Power Measurements of OFDM Signals
Mark Briggs, Juan Martinez, David Bare
Elliott Laboratories
684 W. Maude Ave
Sunnyvale, CA 94085 USA
[email protected]
Abstract— There are many ways to measure the output power
from wireless devices – average, peak envelope power, II. REGULATORY STANDARDS AND MEASUREMENTS
instantaneous peak power and average peak power being
examples. In addition there are multiple instruments and A. Regulatory Requirements
methods available to make these measurements spectrum Regulatory testing for 802.11a devices in the USA is
analyzers, power meters with peak power sensors, average governed by FCC Part 15 Subpart E [2], which requires that
power sensors and diode detectors. the peak transmit power be measured and compared to the
applicable limit.
The dynamic nature of signals that use Orthogonal Frequency
Division Multiplexing (OFDM) is such that these various FCC Part 15 Part 15.247 [2] contains the requirements for
methods can give widely different results depending on the type 802.11g devices and requires that the peak output power be
of measuring instrument used and the power measurement measured, but does not indicate if this is the instantaneous
required. The requirements for output power in the regulatory peak power or mean peak power.
standards can be confusing, leading to incorrect or inaccurate
measurements. Problems can also be created when alternative In Europe, the applicable standard for 802.11a devices is
methods of power measurements are used in the production line EN 301 893 [3], and this requires that the mean power be
to set the output power of individual devices to ensure they measured, corrected for duty cycle and compared to the
comply with National restrictions on output power. applicable limit. This standard is not yet harmonized.
This paper reviews the various methods used to make power EN 300 328 [4] is the applicable harmonized European
measurements of OFDM signal for regulatory purposes and standard for 802.11g devices and requires the average output
recommends suitable methods for production line measurements power, corrected for duty cycle, be measured.
of 802.11a and 802.11g devices.
The regulatory standards have different requirements for
Keywordscomponent; OFDM, 802.11a, 802.11g rf power the actual output power measurement – peak transmit power,
measurements mean power, average power and peak output power.
I. INTRODUCTION B. Regulatory Measurements – FCC Part 15 E
FCC Part 15 Subpart E [2] defines peak transmit power
Wireless devices using Orthogonal Frequency Division
as “the maximum transmit power as measured over an
Multiplexing (OFDM) can have high peak power to average
interval of time of at most 30/B (where B is the 26 dB
power ratios, with the peak power being up to N times the
emission bandwidth of the signal in hertz) or the transmission
average power, where N is the number of carriers [1]. This
pulse duration of the device, whichever is less, under all
high peaktoaverage ratio can create significantly different
conditions of modulation. The peak transmit power may be
results when measuring output power using the various
averaged across symbols over an interval of time equal to the
methods described in regulatory standards most widely used
transmission pulse duration of the device or over successive
for 802.11a and 802.11g devices, namely FCC Part 15
pulses. The averaging must include only time intervals during
Subpart E, EN 301 893 and EN 300 328.
which the transmitter is operating at its maximum power and
This paper compares the measurements made on an must not include any time intervals during which the
802.11g Radio Local Area Network (RLAN) device using a transmitter is off or is transmitting at a reduced power level.”
peak envelope power meter, an average power sensor, a
spectrum analyzer and a diode detector used with a 500 MHz
Oscilloscope. The comparison of results should help identify
the best method for making compliance and productionline
measurements.
Guidelines issued by the FCC for making the output The measurements were corrected for both the 10dB
power measurement [5] allow for three different methods, all attenuator between the rf port and the sensor and for the 95%
using a spectrum analyzer: duty cycle.
1. Spectrum Analyzer – integration over 26dB
B. Average Power Sensor
bandwidth using a sample detector in maxhold mode
(note, many analyzers do not allow maxhold in The rf port of the device was connected to an average
sample mode) provided power sensor via a 10dB attenuator. The power sensor was a
Rohde and Schwarz NRVZ51 (thermal power sensor) and it
Sweep time < transmission time was connected to a Rohde and Schwarz NRVD power meter.
Sweep is gated to transmission The filter setting (measurement time) was altered but had
little effect on the measured level.
2. Spectrum Analyzer – direct reading (assumes
spectrum analyzer measurement bandwidth is greater The measurements were corrected for both the 10dB
than the signal bandwidth, which is rare for a 17MHz attenuator between the rf port and the sensor and for the 95%
wide 802.11a or 802.11g signal) duty cycle.
