1.

Introduction

Semiconductor-optical amplifiers (SOAs) are key future components in the evolution of all-optical networking. They are flexible not only in terms of functionality, e.g., amplification, gating, and switching,^{1, 2, 3} but can be designed to operate over a wide wavelength range from 0.6 to $1.6\mathrm{\mu }\mathrm{m}$ . More recently, the sharp increase in demand for high-speed access solutions for both business and home usage has stimulated opportunities within passive optical networks (PONs) deployment strategies.^{4, 5, 6, 7} In addition to traditional telecoms applications, SOAs are also being used as the active element in sensor and measurement systems,⁸ some of which operate at $1.3\mathrm{\mu }\mathrm{m}$ .

For many applications, it is advantageous to regulate the SOA gain such that the signal power remains constant at the output. This can readily be implemented through the use of a monitor photodiode and a tap coupler positioned at the amplifier output. However, particularly in the case of high gain SOAs, the amplified spontaneous emission (ASE) strength at the output of the SOA can be comparable with the signal strength, and consequently can compromise the control of the signal strength. This work reports on the design of an automatic power control system that compensates for ASE through knowledge of the SOA drive current.

2.

Background and Concept

Automatic power control (APC) can be implemented by measuring a small portion of the amplifier output with a photodiode and using negative feedback to hold the photodiode signal constant (see Fig. 1). However, in addition to the signal of interest, the amplifier emits amplified spontaneous emission noise (ASE). The monitor diode responds to this noise and consequently, the APC controls the sum of signal and noise powers. As a result, control of the signal power is compromised, particularly at low input signal strengths where significant gain is required.

Fig. 1

ASE characterization test bed.

To address this issue, an optical filter may be placed in line with the monitor photodiode to reject the ASE, allowing the APC to operate only in response to changes in the signal strength. However, this increases the cost and fixes the amplifier use to one particular wavelength division multiplex (WDM) channel. A tap and monitor photodiode at the amplifier input can give insight to the incoming signal strength and thus offer a means with which to correct for errors associated with the ASE strength. Again, this represents additional cost, and furthermore the SOA noise figure is increased by an amount corresponding to the tap coupler insertion loss.

This work reports on a means of addressing these limitations using a direct measurement of the SOA drive current in combination with the total output power to estimate the output signal strength. This does not incur additional cost, since the drive current measurement is already implemented as part of the current regulation.

3.

Methodology

The process relies on establishing a relationship between the total output power (signal and ASE) from the SOA and the output signal strength. This is not trivial, since the ASE level is a function of amplifier gain (G), and the amplifier gain depends on population inversion (drive current) and also the input signal strength. The relationship between gain and signal strength is approximated by⁹

The SOA was driven through all possible combinations of input signal strength (in this case, limited to $1550\mathrm{nm}$ ) and drive current and the output signal strength and the total power output recorded for each of these points. These data were used to calculate a 3-D map relating the “desired” signal strength to any given combination of total output power and amplifier drive current. This map was loaded into the microcontroller chip controlling the SOA drive current. The APC process uses the monitor diode power and drive current as pointers to the map to find the corresponding signal strength, and adjusts the drive current based on how that signal strength compares to the user’s setpoint. As a result, the APC responds to the signal alone, as if there were a filter in the power monitor tap. In principle, since there is no optical filter, the process wavelength is transparent. In practice, significant gain variation across the signal band of interest (gain variation can be around $6\mathrm{dB}$ across the c-band for a high gain SOA¹) will cause the amplifier to saturate at different input signal strengths for different wavelengths, and will restrict the range of operation before recalibration is necessary. Interpolation may be used to minimize the characterization required; however, this has not been addressed to date. Once the amplifier is characterized, the process requires only a microcontroller chip of reasonable power and memory capacity to implement the control algorithm. Generally, this could be accommodated with the chip used to perform the SOA current stabilization and temperature control. This work reports on a demonstration of this concept using a low-power-reduced instruction set (RISC) processor, the microchip PIC16E877.¹⁰

4.

Experiment

A test rig was established comprising a signal source, an SOA with monitor diode, and an optical spectrum analyzer (see Fig. 1). The signal source comprised a tunable laser, an erbium-doped fiber amplifier (EDFA); a tunable filter to remove ASE from the EDFA, and a variable attenuator. The signal was fed into an SOA controlled using a driver with an onboard monitor diode. The output of the SOA was recorded using the monitor photodiode as a function of input signal strength ( $-25$ to $+10\mathrm{dBm}$ ) and drive current. For each point, the “desired” or “target” signal strength was recorded using an optical spectrum analyzer (OSA). Thus a map that relates the target output signal strength to a measurement of the total output power at each bias current point was generated. An example of this is shown in Fig. 2.

Fig. 2

Output power map.

To implement that APC, the desired signal strength is input to the controller. This, through the map obtained before, defines the controller set point for any given drive current. The controller then derives an error signal from measurement of the total optical output power and the set point value, and adjusts the bias current until the set point and monitor power values agree. At this point, the signal power will be at the desired value.

5.

Results and Discussion

Results of early stage trials of the APC system are displayed in Fig. 3. The graph shows the difference between APC implemented without processing the output signal to compensate for ASE, and the case where ASE is accounted for through the procedure described before. Here, the effectiveness of the ASE-compensated APC is clearly visible. In all cases, a significant drop off from the target output signal strength is observed without ASE compensation, particularly at low input signal strength, where the ASE contribution becomes more significant.

Fig. 3

APC with ASE compensation performance.

In the case where the target signal power is $+5\mathrm{dBm}$ , the actual signal strength is $0.5\mathrm{dB}$ low when the input power falls below $-9\mathrm{dBm}$ and is $2.5\mathrm{dB}$ low at $-20\text{-}\mathrm{dBm}$ input power. In contrast, the ASE corrected signals are accurate to approximately $0.2\mathrm{dB}$ over an input power range, which extends to $-20\mathrm{dBm}$ . The APC does become nonlinear at lower input powers, but for most applications these would not be relevant.

6.

Summary

The output from an optical amplifier is a combination of the signal strength and the spontaneous emission. We present an automatic power control system that derives a representation of the output signal strength from a measurement of the total output power and the amplifier drive current. This enables an error signal to be derived that compensates for the influence of ASE and eliminates the need for wavelength selective filters and/or power monitors at the input of the amplifier, thus offering a route to lower implementation costs. APC, which is accurate to $0.2\mathrm{dB}$ , is demonstrated over a range of input powers down to $-20\mathrm{dBm}$ . Significant gain variation over the wavelength range of operation may introduce an offset error. This has as yet not been quantified experimentally. However, the controller can in principle compensate for this variation, provided the wavelength response of the amplifier is known and the operating wavelength is specified. We demonstrate APC with a precision of $0.2\mathrm{dB}$ with input signals ranging down to $-20\mathrm{dBm}$ using a low power microprocessor to implement the processing real time.