Spreading Out with 802.11n/ac
Figure 29: Spectrum Density View of OFDM
Overview of Parts 5 and 6
The last two parts of this tutorial will show the evolution from Legacy (802.11 and 802.11b) and Classical (802.11a and 802.11g) WLAN protocols to our current state of the science, which began with the release of 802.11n in 2009 and was continued with 802.11ac in 2013.
When 802.11n was published by the IEEE 802.11 Workgroup in 2009 it brought with it a sea change in how Wi-Fi worked. Included within the 500+ pages of specifications were major changes such as Multiple Input – Multiple Output (MIMO) antenna functions, a more efficient use of OFDM Guard Subcarriers, an improved Guard Interval, channel bonding, beam steering, spatial division multiplexing, and a new modulation coding level, to name only a few. Taken together these changes have heralded the biggest effect on WLAN fundamentals since the move from Barker/CCK to OFDM in 2001. Taken separately, the incorporation of MIMO technologies to improve SNR and allow for spatial multiplexing has had the biggest overhaul effect on network capabilities. MIMO and spatial multiplexing (SM) techniques allow multiple streams of information to be transmitted (seemingly) at the same time. Although the 802.11n/ac amendments have provided many new features to our WLAN technology, this discussion will continue to focus on the features that effect its underlying OFDM mechanism (watch for a future tutorial titled, “Understanding MIMO”).
Next to SM the biggest improvement to Wi-Fi performance that was made possible by 802.11n is the use of “bonded channels”. 802.11n allows the ability to specify normal or wide channel usage by the administrator. When two 20 MHz (normal) channels are bonded together the result is more than the sum of the two parts. That feature was expanded on in 802.11ac which provides the ability to bond four 20 MHz channels (80 MHz) and even eight 20 MHz channels (160 MHz). That magic will be explored in the final part of this series when we explore 802.11n’s channel bonding, improved use of Guard Subcarriers, the addition of a 5/6 Coding scheme, and 802.11ac’s ability to use 256QAM modulation.
This installment, part 5, will first explore how OFDM protects itself from multipath interference with the use of a Guard Interval (GI) and the reasoning behind the Short Guard Interval (SGI) unveiled by 802.11n.
Multipath RF – Public Enemy #1
The basis of wireless communications is the use of electromagnetic energy to carry information signals from one place to another. When an electromagnetic signal is transmitted onto the air medium from an antenna, it immediately begins to expand outward along both the “E” (electrical) and “B” (magnetic) planes. These terms and the physics of electromagnetic wave propagation were originally defined by James Clerk Maxwell in his 1865 treatise titled, “A Dynamical Theory of the Electromagnetic Field” but are usually referenced today just as “Maxwell’s Equations”.
Figure 30: Electromagnetic Field
RF is EM
Electromagnetic energy (EM) is the result of the two primary forces of electricity (E field) and magnetism (B field) being inextricably bound together. The act of binding these two forces creates an alternating current that travels outward always maintaining the perpendicular relationship of the two main forces. The type of EM used with WLANs is called Radio Frequency (RF) and this energy takes the form of an outwardly expanding traveling wave. Because of the perpendicular relationship of E and B fields these RF waves possess directionality, meaning they have an up and a down and a right and a left.
We refer to this directionality by recognizing the two main beam patterns of an RF wave. They are called Azimuth, which refers to the horizontal beam pattern and Elevation which refers to the vertical beam pattern. We also recognize that an RF wave is oriented or polarized according to its directionality at the time it was induced onto the medium. For optimal RF signal performance the transmitting and receiving antennas should maintain the same polarization.
Reflections cause signals to take multiple directions
As the RF wave propagates outward, always expanding, it may encounter obstacles that effect its line of travel. Various types of materials can have different effects on the RF wave. The obstructing material may absorb, reflect, or block the line of travel of the waveform. When the obstruction causes a portion of the wave to be reflected, an interesting effect can be the creation of multipath. The term multipath refers to a condition in which RF signals arrive at a receiver by way of two or more lines of travel. When this happens under just the right circumstances the RF signal can be severely degraded. Since the beginning, RF communications has suffered from the negative effects of multipath. Over the years there have been several technological solutions that were employed in order to counteract the signal corruption caused by multipath. But none of these are 100% effective. Even with diversity receive systems, advanced signal processing, and MIMO technologies, multipath can still cause interference and corruption to an RF signal.
