Abstract

We present the improved structure and operating principle of a spectrophotometric mean that allows us for the recording of the transmittance of a thin-film filter over an ultra-wide range of optical densities (from 0 to 11) between 400 and 1000 nm. The operation of this apparatus is based on the combined use of a high power supercontinuum laser source, a tunable volume hologram filter, a standard monochromator and a scientific grade CCD camera. The experimentally recorded noise floor is in good accordance with the optical density values given by the theoretical approach. A demonstration of the sensitivity gain provided by this new set-up with respect to standard spectrophotometric means is performed via the characterization of various types of filters (band-pass, long-pass, short-pass, and notch).

© 2015 Optical Society of America

1. Introduction

Since the end of the 90’s, the combined use of powerful design software and reliable and stable energetic deposition processes has paved the way towards the manufacture of complex optical thin-film filters, simultaneously characterized by close unity transmittance in their pass-band and high rejection levels in their stop-band (optical density OD > 6). Such figures are made possible by the use of a large number of layers, typically between one and two hundred.

In some cases, the values of these rejection levels are only a consequence of the stack structure defined to fulfill the spectral transmittance specifications, and accordingly, they are not directly related to a stringent requirement defined by the end-user. The performances of standard spectrophotometers, such as the Perkin-Elmer Lambda 1050 or the Agilent Cary 7000, are sufficient in this case to verify whether the blocking performance of an optical filter is better than a minimum reference level, e.g. 4 OD. However sometimes, this rejection level is a key requirement for final application and thus an accurate measurement of its value over a large spectral range is critical to assess the quality of the manufactured filtering function. This is especially the case for notch or edge filters used in enabling applications such as fluorescence microscopy [1], or Raman spectroscopy [2], but in this case, the same standard spectrophotometers provide only noise floor limited data and/or imprecise results, as illustrated in Fig. 1.

 figure: Fig. 1

Fig. 1 Examples of rejection level measurements provided by optical thin-film filters manufacturers [left: long-pass filter from THORLABS, reference FELH0750; right: short-pass filter from SEMROCK, reference FF01-680-SP-25].

Download Full Size | PPT Slide | PDF

The objective of our work was then to define an instrument allowing us to measure, with good accuracy (typically 1%) and an ultra-wide OD range (between 0 and 12), the spectral transmittance of a thin-film filter from 400 nm to 1000 nm. This broadband feature is a clear difference between our approach and that of previous works published during the last twenty years [3–6], for which measurements were achieved over a wide range of optical densities (typically 0 to 12) but only at some discrete wavelengths corresponding to standard laser lines (for instance, 633 nm or 1064 nm). Indeed, this broadband constraint deters us from using heterodyne detection, despite the fact that this coherent scheme provides a large reduction in dynamic measurement because of the proportionality between the amplitude of the photocurrent beat signal and the square root of the filter transmittance [3]. In a recent publication [7], we proposed a new method of transmission measurement using a scientific grade CCD camera and a set of optical densities as a wide range perfect integration detector and demonstrated that this method allows for the fulfillment of our twelve OD range requirements. However, a spectral crosstalk phenomenon limited the lowest measurable OD value to 8 in the case of a rapid spectral change in thin-film filter transmittance.

In this paper, we will analyze the intrinsic accuracy of our method, explain the OD range limitation induced by spectral crosstalk phenomena, show how the implementation of an additional filtering stage allows for overcoming this limitation, and give some examples of rejection levels measurements in the OD range between 0 and 11 achieved on various filtering functions (band-pass, long-pass, short-pass and notch).

2. Description of the set-up

A schematic representation of our new set-up is given in Fig. 2. The broadband emission of a high power supercontinuum laser source from NKT Photonics (Reference EXB-6) is launched into a volume hologram filtering device from Photon etc (Reference LLTF VIS-2), which delivers a 1 mW free-space single-mode beam with a central wavelength that is tunable between 400 and 1000 nm and a FWHM bandwidth of approximately 2 nm.

 figure: Fig. 2

Fig. 2 Schematic representation of the thin-film filter rejection levels measurement set-up (NKT EXB-6: supercontinuum laser source; LLTF: tunable volume hologram filter; OSD: order sorting device; SH: shutter; BS: beam splitter; ODF1: optical density flip 1; ODF2: optical density flip 2; ODFW: optical density filter wheel; ORC: output reflective collimator; FOL1: fiber optic link 1; TRC: transmitter reflective collimator; TFF: thin-film filter; RRC: receiving reflective collimator; FOL2: fiber optic link 2; SR-193i-B1: motorized Czerny-Turner monochromator; FOL3: fiber optic link 3; FOC: fiber optic coupler; PIXIS 1024B: scientific grade CCD camera; OD_R: reference optical density; RC_R: reference reflective collimator; FOL_R: reference fiber optic link; FEMTO OE-200-SI: variable gain photoreceiver).

