Tom,

After making several retransmission attempts and still not getting a packet through, the radio will discard the packet and move on to sending the next packet. In other words, the packet is lost. Often, depending on the radio or the operating system, the number of retransmission attempts made before the packet is discarded is software-configurable. One typical default value for the number of retransmission attempts is eight.

jack


Tom DeReggi wrote:

I just installed a PTP 900Mhz Atheros SR9 StarOSV3 link that had 5% packet loss that I could not get rid of.
(Set 12mbps modulation, and averaged greater than 20db SNR.)

In theory, CSMA/CA should not get PAcket loss, like a TDD system might, as the CSMA waits for acknowledment and re-transmits if it does not get it, Wifi's built-in native ARQ.

I was not surprices to see Latency skyrocket, or retransmisson to sky rocket, but I was surprised to see uncorrectable 5% packetloss. Any ideas on why it occured. Meaning why 802.11 MAC didn't self correct the packet loss with its native re-transmission?

Tom DeReggi
RapidDSL & Wireless, Inc
IntAirNet- Fixed Wireless Broadband


----- Original Message ----- From: "Charles Wu" <[EMAIL PROTECTED]>
To: "'WISPA General List'" <wireless@wispa.org>
Sent: Thursday, December 28, 2006 4:47 PM
Subject: RE: [WISPA] Alvarion Comnet Radios have arrived - regardinginterference - Part 1


I go to see Mickey Mouse for a few days and look where this thread has
gone...wow

So, my 2 cents...

One of the largest concerns in the license-exempt world is the question of a
system's interference robustness.  However, before we can get into further
detail on the pros and cons of Alvarion VL vs Canopy, CSMA/CA vs GPS, etc --
it is necessary to realize that interference as a term is extremely broad
and vague, and can mean just about anything to anyone.  Heck, all radios in
the market have some sort of "interference robustness / avoidance
capability" -- the trick to understanding a system's capabilities is knowing
what TYPE of interference the system can actually handle.  Read on...I'll
talk more about each particular platform when I get some time to write Part
2 =)



WHAT IS INTERERENCE?

In the wireless world, interference, by definition, is a situation where
unwanted radio signals operate in the same frequency channels or bands -
i.e. they mutually "interfere," disrupt or add to the overall noise level in
the intended transmission.

Interference can be divided into two forms, based on whether it comes from
your own network(s) or from an outside source.  If the interfering RF
signals emanate from a network under your control, whether it is on the same
tower or several miles away, it is termed "self-interference."  If the
opposing signals come from a network, device or other source that is not
under your control, it is termed "outside interference."  Thus, the
definition of what type of interference is being combated is not based on
technology, but ownership.

In licensed bands, where spectrum is relatively scarce (due to high costs)
self-interference alone must be taken into account; however given a more or
less known operating environment (the radio spectrum will only have signals
transmitting that are under control by a single entity) proper product
design and network deployment can reduce these interferes to a level where
they do not impact network performance.

Self-interference is not a phenomenon that is confined to licensed band
operations; license-exempt bands must address the same issues.  The
techniques and design elements of a given product that serve to reduce and
tame self-interference in licensed band operations can be applied directly
to license-exempt systems.

THE LICENSE-EXEMPT CHALLENGE OF INTERFERENCE

In the license-exempt bands, not only must self-interference be accounted
for, but, given the nature of the regulations governing these bands,
external interference must be designed for as well.  This can be extremely
challenging, as there is no way of knowing in advance where these outside
signals may be or will be sourced from, or even how strong the interfering
transmissions will be relative to the desired transmission.  This aspect of
the license-exempt bands represents the possible "downside" of
license-exempt network operation.

Yet as potentially damaging and unpredictable as external interference can
be in license-exempt networks, a properly designed and implemented broadband
wireless system can make a significant difference in the performance of a
network under siege from unwanted external radio transmissions.

DEALING WITH COCHANNEL INTERFERENCE: PHY LAYER

1. Modulation & the C/I Ratio

At the most fundamental level, an interfering RF source disrupts the digital
transmission by making it too difficult for the receiving station to
"decode" the signal.  How much noise or interference a digital RF
transmission can tolerate depends on the modulation used.

Fundamentally, modulation is the method whereby zeros and ones are
communicated by varying one of three aspects of radio signal.  The three
portions of an RF signal that can be changed or modulated are phase,
frequency and amplitude. Shirting the properties of any of these parameters
can be used to communicate different "states."  These states, in turn, are
translated to zeros and ones for binary communications.

