BRAN31d034rev4
Page 1 of 12
European Telecommunications Standards Institute
BRAN #31
LS to CEPT JPT5G
Source:
ETSI BRAN HL Spectrum Group
Title:
DFS Detection Threshold
Executive Summary
According to the most recent analyses of sharing with RLANs, some military systems remain a major
concern. Otherwise RLANs can detect radar systems and avoid causing interference to a satisfactory
level using a technically and operationally reasonable DFS threshold of –52dBm. Some countries have
reported certain military systems would require a DFS threshold of –77dBm, –67dBm, or –64 dBm
respectively to allow sharing based on the long term deployment estimates for RLANs. The radars of
concern are deployed in a small number of countries and in geographically limited locations. The DFS
sensitivity level desired by these military agencies renders RLAN operation practically unachievable
with current and expected future commercially viable RLAN technology.
RLAN deployment at 5 GHz is just beginning and market penetration in the next 4 to 5 years will be
low, as will interference.
Therefore, a decision on the DFS threshold for RLANs made now, in preparation for WRC 2003, could
be re-considered at WRC 2007 based on experience gained and technology advances. In this
perspective, setting the DFS threshold now at -52dBm is appropriate.
Introduction
The requirement for DFS to protect radar has been introduced and is under discussion by
administrations participating in the ITU-R JRG 8A-9B preparing for the World Radio Conference 2003
(WRC-03). Simulations and calculations of aggregate interference are being made to determine the
required DFS detection threshold (DFS_Th). However, the ability of RLAN receivers to distinguish a
radar signal from noise and other interference signals is not unlimited, and high false alarm rates will
cause significant RLAN performance degradation – even in an ideal scenario involving only radar
systems and RLANs. Therefore the false alarm rate and the DFS_Th for radar protection are parameters
for which a realistic compromise must be found. This issue was first discussed in BRAN in late 2001
(by Dr M. Pauli (Ericsson) in contribution BRAN25d072), and subsequently contributed to CEPT
JPT5G.
In the meantime, more work has been done by BRAN member companies.
This paper discusses the results known so far and with the objective of arriving at an acceptable value
for the DFS_Th parameter. Such a value should maximize the protection afforded to radar systems
while minimizing the possibly crippling impact on RLAN performance which results from an
excessively sensitive DFS_Th value.
DFS Threshold values calculated for radar protection
At the last ITU-R JRG 8A-9B meeting in September, a methodology was established for simulating
aggregate interference from a deployment of RLANs into a victim radar receiver, ITU-R M.[8A9B/RLAN-DFS]. This methodology was agreed between RLAN industry members and representatives
of US WP8B.
Results of simulations using this methodology are presented in Annex 1 for the relevant radars listed in
ITU-R M.[8B-CHAR], which is a list of radar operational characteristics established by administrations
for the purpose of interference studies. Results with some modifications to the methodology, intended
to show the sensitivity of the results to certain input parameters, are also listed.
The following conclusions may be drawn from the simulations:
BRAN31d034rev4
Page 2 of 12



Rotating radars that regularly scan are protected with a DFS_Th value of –52dBm.
These radars are Types A,C,E,F,G,H,I,J.
Radars P and Q are also protected with DFS_Th of –52dBm.
Airborne radars are typically operated in the range above 5350 MHz and below 5470 MHz
and are therefore not taken into consideration.
Based on the initial simulation results, the radar interests in the US have stated they need a value of -67
dBm for DFS_Th to protect the radar of most concern, radar K, in the specific deployment scenario of
radar K located within or at the edge of a dense urban area and with mainbeam aimed away from the
urban centre.
It has been noted by some European administrations that they also have military installations which
require protection, and they have calculated DFS_Th levels of –64 dBm or even –77 dBm may be
required.
Other administrations may have a different view; in particular, the actual radar and deployment
scenario near dense populations will vary according to the territorial jurisdiction of an administration.
Some parameters in the ITU-R M.[8A-9B/RLAN-DFS] scenario have been modified in simulations run
by ETSI members to see what impact they have:
 Reduction in active device density by half results in a 3dB allowable increase in DFS_Th.
 The radar 0-5 km outside the dense population zone instead of in its centre as in the original
methodology allows a 2dB increase in DFS_Th.