Download Full Size | PPT Slide | PDF

However, at the output of the LLTF, the filtered light source exhibits, in addition to the 2 nm bandwidth line centered at λ0, two types of harmonics, respectively located at 2λ0 and λ0/2. The first one is due to residual non-linearities of the refractive index modulation in the volume of the hologram, whereas the second one is simply the second order of the filtering function. Figure 3 shows two examples of such harmonics, the first one when the LLTF is tuned at 450 nm (non-linear harmonics centered at 900 nm with an intensity of approximately 104 of the fundamental) and the second one when the same LLTF is centered at 950 nm (second order resonance at approximately 475 nm with an intensity close to 1% of the corresponding fundamental). An order sorting device (OSD) is thus placed in front of the beam splitter to remove these parasitic peaks. It consisted of two edge filters (short pass FESH0700 and long pass FELH0600, both from Thorlabs) mounted on a bistable device and whose position is switched following the value of the central wavelength (short pass for wavelengths lower than 650 nm and long pass for wavelengths greater than 650 nm).

 figure: Fig. 3

Fig. 3 Normalized spectral power density of the filtered source at the output of the LLTF (blue line, central wavelength λ0 = 450 nm; red line, central wavelength λ0 = 950 nm).

Download Full Size | PPT Slide | PDF

Then, this so prepared light beam passes through an electromechanical shutter (SH) whose purpose will be explained in Section 3. A non-polarizing beam splitter (BS) reflects a small portion (typically 10%) of the incoming beam toward a reference channel, where it is focused by a parabolic off-axis reflective collimator (RC_R) into the 2a = 50 µm diameter core of a step index all silica fiber (FOL_R, reference fiber optic link). The output end of this reference fiber is connected to the FC receptacle of a FEMTO OE-200-SI variable gain photoreceiver, equipped with a silicon PIN detector of 1.2 mm in diameter. The conversion gain is remotely switchable from 1×103 to 1×1011 V/W, and the output voltage is digitized by a 16-bit National Instruments USB-6211 acquisition module.

The main beam transmitted by the non-polarizing splitter passes through three stages of reflective optical densities. The first two (ODF1 and ODF2, each with an optical density of approximately 3) are installed in two-position, high-speed 0–90° flip mounts, whereas the last (ODFW) is mounted on a 6-slot motorized fast change filter wheel (slot1: no optical density; slot 2: OD 1; slot 3: OD 2; slot 4: OD 3; slots 5 and 6 not used). All these mechanical devices are remotely operated, which permits us to modify, at any time, the value of the optical density present in the measurement arm between zero and nine in steps of one. The attenuated beam is then focused by a silver coated, off-axis, aspheric mirror (ORC, output reflective collimator) into the 2a = 50 µm diameter core of a step index all silica fiber (FOL1, fiber optic link 1). The output end of this FOL1 fiber is connected to the object focal plane of an aspheric mirror (TRC, transmitter reflective collimator) that delivers a collimated, aberration-free beam passing through the thin-film filter (TFF) whose rejection levels have to be measured. The beam transmitted by this filter is focused by the same type of optical element (RRC, receiving reflective collimator) onto the input extremity of a b = 70 µm square core all silica fiber from Ceramoptec (FOL2, fiber optic link 2).

The output end of this square core fiber is connected to the entrance port of a Czerny-Turner monochromator from ANDOR (reference Shamrock SR-193i-B1). A ruled grating with 300 lines per millimeter is located on a motorized turret whose angular position can be remotely controlled through a USB 2.0 link. A last fiber optic link (FOL3, all silica step index fiber, 200 µm core diameter) is used to connect the exit port of this monochromator to the 13×13 µm2 photodiode array of a Princeton Instruments PIXIS:1024B camera, a low-noise, back-illuminated, thermoelectrically cooled, 16-bit scientific-grade CCD imaging system. This connection is achieved through a two-lens fiber optic coupler (FOC), which forms an image of the FOL3 extremity on the photodiode array plane with a lateral magnification γ of approximately 2.3.

3. Measurement method

This section will be devoted to a detailed description of the measurement principle and to an analysis of the main factors that could degrade the accuracy of the determination of these rejection levels over a wide spectral range.

3.1. Operating principle

To achieve the measurement of the transmittance TF(λ) of a thin-film filter at a wavelength λ we chose [7] to replace the direct comparison between the powers of the incident and transmitted light with the locking of the signal provided by the CCD camera to a high, constant level equal to 70% ± 10% of its full well capacity (FWC).

This is achieved through the combined adjustment of two parameters, i.e.

  • the opening duration τλ of the electromechanical shutter SH (in the closed position by default)
  • the effective value OD(λ) of the optical density present in the measurement arm.

This adjustment aims to reach the smallest opening duration τλ, the floor value being here defined by the timing specifications of the shutter (40 ms). The OD changes being discrete (steps of one) means that we have to first define the value of the optical density to be used for the measurement, and then finalize the locking of the effective signal provided by the camera through a fine adjustment of this shutter opening time. However, even if 1 hour of integration time is allowed by the ultra-low dark current level of the camera when its photodiode array is cooled to −70°C, we decided to limit this opening time to 100 seconds to ensure a reasonable duration for a single acquisition.