Caveat: 802.11n, which introduced advanced signal processing techniques along with the use of Multiple Input – Multiple Output (MIMO) technologies, provided a breakthrough in our fight against the negative aspects of multipath and actually harnesses multipath to our advantage in several ways. However, there can still be negative multipath effects even with 802.11n/ac. These good and bad multipath issues will be discussed in more detail in the future tutorial, “Understanding MIMO”.
When RF is used as a communications system there must be both a transmitter and a receiver side to the system. When a clear line-of-sight (LoS) exists between the transmitter and receiver the RF waveform carrying the end user information or signal will arrive directly at the receiver in a straight-line path. However, when there is a multipath component in the topology of the system there may also be a secondary waveform, which is the result of the primary wave being reflected and taking a different line of travel. If the reflection angles are correct this multipath reflection can strike the receiving antenna in addition to the direct or primary wave. Depending on when and how this secondary waveform arrives it may affect the primary wave.
Figure 31: Example of multipath condition
The effects of multipath are location dependant. Much like the angle taken by a bank shot on a pool table, a multipath reflection must be angled precisely in order to cause fading at the receiver. If the distance is too great or if the angle of reflection is not just right, the reflected signal will pass harmlessly. However, if everything is lined up correctly, the reflected signal will arrive at the receiver simultaneously with a primary signal. The effects of multipath under those conditions can include constructive reinforcement, destructive fading, and phase shifting of the original signals.
Historical/Terminology Note: Two types of fading are recognized. When the receiver has a strong Line-of-sight (LoS) to the transmitter allowing a dominant direct path primary signal and in addition experiences strong secondary multipath reflections, the condition is referred to “Rician fading”. The term Rician fading comes from the work published by Stephen O. Rice in 1944 in his treatise, “Mathematical Analysis of Random Noise”. When the multipath environment does not contain a dominant LoS component and is composed of only reflections, the condition is referred to as “Rayleigh fading”. The term Rayleigh fading is taken from references to the the discoveries made by an English physicist named John William Strutt, who was the third Baron of Rayleigh (Baron Rayleigh) and received the Nobel Prize for Physics in 1904.
802.11a/g Introduced Guard Interval (GI)
As mentioned previously, when an RF signal widens outwards from the source transmitter it may encounter obstructions which can reflect back in upon the receiving antenna in addition to signals which are propagating directly to the receiver. These reflected signals will have taken a longer path than the direct signal since the shortest distance between two points is the straight line. The amount of delay between the primary received signal and any additional secondary received signals (measured in nanoseconds) can cause corruption due to ISI between the primary and secondary symbols. Since multipath fading has always been such a problem to data communications, over the years there have been several methods used to counter these problems. When OFDM was introduced an integrated technique for multipath mitigation was included within the signaling method. OFDM, as used in 802.11a/g, provided a feature known as “Guard Interval” (GI) as a means to compensate for the negative effects of multipath by prepending a portion of the OFDM symbol normally used for end-user data with duplicate, “throw away” data copied from the end of the symbol onto the beginning like a nose cone. These sacrificial bits are calculated to comprise the exact timing of a destructive multipath wave from the immediately previous transmission as it arrives simultaneously with the second primary wave. Since this data is not part of the demodulated symbol time its corruption has no adverse effect on data reception.
Figure 32: Example showing delay spread
ISI due to Delay Spread
The key to understanding how the Guard Interval (a.k.a., Cyclic extension (CE) or Cyclic Prefix (CP)) is the realization that the fading caused by multipath is not to the radio signal that caused the reflection but to the next symbols following the leading symbol. The time difference between the first direct signal and a subsequent reflected signal is known as the delay spread. When the delay spread of the reflected signal coincides with a second direct signal the reselt may be corruption to the second signal due to inter-symbol interference (ISI). The use of GI is to protect the majority of the OFDM symbols from corruption due to inter-symbol interference (ISI) caused by delay spread in the transmission path.
