1. 1 ITU-R WP8B Radar Symposium “History and Status of 5 GHz RLAN and Radar Dynamic Frequency Selection (DFS) In the United States” Frank Sanders U.S. Department of Commerce Institute of Telecommunication Sciences www.ntia.doc.gov September 2005

2. 2 Radar/DFS Background and History WRC-03 allocated the bands 5250-5350 and 5470-5725 MHz to the mobile service on a co-primary basis with the existing services. WRC-03 adopted Resolution 229 that provides conditions for use of the bands by the mobile service. Resolution 229 provides limits for the protection of existing services including the use of DFS as a means to protect radiolocation systems. In the ITU-R, Recommendation M.1652 was produced by Joint Task Group (JTG) 8A/9B internationally to facilitate development of the RLAN devices. The recommendation contains 5 GHz radar system characteristics and a description of the RLAN channel move times along with other information. In the United States, the Federal Government in coordination with the RLAN Industry Vendors have been working together for two years to develop certification test plans and procedures for devices that operate in the bands 5250-5350 and 5470-5725 MHz bands. Bench tests with 5 GHz devices from RLAN manufacturers have already taken place at the ITS Laboratories in Boulder, Colorado over the past two years

3. 3 What is Dynamic Frequency Selection? DFS is an interference mitigation/avoidance mechanism in which a Radio Local Area Network (RLAN) device operating in the bands 5250-5350 and 5470-5735 GHz is supposed to automatically “sense” if a radar is operating in its vicinity and vacate that frequency in a timely manner when detection occurs. This allows the RLAN devices to share spectrum with radars operating in those bands by selecting channels not being used by the radars in its local area. The DFS functions in the RLAN device are not user controlled or accessible. The RLAN devices must totally vacate the channel (no emissions) with 10 seconds of radar detection and have 260 ms of time within that to shut the network down. It must not use that channel for 30 minutes and must check the new channel for 1 minute before it uses it.

4. 4 These Bench tests included: Power-on test: No RLAN emissions until after the power-up cycle has been completed and the power-on channel is monitored for 1 minute. Radar detection 6 seconds into initial channel check after power-on cycle completed. Radar detection 6 seconds before end of initial 1 minute check time after power-on cycle completed. In-Service monitoring: This is the most comprehensive test as the RLAN device must detect various synthesized radar waveforms representative of those operating the 5 GHz bands. For In-service tests, a MPEG file is streamed from computer to computer using an Access point (AP) and a Client device to “load” the RF channel. The AP has the DFS functions built-in. 30 minute non-occupancy test: Once a channel has been identified as being used by a radar, the RLAN device must not use it for 30 minutes.

5. 5 Engineers at ITS developed a test bed that has two main sub-systems Radar signal generator and synthesizer Produces bursts of un-modulated and chirped pulses in 5 GHz bandsProduces bursts of un-modulated and chirped pulses in 5 GHz bands Variable and user selectable frequency, # of pulses, pulse width, pri, and chirp bandwidthVariable and user selectable frequency, # of pulses, pulse width, pri, and chirp bandwidth RF Power control on pulsesRF Power control on pulses Uses Agilent Vector Signal Generator and other test devicesUses Agilent Vector Signal Generator and other test devices Timing measurement system Monitors RF activity on Rlan channelMonitors RF activity on Rlan channel Uses Agilent Vector Signal Analyzer and E4440 spectrum analyzer to have fine and coarse measurement of the RF emissions of the Rlan AP and client transmissions over 12 secondsUses Agilent Vector Signal Analyzer and E4440 spectrum analyzer to have fine and coarse measurement of the RF emissions of the Rlan AP and client transmissions over 12 seconds Very accurate as shown on page 9 of this presentation.Very accurate as shown on page 9 of this presentation. The two systems are synchronized so that a press of a button starts an in-service test and collects data for 12 or 24 seconds.

6. 6 Results of First Round of Radar/DFS Bench tests 5 GHz RLAN devices from four different manufacturers were tested at the ITS Laboratories, consisting of Access points (AP’s) and Client devices. Three used 802.11 Wi-fi architectures, and one was a frame based system where the frame talk/listen ratio was user controlled. For the in-service tests, the devices were tested with three radar waveforms: The radar waveform parameters are contained in the 5 GHz Report and Order (see FCC docket 03-122 at http://gullfoss2.fcc.gov/prod/ecfs/comsrch_v2.cgi)The radar waveform parameters are contained in the 5 GHz Report and Order (see FCC docket 03-122 at http://gullfoss2.fcc.gov/prod/ecfs/comsrch_v2.cgi) Two were fixed frequency and one was frequency agile.Two were fixed frequency and one was frequency agile. The tests were based on MPEG video and MP3 audio files streaming from one access point to one client using two computers, aggregate tests were not performed (AP with multiple clients).The tests were based on MPEG video and MP3 audio files streaming from one access point to one client using two computers, aggregate tests were not performed (AP with multiple clients). Access Point had DFS capabilities, not the Client card.Access Point had DFS capabilities, not the Client card. Ad-hoc networks were not tested (client- to-client).Ad-hoc networks were not tested (client- to-client).

