Measurement of microwave ovens radiation between 1 & 18 GHz
in relation with the CISPR standardisation activities
Bernard Despris
FRANCE TELECOM
CNET DMRRMC
38-40 rue du G&&al Leclerc 92 794
Issy Moulineaux Cedex 9 - FRANCE
Abstract : This paper describes a study conducted to
characterise the radiation emitted by microwave ovens
between 1 and 18 GHz, both inside and outside the oven
allocated band (2.4 - 2.5 GHz). Results are presented in
conjunction with the on-going work within a CISPR ad-hoc
group in charge of defining the emission limits for microwave
ovensand other ISM equipment between 1 and 18 GHz, with
the exclusion of the allocated ISM bands. The emphasiseis
on the frequency and time domain characteristics of this
emission,by illustrating the significant differencebetweenthe
peak and the average emitted level. Different weighting
functions are discussed and the corresponding measured
results are presented.
INTRODUCTION
emission limits between 1 and 18 GHz for all ISM equipment
(except inside the allocated ISM bands 2.4 - 2.5 GHz and
5.725 - 5.875 GHz where the emission is unrestricted).The
seconddraft proposing limits has been issuedto the National
Committeesin September1996. Commentswill be available
in the course of 1997 and discussions will take place to
preparethe next draft that will be submittedfor voting.
In conjunction with this standardisation activity,
measurementsand experiments have been performed in
various laboratoriesaround the world (Japan,USA, Germany,
the Netherlands, France...) to characterise the microwave
oven radiation and its effects on radio systems.Results of
measurementsperformed in our laboratory in France are
presentedin this paper. We will seehow theseresultsrelate to
the technical issuesdiscussedwithin the ad-hocgroup.
More and more radio services are operating above 1 GHz,
especially the new mobile radio systems (PCS in the USA,
DCS 1800 in Europe, PHS in Japan...),and they will have to
cope with the electromagneticradiation producedby domestic
microwave ovens. Some other systemssuch as the LEO (Low
Earth Orbit) satellite systems or the RLANs (Radio Local
Area Networks) are allocated inside the ISM band (2.4 - 2.5
GHz).
All results presented here were performed using the
measurementmethod defined by the CISPR ad-hoc group :
measurementdistance of 3 meters in a fully anechoicroom,
resolution bandwidth of 1 MHz and oven loaded with one
litre of tap water.
In the standardisationfield, microwave ovensare classified as
ISM (Industrial, Scientific and Medical) equipment and the
radiated field emission limits are under the scope of SubCommittee B of CISPR (International Special Committee on
Radio Interference). At present, in the relevant standard [l]
(CISPR Publication 1l), limits between 1 and 18 GHz are
under consideration except in the satellite broadcasting
receiving band (11.7 to 12.7 GHz) where the oven effective
radiated power (ERP) is limited to 57 dB@W). This
frequencyband correspondsto the fifth harmonic of the oven
and, in order to achieve such a low emission level, ovens have
to be equippedwith a fifth harmonic chokebut, even with this
additional component, the limit can hardly be met. In
addition, this satellite broadcastingband is not as extensively
used as other frequency bands between 1 and 18 GHz : the
existing limit protects a limited bandwidth may be too
stringently and the rest of the band is left unprotected.
Several papers presented in previous EMC symposia have
describedin details how a microwave oven operates[2], [3],
[4]... We will only briefly describe here the characteristics
that have a direct influence on the leakagefield radiatedby an
oven.
cHAI?ACTERISTICS
OF MICROWAVE
OVENS
The magnetron receives its input power from a half wave
rectified ac. The frequency emitted by the magnetron is
proportional to its input voltage and therefore follows the
mains variations (50 or 60 Hz period).
Due to modulationby the oven stirring device,the modulation
at mains frequency and the changing absorptionof the load,
the disturbancesignal will be amplitude (AM) and frequency
(FM) modulated.An important sourcefor energylossesis the
slit in the circumferenceof the oven door, which behavesfor
the frequenciesconcernedas a phased array antenna due to
the AM and FM modulated signal. The radiated disturbance
To try to improve this situation, an ad-hoc group (chaired by pattern is not constant and the main bundle is moving in all
the author of this paper) has been set up in 1994 to draft some directions as a function of time.
0-7803-4140-6/97/$10.00
118
FREQTJENCYDOMMNMEASUREMENTS
As the emission of an oven is highly fluctuating with time,
measurements in the frequency domain are generally
performed in max hold mode with a peak detector. This
method obviously maximises the emission and is not
representativeof the level presentat a given time, but enables
to identify the frequencies of maximum emission for each
oven.