Finally, let us stress that this feedback approach permits the cancellation of any nonlinearity problem over a wide operating range.

3.2. Data acquisition and processing

Figure 4 shows the time chart followed to achieve the measurement of the light flux transmitted by the sample to be tested.

 figure: Fig. 4

Fig. 4 Time-chart of the light flux acquisition.

Download Full Size | PPT Slide | PDF

Once the wavelength λ is chosen and the set of optical densities inserted in the measurement channel is defined [OD(λ) being their resulting value], the supervision PC begins the acquisition sequence by opening the internal shutter of the CCD camera. Then, it starts the digitizing of the voltage level provided by the photoreceiver (sampling frequency f = 100 kHz) and waits for 100 ms before opening the SH external shutter. After τλ seconds, the SH external shutter is closed and the same time delay (100 ms) is inserted before stopping the digitizing of the photoreceiver data. Then, the internal shutter of the camera is also closed, and the digitizing of the CCD video signal is performed on 16 bits at a 2 MHz clock frequency. Finally, the digitized frame is transferred to the memory of the supervision PC.

At the end of an acquisition, two result files are thus available: the first one is an image file including a magnified view of the extremity of the FOL3 fiber (see Fig. 5(a) whereas the second one is a 1D array representing the time evolution of the amplified current delivered by the photoreceiver (see Fig. 5(b).

 figure: Fig. 5

Fig. 5 Data files obtained at the end of an acquisition: (a) Camera image file, including, near the center, a magnified view of the extremity of the FOL3; (b) Photoreceiver data file.

Download Full Size | PPT Slide | PDF

The method used to process the first array is as follows. One defines a circle, with a radius of approximately 200 pixels, centered on the image of the fiber and surrounded by a square with sides of 400 pixels, as represented in Fig. 5(a). The circle radius is chosen to be much larger than that of the fiber image (which does not exceed one hundred pixels, whatever the fiber type used as FOL3), to relax the constraints on the focusing of the fiber image. One begins by computing the mean value of the digitized signals of all pixels located together inside this square and outside this circle, i.e. in the white area represented in Fig. 5(a); this result corresponds to the mean value DC of the CCD dark signal for this video frame. Then, one computes the sum of the digitized signals of all pixels located inside the circle after subtracting this dark level Dc. Note that this result, called Sc, is obtained through the spatial integration of the number of photo-electrons created in each pixel located inside the summation circle. It is thus directly proportional to the total number of photons transmitted by the sample plus the OD set during the opening duration τλ of the external shutter SH.

To process the second data file, one begins by computing the mean value of the photoreceiver digitized voltage over the first and last 20 ms of the recording. In accordance with the time-chart of the acquisition presented in Fig. 4, the result corresponds to the mean value of the photoreceiver dark current. Then, one subtracts this dark current from all the elements of the data file and computes their sum, henceforth called SR. This result is obtained through a time integration of the number of photo-electrons created in the sensitive area of the photodiode. It is thus proportional to the total number of photons detected by the reference channel during the opening duration of the SH shutter.

The final result F of an acquisition is then simply given by the ratio between SC and SR, corrected by the OD transmittance, i.e.

F(λ)=10OD(λ)SCSR(λ)

In this way, the influence of the shutter opening duration τλ on the final result F is totally canceled. This is the same for the time profile of the SH opening function, especially for its rise and fall times. In fact, this approach ensures the passive synchronization of both detection channels, where τλ plays only the role of a tuning parameter that allows for tightly locking the value of the effective signal recorded by the camera onto a numerical target ST defined by the following relation

ST=0.7×216×N45000×N
where N is the number of pixels inside the image of the FOL3 core and 16 is the number of bits used to digitize the video frame (216 corresponds therefore to the FWC of the CCD array).

Finally, the result of a transmittance measurement of a thin film filter at wavelength λ is given by

TF(λ)=FTFF(λ)FBL(λ)
where the numerator is the final result F obtained when the thin-film filter (TFF) is placed in the measurement channel and the denominator is the same quantity when this TFF is removed (the subscript BL here meaning baseline).

3.3. Performance analysis

3.3.1. Signal-to-Noise Ratio

Measurement channel Let us consider a single pixel of the CCD camera located in the image of the FOL3 core and let us call m¯ the total number of photo-electrons created in this pixel during the opening time τλ, with

m¯=m¯DC+ηC(λ)α(λ)TF(λ)×10OD(λ)P(λ)Nhντλ=m¯DC+m¯λ
where m¯DC/τλ is the dark current IDC of the CCD array, ηC(λ) is its quantum efficiency, α(λ) is a coefficient of transmission describing the operation of the measurement channel, P(λ) is the light power available at the LLTF output, and is the energy corresponding to a single photon of frequency ν (ν = c/λ).

After frame digitizing, the numerical data M associated with the same pixel is defined by

M=216m¯FWC+M0
where M0 is a numerical offset (M0 ∼ 700) used to correctly record the dark current fluctuations in the case of low light level detection.