For example, with frequency modulation, if the sine wave is at frequency
one, it is decoded as a zero.  If the sine wave is shifted slightly to
frequency two, this is decoded as a one.  This type of modulation is
referred to as Binary Frequency Shift Keying (BFSK).  In this example, a
system must only be able to tell the difference between one of two states or
phases.  More complex modulations, such as 16 QAM (Quadrature Amplitude
Modulation), attempt to differentiate among 16 different possible states of
an incoming signal.

The advantage to higher order modulation schemes, like 16QAM, is that
compared to BPSK, 16QAM conveys more information per bandwidth (more
bits/Hz).  The disadvantage of 16QAM lies in the fact that, in order to
distinguish among the 16 different states, the signal must be very clean and
very strong relative to background noise and/or interference.

The ability of a receiving station to decode an incoming signal at the most
basic physical layer is dependent on a factor called the "carrier to
interference ratio," or C/I.  This term means exactly what it says: how
strong the desired signal (the carrier) is relative to the unwanted signals
(the interference).  C/I ratios are based primarily on the modulation used,
with more complex modulations requiring higher C/I numbers than more robust
modulations, such as BFSK.






Modulation
<outbind://93-00000000A1BC9C82EE27CA43A85AEBB3458E5835C444CD00/#_ftn1> [1]

Throughput

(20 MHz Channel)

Rx Sensitivity

C/I (dB)

Co-Channel


BPSK ½

6 Mb

-86 dBm

6 dB


BPSK ¾

9 Mb

-85 dBm

10 dB


QPSK ½

12 Mb

-85 dBm

12 dB


QPSK ¾

18 Mb

-82 dBm

15 dB


16 QAM ½

24 Mb

-80 dBm

17 dB


16 QAM ¾

36 Mb

-76 dBm

21 dB


64 QAM 2/3

48 Mb

-70 dBm

29 dB


64 QAM ¾

54 Mb

-66 dBm

31 dB



 _____

<outbind://93-00000000A1BC9C82EE27CA43A85AEBB3458E5835C444CD00/#_ftnref1>
[1] RedLine Communications 5.8 GHz AN-50 System Specifications



Due to the large amount of available license-exempt spectrum (a 500+ Mhz
chunk in 5 GHz compared to 6 Mhz slivers in 3.5 GHz) one can easily justify
sacrificing bandwidth efficiency in favor of more "interference resistant"
lower-order modulation products.  For example, the Motorola Canopy product
uses BFSK for modulation, and is able to operate properly with an error rate
of 1E-4 at a 3 dB C/I; i.e., the wanted signal need be only 3 dB higher in
power than the unwanted interferes. A system operating with 16 QAM at these
levels would require a C/I ratio closer to 20 dB.



Putting this into perspective, with every 3 dB of additional signal
strength, the power of a signal is doubled. This means that a system with a
3 dB C/I ratio can tolerate an interfering signal that is many times more
powerful than a 16QAM system and still operate at the specified error rate.
Whether the interference is from another cell site on the network or another
network completely, systems employing lower-order modulations systems (like
BFSK) will tolerate substantially higher levels of interference before the
communication stream becomes impacted.  All other PHY layer techniques are
designed to improve this most fundamental measurement of network robustness
and operational effectiveness by sustaining the necessary C/I level.



2. Another Method: Receiver Threshold Dampening



In some instance, in order to avoid a "race to the bottom," certain
license-exempt systems retain high modulation bandwidth efficiencies in
interference heavy environments with a method known as receiver threshold
dampening.  Trango Broadband uses a RF Rx Threshold command to specify the
receiver sensitivity of the radio.  It is a powerful tool for a higher
modulation radio operating in a noisy environment, as it allows the radio to
block out and ignore signals received below the preset RF Rx Threshold.



By creating an artificial receiver threshold below which no RF signals are
processed, the Receiver Threshold Dampening allows for the rejection of
distance interferences and reduces co-location interference at the expense
of a reduced coverage radius.



PHY: DEALING WITH INTRA-SYSTEM INTERFERENCE



TDD Synchronization



License-Exempt systems that use Time Division Duplexing (TDD) for separating
upstream and downstream communications are ideally suited for dynamically
changing asymmetric traffic, like data. The ability to adjust the amount of
bandwidth dedicated for upstream and downstream communications without
changing hardware is a powerful feature.



