deletions | additions       diff --git a/section_Experimental_results_We_configure__.tex b/section_Experimental_results_We_configure__.tex  index 416e2c4..60717d4 100644  --- a/section_Experimental_results_We_configure__.tex  +++ b/section_Experimental_results_We_configure__.tex 
...
We tune our CW source at $λ_s$, attenuated to a count rate of about $3$ kcps per temporal bin ($166$ ps), and by recording, separately, the depletion of the signal and the gain of the idler (figure 4(c) !!), we measure the conversion efficiency $\eta$ as the total pump power varies, while keeping the two pumps balanced and the polarization aligned, as shown in figure \ref{fig:vs_power}.  For $P_1 = P_2 ~ 5$ W (average total pump power $25$ mW), our setup shows conversion efficiency of $99.4$ \% over a $0.6$ ns window ($\vert\alpha\vert^2 \simeq 0.15$ at input of the nonlinear fiber). For larger values of $P$, $\eta$ peaks and then decreases, but conversion deviates from the expected sinusoidal model because higher(cascaded) order frequency conversion start taking place. In figure \ref{fig:vs_power}(b) we show the measurement taken using the DCM to spectrally separate the fields, showing the evolution of the second order signal $\omega_{sII} = \omega_s - \Delta\omega$ and idler $\omega_{iII} = \omega_i + \Delta\omega$.  We tune the single photon source and filters to heralded the presence of photons at $λ_s$ into the BS setup. The set the heralding rate to $~200$ KHz for a detected pair rate of $~6000$ cps. Pumps can indeed be triggered to be generated on the event of a heralded photons, but the conversion efficiency is technically limited by the amplitude fluctuations introduced by the EDFA due to the random time between each pulse. To overcome this limitation, we let the system run independently in condition similar to the classical measurement, and we record the three-fold coincidences between signal, herald and pump: in this condition the detected pair rate is $~20$ pairs/s in a 0.6-nanosecond $0.8$-nanosecond  window. We show the results figure \ref{fig:vs_power}, where we observe a conversion efficiency of ?. Unitary conversion efficiency is limited by the fact that single photon bandwidth ( $\delta \lambda =0.57$ =0.60$  nm FWHM) and the acceptance ( $\delta \lambda_{BS} = 1.17(2)$ nm FWHM) 0.47(2)$ nm)  are comparable: for these values  theoretical calculation calculations  show a maximum efficiency $\eta$ of about $89\%$. To verify that the frequency translation indeed preserve the quantum statistic on the translated output, we measure the second order time correlation $g^{(2)}(0)$ of the output state. We inject the single photon field into a 50/50 fiber beamsplitter, and record heralded singles detection rates $A(0)$ and $B(0)$, and heralded coincidence rate $C(0)$ from the two coupler outputs. Normalization of the $g^{(2)}(0)=C(0)/G^{(2)}(0)$ is obtained from the product of the singles $G^{(2)}(0)=A(0)B(0)$.% Error bars are extracted from statistical error on the count events, and dark counts event are retained.