Figure 33: 800 nanosecond GI (long guard interval)
Long Guard Interval – 802.11a/g
In order to dissect the Guard Interval it is important to think of the signal transmissions from the viewpoint of the Time domain rather than the frequency domain. An entire OFDM symbol, as used in Wi-Fi on standard width channels, requires four microseconds to transmit. But not all of this time is used to transmit end user data. The Guard Interval represents the first .8 microseconds (800 nanoseconds) of the symbol time.
Figure 34: Example of OFDM Transmission showing long GI (800 ns)
Figure 35: Example of OFDM Reception showing effects of ISI due to delay spread
802.11n Introduces Short Guard Interval (SGI)
When 802.11n was released it included an optional Short Guard Interval (SGI) function that reduced the amount of delay spread protection by half, from 800 ns to 400 ns (.4 microseconds) which leaves 3.6 microseconds of the symbol time available for data compared with 3.2 microseconds of data capacity when using the long GI. The overall effect is that the SGI feature provides an 11% increase in data capacity at the expense of some degree of multipath protection.
Figure 36: Example of OFDM transmission showing long and short GI (400/800 ns)
Results of SGI
In order to see the results produced by using the SGI we can compare the 802.11n data rates with, and without, the SGI. If you look at the area inside the red rectangle in Figure 9 under the “No SGI” and “SGI” columns of MCS Index 0, you’ll see an increase of 700 kbps. This accounts for the amount of data speed gained by reducing the Guard Interval from 800 nanoseconds to 400 nanoseconds in 802.11n. This gain is the same when using 802.11ac.
Figure 37: Comparison of Standard Guard Interval rates to Short Guard Interval rates
But SGI Requires More Precision
The use of a guard interval is to protect the majority of the OFDM symbols from corruption due to inter-symbol interference (ISI) caused by delay spread in the transmission path. Shortening the GI has the benefit of allowing more of the transmission time to carry user data but also carries the negative effect of making the signal more vulnerable to the delay spread of multipath channels. The use of SGI, especially where there are long multipath delays, requires more precise synchronization at the receiver.
802.11n Channel Models
During the development of the 802.11n standard a set of 6 channel and path loss models were created based on measurements of impulse responses in both the 2.4 and 5 GHz bands. Each of these 6 channel models represents a different indoor environment. Channel models B, D, and E are the most common. One of the results of testing using these channel models was to determine what effect, if any, that delay spread caused by multipath, has on the quality of the received signal.
Figure 38: 802.11n Channel Models
Understanding the effects of Delay Spread on Receiver Performance
The 802.11n Channel models were the direct result of empirical testing. The results of these tests is published and available for review.
Testing with Long GI
According to test data, when transmissions are made using an 800 ns GI, multipath degradation has no discernible negative effect on the transmission until the delay spread equals or exceeds 300 ns. That level of multipath is unlikely in channel models B, D, and E under normal conditions. However, the use of the 400 ns GI (SGI) revealed that received signal degradation can occur under several scenarios.
SGI with Channel Model B
Using SGI with channel model B, (residential room-to-room) in which typical delay spread levels are up to 15 ns, tests have shown the signals to be unaffected with time offsets of up to 100 ns. But as the offset of the reflected signal is increased up to 200 ns some amounts of signal degradation become apparent. When using SGI with channel model B, a delay spread of 250 ns between the primary and secondary signal components the receiver is likely to fail altogether.
SGI with Channel Model D
Testing of channel model D (areas larger than residential channel model B) with SGI, showed that signal degradation does not become a problem until after the time offsets between primary and secondary wave components differ by more than 50 ns. At 100 ns and above, channel model D experiences slight degradation with a full receiver failure occurring at the 150 ns time offset.
