90 degrees, 180 degrees and 270 degrees. Now that we have multiple shifts we can send multiple bits with one shift. a shift of 0 degrees could be “00”, 90 degrees “01”, 180 degrees “10” and 270 degrees might represent “11”.

Even with DQPSK we will not get particularly fast data transmissions. Instead of creating various names such as D”X”PSK the number of phases is usually prefixed to the term Quadrature Amplitude Modulation (QAM) which can be confusing since it says “Amplitude” but rest assured we are still talking about phase shifts. 16-QAM thus would be capable of detecting 16 phase shifts and 64-QAM would be… 64 phase shifts. These are the two technologies present in OFDM used in the 802.11a/g standards to achieve their fast transfer rates. The ability to detect up to 256-QAM does exist and is in fact part of the 802.11n draft at present. To figure out how many bits we can fit into a number of shifts you can complete the simple equation: bits/shift = log base 2 (# of phases). Using this we see that 256-QAM is able to transmit one byte per shift

Encoding Techniques:

So, we know the different types of spread spectrum, but these technologies are not defined just by their waves, but in how they transmit data. In order to transmit this data you first need to encode it. Some individuals use the terms “encoding” and “modulation” interchangeably, this is wrong. Modulation is the process of sending what has been encoded. In other words, encoding gives you the sequence of ones and zeros and modulation defines how to get them from point A to point B.

In some cases the modulation technique used is the encoding technique, this is probably how the two terms got confused in the first place. For example if you wanted to send “101110” with DBPSK you could simply do a 180 degree shift, 0 degree, 180, 180, 180, 0. Of course, this would not be very resilient to corruption.

To combat the inevitability of signal corruption there are various forms of encoding that take place before the signal is actually sent. DSSS uses something called a “Barker Sequence” of 10110111000 which is XORed against each *bit* of the signal, this results in 2 possible outcomes of the XOR. 1 XOR 10110111000 which equals 01001000111 and 0 XOR 10110111000 which equals 10110111000 (which is just the Barker Sequence). The beauty of this system is that if a bit gets flipped in transit it can be figured out. For example if the sequence 10110110001 came in we see that two bits have been flipped from the sequence for “1” so even though two of the bits were corrupted we still know the original data was a “1”. The process of XORing the signal with the Barker Sequence is called processing gain. To keep things straight, people often call the bits of the resulting signal after processing gain and those of the Barker Sequence as “chips”. The data is in bits, but the encoded signal is in chips, got it?

Processing gain does introduce quite a bit of overhead though as you can see, now 11 chips need to be sent for each bit of data. In order to achieve faster speeds something called Complementary Code Keying (CCK) is employed. CCK uses an 8-chip variable Barker Sequence (64 to be precise). This is what allows 802.11b HR-DSSS to achieve the 5.5 and 11 Mbps transfer speeds. For 802.11b we can use the following rate table to find out how the signal is making its way though the air:

Data Rate Encoding
1 11 chip fixed. 11 chips encoding 1 bit transmitted via DBPSK
2 11 chip fixed. 11 chips encoding 1 bit transmitted via DQPSK
5.5 8 chip variable. 8 chips encoding 4 bits transmitted via DQPSK
11 8 chip variable. 8 chips encoding 8 bits transmitted via DQPSK

The 802.11 implementation of OFDM adds what is called Convolution Coding which basically takes a couple of the sub-carriers and uses them to transmit parity information. This way if a number of the sub-carriers are killed from narrowband interference the data can be re-created, think RAID 5. Pretty simple for such a complex system. Here is the corresponding table for OFDM transmissions:

Data Rate Modulation Bits/Transition R Length of Symbol Bits Encoded by 1 Symbol
6 DBPSK 1 1/2 48 24
9 DBPSK 1 3/4 48 36
12 DQPSK 2 1/2 96 48
18 DQPSK 2 3/4 96 72
24 16-QAM 4 1/2 192 98
36 16-QAM 4 3/4 192 144
48 64-QAM 6 2/3 288 192
54 64-QAM 6 2/4 288 216

R is the fraction of sub-carriers used for data vs. those used for parity information, the rest should be pretty self explanatory. One thing to note is that while 802.11b using DSSS is using DQPSK for transmissions at 11Mbps, due to the enhanced features of OFDM 802.11g is able to use the same DQSK to achieve 12Mbps yet retain a longer range. That is to say using 802.11g you will be able to experience that 12Mbps further from the AP than if you were using 802.11b even thought they are using the same modulation technology (If you are using 802.11a it would be comparable because of the reduced range of the 5Ghz signal).