Factors that affect DFS Threshold values
The RLAN community will need to implement DFS with radar detection in the bands 5250-5350 and
5470-5725 MHz subject to decisions in administrative jurisdictions. ETSI EN 301 893 will set
requirements in CEPT countries, and may follow the ITU-R Recommendation being developed. The
simulations of aggregate interference result in a value or values for DFS_Th as discussed above. It is
also necessary that these values are achievable, and can be implemented without excessive technical or
cost difficulty for RLAN manufacturers. Most importantly, these values should be realistic within the
normal interference environment in which RLANs must operate. The latter issue is the first gating
factor: independently of the cost and technical overhead the ability to operate normally when no radar
signals are present must be guaranteed.
Factors to consider for normal operation of RLANs that must detect radar signals are:
1. The expected range of RSS for associated devices
2. The expected range of RSS for interfering co-channel or adjacent channel RLANs
3. The false alarm rate associated with the DFS_Th value and integration time
Common values of RSS are given in Table 1.
Table 1
Parameter
RSS, associated
minimum
-85 dBm 1
maximum
-30 dBm 2
RSS, associated
RSS, interference
-82 dBm 1
NA
-30 dBm 1
-84 dBm
Notes
1
Hiperlan/2 standard, 6 Mb/s, PDU(54b)< 10%
2
Hiperlan/2 Class 2 receivers
1
IEEE 802.11a standard, 6Mb/s, PDU(1000b)<10%
interferer at 30m, calculated per Annex 2
It can be seen that the RLAN receiver, operating at the lowest data rate, is specified to work down to
–85 dBm, with typical values of adjacent channel interference around the same level.
Hence when selecting an interference/noise floor to use for the calculation of false alarm rate, this RSS
value represents the lowest level which still allows the RLAN to function adequately at its lowest data
rate.
BRAN31d034rev4
Page 3 of 12
At higher data rates, the RSS levels need to be higher and the receiver is specified to operate at least up
to –30dBm to allow the RLAN link to operate in an interference environment while still maintaining
requisite C/(I+N) ratio.
The design of the RLAN systems and the applicable standards, are in fact intended for high system
capacity per unit floor area. In such deployments adjacent channel and co-channel interference, not the
receiver noise, is the limiting factor for systems throughput. The design of the entire RLAN system
and each device, involving the signal structure, spectrum masks, interference rejection properties,
medium access protocols and receiver signal and noise budgets, accounts for this limitation, and radar
detection must also take this into account.
A high false alarm rate will severely restrict RLANs in terms of operating range and/or the minimum
distance between independent RLAN networks, even at the minimum floor levels specified in Table 1.
Summary of simulation results (Annex 3)
False alarm rate
The false alarm rate (FA) in the RLAN receiver with interference/noise floor –84 dBm is 1% for a
receiver detection level (rss_threshold ) of –79 dBm (Fig. 2, Annex 3). The curve falls very sharply,
but in order to ensure acceptable service quality level during RLAN transmission when in normal
operation, the FA should be several orders of magnitude lower. Detection circuit implementation
margins are not included, and the interference floor is assumed to be equal to the best case receiver
sensitivity.
Radar Detection Performance
If only one pulse is detected, with the radar signal level –67dBm, the detection failure rate DF is 1%
with rss_threshold -74 dBm. (Fig.2, Annex 3). Thus there is less than 2 dB only of margin for the FA
and DF for rss_threshold set to a value between the two criteria. In practice this will be impossible to
meet with reasonable cost: the typical tolerance specified for RSS measurements is +/- 5dB in the
RLAN standards.
If several radar pulses are available for detection, the DF improves. (Figure 3, Annex 3).
The effect of increasing the radar signal level that must be detected is shown in Figures 4, 5 and 6 of
Annex 3 for DFS_th of –64dBm and -62 dBm. Figure 6, where 10 pulses are detected with DFS_th =
-64dBm yields improved margin for implementation of around 7 dB.
Impact of (self-)interference
As noted above, RLANS are designed to operate in an interference limited environment and still use
the available spectrum efficiently. This self-interference increases the false alarm rate at any threshold
level: even at –52 dBm, it plays a role. This role is complex and is still being investigated. It is
influenced, for example, by new factors such as a 30 minute moratorium on channel reuse after a DFS
process triggers a move from the channel.
Setting extremely low thresholds to protect (one type of) radar requires that the (self-)interference be
reduced by a factor of 15 dB or more. This requires much lower transmit power levels and, therefore,
reduced range and/or reduced link margins, while the false alarm rate will not decrease. The impact is
that the throughput of RLANs is significantly reduced - by factor 2 to 4, depending on bit rate and cell
separation.
