The optical carrier at the output of a semiconductor laser exhibits fluctuations in intensity, phase, and frequency even when biased with a constant current. The two fundamental noise mechanisms are spontaneous emission and electron-hole recombination (shot noise), although the former is dominant. Each spontaneously emitted photon adds to the coherent field generated by stimulated emission as a small field component with random phase. The spontaneous emission rate is very high (approximately 10¹² s⁻¹), causing the intensity and phase of the emitted light to fluctuate on a timescale as short as 100 ps. Phase fluctuations lead to a finite laser linewidth, while intensity fluctuations lead to a limited signal-to-noise ratio (SNR).
This reduction in OSNR, and the consequent increase in BER, can significantly limit the transmission of both analog and digital signals. Therefore, to ensure the proper functioning of an optical communication link, an accurate characterization of the laser intensity noise is necessary to compare its power level with the signal and optimize performance. In this article, we will discuss the basic principles of laser intensity noise and present methods for its measurement and characterization.
Relative intensity noise.
Relative intensity noise (RIN) describes the contribution of laser intensity fluctuations to the receiver's electrical noise relative to the electrical signal power. Generally, RIN is normalized to a bandwidth of 1 Hz so that it can be made independent of the receiver's bandwidth. Although it is optical noise (fluctuations in the laser's optical power), it is defined in terms of electrical power at the photodetector output. More specifically, it is calculated as:
where N is the electrical noise power, Pf is the received photocurrent power, and B is the noise bandwidth. The corresponding RIN value in dB/Hz is then calculated by applying 10log10(·).
Figure 1 shows the typical RIN noise spectrum of an InGaAsP laser at 1550 nm for different optical power values.
As can be seen in this figure, it is not white noise, but rather exhibits a high-intensity peak at a frequency close to the laser relaxation oscillation frequency. Additionally, it decreases rapidly for frequencies higher than the relaxation oscillation frequency, acting as a bandpass filter for fluctuations due to spontaneous emission.
Finally, the RIN decreases as P-3 for low powers, becoming P-1 for high powers.
Based on the above definition of RIN noise, its measurement requires an unmodulated laser, or a laboratory instrument capable of measuring both powers in the presence of a specific modulation pattern. For this reason, the IEEE 802.3ae standard (“10Gb/s Ethernet Task Force”) defines the RIN OMA as the ratio between the average electrical power of noise normalized to 1 Hz and the average electrical power of a square wave:
The difference between the two definitions of RIN noise is reflected in Figure 2. If the average optical power is maintained and the extinction ratio is very high, then both definitions provide virtually the same result.
RIN Noise Measurement.
Several techniques can be used to measure the RIN noise of lasers. As a first example, one of them is based on the use of a digital oscilloscope. The device under test (DUT) is fed with a square wave signal from a modulation pattern generator. To measure the RIN, the DUT output is connected directly to an oscilloscope with optical input, or to a photodetector and oscilloscope with electrical input. From the captured waveform (Figure 3), the OMA RIN noise level can be calculated as:
'
where N1 and N0 are the optical noise powers of levels “1” and “0”, respectively, and ModAmp refers to the modulation amplitude. Optionally, the measurement can be performed by including a variable reflector in the laboratory setup, thus modeling the worst-case scenario. It should be noted that the RIN noise level of lasers worsens in the presence of external reflections towards the DUT.
From the results in Figure 3, an OMA RIN of -114.8 dB/Hz is obtained. The RIN value can be calculated directly from the values of level “1” of the square wave signal. The dynamic range of the measurement depends on the DUT's output power, the type of O/E conversion, and the level of internal electrical noise (primarily thermal). For example, for an Agilent 86116C 43 Gbit/s receiver module (BN = 33 GHz), characterized by an equivalent noise power of -10 dBm and a maximum average optical input power level of 10 dBm, there is a dynamic measurement range of RIN of -143 to -119 dB/Hz.
Another method for measuring RIN noise uses an electrical spectrum analyzer. The measurement involves analyzing the RF spectrum, estimating the noise and signal levels (CW) to isolate the contributions of the different noise components (thermal, shot, and RIN). For example, the internal thermal noise of the measuring equipment can be estimated by canceling the input optical signal, while shot noise can be calculated from the average optical power. Figure 4 shows an example measurement using the Agilent 71400C/71401C instrument.
Finally, RIN noise can also be estimated from measurements in the optical domain. For example, the BOSA instrument from Aragón Photonics allows for RIN noise measurements of DFB lasers by analyzing their optical spectrum. Additionally, this instrument allows for the measurement of other parameters such as the line broadening factor or the peak frequency of the relaxation oscillations. The laboratory setup is as shown in Figure 5. As before, the DUT is simply connected to the input of the BOSA, and the trace of the optical spectrum is captured. Subsequently, an ideal Lorentzian curve is fitted to the laser's emission peak and subtracted from this trace, obtaining RIN noise spectrum measurements as shown in Figure 6. In this specific case, the influence of the laser's polarization current on RIN noise levels can be observed.
Techniques for Compensating for RIN Noise.
Laser RIN noise limits the quality of optical communication systems. Depending on the application, RIN values on the order of -160 to
-130 dB/Hz are required. Thanks to the laboratory instrumentation presented earlier, the system's RIN noise levels can be accurately estimated, even in the presence of reflections. If these values do not meet our specifications, then a technique must be employed to reduce its effects. Among the best-known is the use of a balanced photoreceptor (Figure 7). This is an optical receiver in which two photodetectors with identical characteristics are integrated and which operate using a 
differential scheme. That is, the outputs of both photodetectors are differentially amplified in order to double the signal level and compensate as much as possible for the noise level. In addition to RIN noise, it also allows for the neutralization of shot noise. With this scheme, RIN noise reductions of up to 20 dB can be achieved.
