Vous êtes sur la page 1sur 2



V. A. Grigor'ev, A. V. Kuzichkin, and V. A. Savichev UDC 621.317.361

A method of measuring the frequency-modulation index is proposed based on a measurement of the amplitudes
of three spectral lines, described by Bessel functions. The method enables any values of the frequency-
modulation index to be measured directly without additional calculations.

Several methods of measuring frequency-modulation index are known based on measurements of the frequency
deviation of the FM signal: the frequency-detector method, the oscilloscope method, the direct method etc. [1,2].
The main drawback of the oscilloscope method and the method based on the use of a frequency dectector is that it is
impossible to make direct measurement of the frequency-modulation index and it is necessary to have a priori information on
the modulation frequency or to make additional measurements to determine it. This complicates the measurement procedure.
There are also systematic errors when measuring the frequency deviation using these methods in view of the infinite nature
of the spectrum of the FM signal. The direct method of measuring the frequency-modulation index based on a determination
of the zeros of a Bessel function of zero order enables only certain values of the frequency-modulation indices to be measured
(2.4, 5.52, 8.65 etc.). In this paper we analyze a method of measuring the frequency-modulation index [5] based on the use
of recurrence relations for Bessel functions. It is shown that the range of measured values is practically unlimited.
As we know, a frequency-modulated signal can be represented in the form of a Fourier series, in which the amplitudes
of the spectral components are described by Bessel functions

S(t) =Arn{lo(m/)coso~ot-b ~ Ik(mf)[cos(oo..b

-[-kf~)t~ (-- 1) kcos(o~0--kf2)t I },

where A m is the amplitude of the high-frequency signal, Io(mf) and Ik(mf) are Bessel functions of zeroth and k-order
respectively, mf is the frequency-modulation index, o:0 is the mean oscillation frequency (the carrier frequency), and fl is the
modulation frequency.
The recurrence relation for Bessel functions can be expressed in the following form [3]:

Zk-1(ms)+Zk+l(m/)----- ~-I Z~(ml)'

and using this the frequency-modulation index is given by the expression

2k'rk(mY) (1)
ms----- Ik_l(my).bIk+l(my)

Hence, to measure the frequency-modulation index one needs to measure the amplitudes of three successive spectral
components. For example, if one uses the zeroth, first and second harmonics, expression (1) can be written in the form

2Ix(mr) (2)
my= Io(mr) +Is (my)

Translated from Izmeritel'naya Tekhnika, No. 9, pp. 50-53, September, 1993.

1036 0543-197,2/93/3609-1036512.50 r Plenum Publishing Corporation


FCI FC2 0,02 1',i //i .


-\ ",, 'J!\ i./
5,7 6.3 7
T- 3q.7r 710,2- 11,7I o7

Fig. 1 Fig. 2

The block diagram of the system for measuring the frequency-modulation index (Fig. 1) contains a frequency converter
FC1, and intermediate-frequency amplifier IFA, which filters the image channel and separates the required part of the spectrum
of the frequency-modulated signal being investigated, a frequency converter FC2, a bandpass filter BPF, which separates the
corresponding harmonics of the signal, amplitude detectors AD, which determine the amplitude of the spectral components,
and a calculation unit CU in which the measured frequency-modulation index is calculated from the measured values of the
amplitudes. The frequency converters contain a mixer M and a heterodyne Het which reduces the frequencies to the required
intermediate frequencies. In frequency converter FC 1 the choice of intermediate frequency should ensure reliable suppression
in the IFA in the image channel and separation of the required part of the FM-signal spectrum, for example, the spectral
components 0 < k < kmax. The intermediate frequency of the converter FC2 is chosen to satisfy the requirement of relaible
separation of the k-th spectral component for a specified level of suppression of the neighboring (k - 1)-th and (k+ 1)-th
spectral components and the known selectivity of the bandpass filters. The calculation unit CU contains an adder ~, which adds
the amplitudes of the ( k - 1)-th spectral components taking the phase relations into account, a multiplier Mul, which multiplies
the amplitudes of the k-th harmonic of the signal by a normalising factor 2k, and divider Div which divides the signal from
the output of the multiplier by the output signal of the adder, thus giving an estimate of the measured frequency-modulation
The circuit described is identical with the circuit of a typical spectrum analyzer and operates over the whole frequency
band using modulation. The range of modulation frequencies is limited by the frequency resolving power of the bandpass filters,
and also by the instability of the center frequency of the FM singal and the heterodyne signals, by means of which frequency
conversion of the signal being investigated is achieved. In practice it is unusually possible to carry out spectrum analysis at
modulating frequencies of 10-30 Hz.
Uniqueness of the measurement of the frequency-modulation index is ensured by satisfying relation (1) for any three
successive spectral components of the FM signal over the whole range of values of mr.
We will estimate the error of this method of measurement using the general rules for determining the errors of indirect
measurements [1]. We will assume in the analysis that the frequency converters operate linearly, that the amplitude frequency
response of the intermediate-frequency amplifier is uniform, and that the condition for the phase-frequency response of the
intermediate-frequency amplifier to be linear in the part of the spectrum being investigated is satisfied.
Consider the errors which occur due to the effect of random noise in the measurement channels, accompanying the
amplitude modulation, the spread in the transfer constants of the amplitudes of the spectral components (due to nonuniformity
of the amplitude-frequency response of the intermediate-frequency amplifier and the difference in the static transfer constants
of the bandpass filters), the nonideal nature of the conversion unit (the band-pass f'xlters and the amplitude detectors), and the
change in the measured quantity during measurements. We will distinguish between the systematic and random components
of the error.
Note that the systematic error is zero. We will estimate the random component of the error using existing rules [1].
The root mean square deviation of the value of the frequency-modulation index mf is given by the expression

V/ Omf 2 2 Omf 2 o2,, { Ornf 2

= ,

where aik, aik_l, aIk+l are the root mean square deviations of the values of the amplitude of the corresponding spectral