3. Spectrum Analyzer – integration over 26dB This test method is equivalent to the method outlined in
bandwidth using video bandwidth averaging and EN 300 328 and EN 301 893.
sample detector in maxhold mode (note, many
analyzers do not allow maxhold in sample mode). C. Spectrum Analyzer
The rf port of the device was connected to an Agilent
C. Regulatory Measurements – FCC Part 15.247 8564E spectrum analyzer. The center frequency of the
FCC Part 15.247 [2] does not define peak output power as analyzer was set to the center frequency of the signal being
either average peak output power (as inferred by 15 Subpart E measured and the span to 30MHz. Resolution and video
above) or instantaneous peak output power. At this time it is bandwidths were set to 1MHz and 3MHz respectively and the
taken that the requirements are for instantaneous peak power. trace was set to maxhold for 30 seconds. The trace detector
was set positive peak.
D. Regulatory Measurements – EN 300 328 and EN 301 893 The power was calculated using the power over channel
EN 300 328 and EN 301 893 both require the use of a function of the analyzer (an integration function), with
wideband power meter with thermocouple detector (or channel bandwidth set to 20 MHz.
equivalent) to measure the output power. This power is then
This test method is the closest equivalent to the method
corrected for transmitter duty cycle, x, by the factor 10
outlined in [5] for FCC Part 15 Subpart E.
log(1/x).
D. Diode Detector and Oscilloscope
III. POWER MEASUREMENT METHODS
The rf port of the device was connected directly to a diode
Measurements were made on an 802.11b/g device that detector. The detector was connected to a Tektronix TDS 520
was transmitting a signal with a duty cycle of ~95% using 500MHz oscilloscope with the input impedance set to 50W.
four different methods/instruments, as detailed below. A LabViewâ script was used to process the trace data from
Measurements were primarily made with the device the oscilloscope to provide the rms, average and peak values
transmitting an OFDM signal (i.e. 802.11g) at a data rate of over a transmission. The voltage recorded by the ‘scope was
6Mb/s. Previous measurements had demonstrated that converted to a power level by applying a CW signal to the
changing the data rate had little effect on the measured detector and adjusting the level of the CW signal until the
power. Additional measurements were made with the device required voltage was read on the ‘scope. The signal level
transmitting using 802.11b; a direct sequence spread from the generator was then measured using the average
spectrum modulation, for further comparison. All power sensor detailed in B.
measurements were made on three different channels. As these measurements were made over the transmission
burst, no additional corrections for duty cycle were required.
A. Peak Envelope Power Sensor
As the video bandwidth of the measurement is determined
The rf port of the device was connected to a peak envelope
by the termination impedance additional measurements were
power (PEP) sensor via a 10dB attenuator. The power sensor
made with the ‘scope’s input impedance set to 1MW to
was a Rohde and Schwarz NRVZ32 and it was connected to
demonstrate the limitations a reduced video bandwidth could
a Rohde and Schwarz NRVD power meter. The filter setting
set on the measurements.
(measurement time) was altered but had little effect on the
measured level.
Table I MEASUREMENT RESULTS – 802.11G (OFDM)
Method Result (dBm)
Peak Envelope Power Sensor 18.0
Average Power Sensor 10.2
Spectrum Analyzer 18.8
Diode Detector – Instantaneous Peak Power
18.2
(50W)
Diode Detector – RMS Peak Power(50W) 12.4 a
a
On one channel the rms peak power measured using the diode detector was 2.2dB higher than the
average power. On the other two channels the rms power measured via the diode detector was within
0.3dB of the average power measured via the average sensor .
IV. POWER MEASUREMENT RESULTS – 802.11G
The results, detailed in Table I, highlight the differences
between the methods used. The diode detector measurements
and timeamplitude plot (Fig. 1 – Diode Detector output into
50 ohms) demonstrate the high peak:toaverage ratio of Figure 2. AmplitudeTime Trace from Diode Dectector into 1MW
5.8dB. Other measurement windows on different channels
gave ratios as high as 7.8dB. As the Spectrum analyzer method uses a much reduced
video bandwidth (3MHz) compared to the signal bandwidth,
The use of a high video bandwidth in making one might be surprised that the output power measured using
instantaneous peak measurements is emphasized by the integration method is very similar to the measured peak
comparing the diode detector plots with the diode detector power. With the analyzer placed in a maxhold function, the
terminated in 50 ohms (Fig. 1) and 1Mohm (Fig. 2). The time varying nature of the signal is ignored. The analyzer is
peak to average ratio is only a factor of 2:1 in terms of the not making a power measurement over a single transmission
output voltage with the 1MW termination, 6:1 with the 50W burst, rather it is capturing the highest power (averaged over
termination. With the particular device under test, the rms a period of 1/video bandwidth, or 0.33ms) in each 1MHz band
power did not appear to significantly change with the over a period of 30 seconds. This results in an overly
termination impedance. conservative estimate of a average power.
