1、Invited paper, OFC, Anaheim, CA, Mar. 2001Frequency response metrology for high-speed opticalreceiversPaul D. Hale, Tracy S. Clement, and Dylan F. WilliamsNational Institute of Standards and Technology, 325 Broadway, Boulder, Colorado 80305(303) 497-5367, haleboulder.nist.govAbstract: We use “standa
2、rd” optical sources along with microwave calibration methodsto accurately calibrate optical receivers to 50 GHz and beyond. The calibrated receiverscan be used to characterize sources and receivers using common electricalinstrumentation.Publication of the U. S. Government, not subject to U. S. copyr
3、ight.OCIS codes: (060.2380) Fiber optics sources and detectors; (120.3940) MetrologyThere are two types of optical sources whose modulation can be measured or calculated from fundamentalprinciples: the heterodyne beat between two single-frequency lasers (frequency-domain) and the short pulsefrom a m
4、ode-locked laser (time-domain). While these sources are essential for receiver characterization,good sources are not all that is required. Since typical optical systems must be characterized over abandwidth as much as 10 times larger than the bit rate, calibrating the electrical instrumentation isex
5、tremely challenging. We will discuss the construction of standard optical sources and the importance ofelectrical calibrations in both the heterodyne and short-pulse measurement methods for receivercharacterization below 50 GHz, and propose measurement strategies for going beyond 50 GHz.Heterodyne m
6、easurementsHeterodyne measurement gives the most accurate estimate of the magnitude of an optical receiversfrequency response. Our implementation of heterodyne measurements uses two tunable single-frequencylasers operating at 1319 nm 1; an external cavity semiconductor laser (ECL) and a Nd:YAG laser
7、(YAG2), to generate a beat frequency, tunable from 1 MHz to several hundred gigahertz. We pass free-space beams from each of the lasers through variable attenuators and combine them in a beam splitter. Thecombined beam passes through a polarizer and is coupled through a single-mode fiber to the rece
8、iverspecimen. Polarizing and combining the beams in a single-mode fiber is critical, as this allows us to exactlymatch the polarization and spatial-mode of the two beams incident on the receiver, and achieve a preciselycalculable modulation depth. With matched received photocurrent from each laser,
9、the optical andelectrical signals have 100 % modulation depth.When the modulation depth is 100 %, the normalized frequency response 2()f of a photodiode is 22 rf21dc r2() ,PfiZ= (1)where Prf is the microwave power the photodiode would deliver when connected to a load Zr =50 atfrequency f, and idcis
10、the total dc photocurrent the photodiode draws from the bias supply. Determining2()f from the power Pmmeasured by a microwave power meter requires accurate microwave calibration.To perform this calibration, we must first measure the scattering parameters Sijof any network between thephotodiode and t
11、he microwave power sensor, and the electrical reflection coefficients pand sof thephotodiode and power sensor. Then Prf is related to Pmbymrf 2sp 1,()| |PPkf r=(2)where k(f) is the sensors calibration factor at frequency f and 321sp11 p 22 s s p 21 12 11 22.1()SrSS SSSS= (3)The last term in (2) corr
12、ects for impedance mismatch and loss in the interconnecting network and reducesto |1- ps|2if the power sensor is connected directly to the photodiode. We illustrate the importance ofthe corrections in Fig. 1, where the symbols correspond to uncorrected measurements of a photodiodesresponse using two
13、 different interconnecting networks. Using the bias T gives high mismatch, and using theattenuator gives high loss. The curves show the corrected response, which has the effects of network lossand mismatch (ripple) removed by application of (2).In applications where the photodiode is to be used only
14、 to measure the modulation depth of a source, atransfer standard 2 consisting of a photodiode and a power sensor can be calibrated as a single unit withlower uncertainty than a calibration using (1). The combined response 2()f! of the transfer standard isfound using (1) with Pmsubstituted for Prf .
