Pulsed, high-power LED illumination for tomographic particle image velocimetry

Abstract

This paper investigates the use of high-power light-emitting diode (LED) illumination for tomographic particle image velocimetry (PIV) as an alternative to traditional laser-based illumination. Modern solid-state LED devices can provide averaged radiant power in excess of 10 W and by operating the LED with short high current pulses theoretical pulse energies up to several tens of mJ can be achieved. In the present work, a custom-built drive circuit is used to drive a Luminus PT-120 high-power LED at pulsed currents of up to 150 A and 1 μs duration. Volumetric illumination is achieved by directly projecting the LED into the flow to produce a measurement volume of ≈3–4 times the size of the LED die. The feasibility of the volumetric LED illumination is assessed by performing tomographic PIV of homogenous, grid-generated turbulence. Two types of LEDs are investigated, and the results are compared with measurements of the same flow using pulsed Nd:YAG laser illumination and DNS data of homogeneous isotropic turbulence. The quality of the results is similar for both investigated LEDs with no significant difference between the LED and Nd:YAG illumination. Compared with the DNS, some differences are observed in the power spectra and the probability distributions of the fluctuating velocity and velocity gradients. These differences are attributed to the limited spatial resolution of the experiments and noise introduced during the tomographic reconstruction (i.e. ghost particles). The uncertainty in the velocity measurements associated with the LED illumination is estimated to approximately 0.2–0.3 pixel for both LEDs, which compares favourably with similar tomographic PIV measurements of turbulent flows. In conclusion, the proposed high-power, pulsed LED volume illumination provides accurate and reliable tomographic PIV measurements in water and presents a promising technique for flow diagnostics and velocimetry.

Appendix

The intensity I of the recorded particle images shown in Figs. 7 and 16 can be expressed as follows

$$ I = I_n + I_f + I_p $$

(3)

where I _n is the noise level of the CCD sensor, I _f the intensity contribution due to flair, reflections and out-of-focus particles and I _p the particle image intensity. The noise level I _n can be determined by recording a black image or from the camera specifications. The background intensity I _b  = I _n  + I _f is assumed to follow a normal distribution and is obtained by fitting a Gaussian model to the left-hand side of the intensity pdf. The particle image contribution is then obtained as I _p  = I − I _b .

Fig. 16

Pdf of the image intensity I (dot dashed line) and its component I _p  (solid line) and I _b (dashed line)

From Eq. 3, the following signal-to-noise ratios can be defined.

$$ \hbox{SNR}_i = \frac{\langle I_p \rangle}{\langle I_b \rangle}; \qquad \hbox{SNR}_s = \frac{\langle I_p \rangle}{I_n} $$

(4)

where $\langle \rangle$ denotes mean intensities. As the strength of the illumination increases both I _f and I _p increase proportionally and SNR _i remains nearly constant. The definition of SNR _i is equivalent to that of image contrast in Adrian and Westerweel (2010), which is independent of the illumination and promotional to the inverse of the seeding density. For the present case, SNR _i ≈ 2 irrespective of pulse width or drive current. The definition of SNR _s is more suitable to characterise the effect of the illumination quality; however, it ignores the contribution of the background intensity. By combining the two definitions, the signal-to-noise ratio expressed as the difference between the particle and background intensity over the noise level is defined as

$$ \hbox{SNR} = \hbox{SNR}_i \left( 1 - \frac{1}{\hbox{SNR}_s} \right) $$

(5)

Note, in the extreme where the particle image intensity tends to zero (i.e. low light conditions), the SNR also approaches zero.