DIARIO DE CAMPO N° 05 INSTITUCION EDUCATIVA: ”33080”

The most direct way of modulating the output of a laser is to vary the strength of the laser excitation. This is useful in the case of semiconductor lasers because the light output is directly proportional to the injection current over a large current range, hence the conversion of the current modulation to light modulation is also linear. Very high repetition rates can be achieved since the photon lifetime within the laser cavity is on the order of picoseconds. Following analysis of the rate equations pertinent to semiconductor lasers the relaxation oscillation resonance can be determinedl03_ The photon number within the laser responds to a step-like current excitation by exhibiting damped relaxation oscillations, or ‘sphdng’ at an angular frequency,

where y = ^

- ( 2)^

'^sp

where Tgp is the spontaneous emission lifetime, is the cavity (or photon) lifetime, and CÙQ is the relaxation oscillation frequency. For continuous excitation, the presence of the relaxation oscillation resonance affects the modulation depth of the output laser signal. The modulation depth remains constant for frequencies well below the resonance, increases significantly in the vicinity of the resonance, then falls markedly at frequencies above the resonance. The resonant enhancement of modulation index can be ~10 times that of low frequency modulation indices, although electrical parasitics lead to a flatter small-signal

modulation response and a smaller enhancement at resonancel04^ The relaxation oscillation fr equency of typical semiconductor lasers is normally several gigahertz, and can be increased, for example, by reducing the photon lifetime by decreasing the laser length or by frustrating the facet reflectivity. Modulation at frequencies of up to 66GHz has been previously demonstrated experimentally!®^ although only small modulation deptlis were obtained.

By applying a strong sinusoidal or short cuiTent pulse to the semiconductor laser the first ‘spike’ of the relaxation oscillation can be isolated!®*». This behaviour in semiconductor lasers is

termed gain-switching and can generate pulses shorter than other laser systems due to the smaller time constants involved. In fact the output pulses can be significantly shorter than the applied drive waveform. The operating principle is considered in figure 3.5. The excitation pulse increases the gain above threshold where the laser turns on and the light builds up in the cavity.

The rapidly increasing light signal can saturate the gain through stimulated emission thus turning

the laser off. Depending on the details of the excitation waveform (and hence the gain) the output can consist of multiple pulses due to gain recovery.

Unsaturated

excitation pulse _G

I Threshold gain light intensity Time Time

Figure 3.5 Pulse generation through gain switching.

In the case of single pulse outputs a rough estimate of the pulse duration can be made from the usual rate equations. Perturbation of the steady state laser emission^®’^ shows that the ringing period is,

At~(x<-i:sp/Ti)i/2

where T| is the fractional excess gain above threshold. Since the width of the pulse in the isolated

spike regime is much less than one period, then At is an overestimate of the pulse duration. For the case of Tl~l, Tc=3ps, and TQ=lns, then At ~60ps.

Another theory^ll® predicts that.

1/2

where fj^ is the modulation frequency (which can be replaced by the inverse duration of the excitation pulse). From this a value of At of around 50-60ps is found for a modulation frequency of IGHz. Therefore both these approaches imply that the resultant gain-switched pulse duration

should be of the order of a few tens of picoseconds.

Gain switching has the advantage of short pulse generation without any other optical components, and experimentally gain-switched pulses of ~20ps duration have been obtained, at multi-gigahertz repetition rates^®*. Analysis has indicated that for an inversion ratio, r, of 2 the pulsewidth is -5 times the photon lifetime (tc-2-3ps), but in practice typical lasers can only be driven to r~l,5 resulting in pulse durations in the range 20-30ps. Short cavity lasers can further

reduce the duration of the gain-switched pulses due to a decrease in the photon lifetime. Because

the build-up starts from spontaneous emission noise, many axial modes are excited with random phases, and the output is essentially a noise burst. Therefore gain-switched pulses suffer from lai'ge amounts of amplitude and timing jitter, and the pulse duration and wavelength are normally

ill-defined. In principle, relaxation oscillations can be a single mode event, and this situation can be promoted by injection seeding the pulsing laser^®^, or alternatively by using a laser with

inherent mode control such as a distributed feedback (DFB) laserf f®.

Gain-switched laser pulses have a substantial frequency chirp across the pulse envelope originating through gain saturation produced by the strong optical field. Transmission through long lengths of dispersive optical fibre has been used to dechirp and reduce the duration of gain- switched laser p u lses^. Such techniques tend to defeat the basic simplicity of these lasers and

so the shorter, cleaner pulses produced by modelocking become a more attractive option. In the following section the active modelocking of semiconductor lasers will be described with particular emphasis being placed on quaternary InGaAsP devices that operate at wavelengths around 1.5p.m.