The Mayo SPPDG has a long history of research in the field of high speed electronics, but since the mid-1990's, the group has become active in the field of optoelectronics as well. The following sections describe some of Mayo's measurement capabilities and projects in this area.
Digital Communications
In 1997 the SPPDG began researching optoelectronic components for high speed digital communication applications. The early research efforts focused on systems that utilized vertical cavity surface emitting lasers (VCSELs). VCSELs offer many advantages over standard edge-emitting lasers including lower power, lower cost, and better ease of integration with integrated circuits. SPPDG developed several new packaging techniques to demonstrate the technical feasibility of a low-cost replacement for the standard transistor outline packages ("TO" cans) which were commonly used for packaging VCSELs at that time. The SPPDG-designed packages did in fact demonstrate a cost efficient method to "break through" the 1 GHz performance barrier imposed by the TO can packages.
In 1998, SPPDG assisted industry collaborators in the design of packaging eight or more parallel optical channels, with each channel operating at rates of 2.5 gigabits per second (Gbps). These parallel channels allowed for aggregate throughput rates exceeding 10 Gbps. Modulation of the VCSELs for the transmit channels and photodetectors on the receiving channels was accomplished with integrated circuits fabricated in cost-effective bulk CMOS technologies.
In 2000, SPPDG designed transmit and receive boards capable of 10 Gbps transmission. These boards used components with the highest performance VCSELs and photodiodes available at that time. The combination of the matching transmitter and receiver boards demonstrated the goal of 10 Gbps, with very low bit error rate in a very high shock and vibration environment.
In 1998 SPPDG began research into so-called "free space" implementations in which the light from VCSELs is directed to photodetectors by means of lenses and mirrors. This technique requires that the VCSELs and photodetectors be in close proximity to one another. However, the free space technique appeared to offer much more performance over all-electrical implementations for such systems as fully connected, non-blocking crossbar switches. The SPPDG collaborated with Honeywell, the University of Delaware, George Mason University, and IBM to construct two generations of demonstration systems of high speed free space optical interconnect crossbar switches. Unfortunately, the projects did not progress to the point at which these free-space optical interconnect systems could be fielded as communications modules in real-world systems.
Conceptual drawing of high speed free space optical interconnect demonstration with 10 Gb/s per channel target data rates.
Analog-to-Digital Conversion
The development of higher bandwidth, higher dynamic range commercial electronic analog-to-digital converters (ADCs) has been historically slow, in part due to the limitations of sample timing uncertainty (i.e., "jitter"). Some of this "jitter" is generated by the signal source used to clock the ADC components, but some is generated by the ADC internally, by the clock distribution and sample-and-hold circuitry. Over the past two decades, researchers in the field of optoelectronics have demonstrated that an optoelectronic modulator can be used to perform sampling of an electrical signal onto a train of optical pulses, which serve as the clock. As optical pulse trains with very low jitter (tens of femtoseconds) have been demonstrated, the feasibility has been investigated of using optical A/D conversion to overcome the limitations of electrical sampling techniques used in conventional ADCs.
Schematic of Mach-Zender interferometer (a) and output path optical intensity (b).
Although several methods have been proposed for performing optical sampling, the Mayo group worked with MIT Lincoln Laboratory on one specific method described here. A Mach-Zehnder interferometer (a particular type of optoelectronic modulator) modulates the intensity of light in two separate optical output paths; the output signal is determined by the intensity of the optical input and the voltage on the input electrodes. Therefore, an optical pulse that enters the modulator is split between the two optical outputs in a ratio that depends on the voltage on the electrodes at the time the pulse traversed the modulator. As the optical pulses are very short (as short as 1-10 ps), the output amplitude of each optical pulse provides a precise sampled "snapshot" of the input voltage at the input electrodes at the time this pulse passed the electrodes. At this point, the sampling is complete, but the light pulses must be converted back into electrical voltages so that they may be quantized by commercial electrical ADCs. Note that these ADCs no longer need to sample the actual input signal, and as a result, they may have somewhat less restrictive jitter specifications. In addition, before the light pulses are quantized, they can be demultiplexed using additional modulator stages to reduce the data rates, further reducing the strain on the electrical ADCs.
Photonic integrate and reset (PHIR) test board with demux.
Such an optical sampling system relies critically on the component that converts the light pulses to electrical voltages to be quantized. This component must convert the total energy of each pulse to a voltage with very high precision (8-12 effective bits) at the incoming pulse rate (tens to hundreds of MHz). SPPDG's primary task in collaboration with MIT Lincoln Laboratory, was the design, simulation, fabrication, and testing of these components. We chose a simple approach, using a photodiode to convert the optical pulses to current, which was integrated on a capacitor and buffered with an output driver. One printed circuit board implementation of this photonic integrate and reset (PHIR) approach also contained a stage of 1:2 electrical demultiplexing, halving the necessary clock rate of the back-end electrical quantizer.
Photograph of laboratory hardware for testing PHIR circuits.
Testing the PHIR circuits required well-timed, well-characterized input optical pulses equivalent to those used in an optical sampling system, and it was also necessary to capture the analog electrical output with the precision desired in the optical sampling system. In short, it was necessary to build a complete optical sampling system to test the PHIR circuit component. A relatively simple gain-switched laser amplified with an Erbium doped fiber amplifier (EDFA) was used to produce the optical pulses, which were sampled with a commercial optoelectronic modulator with a sine wave input. The PHIR output was then quantized using a commercial electrical ADC board.
Demonstration of far upper-Nyquist optical sampling measures single tones in first, eleventh, and sixty-first Nyquist bands.
Single-tone measurements of the PHIR circuits demonstrated an interesting feature of this optical sampling technique: upper Nyquist sampling. ADCs capture information in a bandwidth equal to the sampling rate divided by two (the Nyquist frequency); however, input test tones of higher frequency than Nyquist may be sampled, as they are "aliased" down into the sampled bandwidth. Therefore, with appropriate input filtering to prevent "aliasing", one may utilize the sampled bandwidth at a high frequency. However, in most electrical ADCs this approach is not practical, since the analog input bandwidth is typically limited by the same mechanisms that limit the sample frequency. However, in optical sampling systems, the sample rates are limited primarily by the back-end electrical quantizers, while the analog input bandwidth (the bandwidth of the optoelectronic modulators) can be much higher -- perhaps tens of GHz. This characteristic of the optical sampling systems makes far upper Nyquist sampling (e.g., sampling at frequencies far higher than the Nyquist frequency) practical. Measurements are shown of fast Fourier transform (FFT) data for single tones in the first, eleventh, and sixty-first Nyquist bands respectively. Phase noise in our system limited the performance at the higher frequencies, but the use of more advanced optical pulse sources (such as mode-locked lasers) could improve this performance significantly.
Device Characterization
The SPPDG's laboratory has established capabilities to measure the optical-to-electrical performance of transducers such as PIN and avalanche photodiodes (APSs), as well as optical transmitters (VCSELs). Our laboratory can perform measurements on discrete and array devices in either bare die form, or as packaged parts. Characterization of the optical-to-electrical and electrical-to-optical conversion performance of the devices can be performed from DC to over 25 GHz pulse rates at optical wavelengths of 500 nm through 1630 nm. Spectral responsivity measurements can be performed from 300 nm to 10 mm. All of these measurements are also performed in a low noise environment. Examples of an APD array test suite, and a VCSEL characterization test station in the laboratory, appear respectively in the two figures below.
Laboratory setup for testing optical-to-electrical transducer arrays.
Mayo lab test setup for characterization of VCSELs from Colorado State University, designed for within the C20I Program.