SGI with Channel Model E
The largest indoors channel model area that was considered during these early tests was channel model E, which accounted for large office areas such as multi-floor buildings. Channel model E normally exhibits root mean square (RMS) delay spreads of 100 ns but as long as the time offsets of multipath components is 50 ns or less there is little noticeable signal corruption. However, this channel model experiences higher degradation than channel model D when the time offset approaches 100 ns, suffers high degradation with 150 ns delays, and fails completely when delay spread offsets occur at 200 ns or higher.
In a nutshell, multipath delay spread can have a negative effect of signal transmission even when using 802.11n/ac. The long Guard Interval (GI) and short Guard Interval (SGI) are functions built into OFDM that help to mitigate these effects. The long GI successfully protects OFDM transmissions under most conditions. The SGI allows for more user data by reducing the long GI data area, but SGI is more susceptible to errors caused by delay spread and must have a higher degree of timing synchronization in order to operate correctly. In larger coverage areas its more likely that multipath reflections will cause data corruption when SGI is in use. When SGI is used the three most commonly used channel models (B, D, and E) experience receiver failures when timing between the primary and reflected components of the OFDM symbols are offset between 150 and 250 ns. Since the fifty-two (data populated) 802.11 OFDM subcarriers are sampled and transmitted over a 4000 ns interval, the amount of tolerance is between 2 and 3 subcarriers, which means that there is not much room for error.
An OFDM symbol, which exists for 4000 ns (4 microseconds), when divided by 52 allows about 77 ns (4000 / 52 = 76.9) per subcarrier.
- Channel model B fails when timing is offset by 3 subcarriers (250 ns / 77 ns = 3.25 sc).
- Channel model D fails when timing is offset by 2 subcarriers (150ns / 77 ns = 1.95 sc).
- Channel model E fails when timing is offset by 3 subcarriers (200 ns / 77 ns = 2.60 sc).
This concludes Part 5.
Next: Understanding OFDM – Part 6 “The Changing of the Guards”
Great series, I watched the video of your presentation at WLPC 2016 on this topic while following along through the blog posts. This certainly helped my understanding of OFDM as I review for the CWNA exam. Thanks Rick!
Great blog series! Looking forward to Part 6. In the meantime I have a few questions:
1. As I read your blogs and learn about the sub-carriers used in OFDM and the different modulation types I was wondering how a receiver in a 40 or 80MHz channel managed to lock onto the right set of sub-carriers. If I was an 802.11ac 80MHz receiver I am using 4 x 20MHz channels but for any given message that comes my way I don’t know in advance whether it is going to be a message on my primary 20 MHz channel, a data message using all available subcarriers across the 80MHz bandwidth, 4 simultaneous control messages (RTS) on each individual 20MHz channel, a 20MHz data message on an OBSS etc. How does this work? Will it be covered in Part 6?
2. Related to the above, aren’t the way that subcarriers are allocated and used, different for 20 Vs 40 Vs 80 Vs 160 MHz bandwidths. Does a transceiver dynamically adjust its use of the subchannels accordingly? Is that how a transmitter is able to send RTS’s on each 20MHz sub-channel of an 80MHz channel in parallel?
3. In the section that talked about noise – how do STA’s sold for residential use measure noise? Do they just use a preset value and then if they have too many retries change the modulation type to try and compensate or do they have some built in noise measurement?
4. In Part 5 under discussion of SGI how would one know what the delay spread was in a specific environment? Is the bottom line recommendation that SGI is OK to use in residential deployments or is not using SGI still best practice?
Rick Murphy says
Good questions, Michael. Most of what you ask will be better answered in Part 6, but let me give you some quick replies here to point you the upcoming direction.
