Frequency response indicates optical-receiver performance

Marcia Stellpflug

The increase in channel bit rates places increasing demands on the receivers used in the optical-to-electrical (O-E) conversion process. This trend requires the development of appropriate test procedures and instruments to ensure that all outgoing devices work as expected.

To ensure the performance of a 40-Gbit/s optical receiver, the frequency response must be measured to at least 60 GHz. The dominant frequency components are near 20 GHz, while their third harmonics near 60 GHz are essential to reproduce the signal with good fidelity. Even higher frequencies are needed for more complete characterization.

Three main factors limit the speed of optical receivers: the diffusion of carriers, the drift transit time in the depletion region, and the capacitance of the depletion region. The slowest of these three processes is carrier diffusion. Carriers generated outside the depletion region slowly diffuse into it. Since diffusion is significantly slower than the drift transit time, carriers should be generated at or near the region. In the depletion region, carriers can be assumed to be moving at a constant speed if the reverse bias across the region exceeds the velocity saturation value.

The longest transit time is for the carriers that need to traverse the full depletion-layer width. Therefore, the desired transit time limits the thickness of the depletion region. This constraint in turn places limits on the capacitance of the optical receiver for a given geometry and dielectric value of the depletion layer material, because the maximum response of the receiver will occur when the transit time is comparable to the resistance-capacitance time constant. Additional factors that affect the speed of optical receivers include charge trapping and parasitic capacitance.

All of these factors affect the bandwidth and frequency response of optical receivers. Frequency-response measurements are a way of determining the effect of these factors on the performance of high-speed receivers. Three methods or tools may be used to measure frequency response: the impulse-response measurement, an electrical network analyzer, and the optical heterodyne technique.

IMPULSE-RESPONSE MEASUREMENT
Impulse-response measurement uses a short-pulse modelocked laser to infer an optical receiver's frequency response. The response of the receiver to all modulation frequencies is acquired simultaneously in the time domain. However, time-domain methods are sensitive not only to the magnitude response of a receiver but also its phase response (the relative delay for various RF components).

Assumptions about the phase response need to be made to extract accurate amplitude information; otherwise, the bandwidth of the optical receiver may appear artificially narrow. Additionally, one needs to know the spectral shape of the laser pulse, and its bandwidth should be at least 10 times larger than the required receiver bandwidth. For these reasons, the uncertainties associated with impulse-response techniques can be significant.

Frequency resolution is limited to the inverse of the time-record length. Because the maximum sampling interval is determined by the requirement that fast transitions be adequately sampled, the maximum number of points that can be conveniently handled by the measurement system limits the maximum time-record length and, therefore, the frequency resolution.

ELECTRICAL-NETWORK ANALYZER
Another technique is to use an electrical- network analyzer (ENA) with a high-speed O-E converter (modulator) to generate an optical microwave signal. One advantage of this technique is that it measures a receiver response directly in the frequency domain.

The ENA can generate both amplitude and phase information; however, the signal processing that is required to extract the phase information becomes increasingly difficult as the radio frequencies under test reach 60 GHz and beyond. Furthermore, the O-E converter must be calibrated for its own frequency and phase response. As frequencies become higher, the frequency response of the modulator becomes increasingly more complex and variable from unit to unit.

OPTICAL HETERODYNE TECHNIQUE
Heterodyne measurements also give information in the frequency domain by generating broadly tunable microwave signals directly in the optical domain.

Paul Hale, project leader in the Opto-electronics division, and coworkers at the National Institute of Standards and Technology (Boulder, CO) have used an Nd:YAG heterodyne system with low inherent optoelectronic-response uncertainty to develop frequency-response calibrations that are traceable to national standards.¹

The Nd:YAG system calibrations are not only well understood but also are independent of the frequency-response measurement. Furthermore, the excitation of the detector can be calculated from first principles.

For this reason, the optical heterodyne approach gives the most accurate estimated amplitude of an optical receiver's frequency response. An electronic locking accessory on a single-frequency Nd:YAG laser can generate broadly tunable optical microwaves via an optical heterodyne technique.² Optical microwave signal generators are commercially available.

An optical heterodyne system includes two main components: a pair of single-frequency lasers—the outputs of which are mixed together to produce a difference frequency—and a method of precisely controlling these lasers to give the desired frequency. The generation of optical microwaves is accomplished by combining the output from two lasers in a fiberoptic coupler (see Fig. 1).

The control electronics can independently adjust the oscillation frequencies of the two lasers and thereby tune the difference frequency from DC to greater than 80 GHz. The resulting frequency difference is controlled by making small adjustments in the frequency of one or both of the lasers. A

frequency-difference ramp can be generated by simply thermally tuning—or sweeping—the laser frequency. A DC signal corresponding to the frequency difference is output so the receiver response can be recorded as a function of the frequency difference.

The combined laser beams pass through polarization-maintaining fiber so the signal from each laser is in the same polarization state when it reaches the device under test. The actual mixing process occurs at the optical receiver when the optical waves are converted to an electrical signal.

Matching the polarization and spatial modes of the two beams incident on the receiver is necessary to achieve a stable modulation depth. With matched received photocurrent from each laser, the optical and electrical signals have 100% modulation depth. At modulation depths less than 100%, a bias current is always present on the receiver. This bias is undesirable because both the receiver and the electronics have a limited capacity for power handling and can be saturated. Another consideration is that if a bias current is present on the optical receiver, it is important to know the magnitude of the current and the constancy of the signal over time.

The frequency resolution of a phase-locked heterodyne system can be determined to a fraction of a Hertz and will depend upon the accuracy of the reference signal. The accuracy of frequency signals in a heterodyne system running in open loop is contingent upon the temperature stability of the lasers. At higher frequencies, the frequency resolution is determined by the choice of hardware and design. Keeping the same relative resolution (less absolute resolution at higher frequencies) allows faster sweeping through broad frequency ranges.

The Nd:YAG systems operate at a wavelength of 1319 nm. Indium gallium arsenide (InGaAs) optical receivers behave similarly at 1319 and 1550 nm. Typically, the dominant limitations on the frequency response of a receiver are either inherent properties of the device—such as parasitic capacitance and carrier migration velocities—or packaging-related parasitics. Optical receivers for 1550 nm do have a longer absorption depth, which changes the carrier distribution geometries and ultimately carrier migration time. This effect typically is not dominant and the frequency response at 1550 nm can be quite accurately inferred from the frequency response at 1319 nm with knowledge of the construction geometries and parameters. Of course, for greatest accuracy, high-speed optical receivers optimized for 1550 nm should be characterized at the application wavelength. Commercial systems are under development that will measure the frequency response at this wavelength.

FUTURE RESPONSE
Although the techniques under development for generating optical input to high-speed receivers have the potential to scale to higher frequencies, measuring the response of the receivers is limited by the availability of high-frequency spectrum analyzers and RF power meters.³ It is possible, but expensive, to generate high-frequency sources and mixers for spectrum analyzers above 50 GHz. Qualified cables and connectors are also more difficult to acquire, while commercial RF power meters are available up to 100 GHz.

REFERENCES

1. T. Olson and F. Adams, Microwave Journal 22, ( August 1993).
2. P. Hale et al., J. Lightwave Technology 14, 11 (1996).
3. P. Hale, OFC 2001, WQ1-1.

Marcia Stellpflug is product marketing manager at Lightwave Electronics, 2400 Charleston Rd., Mountain View, CA 94043. She can be reached at 650-526-1296 or [email protected].