TDD systems operate by transmitting downstream (from the AP to the SU) for a
period of time - 1 ms for example.  Following a short guard time, the SMs
then transmit on the same frequency in the upstream.  For a cell site with
more than one radio operating in TDD mode, it is important that all the
sectors of the cell transmit and receive at precisely the same time.
Otherwise, if sector 1 is transmitting when sector 2 is receiving, sector 2'
incoming transmission can be interfered with because the sector 1 signal is
so close that it is strong enough to "flood" or overwhelm the "front end" in
sector 2.



When deploying a TDD system in a cellular topology, it is desirable to be
able to use the same frequency in each cell site even though those cell
sites are possibly several miles away.  This means that sector 1 from AP A
may interfere with sector 1 of AP B.  The frequency planning diagram below
shows how such signals might interfere.  In this case, inter-cellular
synchronization is required, making sure that all the sectors in all the
cell sites are properly timed and synchronized in terms of downstream and
upstream communications



Delivering tight synchronization across potentially hundreds of square miles
can be a challenge.  Some systems utilize TDD synchronization via GPS to
solve this issue.  These precise satellite signals are used for timing and
ultimately, transmit/receive synchronization, thus tying all sectors within
the system to the same "clock."



Another Method: Antenna F/B Ratio & ATPC



When a BWA signal is followed from end to end, it leaves the radio and
travels first through a transmitting antenna, over the air to a receiving
antenna, and into the radio.  The antenna, an important component in the RF
chain, can also have an impact on how well the network tolerates
interference, both internal and external.



Antenna performance is specified in a variety of ways, but for purposes of
this discussion, we focus on the front-to-back ratio.  The front-to-back
ratio of an antenna indicates how much of an incoming signal will be
absorbed coming into the front of the antenna as compared to how much of a
signal arriving at the back of the antenna is absorbed.



When deploying networks in a cellular topology, the performance of the
antenna in rejecting unwanted signals from behind is an important feature.
In some cases, metal RF shields can be mounted on an antenna, augmenting and
further increasing an antenna's F/B ratio.



Automatic Transmit Power Contrrol (ATPC) is a feature that allows the system
to self-optimize the transmit power and provide for the best overall link
performance.  The ATPC function automatically will adjust the output power
level of remote-end systems to match a pre-specified signal strength value.
When ATPC is enabled, the system will attempt to establish the wireless link
and exchange performance information.  Once the wireless link is
established, the master-end system will dynamically adjust the remote-end
systems transmit power to maintain optimum link characteristics while
minimizing power output.  In short, ATPC optimizes the transmission power
for best operation, while minimizing excess power and interference with
other devices.



DEALING WITH INTERFERENCE: MAC LAYER



Frame/Slot Size



A typical MAC frame for a TDD system is shown below.  As can be seen, the
upstream and downstream portions of the frame are divided into slots, each
slot carrying what can be termed a "radio data packet," or RDP.  The
original data, an IP packet datagram, for example, is segmented into packets
that fit into a RDP.



Despite all the best system deployment designs, there will be instances
where interference will overcome these measures and corrupt a MAC frame or a
portion of a MAC frame.  When this happens, the corrupted data must be sent
again.  If the MAC frame is designed for large RDPs on the order of several
hundred bytes, the entire slot must be re-transmitted even if only a small
amount of this packet is damaged.



The impact on network throughput as a result can be large, with a few bytes
in error causing hundreds of bytes to be re-sent.  By using a smaller RDP
size, the re-transmission can be contained to only those bytes that were
damaged, thus avoiding the re-send of large chunks of valid data.  However,
as RDP size decreases, the slot header which is fixed becomes a more
significant portion of the packet data, hence increasing the MAC layer
overhead.





Automatic Retransmission Request



In wireless broadband systems, small amounts of interference can have large
impacts on end-to-end network performance.  This is tied to the way TCP/IP
networks were designed to operate in the wired world.



TCP/IP was designed to operate over wire, where interference was assumed to
be negligible.  The protocol design calls for positive acknowledgement sent
from the receiving station to the sending station for every IP packet sent
out.  If the sending station does not receive the TCP ACK in a certain
amount of time, it is assumed that the cause was congestion of the network -
not an error resulting from transmission impediments.  When encountering
congestion, TCP responds by dramatically slowing down the transmission and
then increasing transmission speed slowly.



