Ultradense wavelength-division multiplexing (DWDM) filters can potentially double, triple or even quadruple spectral efficiency by reducing the space separating channels to 50, 25 and 12.5 GHz, respectively. Thus far, however, the performance of these advances has not delivered a corresponding improvement in cost.
A 25-GHz transceiver, for instance, provides twice the channel count of a 50-GHz device, but it imposes a 10 to 20 percent higher cost per channel. A significant part of the additional per-channel outlay arises from the advanced optical monitoring technology needed to measure more voluminous and more closely spaced channels. Such monitors need much higher resolution than the channels they measure and have become increasingly challenged as the prerequisite resolution drops from 60 picometers for channels spaced 100 GHz apart to 10 or 15 pm for channels separated by 12.5 GHz.
Optical monitoring instruments work by drawing a percentage of signal power from a fiber-optic network, demultiplexing it into individual wavelengths and measuring each channel's optical power, spectral accuracy and signal-to-noise ratio (SNR). The cost of these instruments is linked closely to the complexity of their demultiplexing components, which rely either on diffraction gratings or Fabry-Perot interferometers. Most current instruments-those monitoring channels separated by 100 GHz or more-rely on a combined diffraction grating and photodetector array. But such devices become problematic as channel counts increase. For starters, they cannot easily demultiplex more finely spaced channels, because most detector arrays have resolutions limited to 512 pixels. Also, grating-based devices require a separate photodetector for each channel under test. This requires hundreds of detectors.
Limited tuning range
One approach is to demultiplex signals using Fabry-Perot interferometers, which require only a single-element photodetector to measure multiple channels separated by 25 GHz or less, using a resolution bandwidth corresponding to a finesse of around 2,500.
Different filters apply different tuning methods. One approach uses a resonant cavity, filled with an electro-optic material such as liquid crystal, between fixed plates. Applying a voltage to the material alters its refractive index and allows the filter to tune across a wavelength band. Generally, however, the small value of the electro-optic coefficient requires high voltages and provides a relatively small tuning range. It also results in large, bulky devices. A more versatile approach uses microelectromechanical-systems technology to alter the separation between a fixed and a movable resonator mirror. The mirror spacing establishes a single peak within a given wavelength band. Applying electrostatic forces to the movable mirror alters the spacing, allowing the filter to tune across its free spectral range. A change in separation of 1 micron, for example, enables tuning of a single resonance peak across 100 nanometers or more-enough to cover signals in the C-band and L-band.
The resonant-cavity structure and detector configuration reduces the footprint and cost of Fabry-Perot filters, but the potential trade-off to this simplicity has been the challenges associated with silicon's high rigidity. That rigidity, and the need to operate the micromirrors at moderate voltages, requires thin, delicate structures for the spring hinges. Such designs use a confocal cavity, where one of the mirrors is concave, in order to overcome problems in keeping the mirrors parallel. However, this design causes side modes on the filter profile that vary with wavelength, resulting in power measurement difficulties, particularly with SNR.
An alternative approach, which allows the use of flat mirrors, incorporates compliant elastomeric materials to support the movable mirror, hence the name C-MEMS. These materials are as much as six orders of magnitude less stiff than silicon and can be deposited in a much broader range of layer thicknesses. Unlike carbon-based elastomers, the materials used in C-MEMS have a Si-O-Si backbone, giving them excellent mechanical, chemical and thermal stability. Compared with silicon-based MEMS, elastomer-based C-MEMS achieve a given mechanical deflection, and their mechanical range of motion is much larger. Electrostatic force is enough to drive the mirrors and keep them parallel over the lifetime of the device, regardless of ambient vibrations, shocks or fluctuating temperatures. Counterbalance design
A C-MEMS Fabry-Perot filter consists of a micromachined silicon chip set comprising three or four individual chips. One chip has high-reflectivity and anti-reflective coatings on each side of its optically active area, which is surrounded by a large electrode. An elastomeric ring provides the filter's flexible element and provides a passive counterbalance to the electrostatic force. A second chip incorporates the drive electrodes, spaced several microns away from the movable mirror. Bonding the two pieces together with a reference mirror forms a basic mirror-driving unit with a reasonable voltage budget. An alternate configuration uses four chips to create two identical mirror-drive units.
In either case, C-MEMS requires a low voltage budget to move the mirror along its normal optical axis and tune over the full range of channels in the C- or L-band at scan rates of at least 10 Hz. More important, it ensures a high degree of parallelism and a consistent filter shape with peak wavelength insertion losses under 2 dB.
As a result of conditioning treatments, C-MEMS Fabry-Perot tunable filters have a simple calibration process. They only require a simple curve-fit method that yields accuracies of picometers. Such filters have demonstrated wavelength repeatability of 10 pm over the C-band tuning range.
Like the optical networks they constitute, optical components confront diverse environmental conditions under which they must deliver consistent and predictable performance. The most common environmental factors are vibration and temperature change. Together, these factors present major design and packaging challenges for all devices placed in the optical path.
Passive isolation-the common solution-proved insufficient to protect Fabry-Perot filters from residual environmental changes. A better approach is to configure filters with dual mirrors suspended in identical driving units. This enables common-mode cancellation principles to maintain zero relative differential movement for the mirror pairs-nulling environmental-vibration effects and yielding a stable filter peak position within several picometers when subjected to environmental vibration. This design has made C-MEMS tunable Fabry-Perot filters immune to the vibration encountered in normal operation.
One effective design for C-MEMS Fabry-Perot filters maintains the core tunable filter chip set under a constant temperature condition. Heating the device to a higher temperature than the specified environmental extreme keeps the core filter chip set temperature stable.
The supporting package temperature is still subject to environmental change and, if not properly isolated, mechanical thermal stress or stretch can impose wavelength drift. The solution: zero-stress or quasi-zero-stress attachment methods that minimize thermal stress spread into the optically active area. The approach enables reduction of peak wavelength temperature sensitivity to below 4 pm/ degrees C.
If the device is configured to reoptimize every 30 minutes, it will lock the wavelength drift to less than plus/minus 30 pm under temperature-changing rates of 0.5 degrees C/minute. Performed more frequently, reoptimization can help reduce drift to less than 10 pm.