In accordance with the principle of operation described in section 3.1 and with the data processing detailed in section 3.2, the signal recorded by the measurement channel is given by

SC={STforTFTcSTTFTcforTF<Tc
where Tc is a critical transmission value corresponding to the lowest filter transmission for which the numerical target ST can be reached, i.e. with OD = 0 and τλ = 100 s. This parameter can be expressed in a simple way with the help of two quantities, i.e. the shutter opening time [τλ]BL and the optical density [OD]BL used for the baseline recording at the wavelength λ
Tc(λ)=[τλ]BL10010[OD]BL

On the other hand, the fluctuations of the number of photo-electrons detected at the pixel level are defined by

σm2=m¯+IDCτλ+σRN2=m¯+σ02
where the first term is the photon noise, the second is the dark current contribution and the third is the camera read noise. Thus, the variance of the signal at the camera level is given by
σSC2=(216FWC)2[Nm¯+Ncσ02]
where Nc is the number of pixels in the summation circle (Nc ∼ 125 000) defined in section 3.2.

For the scientific grade CCD camera we selected (PIXIS 1024B), whose mean features are as follows

FWC=100000eIDC=0.001e/s@70°C operationσRN=9erms@2-MHz digitization
the dark current contribution can be fully neglected, and thus the signal-to-noise ratio (SNR) of the camera channel is given by
SNRC=SCσSC=N×0.7FWCScSTN×0.7FWCSCST+NcσRN2

Figure 6 shows the evolution of the measurement channel SNR with respect to the TFF optical density. This SNR is constant and close to 104 when the TFF transmittance is less than the critical value (2.66 × 108), and then it decreases with a slope of 1 in semi-logarithmic units. A SNR of 1 is obtained for optical density of approximately 12.

 figure: Fig. 6

Fig. 6 Evolution of the measurement channel SNR with respect to the TFF optical density (semi-logarithmic units).

Download Full Size | PPT Slide | PDF

Reference channel The instantaneous voltage V delivered by the FEMTO photoreceiver used in the reference channel is given by

V=GS[IDC+Sβ(λ)P(λ)]=GS[IDC+Iλ]withS=eηR(λ)hν
where G is the conversion gain of the amplifier (switchable from 1×103 to 1×109 V/W), IDC is the dark current of the photodiode (typically 2 pA), ηR is its quantum efficiency, S is its sensitivity (typically 0.6 A/W @ 850 nm), and β(λ) is a coefficient of transmission describing the operation of the reference channel. The gain G is chosen to obtain the highest voltage that remains in accordance with the digitizing range of the NI USB-6211 acquisition module, i.e. 0–10 volts.

The fluctuations of this voltage are characterized by the following variance σV2

σV2=(GS)2σI2=(GS)2[S2NEP2B+2eIλB]=G2[NEP2+2eβ(λ)P(λ)S]B
where NEP is the noise equivalent power of the photoreceiver (typically 800fW/Hz for a standard gain of 105) and B is the detection bandwidth (in our case, 400 kHz).

According to the data processing presented in section 3.2, the signal recorded by the reference channel is defined by

SR=fτλGIλS=fτλGβ(λ)P(λ)
where f is the sampling frequency of the NI USB-6211 acquisition module (100 kHz).

The fluctuations of the reference signal are therefore characterized by a variance σSR2 defined by

σSR2=G2f[(τλ+0.2)NEP2+τλ2eβ(λ)P(λ)S]B

Thus the reference channel SNR is given by

SNRR=fτλβ(λ)P(λ)B(τλ+0.2)NEP2+τλ2eβ(λ)P(λ)S

This expression is an increasing function of τλ, which means that its lowest value is obtained for the shortest opening time, i.e. 40 ms. Moreover, the light power β(λ)P(λ) describing the operating of the reference channel is approximately 80 µW @ 700 nm. Thus, the reference channel SNR is always greater than 1.2 × 106, which shows that the relative accuracy of the measurement is, from a theoretical point of view, entirely defined by the inverse of the measurement channel SNR.

Noise Floor Figure 7 shows a comparison between the noise floor experimentally recorded with our set-up and the theoretical value of the TFF optical density for which the measurement channel SNR is equal to 1.

 figure: Fig. 7

Fig. 7 Comparison between experimental and theoretical noise floor (blue circles: experimental data; continuous red line: theoretical data corresponding to a SNR of 1).

Download Full Size | PPT Slide | PDF

Even if the theoretical prediction is slightly below the experimental results (approximately 1 OD), its spectral dependence is nevertheless in good accordance with the experimental one. This small level discrepancy can be explained by the residual influence of the large intensity noise of the high power supercontinuum laser source. Indeed, one can see in Fig. 5(a) a demonstration of this noisy behavior; the fluctuations of the voltage delivered by the FEMTO receiver are much larger than the theoretical ones estimated in the previous paragraph (typically more than 101 compared to 106). The integration scheme used for both measurement and reference channels allows for the cancellation of most of the influence of this intensity noise, but a residual contribution nevertheless remains because of the slight difference in the time treatment performed on each channel. On the measurement channel, the CCD camera achieves a perfect analogical integration of the number of photo-electrons created during the acquisition time, whereas on the reference channel, the supervision PC performs a summation of the numerical data obtained by sampling the photoreceiver voltage at the clock frequency f.