In a BWA network, a lost or corrupted packet can occur from interference
<outbind://93-00000000A1BC9C82EE27CA43A85AEBB3458E5835C444CD00/#_ftn1> [1].
However, the TCP protocol has no knowledge or ability to account for higher
error rates and responds by slowing down the end-to-end data rates. This is
a phenomenon that can multiply a small amount of RF interference into
significant network degradation.
<outbind://93-00000000A1BC9C82EE27CA43A85AEBB3458E5835C444CD00/#_ftn2> [2]



In most license-exempt broadband wireless systems, this is not a factor.
Automatic Retransmission request (ARQ) is a feature that addresses this
issue.  ARQ inspects the RDPs that come in and looks for errors.  If an
error is detected, the system will send a request to the sending entity to
re-send the RDP.  All of this is accomplished two layers below TCP in the
protocol stack. The net effect is that as far as TCP is concerned, it never
receives a packet of data with an error as a result of the wireless portion
of the network, thus preventing TCP from invoking the slow start algorithm
and keeping the end-to-end data rates at the peak or just slightly below
peak operational rates.
<outbind://93-00000000A1BC9C82EE27CA43A85AEBB3458E5835C444CD00/#_ftn3> [3]



Centralized Transmission Control



The popularity and cost-economics of the IEEE 802.11 standards make it a
potentially attractive BWA solution.  However, it is worth noting that the
IEEE 802.11 MAC operates in what is known as a distributed control manner
(CSMA/CA).  In this "contention-based" scenario, each SU has the ability to
send a packet at its own discretion. Typically in this scenario the SU will
"listen" and if it does not hear any transmission, it will assume that the
channel is clear and will send its data.



In high density or high traffic deployments, the problem arises due to the
fact that (1) license-exempt wireless communications are typically half
duplex and (2) the sending SU typically cannot hear other SUs in the system
"hidden node" problem.  In this instance, two or more SUs may send a packet
at the same time, corrupting both and causing a retransmission.  Unlike
contention-based Ethernet networks (CSMA/CD), the half-duplex nature of
license-exempt wireless communications precludes the sending of an "error
notice" until initial transmission is complete, further delaying the
correction.  In addition to this "self-induced" interference, external
sources can also block SUs from hearing each other to the same effect.




 _____

<outbind://93-00000000A1BC9C82EE27CA43A85AEBB3458E5835C444CD00/#_ftnref1>
[1] It is also worth noting that ARQ is also applicable in licensed bands,
as atmospheric effects, such as multipath, fading and ducting effects, are
also a cause for lost or corrupted packets within BWA network operations

<outbind://93-00000000A1BC9C82EE27CA43A85AEBB3458E5835C444CD00/#_ftnref2>
[2] Reference Layer-2 Frame Loss Concealment Discussion in our 3.5 GHz
Wireless DOCSIS Review for a more detailed analysis of this phenomenon

<outbind://93-00000000A1BC9C82EE27CA43A85AEBB3458E5835C444CD00/#_ftnref3>
[3] It is worth noting that not all ARQ implementations are created equally.
For example, 802.11 implements a form of ARQ known as the stop and wait
protocol. Essentially, the transmitter sends the packet, and then waits for
an ACK from the receiver before it tries to send the next packet.  If the
transmitting station is the AP, this delay will impact all of the SUs
because they will have to wait to send any transmission destined for the AP.
If the stop and wait ARQ is combined with the RTS/CTS protocol, even each
ACK will have to be preceded by an RTS and followed by a CTS, thus slowing
down the network significantly.







-------------------------------------------
WiNOG Wireless Roadshows
Coming to a City Near You
http://www.winog.com




--
Jack Unger ([EMAIL PROTECTED]) - President, Ask-Wi.Com, Inc.
Serving the License-Free Wireless Industry Since 1993
Author of the WISP Handbook - "Deploying License-Free Wireless WANs"
True Vendor-Neutral WISP Consulting-Training-Troubleshooting
Newsletters Downloadable from http://ask-wi.com/newsletters.html
Phone (VoIP Over Broadband Wireless) 818-227-4220  www.ask-wi.com



--
WISPA Wireless List: wireless@wispa.org

Subscribe/Unsubscribe:
http://lists.wispa.org/mailman/listinfo/wireless

Archives: http://lists.wispa.org/pipermail/wireless/

Reply via email to