Although every type of microwave oven has its own frequency
(( signatureD, it is generally around the samefrequenciesthat
thesemaximum occurs. We will give here a few examplesof
emission obtainedbetween 1 and 18 GHz, basedupon which
somegeneralcharacteristicscan be deduced.
abovethe noiselevel, i.e. between2.1 and 2.8 GHz. The main
characteristic here, that is common to every existing oven is
that, although the oven allocation (without any limitation of
the emission)is 2.4 - 2.5 GHz, the maximum emissionoccurs
between approximately 2380 and 2480 MHz. This has two
consequences
:
- just below the allocatedband, i.e. at 2.4 GHz, where CISPR
has to set emission limits, the emitted levels are very high
(between 100 and 110 dBpV/m) and could disturb future
mobile radio servicesthat would be allocatedhere ;
- in the upper part of the allocatedband, emissionsare rather
limited. For this reason and as new spectrum allocations
becomes more and more difficult to find, the World
Administrative Radio Conference(WAX) has decided to
allocate the band 2483.5 - 2500 MHz for Low Earth Orbit
(LEO) satellite systems.
In addition, recently, the other part of the oven band (2400 2483.5 MHz) has been allocated to Radio Local Area
Networks (RLANs). These systems use spread spectrum
modulation techniques that are felt more immune to
disturbancesproducedby microwave ovens than traditional
modulation techniques. However, if they are used in close
proximity to a microwave oven, they will experience
interference that will at least increasethe bit error rate and
therefore prolong the transmission time, or even drop the
link.
Figure 1 : Emission of a microwave oven between 1 and 2
GHz - Max hold during 2 minutes with a peak detector
In general, out of band emissions between 1 and 2 GHz are
rather limited. However, some of the ovens we have tested
exhibited an emission peak at a frequency correspondingto
half of the fundamental frequency (1.225 GHz), as shown on
figure 1.
4”1
Figure 3 : Emission of a microwave oven between3.6 and
8.1 GHz - Max hold during 2 minutes with a peak detector
Between 3 and 10 GHz, ovens generally show significant
emission levels only in the frequencyrange shown on figure 3
above, i.e. between 3.6 and 8.2 GHz. As the oven second
harmonic (4.9 GHz) and third harmonic (7.35 GHz) fall
within this range, it is not surprising to register high
emissions at these frequencies.More interesting is the fact
Figure 2 : Emission of a microwave oven between 2.1 and that significant levels occur also at frequenciesin between
theseharmonics ; they are different for eachoven, an example
2.8 GHz (including the ailocated band 2.4 - 2.5 GHz) is given on figure 3. They may be due to the changing of the
Max hold during 2 minutes with a peak detector
load during the oven operation, and they are producedless
Between2 and 3 GHz, the emission is important only around frequently than the harmonics. Another characteristic is that
the oven allocatedband. In figure 2, we have shown only the the fourth harmonic (9.8 GHz), not shown here, did not
part of this frequency range where some emission occurred appearabovethe noiselevel for all the ovenswe havetested.
119
Figure 4 : Emission of a microwave oven between 10 and
18 GHz - Max hold during 2 minutes with a peak detector
Figure 6 : Time domain measurement at 2000 MHz during
100 ms (5 mains periods)
Between 10 and 18 Gl3z (figure 4), three harmonics
frequencies are met : 5th (12.25 GHz), 6th (14.7 GHz) and
7th (17.15 GHz). They all appear on the graph, but only the
level of the sixth is significant. This is due to the fifth
harmonic choke that is routinely installed on all magnetron to
meet the only existing emission limit, that suppressthe fifth
and also the seventh harmonic, but that has the disadvantage
to increasethe level at other harmonics (especiallythe 6th).
Figures 6 and 7 show the time variations at two randomly
chosen frequencies, one above and one below the ISM band.
It should be mentioned first that, at these frequencies not
corresponding to a specific high emission (harmonic...) the
levels encounteredare very low and are in fact lower than the
noise floor of the frequency domain measurementsperformed
in max hold mode and presented before. Here again, the
influence of the mains frequency modulation can be clearly
seen,with also two bursts every mains period. But, additional
randomly appearing bursts are also experienced(for example
at 5 ms on figure 6 or at 90 ms on figure 7).
Due to the way the energy is produced inside a magnetron,
the knowledge of the frequency spectrum is not su.%cientto
evaluate the interference potential of a microwave oven
towards radio systems. For this reason, time domain
measurementswere also performed and results are presented
here. They are obtained with a spectrum analyser set with a
null frequency span.