This excess noise also has an influence on the accuracy of the TFF transmittance measurements achieved at low OD values and limits, in this case, the relative uncertainty to approximately 1%.

3.3.2. Influence of the line spectral profile

The main difference between the transmittance measurement of a neutral density filter and that of a thin-film filter is that for the latter, there is the presence of quick changes of its transmission with the wavelength. Due to these rapid variations, the result of a spectral transmittance measurement can be influenced by the effective spectral profile of the filtered source.

In our previous publication [7], we showed that the spectral profile of the light power density P(λ) delivered by the tunable filter is characterized by relatively wide tails (typically 106 of the maximum at 25 nm from the central wavelength; see Fig. 8). The presence of such tails induces a spectral crosstalk phenomenon that limits the lowest measurable OD value to 8 in the case of rapid spectral changes in the transmittance of a thin-film filter.

 figure: Fig. 8

Fig. 8 Power spectral density of the filtered source: with LLTF only (continuous red line); with LLTF and Shamrock monochromator (blue circles).

Download Full Size | PPT Slide | PDF

This is why we upgraded our first set-up by adding a new filtering device, i.e. a Czerny-Turner monochromator from ANDOR (reference Shamrock SR-193i-B1) equipped with a 300 l/mm ruled grating to drastically reduce the amplitude of these tails. One can see in Fig. 8 the effect of this additional filtering stage on the spectral profile of the source. This new spectral profile can be modeled by a Gaussian one with a 1.8 nm Full Width at Half Maximum (FWHM) bandwidth.

By using a MATLAB program that accounts for the spectral dependence of the optical properties of all components in each setup and performs a computation of a thin-film filter transmittance by calculating the ratio between the results of the spectrally integrated data obtained with and without this thin-film filter, it is possible to predict, with very good accuracy, the result of such a TFF transmittance measurement [7].

In Fig. 9, one can see the results of this modeling for both set-ups (Set-Up 1: LLTF only; Set-Up 2: LLTF + Shamrock) in the case of a dedicated thin-film filter manufactured using the Leybold Optics HELIOS deposition machine of the Institut Fresnel (number of layers: approximately 100). This approach shows that the limitation to 8 of the lowest measurable OD characterizing our previous set-up should be completely removed by the use of a Shamrock monochromator as a second filtering stage.

 figure: Fig. 9

Fig. 9 OD spectral dependence of the TFF used to qualify both setups (light blue line, design data; red line, computed data for LLTF only; dark blue line, computed data for LLTF + Shamrock).

Download Full Size | PPT Slide | PDF

4. Experimental results

This section will be devoted to the presentation of the results of transmittance measurements performed on four different types of thin-film filters with the help of the Set-Up 2 (i.e. including the monochromator Shamrock 193i as final filtering stage; see Fig. 2). For each filter, we will give also the transmittance data obtained by using a standard spectrophotometric mean.

4.1. FRESNEL band-pass filter

As indicated above, this filter has been manufactured on the premises of the Institut Fresnel using a Leybold Optics HELIOS deposition machine. It is the only one for which we know the spectral dependence associated with the theoretical design. The standard spectrophotometric mean used here is a Perkin-Elmer Lambda 1050.

In Fig. 10, one can see the fairly impressive agreement between the theoretical data and the result of the measurement achieved with Set-Up 2, as well as the large discrepancy between these same data and those recorded with the standard spectrophotometric mean (up to 4 OD at approximately 700 nm). However, one can notice that above 6 OD in the visible and 5 OD in the near-infrared, both last curves are in perfect accordance. This is especially the case for the position of the sharp peak located at approximately 690 nm and that of the near-infrared edge whose positions are slightly shifted with respect to the theoretical predictions. This confirms that these small deviations can be explained reasonably as being the results of minor deposition errors occurring during the manufacturing of the filter.

 figure: Fig. 10

Fig. 10 Experimental results obtained on the FRESNEL band-pass filter (light blue line: theoretical data; green line: Perkin-Elmer Lambda 1050; blue circles: Set-Up 2; black circles: Set-up 2 noise floor).

Download Full Size | PPT Slide | PDF

Moreover this perfect agreement between these two experimental curves remains when both data files are represented in linear units, as shown in Fig. 11. This confirms the accuracy of the 1% mentioned in the last paragraph of section 3.3.1.

 figure: Fig. 11

Fig. 11 Experimental results obtained on the FRESNEL band-pass filter (Perkin-Elmer Lambda 1050; blue circles: Set-Up 2).

Download Full Size | PPT Slide | PDF

4.2. THORLABS long-pass filter

This long-pass filter has been supplied from Thorlabs (reference FELH0750). Figure 12 presents the transmittance data provided by the manufacturer and the results of the measurements performed with Set-Up 2.

 figure: Fig. 12

Fig. 12 Experimental results obtained on the THORLABS long-pass filter (green line: manufacturer data; blue circles: experimental data recorded with Set-Up 2; black circles: noise floor of Set-Up 2).

