Look-Up Table Generation

The aim of the calibration process is to obtain a LUT in which stable wavelengths and their corresponding current triplets are stored. How this LUT is generated is described in the next section.

At the beginning of the calibration for each of the three phase currents, a full char-acterization has been done, yielding to maps of wavelength, power, SMSR, etalon and reference signals. To overcome hysteresis effects, each measurement has been carried out three times, two times with the positive slope of the tuning vector and one time with the negative slope. This measurement data sets are fed to the processing software (red arrow) which flow is illustrated in figure 3.19. The generation program is divided into eight independent functions, after each step the LUT evolves in precision and stability.

These functions are described in more detail in the next paragraphs.

Transform

Figure 3.19: Flowchart for LUT generation for a single phase current of the laser. The different functions are described in the next chapter.

Mode-Jump Filter As it can be seen in the wavelength map of figure 3.17 on the left, the wavelengths are concentrated on island divided by a common border. The islands across a wavelength band (in diagonal direction from bottom left to top right) have nearly the same wavelength value. When a supermode hop occurs the wavelength band is changed, resulting in a high variation of the output wavelength. The wavelengths located at the border of the wavelength islands are not as stable as the wavelengths located in the middle of a wavelength island. For that a special filter has been designed to remove these points and their corresponding current triplets.

For this filter a kernel based algorithm is applied to the wavelength map. The kernel consists of nine pixels and has a quadratic shape. An illustration of the filter kernel is given in figure 3.20. The center pixel’s wavelength is compared with its eight adjacent pixels by calculating the absolute wavelength difference. If a pixel is identified having a difference bigger than 100 pm, the center pixel is declared as unstable and is then

removed from the map. By applying the filter to all pixels of the map the border regions are deleted.

Figure 3.20: Left: Filter kernel with eight adjacent pixels surrounding a center pixel.

Right: Applying the mode-jump filter kernel to a wavelength, schematic view.

The results when applying the mode jump filter can be seen in figure 3.21. On the left side the initial wavelength map is shown for which the mode jump filter kernel is applied, resulting in a wavelength map shown in the right of figure 3.21.

Figure 3.21: Left: Initial wavelength map. Right: New wavelength map with removed unstable wavelengths on the border between wavelength islands.

Masking Filter By investigation of the wavelength maps it is obvious that the same wavelength is generated by multiple combinations of the current vector. In fact, complete wavelength bands have repeating patterns which do not contribute to a stable and detailed LUT. If the laser currents are changed by a high amount the internal processes of the MG-Y structure is disturbed resulting in wavelength fluctuations. In addition the laser’s temperature controller must be able to deal with such high fluctuations. For better stability strong current slopes shall be avoided. In order to delete repeating patterns in the wavelength map, a masking filter is applied to the map. This is illustrated in figure 3.22 by the red shadowed area in the left picture. In the right picture the repeating pattern is deleted resulting in a LUT which allows a more smooth tuning of the laser.

Figure 3.22: Left: Wavelength map with repeating patterns (marked with red and black color). Right: Repeating patterns are removed.

Wavelength Variation Filter This function is applied to the wavelength map to delete points which are identified to be affected by hysteresis effects. As mentioned before, the maps are measured with the positive and negative slope of the tuning vector. By com-paring these maps, points with identical current triplets but with a wavelength difference higher than 10pm are declared as hysteresis affected and are deleted. The remaining points have almost no significant hysteresis.

Power and SMSR Threshold Filter As a last data consolidating step, the remaining wavelength maps are checked point by point for their power level and SMSR. Too low power and a poor SMSR will negatively affect the accuracy of the measurement system.

Therefore points with power lower than 7.5 dBm and SMSR lower than 25 dB are deleted.

In figure 3.23 the power (left side) and the SMSR (right side) measured by the wavelength meter are illustrated. In addition also the threshold levels are indicated by dashed lines.

Figure 3.23: Left: Lasers output power for different sample points with indicated thresh-old level of 7.5 dBm. Right: Side-Mode Suppression Ratio (SMSR) also with indicated threshold level of 25 dB.

Wavelength Map to LUT Transformation After all fit filters have been applied to the measured wavelength maps, the transformation into a LUT for laser control holding

the current vector can be done. In addition also the etalon and reference photodiode values will be stored in the LUT for later laser stabilization. The wavelength values are sorted in ascending order. The control wavelength triplet as defined by equation 3.2 as well as the etalon and reference diode voltage are located within the same row of the corresponding MG-Y laser wavelength. An excerpt of the LUT for a phase voltage of 0.72 V is shown in table 3.3.

Wavelength [nm] RR[V] LR[V] PH[V] Etalon[V] Ref[V]

1526.039 1.809494 0.493656 0.72 0.766602 1.77887

1526.072 1.780185 0.493656 0.72 0.508881 1.782227

1526.099 1.751494 0.493656 0.72 0.390472 1.782227

1526.463 1.723397 0.425902 0.72 0.488129 1.777191

Table 3.3: Excerpt of a LUT for a constant phase value of 0.72V.

After generating the LUT and caring out the first wavelength scan, results as given in figure 3.24 on the left are obtained. The laser’s wavelength bands are indicated by the triangular like shape of the red and blue curve. The high fluctuations occurs because on the border region of two different wavelength bands the wavelength overlaps (see red marked area in figure 3.25 on the left). As mentioned before, such high amount of current change is not good for the system stability and has to be avoided. To overcome this problem, a polynomial curve is fitted onto each of the triangular like shaped wavelength bands. Afterwards each right- and left reflector value which differs for more than 200 mV from the fitted value is deleted. The same measurement is carried out with the fitted and cleaned LUT as can be seen in figure 3.24 on the right resulting in a more smooth tuning.

Another option would be the interpolation between entries instead of deleting elements.

Figure 3.24: Left: Spectral scan with high fluctuations in reflector values occur due to simple stacking of values. Right: Smooth LUT where values with a certain distance with respect to fitted values are deleted.

In figure 3.25 on the left side the initial wavelength map corresponding to the non-smoothed LUT is given. The high fluctuations occur at the boundaries indicated by the red circle. The picture on the right shows the LUT where the high fluctuating areas are deleted.

Figure 3.25: Left: Wavelength map before smoothing. Right: Smooth wavelength map with deleted boundary points which cause high fluctuations in reflector values.