7. 7 Test Signals used for first round of 5 GHz Radar/DFS bench tests at the ITS Laboratories Radar test signalPulse repetition frequency PRF [pps] Pulse width W [µs] Burst length L [ms] / No. of pulses (Note 1) Burst Period B [sec] (Note 2) Hopping Rate (Note 4) Fixed Frequency Radar signal 1700126 / 1810 Na Fixed Frequency Radar signal 2180015 / 102 Na Frequency Hopping Radar30001100/30010 1 kHz Note 1: This represents the number of pulses seen at the unit under test (UUT) per radar scan N = [{antenna beamwidth (deg)} x {pulse repetition rate (pps)}] / [{scan rate (deg/s)}] Note 2: Burst period represents the time between successive scans of the radar beam B = 360/{scan rate (deg/s)} Note 3: Radar bandwidth is less than that of the unlicensed U-NII device. Note 4: The characteristics of this frequency hopping radar do not correspond to any specific system. It can hop across the 5250-5725 MHz band. The frequencies will be selected by using a random without replacement algorithm until all 475 frequencies have been used. After all have been used, the pattern is reset and a new random set is generated.

8. 8 Results of First Round of Radar/DFS Bench tests The results of the initial bench tests showed that the 5 GHz devices needed more development on their detection algorithms to achieve a good rate of radar signal detection. Overall, between all the manufacturers the radar detection capabilities of the devices tested was moderate at best and the radar detection was highly dependent upon the RF loading of the channel. That is, detection occurred at a higher rate when the audio file was being streamed. A key finding is that the devices were not able to detect radar pulses that were comparable in length to a typical 802.11 data packet. The devices had no way to determine if the long radar pulse was a true radar signal or a corrupted 802.11 data packet. To eliminate false “detections” and unnecessarily vacate a channel, was the challenge to the RLAN Industry in developing proper algorithms. Radars that use longer pulse widths and their characteristics are contained in ITU-R M.1652 and these radars must be protected and detected in a timely manner. Similar Rlan/radar DFS tests performed by other Administrations have also drawn similar results and conclusions. Their tests used similar radar test signals that were used in the NTIA bench tests.

9. 9 Sample Data from Radar/DFS Bench tests using SA and VSA Radar burst at start of 1 min. check time (SA) Radar burst at end of 1 min. check time (SA) 1 minute power-on test (SA) In-Service test with MPEG file (VSA)

10. 10 Typical data flowing/radar burst/channel move sequence

11. 11 To move things forward, the Federal Government and Industry did the following Developed a set of radar signal parameters, including those with long pulses, that are representative of radar systems operating in the 5 GHz band for type acceptance compliance tests. Guarded against specific radar signal pattern recognition by having a wide variance in the characteristics, i.e., pulse width, pri, # of pulses per burst, and chirp bandwidth. Performed another round of bench tests at the ITS laboratories in August 2005 with the new set of radar signal parameters and updated 5 GHz devices provided by the RLAN Industry. Results are still being analyzed. Develop rules to prevent any end user from accessing the RLAN device algorithms and extracting ANY information about the radar signal that was detected. Use the results of the bench and field tests to validate the radar signal test parameters, the test procedures, and true proof of concept. (pending) Publish a final set of type acceptance rules and test procedures for companies that want to market and sell these devices. (pending)

12. 12 Parameters for the radar signal characteristics for recent Bench tests Table 1: Fixed System Radars (no modulation) Fixed Radar Set Pulse Width (µsec) PRI (µsec) # of Pulses Per burst 1 - fixed1142818 2 - variable1-5150-23023-29 3 - variable6-10200-50016-18 4 - variable11-20200-50012-16 Long Pulse Radar Set Pulse Width (µsec) PRI (µsec) Chirp Bandwidth (MHz) # of Pulses Per burst # of Bursts per 12 seconds 1 [1] [1] 50-1001000- 2000 5-20 1-38-20 Table 2: Long Pulse Radar signal with linear FM Chirp

13. 13 Parameters for the radar signal characteristics for recent Bench tests Parameter Type Pulse Width (µsec) PRI (msec) Pulses per HopMinimum Trials 1 - Fixed1333910 The test signal parameters shown in Table 1 represent the first set of test signals to be used to perform the conformance test procedures. The percentage of detection (as shown in figure 1) calculated by: Formula 1: In addition an average probability of detection is calculated as follows: Formula 2: Table 3: Frequency Hopper (no modulation)

14. 14 Parameters for the radar signal characteristics for recent Bench tests The minimum step values for each of the variable radar parameters shown in Tables 1 and 2 is an interval of 0.1 µsec for Pulse Width, a 1 µsec step interval the PRI, and a step interval of 1 for the number of pulses. Exact values for acceptable pass/fail percentages are still being coordinated between U.S. Government and RLAN Vendors. The parameters in Table 3 are fixed and include that a minimum of 10 trials per set be run with a minimum probability of detection calculated by:

15. 15 Summary of 5 GHz Radar/DFS Activities Bench tests with new set of radar characteristics were performed in August 2005 at ITS Laboratories in Boulder, Colorado with three Rlan vendors supplying devices. Each vendor had up to 1 week of laboratory time and was allowed some modifications of their equipment prior to actual tests with some experimentation with the radar test signals. The results of the August 2005 bench tests showed that the RLAN manufacturers have greatly improved their devices ability to detect the radars (simulated) that are listed in ITU-R M. 1652, over the results from the previous bench tests. The target date to allow these devices to share spectrum in the 5 GHz band is in early 2006, pending additional work and coordination between the U.S. Government and the RLAN Industry on some key issues.

16. 16 Overview of test set-up Rack Timing measurement System Radar Signal Generator Horn 3 meters Access Point Omni Spectrum Analyzer Desktop PC with MPEG and.wav files Ethernet Ant Log Antenna Laptop PC Client Card Control computer With telnet

17. 17 Fixed Frequency Radar Simulator

18. 18 Frequency Hopping Radar Simulator