Future developments
RLAN technology has developed from the simple 1 Mbit/s systems of the early 1990’s to highly
sophisticated devices that support 54 Mbit/s data transfer. At the same time the cost has dropped
significantly. The end of technological evolution has not been reached and new techniques are likely to
emerge that will allow RLANs to operate at different set points for range and data rate. Such future
development could open the way for maintaining RLAN performance while offering more protection to
sensitive radar systems. These advances will have an impact in the future when predicted RLAN
deployment densities used in the ITU-R JRG 8A-9B methodology may be reached (i.e. 5 to 10 years).
BRAN31d034rev4
Page 4 of 12
Setting an unnecessarily low DFS threshold now will reduce or even cripple RLAN performance and
seriously inhibit market acceptance and market growth.
Similarly, radar (ECM) technology continues to evolve and may well result in a wider margin of
tolerance for RLAN interference. In any case, having field data available to validate or correct the
model simulations done to date, would be a benefit to all concerned.
With the 5GHz RLAN market just starting its development, market penetration in the next 4 to 5 years
will be low enough to avoid significant interference, even to sensitive radar systems. Therefore, it
should be considered to set a DFS threshold now that protects the majority of radars and to review this
in 4 years and tighten or relax this requirement as appropriate in the light of new information on RLAN
and radar operational impact and technological advances.
National Restrictions
Administrations that feel strongly about protecting their military systems, could consider the option of
limiting the transmit power levels of RLAN devices operating in their jurisdiction. Such a choice
would disadvantage the user community but that is the price of protecting the military.
Since RLAN client devices are controlled by Access Points that are fixed in one location, standard
RLAN configuration tools can be used to set operating parameters including power levels and
operating channels, of client devices without the need for human intervention.
Further, starting up ad-hoc networks of portable devices – which happens only in exceptional
circumstances, can be constrained to the non-radar bands in order to eliminate interference in this mode
of operation.
Thus the risk of cross border movements of RLAN endangering radar operations is minimal.
Conclusions

Most radar types are protected by a detector set to detect radar signals down to –52 dBm. This
allows enough margin for implementation as well as for raising of the interference floor to
levels that provide good RLAN system capacity.

A requirement to detect –67 dBm or lower provides no opportunity for implementation margin
or for an interference floor above the absolute minimum, and may be so rigorous as to be
unachievable in production devices. Certainly, as a minimum, RLAN performance will be
severely curtailed.

It is reasonable to anticipate improvements that allow for better radar protection in the future
that will counteract the expected growth in market penetration by RLANs.

A better understanding of the practical interference impact will help to set better protection
criteria.

The RLAN market should be enabled by setting a reasonable threshold now (at –52dBm) and
revisiting this in 4 to 5 years. WRC 2007 provides a convenient synchronization point for this
work.

Administrations that favor protecting military radar over broadening public use of RLANs
may consider restricting RLAN power levels. Standard RLAN management tools can be used
to assure that all client devices follow a local or national policy.
BRAN31d034rev4
Page 5 of 12
Annex 1
Radar Characteristics from ITU-R M[8B-CHAR] and DFS Detection
Threshold according to simulation results
Radar type per WP8B M.[8B-CHAR]
Function
Platform
Tuning range (MHz)
Tx Power into antenna, peak (kW)
Antenna gain dBi
Peak EIRP (dBm)
RF Bandwidth ( MHz)
PRF (pps)
Pulse Width (microsec)
Scan rate horiz (deg/sec)
DFS threshold to meet interference criteria
A) JRG 8A-9B Sept M.[8A-9B-DFS] method dBm
As above, 1/2 user density
dBm
As A), but radar 0-5 km outside pop limit
dBm
Radar type per WP8B M.[8B-CHAR]
Function
Platform
Tuning range (MHz)
Tx Power into antenna, peak (kW)
Antenna gain dBi
Peak EIRP (dBm)
RF Bandwidth ( MHz)
PRF (pps)
Pulse Width (microsec)
Scan rate horiz (deg/sec)
A
C
E
F
G
H
I
J
K
Meteo
Meteo
Meteo
Meteo
Meteo
Meteo
Meteo
Meteo
Instrumen
tation
Ship/Groun
d
Ground
Ground
ground
Ground
Ground
Ground
Ground
Ground