Figure 7 : Time domain measurement at 2600 MHz during
100 ms (5 mains periods)
In general, the time domain behaviour of a microwave oven
can be summarisedas follows :
- the emission is impulsive and (( modulated)) by the mains
signal : 20 ms (50 Hz) periodicity of the emission ;
Figure 5 : Time domain measurement at 2450 MBz (oven - around the operational frequency, the emission is nearly
centre frequency) during 100 ms (5 mains periods)
present half of the time, whereas, as we move away from
this band, the emission becomesmore and more scarce;
Figure 5 show the time variations of the emission at the - the amplitude of the pulsesrandomly varies with time.
nominal operating frequency. Quite logically, the emission is
present only half of the time (half wave rectified a.c.) and two The consequence is that, when frequency domain
main emission bursts appear during one mains cycle. If measurementsare performed, the result very much depends
observed during a longer time, we would see that the on the type of detector used : the peak level is significantly
maximum level of these bursts also change with time (for higher than the average level, especially as we move away
example between 100 and 120 dBpV/m) depending on the from the oven band. This has an influence on the interference
variations of the load inside the oven.
potential towards radio systems : for the same peak level,
120
microwave ovens are much less disturbing than a CW-like
source(Information Technology Equipment for example).
again the fluctuating nature of the oven emission. In f&t, the
samefrequency is emitted so rarely that, except in the middle
of the allocated band, the (( average logarithmic level )) falls
For measurement above 1 GHz for all kind of non radio within the noise. Therefore, this method cannot be used to
equipment, CISPR is currently defining the suitable detector, characterisethe emission outside the oven band.
and the solution chosen is to use a peak detector that is quick
and easy to use. But, for microwave ovens, the ad-hoc group
concluded that there is a need for an additional measurement
more representativeof the averageemission level.
DIXXJSSION OFAWEIGHTINGMETHOD
The ad-hoc group deliberately investigated only weighting
functions that ,would be easy to use and commercially
available on most of the existing spectrum analysers.The aim
was certainly not to design a new detector, as it was done in
the past for the CISPR quasi-peakdetector.
For this reason,only the following options were considered:
- use of the trace (or video) averaging function : the displayed
level is the average of a series of individual sweeps
petiormed with a peak detector ;
- a reduction of the video bandwidth (VBW) on the spectrum
analyser : it reduces and filters the brief isolated emissions
and retains only the more permanent emissions.
I
Figure 9 : Measurement of a microwave oven in the
frequency domain with a peak detector (upper curve) and
a reduction of the video bandwidth to 1 kJ3z (lower curve)
Figure 9 show the result obtained with the second method
(reduction of the VBW) as compared with a classical max
hold result. The noise level is highly reduced by the
weighting, as well as the frequency of emission of the oven,
For both of these methods, one drawback was identified : if also showing that these emissions are not permanent (if the
the measurementsare performed in logarithmic display (as it source was CW like the clock of an Information Technology
is the case for all the curves here), both the trace and the Equipment, both methods would give nearly the same level).
video bandwidth average the logarithmic levels and not the It can also be seen that this weighting technique is more
linear ones. As a consequence,for example if a squarepulsed useful to assesslevels outside the allocatedband.
signal (half on, half off) at a level of 60 dBpV/m is measured,
the level indicated by both detectors will be 30 dBuV/m, Based on these results, the reduction of the VBW was chosen
whereasthe true averagelevel would be 54 dBpV/m. And, the as the weighting method for measurementsof ISM. The admeasurementof a microwave oven emission in linear display hoc group then discussed what value of the VBW was the
is nearly impossible to perform. For this reason, the term most appropriate. Considering that the oven is modulatedby
(( weighted )) was chosen for this type of measurementrather the mains frequency (50 or 60 Hz), it was concludedthat a
value lower than the latter was neededin order to obtain the
than (( averageD.
average (in logarithmic units) of the emission. For this
reason,the value of 10 Hz was chosen.
LATEST PRO~SATS FOR LIMITS
We will summarise in this section the content of the latest
draft distributed to the CISPR National Committees
proposing limits between 1 and 18 GHz [5]. The rationale
upon which the values of the limits themselves have been
derived is not explained, only the main principles and the
measurementtechnique to be used will be detailed.
We have seenthat for microwave ovens, high peak levels can
be tolerated, as the averageemission level is very much lower.
But, when elaborating the proposal, one element to take into
account is that microwave ovens are only one specific kind of
ISM (although the most widely spread) and that it may exist
Figure 8 show the result obtained with the first method (trace CW-like ISM sources (some prototypes are already under
averaging) as compared with a classical max hold result. As development), for which high peak levels would surely
can be seen,the two curves drastically differ, illustrating once disturb the radio servicesto be protected above 1 GHz.