1. – The 802.11 puzzle comprises many interlocking pieces. In my “Understanding OFDM” series, I’ve focused only on the OFDM multiplexing piece of the puzzle. But in order to see the bigger picture, its important to realize that the purpose of the OFDM subcarriers is to induce traveling waveforms of information (symbols) which can be received by all the receivers in a basic service area (BSA). The leading edge of information carried by these symbols makes up the physical layer preamble which is read by all devices capable of decoding this portion of the transmission. With the early forms of OFDM that were used in 802.11a/g, the legacy preambles were contained within 20 MHz wide channels only. But with 802.11n, a channel bonded, 40 MHz wide version was added. And again with 802.11ac, 80 MHz and 160 MHz wide adaptations were added. In 802.11n and 802.11ac, where transmissions are sent across wider channel ranges, the PHY layer continues to send the beginning portions of the preamble as if it were a legacy transmission, segmented into discrete 20 MHz wide preamble fields that can be understood by all listeners in each 20 MHz wide channel. Following the legacy portion of the Physical Layer Convergence Protocol (PLCP) header, the newer technologies insert a number of 802.11n (HT) or 802.11ac (VHT) Signal fields in 20 MHz wide increments to be read by devices capable of those mechanisms. The Channel Bandwidth field (CBW) in either the HT or VHT Signal field is the indicator to a potential bonded channel receiver that the remainder of the Physical Layer Protocol Data Unit (PPDU) will expand to fill the bonded channel bandwidth. With 802.11n, the single “CBW 20/40” bit is left at 0 for 20 MHz transmissions or set to 1 to indicate the remainder of the PPDU will be 40 MHz wide. In 802.11ac the two bit “Bandwidth” field is set to 0 for 20 MHz, 1 for 40 MHz, 2 for 80 Mhz, and 3 for 160 or 80 + 80 MHz OFDM. More details on the actual waveforms and subcarrier divisions wil be included in Part 6.
2. The way the subcarriers are ordered is dependent on the channel bandwidth in use. Immediately after the HT or VHT Signal fields, referenced above, comes the Short Training Field (STF). This is the first field that spans across the bonded channel width. At this time the OFDM subcarriers are dynamically reconfigured. Remember that legacy OFDM channels that use 20 MHz wide channels are called OFDM 64, while 802.11n 20 MHz channels are called OFDM 64 HT. With the
802.11n 40 MHz channels we refer to the OFDM subcarrier assignment as OFDM 128. Similarly, 802.11ac 80 MHz channels are known as OFDM 256, and 160 MHz wide channels are called OFDM 512. These new OFDM configurations require changes to the the placement of Guard Subcarriers, Pilots and DC. This subcarrier reordering will be explained graphically in Part 6.
3. This question does not have a clear and simple answer because the quality of the consumer grade client devices varies greatly. But generally, all 802.11 stations (STA) have the ability to detect and evaluate energy at the lowest level of the Physical Layer (PHY). The evaluation is a measurement of the ambient energy which includes the radiation given off by the station itself (noise figure) and the background energy in the environment, over time. This composite measurement
is termed the noise floor but is only accurate to the device which is doing the measuring. When signal information, in the form of OFDM transmissions, are present on the air medium, the energy readings at the listening device will show bursts of higher amplitude, relative to its measurements, in the frequencies that contain the information. If these energy bursts can be decoded into OFDM symbols, and then into information bits, then the listening STA may be able to successfully receive the transmitted frame. But the measurements made of the noise floor and signal levels are immediately discarded by the PHY. They’re not retained in any way, as part of the normal operations of an 802.11 STA. There’s no simple or inexpensive way to store or reuse these measurement levels. You can purchase specialized, purpose-built (read “expensive”) receive-only devices that
have been designed to do this. They’re known as Spectrum Analyzers.
4. Generally the use of SGI is safe. Studies following the release of 802.11a show that the 800 nanosecond GI (normal or long GI) was overkill. The 400 nanosecond GI (Short GI) is sufficient in most cases to counteract the effects of delay spread caused by multipath. Most AP and client configuration utilities have SGI turned on by default. In some cases, for instance, in highly reflective environments or in situations where the antennas have been indiscriminately tuned, you may have better results by turning off SGI. But that should be explored case-by-case.
Hope these short answers are of help. As mentioned, Part 6 should give a better and more graphical description of these items. Thanks for your input and I’ll also address the typos you mentioned separately, in the near future.
Rick – thanks for the response. Absolutely cleared up my confusion.
Any ETA on Part 6?