5300-5700 5600-5650 5600-5650 5300-5700 5600-5650 5600-5650 5600-5650 5250-5725
250
250
39
44
123
128
0.5
20
50/250/1200 0-4000
2
0.05-18
0.65
0-36
-52
> -52
> -52
L
-52
> -52
> -52
M
250
250
250
250
0.0225
250
50
134
0.91
2000
1.1
21-24
40
124
0.6
250-1180
0.8-2
30-48
40
124
0.5
259
3
30-48
50
134
2
250-1200
0.8-5
6-18
50
134
1.5
50-1200
0.8-6
6-18
40
84
10
100000
0.1
1.2
38.3
122
4
3000
1
na
-52
> -52
> -52
N
-52
> -52
> -52
O
-52
> -52
> -52
P
-52
> -52
> -52
Ground
Ground
Ground
5350-5850 5350-5850 5400-5900 5400-5900
2800
1200
1000
165
54
148
4
160/640
0.25/1/5
na
47
138
4
160/640
0.25/1/6
na
45.9
136
4
20-1280
0.25-1
na
42
124
8
320
100
na
DFS threshold to meet interference criteria
A) JRG 8A-9B Sept M.[8A-9B-DFS] method dBm
As above, 1/2 user density
dBm
As A), but radar 0-5 km outside pop limit
dBm
Notes
Gaussian distribution path loss makes <0.5 dB difference
-58
-55
-55
-62
-58
-60
-56
-53
-55
-52
> -52
> -52
Q
Instrumen Instrumen Instrumen Instrumen Surface/
Surface/
tation
tation
tation
tation
air search air search
Ground
5300
250
-52
> -52
> -52
R
-67
-64
-65
R2
Research/ Research/
Earth
Earth
Imaging
Imaging
Airborne
Airborne
S
Search
Ship
Ship
5300
5450-5825
5300
5300
5250-5725
Airborne
360
285
16
1
0.1
28
30
26
26
35
114
115
98
82
85
1.5
10
90.00
1.00
1.00
500 2400/1200/750
1000-4000 1000-4000 200-1500
20
0.1/0.25/1
7/8
7/9
1
36/72
90
na
na
20
-70
-67
-69
-52
> -52
> -52
-52 NA
> -52
> -52
NA
NA
BRAN31d034rev4
Page 6 of 12
Annex 2
Interference Signal Strength, Neighbouring RLAN
This annex presents a simple calculation of interference level from a neighbouring RLAN to obtain an
order of magnitude interference signal level. Neighbouring RLAN interference sets the typical
environment in which an RLAN will be expected to function correctly.
It is assumed that there is no co-channel RLAN operating in the vicinity: if there is a radar in operation
on the chosen channel, other RLANs should not have started operation on that channel.
Assume an RLAN device with power 100mW is operating at a distance of 30m from the device about
to start an ACT. Then we have the following parameters for calculation of interfering signal level:
Interfering transmitter power EIRP
Path loss 30m indoor model ITU-R P.1238, prop index =3
Path loss 30m free space
Adjacent channel transmit power mask
Interfering signal, indoor model
Interfering signal, free space
20 dBm (100mW)
91 dB
76 dB
-28 dB
-99 dBm
-84 dBm
BRAN31d034rev4
Page 7 of 12
Annex 3
Impact of radar detection threshold on False Alarm rate and
Radar Detection Failure
Introduction
This Annex aims at responding to the need to know if radars operating in the 5 GHz band can be
detected by RLAN devices without extreme constraints on the RF front-end design or on the system
capacity of the RLAN.
Radar detection may only be possible during the time there is no active transmission in the RLAN cell.
During Channel Availability Check Time (ACT) there is no transmission for a significant time [e.g. 60
s]. During in-service monitoring the detection can take place during moments of no uplink or downlink
traffic.
The detection of a radar pulse of power above the so-called DFS threshold shall trigger a DFS
procedure, and the AP must leave the channel within a short delay. In this Annex the interplay of DFS
threshold (DFS-Th), RLAN False Alarm Rate and Rader Detection Failure Rate are presented and
discussed from the point of view of implementation of radar detection in a RLAN receiver.
Assumptions and methodology
It is proposed to characterize the radar detection by two performance metrics:
Detection Failure probability (DF)
This is the probability that a radar is present in the channel but no radar pulse is detected by the RLAN
device. The DF probability must be as low as possible not to interfere with radars. A value DF=1% is
often set as a target.
False Alarm probability (FA)
This is the probability that a radar is not present in the channel but a pulse is detected by the RLAN
device, triggering a useless DFS procedure. This can happen when a noise or interference burst is
mistaken for a radar pulse. A high Interference/Noise Floor in the RLAN receiver will obviously
increase the FA but this probability should be low as possible not to trigger unnecessary DFS, which
would severely limit the RLAN capacity especially in dense deployment scenarios. During ACT it is
sufficient to be sure that the channel is not measured as occupied for some small percentage of the
time. It will be seen that this percentage is a very sharp function of the RLAN receiver RSS threshold
setting, with a few dB altering the FA probability by many orders of magnitude.
