over the entire optical gain-bandwidth (A f ) of the am- plifier. ... izyl can be set equal to e A f. With this value of .... 196, 1984. [2] C. A. Bums, J. E. Bowers, and R. S. Tucker, âImproved very-high- .... William Rideout was born in East Orange, NJ,.
912
JOURNAL OF LIGHTWAVE T E C H N O L O G Y VOL 8 NO 6 J U N E 1990
Wide-Bandwidth ReceivedPhotodetector Frequency Response Measurements Using Amplified Spontaneous Emission from a Semiconductor Optical Amplifier ELLIOT EICHEN, JOHN SCHLAFER, MEMBER,IEEE, WILLIAM RIDEOUT, AND JOHN McCABE
Abstract-The white optical noise (spontaneous-spontaneous heat noise) generated by amplified spontaneous emission from a semiconductor-optical amplifier is used to measure the frequency response of very wide-bandwidth photodetectors and optical receivers. This novel technique can be used to characterize optoelectronic components of arbitrarily wide bandwidths.
I. INTRODUCTION
V
ERY wide-bandwidth optical receivers and photodetectors are becoming increasingly important for high-speed digital fiber-optic communications systems, and for optical transmission and signal processing of microwave and radar signals. As the bandwidth of these systems reaches beyond 20 GHz, it becomes increasingly difficult to measure the frequency response of receivers for these systems, primarily because it is difficult to separate the response of the photodetector from the response of a modulated source used to test the detector. While a number of methods have been successfully implemented-directly modulated high-frequency sources or external modulators [ 11, picosecond optical pulses [2], interfering two narrow linewidth lasers [3], measurement of photocurrent shot-noise spectrum [4], [5], or interferometrically demodulating the FM sidebands of a modulated semiconductor laser [6]-none has been completely satisfactory. In this paper we discuss using the white noise generated by amplified spontaneous emission from a semiconductoroptical amplifier to measure the bandwidth of optical receivers and photodetectors. The relatively large electrical-spectral noise generated by the interference between electric field components of the spontaneous emission (spontaneous-spontaneous beat noise), and the fact that this noise extends over the optical bandwidth of the amplifier ( - 50 nm or 10" Hz) allows for a straightforward measurement of frequency response with only an instrumental limit to the highest frequencies that can be measured. Manuscript received October 20, 1989; revised December 18, 1989. The authors are with GTE Laboratories Inc., Waltham, MA 02254. IEEE Log Number 9034615.
11. THEORY Semiconductor-optical amplifiers are lasers whose facets have been anti-reflection coated to suppress optical feedback. They are able to operate at extremely high current densities without lasing, and are generally used to amplify entering optical signals [7]-[ 101. In addition to providing large-signal gain, these devices can generate milliwatts of (incoherent) spontaneous emission spread over the entire optical gain-bandwidth ( Af ) of the amplifier. Each spectral component of this spontaneous bandwidth can be thought of as a sinusoidally varying electric field with a stochastically varying phase. When detected with a square law detector these field components interfere (beat), producing difference frequencies as large as the width of the spontaneous emission optical spectrum. The electrical-spectral noise density generated in this fashion, called the spontaneous-spontaneous beat noise ( i:p-sp), extends from dc to Af. If the optical spectrum is taken as rectangular then the calculated shape of i:p-sp will be triangular [9]. However, for even the high100 GHz), we are conest detector bandwidths (i.e., cerned only with the value of i fp - sp over the first few percent of its spectrum because of its enormous breadth. For example, the width of i:p - sp for an amplifier whose spontaneous emission optical bandwidth is 50 nm centered around h = 1.55 pm is 5 x 10l2 Hz. Measurements up to 100 GHz would utilize only the first 2 % of the width of if - sp. Thus, the exact shape of the spontaneous-spontaneous beat-noise spectrum is not important, and for all practical frequencies i:p-sp is flat. The value of ifppsp ( A2/Hz) in this (low frequency) vicinity is given by [7][lo1
-
.L lsp-sp
=
[ e @ - 1)%psc4e12"rAf
(1)
where e is the electronic charge, G is amplifier gain, m, is the number of transverse modes of the amplifier, qsp is population inversion coefficient, vc is the total coupling efficiency between amplifier and detector, and qe is the detector quantum efficiency. The dc photocurrent generated by amplified spontaneous emission idc (A), is given
0733-8724/90/0600-0912$01.OO O 1990 IEEE
913
EICHEN er al. : WIDE-BANDWIDTH RECEIVERIPHOTODETECTOR FREQUENCY RESPONSE MEASUREMENTS
Amplifier Internal Gain (dB)
[91 by idc
=
e(G -
l)qspqcqemrA$
For typical values of qsp (between 1.5 and 5 ) , G( 10-30 dB), Af ( -30-60 nm), qc( -3 dB), qe(0.7),and m, = 2 (an amplifier which supports only a single spatial mode and both polarization states), i:p-ap is between - 130 and - 160 dBm/Hz. This is several orders of magnitude above the noise floor of conventional microwave electronic electronic instrumentation which is set by thermal noise ( - 174 dBm/Hz for a 50-0 environment). The implication is that the spontaneous-spontaneous beat noise can be measured using conventional microwave instrumentation. The method discussed in this paper of measuring frequency response (i.e., the absolute magnitude of the frequency response) does not allow for a direct measurement of the device's phase response. Both the magnitude and phase information are required to exactly predict the response of the detector in the time domain. While this limitation is shared by most wide-band techniques, under certain conditions the phase response can be obtained as the Hilbert transform of the frequency response [4]. It is important to distinguish between measurements of photodetector bandwidth using the spontaneous-spontaneous beat noise, and previous measurements made using the roll-off in the frequency of shot noise generated by a dc photocurrent [4], [5]. With idkas the photodetector dark current, the shot-noise spectral density ( A2/Hz), i:hot, is given by [ 111
Rewriting the spontaneous-spontaneous spectral density from (1) and (2) as 2i&/Af and equating the shot noise and spontaneous-spontaneous beat-noise spectral densities yields the photocurrent (izyl)at which these two noise sources are roughly equal, or iequdl dc
= eAf
+ idk.
0.5 3.6 5.4 6.7 7.7
(2)
(4)
For almost all reasonable detectors idk
