Professional Documents
Culture Documents
3, MAY/JUNE 1998
597
I. INTRODUCTION
598
(a)
(b)
(c)
Fig. 1. Normalized amplitude spectra of (a) an unmodulated wave ( = 0) and two angle-modulated waves (b)
(1)
or
contains the
where the instantaneous phase
modulating signal
is a constant, represents time, and
is the angular frequency of the carrier, in this case, lamp
voltage (or lamp current) [16][19].
Phase and frequency modulations are two closely related
angle-modulation methods. In the case of phase modulation,
the modulating signal is directly proportional to the instantaneous phase
or
, and for frequency modulation, the
modulating signal is directly proportional to the first derivative
of
or
.
To study spectral behavior of the modulated wave
in
steady state, we shall define its amplitude spectral density
and its average power spectral density
, assuming
that, in mathematical terms,
is a real function, defined over
[16],
is defined as
:
(2)
(4)
where
for
for
or
(a)
599
(b)
(c)
Fig. 3. Limiting spectra for (a) sine wave, (b) square wave, and (c) symmetrical sweep or saw tooth.
and
, the
(5)
(6)
where
and
represent constants. Typical spectra for
several modulation indices are shown in Fig. 1. Observe the
relationship between amplitude of the power spectrum to the
modulation index.
The average power of the modulated wave
(7)
is independent with angle modulation; however, carrier and
sidebands, spaced at
, can vary. According to (4),
,
the magnitude of the spectral terms is determined by
denoting Bessel functions of the first kind. Fig. 2 illustrates
the impact of the modulation index increase on the maximum
spectral term
.
The maximum spectral term, which can excite an acoustic
. For the
resonance, can be minimized by making
theoretical limit case [17], or
, the power spectra is
(8)
(9)
(10)
(11)
600
(a)
(b)
(c)
(d)
(e)
(f)
Fig. 4. Experimental lamp current waveforms: (a) and (b) for sine-wave modulating signal; (c) and (d) for square-wave modulating signal; and (e) and
(f) for symmetrical sweep. (For all time-domain waveforms horizontal scale: 2 ms/div and vertical scale: 1 A/div; for all frequency-domain waveforms
horizontal scale: 5 kHz/div and vertical scale: 0.1 A/div.)
601
Angle-modulated spectrum, produced through intermodulation by carrier and periodic modulating signal, as described
in (4) for sine-wave modulations (Fig. 1), has a discrete
spectral density. When random noise is the modulating signal,
which has a continuous spectrum, in addition to carrierrelated discrete term, the principal part of the angle-modulated
spectrum is continuous.
Strictly speaking, amplitude and power spectral densities
and
, as defined in (2) and (3), do not exist for
random processes, since integral in (2) does not converge as
. However, this can be circumvented in mathematical
terms by utilizing the WienerKhintchine relationship between
of a random process
,
autocorrelation function
defined as
(13)
where
denotes the expected value operator and power
spectral density function
. According to this relationship,
(14)
Power spectral density function
is defined as the Fourier
transform of the autocorrelation function
.
While the mathematical tools for analysis of the angle
modulations by random processes change from before, as
noise is described in terms of statistical properties, essential
properties, such as one estimated by Carsons rule, remain the
same.
According to the principle of adiabatic frequency sweeps
[17], for large modulation indices, the power density of the
angle-modulated wave is proportional to the first-order probability density of the frequency-modulating process. Hence, for
phase modulation by a random noise
, the modulated
wave spectral density is
(12)
(15)
represents the maximum frequency deviation,
where
is the modulating signal frequency. Accordand
ing to (12), the modulation index increase is proportional to the
maximum amplitude decrease; unlike for sine-wave-modulated
spectra, this relationship is linear.
Experimental results, illustrated in Fig. 4(e) and (f), show
a 50% reduction in bandwidth over that of sine-wave modulation.
where
and
represents the density distributions of the derivate
of the modulating noise.
602
(a)
(a)
(b)
Fig. 8. (a) Experimental modulating noise time-domain waveform (horizontal scale: 20 ms/div, vertical scale: 100 mV/div), and (b) its spectrum
(horizontal scale: 25 Hz/div, vertical scale: 50 mV/div).
(b)
Fig. 9. (a) Experimental band-limited white-noise-modulated lamp current
(horizontal scale: 2 ms/div, vertical scale: 1 A/div) and (b) its fast Fourier
transform (FFT) (horizontal scale: 5 kHz/div, vertical scale: 0.1 A/div).
Fig. 10.
603
(a)
(b)
(c)
Fig. 11. Arc appearance with (a) 60-Hz magnetic ballast, (b) high-frequency electronic ballast with acoustic resonances, and (c) high-frequency electronic
ballast without acoustic resonances.
604
Laszlo Laskai (M87SM96) received the Dipl.Ing. degree from the University of Novi Sad, Novi
Sad, Yugoslavia, in 1982 and the Ph.D. degree from
Texas A&M University, College Station, in 1994,
both in electrical engineering.
From 1983 to 1986, he was with Investproject, Novi Sad, Yugoslavia, working on standby
power generation and power distribution for large
industrial consumers. From 1986 to 1990, he was
with Chronar Corporation, Princeton, NJ where he
was involved in the development of power conversion equipment for photovoltaic applications and high-frequency ballasts for
gaseous discharge lamps. Since 1994, he has been with Corporate Research
and Development, General Electric Company, Schenectady, NY. His current
research interests are in lighting and medical electronics.
Dr. Laskai currently serves as Chairman and Transactions Editor for
the Production and Application of Light Committee of the IEEE Industry
Applications Society (IAS). He is also an active member of the IAS Industrial
Power Conversion Committee.
605