Figure S : Measurement of a microwave oven in the
frequency domain with a peak detector in max hold mode
(upper curve) and in trace averaging mode (lower curve)
121
For thesereasons,a two-tier approachwas defined :
- a first measurementis performed in peak max hold mode. If
all emissionsbetween 1 and 18 GHz (except inside the ISM
bands) are below a level of 70 dB uV/m, then no additional
measurementis necessary.If this level is exceeded(it is the
case for all existing microwave ovens), then the measured
result is comparedto the limits given in Table 1 below. The
equipment is declared compliant if these limits are met and
provided that :
- a second measurement is performed with the weighted
method discussed above (VBW reduced to 10 Hz) at two
spot frequencies with regard to the limit given in Table 2
below.
CONCLUSION
Past standardisation efforts to set emission limits for
microwave ovens have not been successful due to the
following paradox : microwave ovens exhibited extremely
high peak emission levels but, at the same time, when
experimentsof their effects on radio systemswere made,very
limited disturbanceswere experienced,
Since a few years, more in depth technical studies,as the one
described in this paper, enabled to understand better the
technical reasons of this paradox. As explained here, the
main reason is the fluctuating nature of the oven emission
with basically, at a given frequenq, two major pulsesemitted
every mains period.
To summarise, for CW-like sources, only one peak
measurement is necessary and for fluctuating sources, one Therefore, the ratio between the peak and the averagepower
peak and one weighted measurementare required (and both is very high in the case of microwave ovens. The resulting
Table 1 and Table 2 limits shall be met).
effect on radio systems(nearly all digital above 1 GHz) very
much dependson the maximum tolerable bit error rate of the
Table 1 - Electromagnetic radiation disturbance peak limits
latter : for systemslike fixed links that can drop the link at
for Group 2 Class B ISM operating at frequencies above
error rates of lo“ - 105, interferences are likely to occur,
500 MHz beak measurements with a resolution bandwidth
whereas for modern mobile services like cellular telephones
of 1 MHz and a video bandwidth higher or equal to 1 MHQ
that can withstand error rates as high as 1U2,the threat is in a
great part reduced or will occur for a higher level of the
Field strength at a
Frequency range
interferer.
measurement distance of
WW
3 metres (dB&V/m)
As it was not realistic to define a specific limit and
92
- 2.3
1
measurementmethod (weighting detector) for each and every
110
2.3 - 2.4
existing and planned radio service allocated between 1 and 1X
92
2.5 - 5.725
GHz, the CISPR ad-hoc group in charge of this work choosea
92
5.825 - 11.7
medium solution that is presented in the last part of this
73
11.7 - 12.7
paper. We can reasonably hope that it should as a first step
solve most of the potential interference situations outside of
92
12.7 - 18
the
allocated ISM bands. For systems allocated inside the
Note
.
Limits
of
this
table
were
derived
considering
fluctuating
-’
oven
band (Low Earth Orbit satellites and Radio Local Area
sourceslike magnetrondrivenmicrowaveovens.
II
Networks), further studies are necessary.
Table 2 - Electromagnetic radiation disturbance weighted
limits for Group 2 Class B ISM operating at frequencies
above SO0 MHz (weighted measurements with a resolirtion
bandwidth of 1 MHz and a video bandwidth of 10 H@
Field strength at a
measurement distance of
3 metres (dB@V/m) ;
VBW=lOHz
60
- 2.4
1
60
2.5 - 5.725
60
5.825 - 18
To check the limits of this table, measurementsneed only
to be performed around two centre frequencies: the highest
peak emission in the 1005 MHz - 2395 MHz band and the
highest peak emission in the 2505 - 17995 MHz band
(outside the band 5720 - 5830 h&Iz). At these two centre
frequencies,measurementsare performed with a span of 10
MHz on the spectrum analyser.
Frequency range
W-J4
RJB’ERENCES
[l] CISPR Publication 11 : Limits and methods of
measurement
of
electromagnetic
disturbance
characteristicsof ISM radio frequency equipment, 1990
[2] Effect of microwave oven interferencesto the performance
of digital radio communication systems- S. Miyamoto and
N. Morinaga - URSI General Assembly - Lille, Sept. 1996
[3] Radio disturbancesin between 1 GHz and 18 GHz caused
by microwave ovens - M. Vrolijk and B. Despres International Wroclaw symposium on EMC - June 1996
[4] Statistical parameter measurementof unwanted emission
from microwave ovens - Y. Yamanaka and T. KobayashiIEEE Internationla Symposium on EMC - Atlanta, August
1995
[S] CISPR/B/175/CD - Proposal for emission limits from 1 to
18 GHz - August 1996
122