(Editor's note: Calculations have been performed with Matlab code. The same notation is used here as
in the Matlab code).
The RLAN device performs power measurements averaged over a duration rss_meas_duration, the
measurements signal_rss being spaced by the averaging duration. Therefore, during ACT, for example,
the number of measurements performed is
nb_trials =round(ACT/ rss_meas_duration)
The radar signal is received at a level of radar_pulse_power (dBm), (equivalent to DFS_Th in above
notation) and pulses have a duration of radar_pulse_duration (ns). The radar signal bandwidth,
measured in MHz, is denoted by B. Many radars send pulses periodically, at a Pulse Repetition
Frequency (PRF) rate. For a rotating radar, the RLAN device will periodically receive a burst of
pulses, such that the number of pulses occurring in the Availability Check Time (ACT), denoted by
nb_RADAR_pulses_per_ACT will be greater than one. The 5 GHz OFDM RLANs sampling period is
T=50ns.
Detection criteria
The detection criterion used here is based on absolute signal power measurements. If the measured
power exceeds a certain threshold, then a radar is assumed to be detected.
BRAN31d034rev4
Page 8 of 12
It is assumed that the noise samples are AWGN. Therefore, a measurement performed during
rss_meas_duration consists of (rss_meas_duration/T) independent complex Gaussian variables. Thus,
the probability density function (PDF) of the noise is Chi-squared centred on noise_floor with
noise_free_deg = 2rss_meas_duration/T
degrees of freedom.
There is no data available on received signal characteristics or statistics of a radar pulse, propagated
over an urban, rural or other environment. In some cases, the radar pulse duration can be shorter than
rss_meas_duration. The radar signal bandwidth is normally smaller than the OFDM signal bandwidth
of 20 MHz, so that radar pulse samples collected during the power measurement will be correlated. If
the radar pulse was sampled at a rate of T_RADAR=1/B, the samples would be uncorrelated. It is
assumed here that the PDF of the power when measuring a radar pulse is chi-squared centred on
meas_avg_radar_power =
radar_pulse_power+10*log10(min(radar_pulse_duration,rss_meas_duration)/
rss_meas_duration)
with degrees of freedom equal to
radar_free_deg = 2.round(min(radar_pulse_duration/T_RADAR,rss_meas_duration/T_RADAR))
The PDF of the power measurements noise_rss_pdf and radar_rss_pdf are plotted in dBm in Figure 1
for a radar of type K received at a level of –67 dBm, and an Interference/Noise Floor of –84 dBm (see
next section).
Once the two distributions noise_rss_pdf and radar_rss_pdf are known, the DF and FA probabilities
can be computed.
DF = (prob(radar_rss_pdf rss_threshold)))nb_RADAR_pulses_per_ACT
FA = 1-(1- prob(noise_rss_pdf  rss_threshold))nb_trials
Note that rss_threshold is the RLAN receiver RSS threshold setting, whose value should be set
between noise_floor and meas_avg_radar_power such as to meet the required DF and FA probability
criteria.
Results
The following values are used in the calculations:
rss_meas_duration = 1 µs
radar_pulse_power = -67 dBm, -64 dBm, -62 dBm
radar_pulse_duration = 1µs
interference/noise_floor = -84 dBm
B = 4 MHz
ACT = 60s
nb_RADAR_pulses_per_ACT = 1, 10
The Figures below plot FA and DF using various values chosen from the above list.
It can be seen that the form of DF steepens as the number of pulses increases. Increasing DFS_th by a
number of dB moves the DF curve the same number of dB along the x-axis.
It should be noted that an interference/noise floor increase will move the FA curve horizontally to the
right also.
BRAN31d034rev4
Page 9 of 12
Figure 1: PDF of the power levels of Radar Pulses and Noise Floor
Figure 2: DF and FA, single radar pulse, DFS_th = –67dBm
BRAN31d034rev4
Page 10 of 12
Figure 3: DF and FA, 10 radar pulses, DFS_th = –67dBm
Figure 4: DF and FA, single radar pulse, DFS_th = –64dBm
BRAN31d034rev4
Page 11 of 12
Figure 5: DF and FA, single radar pulse, DFS_th = – 62dBm
BRAN31d034rev4
Page 12 of 12
Figure 6: DF and FA, 10 radar pulses, DFS_th = – 64 dBm