2.5 inversion

O ••S 2.0 inversion 1.0 30 10 20 0 -20 -10 Frequency Detuning GHz

Figure 4.7: The dependence o f the on/off ratio o f the converted data without optical filtering with detuning o f the input optical wavelength fo r two input powers.

■EilBB

75nW

0

a) b)

Figure 4.8: Optical inversion in the converted data by a) detuning and b) altering the optical power o f the 1.56 pm signal: (50ns/div).

Optical inversion was also obtainable by altering the input power while keeping the detuning constant, and Figure 4.8b shows the output signal relative to the zero level for input signal powers of -75 |iW and -60 jiW. Optical inversion here arises due to the change in FP mode position caused by the change in the level of gain saturation dispersion. It is also clear

Ch 4 Dependence of Wavelength Conversion on Input Signal Power and Wavelength.

from Figure 4.8b that the contrast ratio differs significantly for the two cases.

Unsaturated FP Mode

3.1

3.0

on/off ratio at output -|— I I I I

5 10 Frequency Detuning

Figure 4.9: The variation o f converted data pattern with detuning around the FP mode. The measured contrast ratio is recorded adjacent to the measured traces: (lOnsldiv).

The detuning dependence was investigated in more detail and Figure 4.9 shows the output modulation at different detunings relative to the unsaturated FP mode of the NLOA. The input power was 10 |iW and the input wavelength was 1.56 |im. A high resolution (13 pm) optical spectrum analyser measured the separation, Of, of the input signal wavelength for optimum contrast from the unsaturated FP mode as being Of = +5 to +10 GHz (or ~ -0.1 nm). In Figure 4.9 the output modulation is plotted for various positions on the unsaturated FP mode. These traces are not DC coupled and the on/off ratio at each detuning is given next to each trace. It is clear that absorptive saturation is dominating the switching behaviour with the optimum input signal at a shorter wavelength than the unsaturated FP mode. At a detuning of +5 GHz the output contrast ratio was similar to that at 8f = +10 GHz except the modulation component was much stronger. This is to be expected since the slope of the FP resonance may be greatest here.

Ch 4 Dependence of Wavelength Conversion on Input Signal Power and Wavelength.

For a detuning very near the FP mode resonance inverted operation occurs. This is also to be expected since the saturation of the absorber region will shift the FP to shorter wavelengths thereby reducing the signal gain and output for ‘I ’s compared to *0*s.

The modulation observed for a detuning 9f = 3 GHz is most interesting and shows the output having a series of peaks corresponding to the rising and falling edges of the data bits. One possible explanation of this is as follows. As the input power at 1.3 p,m increases at the beginning of a data bit the mode starts to move to shorter wavelengths. The output power at 1.56 |im starts to increase as the gain increases as the mode moves towards the signal wavelength. As the 1.3 jim signal power continues to increase the saturation pulls the FP mode past the signal wavelength and the 1.56 jum power starts to reduce as the gain reduces. When the

1.3 jim power reduces again, at the end of the data bit, then the mode relaxes back towards its unsaturated wavelength passing through the 1.56 |im signal again causing a second peak in output power. This trace combined with the measured detuning seems to confirm the absorption saturation as being the dominant cause for the wavelength conversion behaviour.

We have seen how the input power at 1.56 }xm affects the conversion and how important it is to ensure that the power and signal wavelength are adjusted to maximise the contrast ratio. The power at 1.3 |im also affects the wavelength conversion behaviour. Increasing the mean input power of the 1.3 |im data improved the obtainable contrast ratio but increasing it too far can result in switch on of the NLOA. Above a critical input power the 1.3 (j.m data signal is sufficient to induce lasing of the NLOA and the device no longer behaves as an amplifier for the 1.56 |im signal. This threshold power is lower if the device is operated with a higher gain current closer to threshold. With the device biased at a gain current Ig = 70.8 mA (I^ = 73 mA) an input power at 1.3 jim of P^ = 130 jiW was sufficient to induce lasing of the NLOA device at A. - 1.553 fim. The on/off contrast ratio without optical filtering was 10 dB but was critically dependent on Pj^. Figure 4.10 shows the variation of output on/off ratio with 1.3 }im input power, P^^. For P^ < -128 jiW the on/off ratio is very low (-3:1) and is only the modulation in spontaneous emission from the device and an optical filter would reduce this significantly. For -128 jiW < P^^ < -136 \xW the output was laser emission but was unstable

Ch 4 Dependence of Wavelength Conversion on Input Signal Power and Wavelength.

with large amplitude variations in the data particularly between single 'I's (patterning) and long sequences of 'I's. Only when P-^> -136 |iW was the output data pattern stable with little amplitude modulation of the '1' level, see Figures 4.11 a, 4.1 lb. When stable the output contrast ratio was -10 dB but the lasing emission was multi-line and would cause significant chromatic dispersion problems if used in systems using standard fibre. From Fig 4.10 we observe that a change in input power of 10% is sufficient to cause a large change in output contrast.

a X o 12 10 unstable 8 6 on 4 2 0.12 0.13 0.14 0.15 Input Power mW

Figure 4.10: Variation o f lasing on!off ratio with input pow er at 1.3 pm.

0 VZR B. 14jl J jsemU