Co-location:

The ability to co-locate APs is important in any networking environment. Even if you only plan on deploying a single AP, a site survey may reveal the presence of several others in the area which may interferer with your signal. Logic dictates that two signals can not be transmitting on the same frequency at the same time or they will corrupt each other.

When we look at co-location strictly from the standpoint of how many APs we can employ FHSS wins hands down. FHSS utilizes 79 discreet channels, if synchronizing technology is used one could theoretically have 79 co-located APs with hopping patterns which prevent any one from stepping on another’s toes. Now if synchronizing is not available you can still get away with about 26 co-located devices, but if you happen to be creating a high traffic LAN you probably will not want to go over 15 APs. However, cost makes this prohibitive. When we consider the maximum throughput of FHSS taps out at 2Mbps it would take 16 APs to equal the through put of 3 DSSS systems.

This takes us quickly into the next realm: DSSS/OFDM. Due to the fact that 802.11b/g operate in the same section of the frequency spectrum they may use different technologies, but they are operating on the same channels. We also remember that these channels are ambiguous specifying only a center frequency. We recall that at most we have 3 channels for co-location, however this is still only in true in the land of theory. The fact is that even if you have two APs, one on channel 1 and one on channel 11 they are liable to interfere with one another.

The interference from co-located systems is not really all that detrimental until we consider mixing 802.11b and 802.11g. If there is one thing you should take away from this section it is to *NEVER* co-locate 802.11b and 802.11g systems, especially APs. The problem with these two technologies is that while they are backwards compatible the DSSS systems can not hear the OFDM signals and visa-versa. In order to keep things working 802.11g systems must employ what is known as protection mode. This mode sends the headers in DSSS so that the 802.11b systems know not to transmit and then uses OFDM for the payload, however this overhead has a dramatic impact on the 802.11g system giving it an effective maximum throughput of 12Mbps. All of a sudden your top of the line 802.11g system is reduced to speeds comparable to that of the 802.11b systems you are trying to get away from. A really good explanation of this problem can be found in the white paper by Devin Akin, a link to which is provided at the end of this paper.

One nice thing about 802.11a systems is that their channels do not overlap. 802.11a which operates in the 5Ghz spectrum has 8 channels which just do not overlap by design. Technically each of the two bands that 802.11a wireless LANs use has 4 channels giving you a total of 8 channels. If you have preformed a site survey and found the area to be saturated with 2.4Ghz signals it might be wise to consider purchasing 802.11a systems, just know your range will be significantly reduced, but if you are interested in providing coverage for a living room for example that is not really an issue.

The FCC/Politics:

Have you ever wondered why 802.11a and 802.11g use the exact same technology but operate in different frequency bands? Well wonder no more! The FCC provides two different license free bands, the 2.4-2.4835GHz Industrial Scientific Medical (ISM) band and the 5.15-5.825Ghz Unlicensed National Information Infrastructure (UNII) band (which is actually broken into UNII-1, UNII-2 and UNII-3). Technically the UNII-3 band is usually used for outdoor use, in point-to-point links, so for the home user we can discount that and restrict the UNII band down to 5.15-5.35Ghz.

Remember way back when we were discussing OFDM? Well the FCC does not consider OFDM a spread spectrum technology, and therefore it was not allowed to be used in the 2.4Ghz ISM band which specifically limited transmissions to FHSS and DSSS. The wireless networking people said… fine, be that way and started to used OFDM in the UNII band which did not impose such limitations.

Eventually the FCC was persuaded to allow provision for the use of OFDM within the ISM band, but it still does not consider it a spread spectrum technology. It was just “close enough” so an addendum was added and thus 802.11g was born so as to allow backwards compatibility with the 802.11b standard it was intended to replace thereby allowing individuals to upgrade gradually.

Closing:

At this point you should know more about wireless than you ever wanted to. If however, you feel you have just started (and to be honest you really have) I would highly recommend taking a look at some of the resources provided in the works referenced section. While information on 802.11n is still hard to come by it is definitely something to look into and was not covered much in this paper. One really interesting aspect of 802.11n is the use of Multiple In Multiple Out (MIMO) to increase speeds. As it stands now all 802.11 technologies are half duplex, which means a radio can only receive or transmit at any time. I honestly hope you have enjoyed this paper as much as I had writing it (maybe more ;).

Works Referenced:

Eric Jackowski
Previous

Establishing long distance 802.11 links
Next

The Joy of Mounting