Download Full Size | PPT Slide | PDF

As previously underlined, there is a really large discrepancy between the data recorded with a standard spectrophotometric mean and those obtained with our set-up (up to 6 OD between 600 nm and 700 nm). Another important issue is the quality of the determination of the steepness of the edge provided by our set-up, in good accordance with the manufacturer data above 5 OD. We can also notice a slight spectral shift between the band edge position corresponding to these two recordings, but this discrepancy can be explained by the fact that the manufacturer data corresponds to typical measurement results and not to those obtained on this filter item itself.

The last key point concerns the OD values obtained between 580 and 710 nm: a comparison with the experimental noise floor of our set-up shows that they are entirely determined by this noise level.

4.3. SEMROCK short-pass filter

This short-pass filter is manufactured and commercialized by Semrock under the reference FF-01-680-SP-25 and one item was lent to us by this company for the duration of our tests. Figure 13 presents the transmittance data provided by the manufacturer and the results of the measurements performed with Set-Up 2. The conclusions are nearly the same as for the long-pass filter from Thorlabs (OD improvement magnitude with respect to standard measurement results, edge steepness determination), but the wavelength range for which the results obtained with Set-Up 2 are noise floor limited extends from 700 nm to 1000 nm, except between 860 and 920 nm where the signal-to-noise ratio is slightly greater than 1.

 figure: Fig. 13

Fig. 13 Experimental results obtained on the SEMROCK short-pass filter (green line: manufacturer data; blue circles: experimental data recorded with Set-Up 2; black circles: noise floor of Set-Up 2).

Download Full Size | PPT Slide | PDF

4.4. SEMROCK notch filter

To complete the qualification of our measurement set-up, we chose to perform a last test with a notch filter, manufactured and commercialized by Semrock under the reference NF03-785E-25 and again loaned to us for the duration of our experimental studies. In this case, the results we obtained with Set-Up 2 are in perfect accordance with the Semrock data (see Fig. 14). The large gap between the lowest measured value (approximately 6.2 OD) and the corresponding noise floor (below 10 OD) proves that these results provide an accurate determination of the rejection level of this notch filter at its centering wavelength.

 figure: Fig. 14

Fig. 14 Experimental results obtained on the SEMROCK notch filter (green line: manufacturer data; blue circles: experimental data recorded with Set-Up 2; black circles: noise floor of Set-Up 2).

Download Full Size | PPT Slide | PDF

5. Conclusion

In this paper, we described a spectrophotometric apparatus allowing us to measure, to our knowledge for the first time, the transmittance of thin-film filters in the visible and near-infrared spectrum over an optical density range from 0 to 11. This ultra-wide range should be furthermore increased up to 12 by replacing the amplified photoreceiver used on the reference channel with a second scientific grade CCD; this would indeed allow the cancellation of the residual effect of the large intensity noise of the supercontinuum laser source by making identical the time treatment of both detection channels.

Acknowledgments

This work was partly funded by the DGA (Direction Générale de l’Armement) and by the CNES (Centre National d’Etudes Spatiales). We acknowledge the Semrock Company, and especially Prashant Prabhat, for the loan of filters dedicated to our qualification tests, as well as for fruitful exchanges.

References and links

1. J. W. Lichtman and J. A. Conchello, “Fluorescence microscopy,” Nat. Methods 2, 910–919 (2005). [CrossRef]   [PubMed]  

2. T. Erdogan and V. Mizrahi, “Thin-film filters for Raman spectroscopy,” Spectroscopy 19, 113–116 (2004).

3. A. L. Migdall, B. Roop, Y. C. Zheng, J. E. Hardis, and G. J. Xia, “Use of heterodyne detection to measure optical transmittance over a wide range,” Appl. Opt. 29, 5136–5144 (1990). [CrossRef]   [PubMed]  

4. Z. M. Zhang, L. M. Hanssen, and R. U. Datla, “High-optical-density out-of-band spectral transmittance measurements of bandpass filters,” Opt. Lett. 20, 1077–1079 (1995). [CrossRef]   [PubMed]  

5. Z. M. Zhang, T. R. Gentile, A. L. Migdall, and R. U. Datla, “Transmittance measurements for filters of optical density between one and ten,” Appl. Opt. 36, 8889–8895 (1997). [CrossRef]  

6. M. Ziter, G. Carver, S. Locknar, T. Upton, and B. Johnson, “Laser-based assessment of optical interference filters with sharp spectral edges and high optical density,” Surf. Coat. Tech. 241, 54–58 (2014). [CrossRef]  

7. S. Liukaityte, M. Lequime, M. Zerrad, T. Begou, and C. Amra, “Broadband spectral transmittance measurements of complex thin-film filters with optical densities of up to 12,” Opt. Lett. 40, 3225–3228 (2015). [CrossRef]   [PubMed]  

