Professional Documents
Culture Documents
9790 - Poster - A. Levick
9790 - Poster - A. Levick
9790 - Poster - A. Levick
Andrew Levick1, Marin Dury1, Jane Ireland1, Emma Woolliams1 and Nigel Fox1
1
National Physical Laboratory, Hampton Road, Teddington, Middlesex, TW11 0LW, UK Corresponding e-mail address: Andrew.Levick@npl.co.uk Author email address 24/26pt
Abstract
We have used the Allan deviation method to measure the noise and drift in the power of a supercontinuum laser with and without power-stabilisation feedback control. The Allan deviation method clearly shows that the feedback control removes drift. The random uncertainty in the mean stabilized laser power, determined using the Allan deviation method, is <0.01 % for averaging times longer than 100 s.
Introduction
All measurement systems have a resolution limited by random fluctuations (noise) or by variations in parameters with time (drift). Noise and drift can take a variety of forms having different measurement times or frequency dependencies. Gaussian or white noise, flicker or 1/f noise, and random walk drift are examples commonly met in electronic measurement devices. The noise in a signal is usually quantified by computing the standard deviation; however this is a single value, and gives no information about the type and timescales of the noise and/ or drift. The mean and standard deviation are only meaningful quantities when white noise dominates. When there is significant drift, the mean and standard deviation are not relevant quantities and averaging does not reduce the random uncertainty.
The laser consists of three main sub-systems: a passively modelocked low power fibre laser at 1064 nm, a high power claddingpumped fibre amplifier and a high nonlinearity supercontinuum generator. The latter generates supercontinuum radiation in the range 450 nm 2200 nm. The laser generates up to 4 W of power over this wavelength range. Figure 3 is the spectrum of the supercontinuum laser (Fianium FemtoPower SC450) measured using Si and InGaAs array spectrometers. For power-stabilisation feedback control, the laser was actively power-stabilised using a feedback control from one of the filter radiometers (FR1) to the pumped fibre amplifier within the laser. The other filter radiometer (FR2) was used as an independent monitor. The signals from the two filter radiometers were continuously monitored and then analysed using the Allan deviation method.
Filter radiometer 2 (FR2) Detector, lter, housing
Results
Figure 4 shows the filter radiometer signals with the stabilisation feedback control applied. Neither the feedback radiometer signal, nor the independent monitor radiometer signal shows drift, demonstrating the absolute stability of the stabilisation. The signals have noise with standard deviation of 0.1 %. Figure 5 shows the relative Allan deviation plots for the signals in figure 4. Both plots have gradients approaching 0.5, indicating that white noise is dominant and drift is negligible. This means that averaging is appropriate and figure 5 shows that the random uncertainty in the mean signal value is <0.01 % for averaging times longer than 100 s. In contrast, figure 6 shows the filter radiometer signals without the stabilisation feedback control applied. Both signals show drift, which is correlated and thus associated with the laser. This drift is 0.9 % over time period of 5000 s in this case, and can be much larger especially when the laser is stabilising after start up. Figure 7 shows the Allan deviation plots for the signals in figure 6. Both plots have gradients >0, indicating drift and/or random walk. It would not be appropriate to average these signals.
Figure 4. Signal (% deviation) versus time for filter radiometers (FR1 left graph, FR2 right graph) with stabilisation.
where and is the average of a series of measurements y(t) over time , and is computed for: 1t ,2t, 4t, 8t.
The plot of log(y()) versus log() is the Allan deviation plot, and the slopes in the plot are interpreted as according to table 1. Noise or drift model Linear drift Random walk Flicker or drift Gaussian or white noise Slope 1 0.5 0 -0.5 Filter radiometer FR2 Filter radiometer FR2
Figure 6. Signal (% deviation) versus time for filter radiometers (FR1 left graph, FR2 right graph) without stabilisation.
Integrating sphere
Figure 2. Photograph of the experiment Figure 7. Relative Allan deviation plots for signals in figure 6.
Conclusion
We have used the Allan deviation method to measure the noise and drift in a supercontinuum laser with and without power-stabilisation feedback control. The Allan deviation plots indicate that the stabilisation clearly removes the drift. The Allan deviation gives more information than the standard deviation value alone.
Figure 3. Spectrum of supercontinuum laser measured using Si and InGaAs array spectrometers
2. J. T. Woodward, A. W. Smith, C. A. Jenkins, C. Lin, S.W. Brown and K. R. Lykke, Supercontinuum sources for metrology, Metrologia, 46, S277-S282, 2009.
www.npl.co.uk
1. D. W. Allan, Statistics of atomic frequency standards, Proc IEEE, 51, 221-30, 1966.
9774/0911
References