With the particular product tested the different methods of
determining the peak power all produced similar levels V. POWER MEASUREMENT RESULTS – 802.11B
(0.8dB of variance between the PEP sensor, spectrum
The results are detailed in Table II. The peak:average
analyzer and diode detector methods). A much larger
power ratio of ~2dB is significantly smaller than the results
difference between the average power as measured with the
for OFDM
average sensor and rms power as measured with the diode
detector is observed in the test data, but this large difference
was only observed on one of the three channels tested. Table II MEASUREMENT RESULTS – 802.11B
Method Result (dBm)
Peak Envelope Power Sensor 19.2
Average Power Sensor 17.5
Spectrum Analyzer 20.2
Diode Detector – Instantaneous Peak Power
19.5
(50W)
Diode Detector – RMS Peak Power(50W) 17.4
Figure 1. AmplitudeTime Trace from Diode Dectector into 50W
VI. IMPLICATIONS FOR PRODUCTION LINE TESTS
Production line output power tests are typically combined
with other functional tests, such as receiver sensitivity or
protocol evaluation measurements. In these particular tests,
the device may be configured to send randomized packets at a
low duty cycle. An example of such traffic is given in Fig. 3
that was captured from an Access Point transmitting a
streaming video signal. The duty cycle was ~10% with the
widest signal only 50mS.
Without a fixed duty cycle the use of an average power
sensor is limited as no duty cycle correction factor can be
applied and the power cannot be accurately determined. Figure 3. RLAN Traffic For Streaming Video
Similar issues may be encountered with a peak envelope
power sensor. The combination of meter and PEP sensor VII. SUMMARY
used for this evaluation started to give significantly reduced
readings when a pulsed waveform with low duty cycle was The correct measurement of output power for OFDM
presented. signals is essential when making regulatory compliance
measurements. These measurements are, however, typically
The implication is that the production line test must take made under controlled conditions of known duty cycle and
into consideration the type of transmissions that are used data pattern which allows for repeatable results with a known
when making power measurements. The use of a diode correlation between the different test methods.
detector and oscilloscope, with some software control, could
easily be used to give accurate readings of instantaneous peak Production line tests, where the data pattern and duty
and rms power over any transmission burst. As the power cycle may not be constant, require more attention to ensure
over a single, short burst may not provide a true indication of that power measurements made have a practical application
the devices performance, the software must be able to take the in calibrating output power to meet international power
data over a number of such bursts. This would be the only limits.
way calibrate the device under test’s output power to ensure
compliance with the various national regulatory limits. ACKNOWLEDGMENT
An approach being implemented by a leading The authors would like to thank Robert Paxman and Jim
manufacturer of RLAN devices is the use of a peak power Baer for their ideas and support.
analyzer, capable of providing cumulative distribution
functions for the peak power over time. This allows the
manufacturer to calculate both instantaneous and rms peak
power, even for nonperiodic transmissions. REFERENCES
[1] E. Lawrey, C. J. Kikkert, “Peak to average power ratio reduction of
Knowledge of the peak and average output power could OFDM signals using peak reduction carriers,” Fifth International
lead to the identification of products with unexpected peakto Symposium on Signal Processing and its Applications, ISSPA ’99,
average power ratios caused by modulation or amplifier Brisbane, Australia, 2225 August, 1999
response within the radio circuitry. It is also key to [2] Federal Communications Commission, “CFR 47 Part 15 Subpart E
Unlicensed National Information Infrastructure Devices,” December 8,
identifying the correct operation of any algorithm described 2003
by Lawrey and Kikkert [1]. [3] ETSI EN 301 893 V1.2.3 (200308), “Broadband Radio Access Networks
(BRAN); 5 GHz high performance RLAN; Harmonized EN covering
essential requirements of article 3.2 of the R&TTE Directive
[4] ETSI EN 300 328 V1.4.1 (200304), “Electromagnetic compatibility and
Radio spectrum Matters (ERM); Wideband Transmission systems; Data
transmission equipment operating in the 2,4 GHz ISM band and using
spread spectrum modulation techniques; Harmonized EN covering
essential requirements under article 3.2 of the R&TTE Directive”
[5] Federal Communications Commission Public Notice DA 022138,
“Measurement Procedure Updated for Peak Transmit Power in the
Unlicensed National Information Infrastructure (UNII) Bands,” August
30, 2002