References

  • View by:
  • |
  • |
  • |

  1. J. W. Lichtman and J. A. Conchello, “Fluorescence microscopy,” Nat. Methods 2, 910–919 (2005).
    [Crossref] [PubMed]
  2. T. Erdogan and V. Mizrahi, “Thin-film filters for Raman spectroscopy,” Spectroscopy 19, 113–116 (2004).
  3. A. L. Migdall, B. Roop, Y. C. Zheng, J. E. Hardis, and G. J. Xia, “Use of heterodyne detection to measure optical transmittance over a wide range,” Appl. Opt. 29, 5136–5144 (1990).
    [Crossref] [PubMed]
  4. Z. M. Zhang, L. M. Hanssen, and R. U. Datla, “High-optical-density out-of-band spectral transmittance measurements of bandpass filters,” Opt. Lett. 20, 1077–1079 (1995).
    [Crossref] [PubMed]
  5. Z. M. Zhang, T. R. Gentile, A. L. Migdall, and R. U. Datla, “Transmittance measurements for filters of optical density between one and ten,” Appl. Opt. 36, 8889–8895 (1997).
    [Crossref]
  6. M. Ziter, G. Carver, S. Locknar, T. Upton, and B. Johnson, “Laser-based assessment of optical interference filters with sharp spectral edges and high optical density,” Surf. Coat. Tech. 241, 54–58 (2014).
    [Crossref]
  7. S. Liukaityte, M. Lequime, M. Zerrad, T. Begou, and C. Amra, “Broadband spectral transmittance measurements of complex thin-film filters with optical densities of up to 12,” Opt. Lett. 40, 3225–3228 (2015).
    [Crossref] [PubMed]

2015 (1)

2014 (1)

M. Ziter, G. Carver, S. Locknar, T. Upton, and B. Johnson, “Laser-based assessment of optical interference filters with sharp spectral edges and high optical density,” Surf. Coat. Tech. 241, 54–58 (2014).
[Crossref]

2005 (1)

J. W. Lichtman and J. A. Conchello, “Fluorescence microscopy,” Nat. Methods 2, 910–919 (2005).
[Crossref] [PubMed]

2004 (1)

T. Erdogan and V. Mizrahi, “Thin-film filters for Raman spectroscopy,” Spectroscopy 19, 113–116 (2004).

1997 (1)

1995 (1)

1990 (1)

Amra, C.

Begou, T.

Carver, G.

M. Ziter, G. Carver, S. Locknar, T. Upton, and B. Johnson, “Laser-based assessment of optical interference filters with sharp spectral edges and high optical density,” Surf. Coat. Tech. 241, 54–58 (2014).
[Crossref]

Conchello, J. A.

J. W. Lichtman and J. A. Conchello, “Fluorescence microscopy,” Nat. Methods 2, 910–919 (2005).
[Crossref] [PubMed]

Datla, R. U.

Erdogan, T.

T. Erdogan and V. Mizrahi, “Thin-film filters for Raman spectroscopy,” Spectroscopy 19, 113–116 (2004).

Gentile, T. R.

Hanssen, L. M.

Hardis, J. E.

Johnson, B.

M. Ziter, G. Carver, S. Locknar, T. Upton, and B. Johnson, “Laser-based assessment of optical interference filters with sharp spectral edges and high optical density,” Surf. Coat. Tech. 241, 54–58 (2014).
[Crossref]

Lequime, M.

Lichtman, J. W.

J. W. Lichtman and J. A. Conchello, “Fluorescence microscopy,” Nat. Methods 2, 910–919 (2005).
[Crossref] [PubMed]

Liukaityte, S.

Locknar, S.

M. Ziter, G. Carver, S. Locknar, T. Upton, and B. Johnson, “Laser-based assessment of optical interference filters with sharp spectral edges and high optical density,” Surf. Coat. Tech. 241, 54–58 (2014).
[Crossref]

Migdall, A. L.

Mizrahi, V.

T. Erdogan and V. Mizrahi, “Thin-film filters for Raman spectroscopy,” Spectroscopy 19, 113–116 (2004).

Roop, B.

Upton, T.

M. Ziter, G. Carver, S. Locknar, T. Upton, and B. Johnson, “Laser-based assessment of optical interference filters with sharp spectral edges and high optical density,” Surf. Coat. Tech. 241, 54–58 (2014).
[Crossref]

Xia, G. J.

Zerrad, M.

Zhang, Z. M.

Zheng, Y. C.

Ziter, M.