15、The second and last terms in (2) are not required for thecalibration, eliminating the need for calibrations of microwave power and measurements of scatteringparameters. The transfer standard can be used to determine the fractional modulation depthmax min max min()/()OMPPPP= + of a source as 22 m221d
16、c r21,()OPMiZ f=!(4)where Pmaxand Pminare the maximum and minimum optical powers of the modulated source.Heterodyne measurements above 50 GHzBelow 50 GHz, we can easily measure the beat frequency with an electrical spectrum analyzer andharmonic mixers, and the microwave power with a coaxial diode po
17、wer sensor. Making measurementsabove 50 GHz is much more difficult, due partly to the lack of commercially available equipment.The frequency-measurement portion of the system is complicated by the lack of readily availablephotodiodes with bandwidths larger than 50 GHz and the high conversion loss of
18、 harmonic mixers,required to down convert the signal into the frequency range of a spectrum analyzer. The combinedconversion loss gives a poor signal-to-noise ratio. Our solution to the frequency-measurement problem isto use a high-power single-frequency Nd:YAG laser (YAG3) with a frequency (wavelen
19、gth) betweenYAG2 and ECL. We measure the YAG2/YAG3 beat frequency with a counter and the YAG3/ECL beatfrequency with a microwave spectrum analyzer. The frequency stability of the ECL in a heterodyne systemis adequate to give sub-megahertz resolution 4, but will not accurately trigger our counter. We
20、 send thebeat signal between YAG2 and ECL, whose frequency is the sum of the YAG3/ECL and YAG2/YAG3beat frequencies, to the receiver specimen.Measurements of microwave power above 50 GHz are also complex. Calibrated coaxial diode powersensors capable of measuring powers of 1 nW to 1 W are not availa
21、ble. Waveguide power sensors thatFigure 1. Raw measured response using 3 dBattenuator (+) and bias T (o) as connectingnetwork. Overlapping curves show correctedresponse.Figure 2. Effect of time-base distortioncorrection on mismatch correction. Thick andthin lines show mismatch-corrected responsewith
22、 and without time-base distortion correction.0 1020304050Frequency, GHz-16-12-8-40Response,dB0 1020304050Frequency, GHz-18-15-12-9-6-30Response,dBhave adequate sensitivity in the 50 to 75 GHz (WR-15) and 75 to 110 GHz (WR-10) frequency ranges areavailable, but must be connected to the receiver throu
23、gh an adapter. Furthermore, the VSWR of thesesensors can be as high as 2.5:1, possibly giving 8 dB (peak-peak) ripple in the measurement of a high-impedance photodiode. By including a 10 dB attenuator as part of the sensor the ripple can be reduced toabout 0.8 dB, small enough that it can be adequat
24、ely corrected using (2). This gives a sensor that is robust,but at the expense of sensitivity. The resulting microwave sensor is nevertheless a good candidate for partof a combined transfer standard.Time-domain measurementsTime-domain methods determine not only the magnitude response of a receiver,
25、but also its phase response.However, the uncertainties are not as well understood as those of the heterodyne method. We use a 50GHz sampling oscilloscope for time-domain measurements. The response measured with an oscilloscopecan be broken into the convolution of three effects: (1) the response of t
26、he oscilloscope to a matched 50 source, (2) a contribution due to the impedance mismatch and loss in the network connecting the receiver tothe oscilloscope, and (3) the effect of the finite duration of the optical impulse stimulus.Calibration and correction of oscilloscope response is not well under
27、stood at high frequencies and includesseveral non-ideal effects. We use the nose-to-nose procedure to calibrate the oscilloscope phase response5. However, we have been unable to verify the accuracy of the nose-to-nose phase calibration, possiblyintroducing large errors into the measurement. We use s
28、wept-sine measurements for calibrating theoscilloscopes magnitude response 5. Drift and jitter are random variations in the sampling time thatoccur over long and short time scales relative to one complete sweep of the display. To correct for drift andto achieve a low noise level, we store 500 to 100
29、0 waveforms and then align and average them 6. Time-base distortion (TBD) is caused by repeatable errors in the oscilloscope delay generator that determineswhen the oscilloscope takes a sample. We use a nonlinear least-squares fit to sinusoidal input waveformsto estimate the TBD, and then use a regr
30、ession spline interpolation of the impulse waveform to correct forthe TBD 5. The corrections for drift and TBD are nonlinear processes, so we perform them in the time-domain first. We then multiply by the linear corrections for mismatch 7, oscilloscope response, and jitter6, in the frequency-domain,
31、 as in (2).Not correcting for the TBD ruins the mismatch correction. Fig. 2 compares mismatch-corrected responsesboth with and without prior correction for TBD, and illustrates the necessity of the TBD correction.Commercially available oscilloscopes operate up to only 50 GHz. However, we have found
32、that foraccurate measurement of time-domain properties such as rise- and fall-time, the bandwidth of themeasurement system should be about 10 times larger than the signal bandwidth. Thus we believe that a 40Gb/s system should be characterized to at least 200 GHz, and preferably to 400 GHz. Both the
33、heterodyneand oscilloscope approaches presently being implemented fall far short of this frequency requirement. Weare currently investigating time- and frequency-domain methods to extend receiver characterization to thesehigh frequencies.1. Some high-speed photodiodes use a guided-wave design optimi
34、zed for 1550 nm and should be characterized at the ultimateapplication wavelength for the highest accuracy.2. P. D. Hale, C. M. Wang, R. Park, and W. Y. Lau, “A transfer standard for measuring photoreceiver frequency response,” J.Lightwave Technol. 14, 2457-2466 (1996).3. “S-parameter design,” Hewle
35、tt Packard Application Note 154, April 1972, 9-13.4. P. D. Hale and C. M. Wang, “Heterodyne system at 850 nm for measuring photoreceiver frequency response,” Symposium onoptical fiber measurements, (Sept. 2000).5. P. D. Hale, T. S. Clement, K. C. Coakley, C. M. Wang, D. C. DeGroot, and A. P. Verdoni
36、, “Estimating the magnitude and phaseresponse of a 50 GHz sampling oscilloscope using the nose-to-nose method,” 55thARFTG Digest, (June 2000).6. T. S. Clement, P. D. Hale, K. C. Coakley, and C. M. Wang, “Time-domain measurement of the frequency response of high-speedphotoreceivers to 50 GHz,” Symposium on optical fiber measurements, (Sept. 2000).7. Time windowing of the measured waveform, a common technique for correcting mismatch, can be arbitrary and canbe a source of uncontrolled measurement errors.