M. Ziter, G. Carver, S. Locknar, T. Upton, and B. Johnson, “Laser-based assessment of optical interference filters with sharp spectral edges and high optical density,” Surf. Coat. Tech. 241, 54–58 (2014).
[Crossref]

Cited By

OSA participates in Crossref's Cited-By Linking service. Citing articles from OSA journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (14)

Fig. 1
Fig. 1 Examples of rejection level measurements provided by optical thin-film filters manufacturers [left: long-pass filter from THORLABS, reference FELH0750; right: short-pass filter from SEMROCK, reference FF01-680-SP-25].
Fig. 2
Fig. 2 Schematic representation of the thin-film filter rejection levels measurement set-up (NKT EXB-6: supercontinuum laser source; LLTF: tunable volume hologram filter; OSD: order sorting device; SH: shutter; BS: beam splitter; ODF1: optical density flip 1; ODF2: optical density flip 2; ODFW: optical density filter wheel; ORC: output reflective collimator; FOL1: fiber optic link 1; TRC: transmitter reflective collimator; TFF: thin-film filter; RRC: receiving reflective collimator; FOL2: fiber optic link 2; SR-193i-B1: motorized Czerny-Turner monochromator; FOL3: fiber optic link 3; FOC: fiber optic coupler; PIXIS 1024B: scientific grade CCD camera; OD_R: reference optical density; RC_R: reference reflective collimator; FOL_R: reference fiber optic link; FEMTO OE-200-SI: variable gain photoreceiver).
Fig. 3
Fig. 3 Normalized spectral power density of the filtered source at the output of the LLTF (blue line, central wavelength λ0 = 450 nm; red line, central wavelength λ0 = 950 nm).
Fig. 4
Fig. 4 Time-chart of the light flux acquisition.
Fig. 5
Fig. 5 Data files obtained at the end of an acquisition: (a) Camera image file, including, near the center, a magnified view of the extremity of the FOL3; (b) Photoreceiver data file.
Fig. 6
Fig. 6 Evolution of the measurement channel SNR with respect to the TFF optical density (semi-logarithmic units).
Fig. 7
Fig. 7 Comparison between experimental and theoretical noise floor (blue circles: experimental data; continuous red line: theoretical data corresponding to a SNR of 1).
Fig. 8
Fig. 8 Power spectral density of the filtered source: with LLTF only (continuous red line); with LLTF and Shamrock monochromator (blue circles).
Fig. 9
Fig. 9 OD spectral dependence of the TFF used to qualify both setups (light blue line, design data; red line, computed data for LLTF only; dark blue line, computed data for LLTF + Shamrock).
Fig. 10
Fig. 10 Experimental results obtained on the FRESNEL band-pass filter (light blue line: theoretical data; green line: Perkin-Elmer Lambda 1050; blue circles: Set-Up 2; black circles: Set-up 2 noise floor).
Fig. 11
Fig. 11 Experimental results obtained on the FRESNEL band-pass filter (Perkin-Elmer Lambda 1050; blue circles: Set-Up 2).
Fig. 12
Fig. 12 Experimental results obtained on the THORLABS long-pass filter (green line: manufacturer data; blue circles: experimental data recorded with Set-Up 2; black circles: noise floor of Set-Up 2).
Fig. 13
Fig. 13 Experimental results obtained on the SEMROCK short-pass filter (green line: manufacturer data; blue circles: experimental data recorded with Set-Up 2; black circles: noise floor of Set-Up 2).
Fig. 14
Fig. 14 Experimental results obtained on the SEMROCK notch filter (green line: manufacturer data; blue circles: experimental data recorded with Set-Up 2; black circles: noise floor of Set-Up 2).

Equations (16)

Equations on this page are rendered with MathJax. Learn more.

F ( λ ) = 10 OD ( λ ) S C S R ( λ )
S T = 0.7 × 2 16 × N 45000 × N
T F ( λ ) = F TFF ( λ ) F BL ( λ )
m ¯ = m ¯ DC + η C ( λ ) α ( λ ) T F ( λ ) × 10 OD ( λ ) P ( λ ) N h ν τ λ = m ¯ DC + m ¯ λ
M = 2 16 m ¯ FWC + M 0
S C = { S T for T F T c S T T F T c for T F < T c
T c ( λ ) = [ τ λ ] BL 100 10 [ O D ] BL
σ m 2 = m ¯ + I DC τ λ + σ RN 2 = m ¯ + σ 0 2
σ S C 2 = ( 2 16 FWC ) 2 [ N m ¯ + N c σ 0 2 ]
FWC = 100000 e I DC = 0.001 e / s @ 70 ° C operation σ RN = 9 e rms @ 2 - MHz digitization
SNR C = S C σ S C = N × 0.7 FWC S c S T N × 0.7 FWC S C S T + N c σ RN 2
V = G S [ I DC + S β ( λ ) P ( λ ) ] = G S [ I DC + I λ ] with S = e η R ( λ ) h ν
σ V 2 = ( G S ) 2 σ I 2 = ( G S ) 2 [ S 2 NEP 2 B + 2 e I λ B ] = G 2 [ NEP 2 + 2 e β ( λ ) P ( λ ) S ] B
S R = f τ λ G I λ S = f τ λ G β ( λ ) P ( λ )
σ S R 2 = G 2 f [ ( τ λ + 0.2 ) NEP 2 + τ λ 2 e β ( λ ) P ( λ ) S ] B
SNR R = f τ λ β ( λ ) P ( λ ) B ( τ λ + 0.2 ) NEP 2 + τ λ 2 e β ( λ ) P ( λ ) S

Metrics