Technical Papers

» All-Fiber Optic Coherence Domain Interferometric Techniques

Jeff Bush, Pepe Davis
Optiphase Inc.
7652 Haskell Ave.
Van Nuys, CA 91406
Phone: 818-782-0997
Fax: 818-782-0999

Michael A. Marcus
Eastman Kodak Company
B-81, Kodak Park, MC 02036, Rochester,
NY 14650-2036
Phone: 716-477-6503
Fax: 716-722-2327

Abstract

Traditional white-light scanning interferometers utilize bulk optic components mounted on a mechanical scanning mechanism. Many emerging applications for these interferometers require fiber optic probes. By design, such fiber compatible instruments are expensive and are limited to slow scan rates. A new "all-fiber" design approach is presented, which reduces the cost of the design and enables higher scan rates.

Keywords: White-Light Interferometry, Coherence Domain Interferometry, OCR, OCDR, OCT

1. BACKGROUND

White-light Interferometry combined with fiber optic probing techniques offers convenient means to perform non-contact measurements serving a wide variety of applications. Popular applications include:

• Medical;
• Intravascular monitoring;
• Tissue, Bone, Plaque discrimination;
• Industrial;
• Thickness Gauge;
• Displacement Locator;
• Optical Inspection;
• Optical Coherence Tomography (OCT);
• Involves translation of the probe over a 2 dimensional surface.

Current methods of manufacture of white light instruments involve fabrication of an interferometer made with bulk optic components where a scanning mirror (or retro-reflector) is mounted on a mechanical scanning element.

The canonical Michelson implementation of the interferometer is shown in block diagram form in figure 1. The "reference" leg invokes a moving mirror to implement a scan. The "sensor" leg contains the device or sample being tested.

 

Common optical sources used for white-light interferometers are: 1) Erbium fiber sources, 2) Super luminescent Diodes (SLD); and 3) Edge Light Emitting Diodes (ELED). Typical spectra and coherence functions for these type of sources are shown in figure 2. All plots are normalized.

If the scanning interferometer of figure 1 was configured with a front surface mirror for the sample and an ELED for the source, the resulting interferogram from a scan would look like that shown in figure 3.

The simplest (and lowest resolution) approach for processing the returns of the scanning white light interferometer is shown in figure 4. Here the output of the optical receiver is electronically filtered. A high-pass or band pass filter is used prior to envelope detection of the interferograms. For this process, we assume the envelope detector produces RMS readings. The processor typically implements an analog to digital converter and provides control for the reference mirror scanning.

Using the processing approach in figure 4, some simple representations of measurement applications may be made for the expected signals for various samples using the interferometer in figure 1.

The simple mirror (top left) produces a single lobe. The rear surface reflector (top right) assumes a partial reflection at the input boundary and a full reflection at the rear surface. The film, plastic or paper sample assumes lossless media and produces a double lobe of equal intensity. Note: most media has some type of attenuation and the second lobe is typically at a lower level. The biological sample shows a return with a reflection or change in backscatter where a boundary is detected, such as that from blood (or other fluid) to an organ wall.

The measurement resolution of white light interferometery is dependent on many factors, such as the optical source power, self-noise and spectral characteristics; the optical return loss of the light from the "sample", the receiver bandwidth and noise, and the processing electronics. Generally, white light systems are configured to provide displacement resolutions from the few nm to many um range, depending on the application.

2. MICHELSON WHITE LIGHT INTERFEROMETERS USING FIBER PROBES

2.1. Bulk Optic Scanning Assembly

The Michelson scanning interferometer from figure 1 can be implemented as shown in figure 6. We show a PM configured interferometer to eliminate polarization fading caused by SMF interferometers. The reference leg with the scanning mirror has a non-reflective fiber termination. For the sensing leg, one may use a partial reflective termination to enable a reference distance to the sample. Conversely, the sensing leg fiber termination may be non-reflective, where a reference reflection may be incorporated into the sample.

There are a number of ways to implement the mechanical scanning approach. The typical techniques utilize a corner cube or retro-reflector mounted on either a motorized linear slide, rocker assembly or beam deformer. Such techniques have proven to be effective in providing appropriate scan ranges, but have a number of undesirable features when being considered for instrument production. These are listed below.

 

• High Cost                     launch and collection optics, specialized motor controllers

• Low Speed Scan Rates inertia limits scan rate, rocker assemblies limit range through angle changes

• Periodic Maintenance   optical alignment and cleaning

• Limited lifetimes           inherent in mechanical systems with moving parts

• Package limitations      compact packages are delicate, size and mass must be increased for robustness

Given the limitations cited, alternatives to the bulk optic scan mechanism are worth consideration.

2.1. All-Fiber Scanning Assembly

The concept of an all-fiber scanner for white light interferometry is highly appealing in that the light is self contained and eliminates the packaging complexities associated with integration of bulk optics with fiber waveguides. Optiphase, Inc. looked into the feasibility of such an approach and evolved the configuration shown in figure 7. This is a PM fiber arrangement using piezoelectric fiber stretchers to implement the scanning function.

We show the PM fiber arrangement in figure 7 as the design of choice. Experiments were conducted using lower cost single mode fiber scanners with PZT modulators and we found that birefringence modulation caused by the modulation process would broaden the coherence response which reduces the instrument resolution. Although there are some interests here for very low cost (SMF) applications, they were not pursued in this development effort.

2.2. The Modulators

The key to enabling the all-fiber scanner for white light applications was to develop efficient piezoelectric fiber stretchers (or modulators). A design investigation was performed which took into account the PZT materials available and the allowable field strengths which can be used; the dimensions and poling of the PZT which enhance the modulation properties; and the method of winding the fiber using multiple wound fiber layers without degrading loss and polarization holding performance. Given we use the optical configuration shown in figure 7, the investigation led to down-selection of optimized design parameters. Generally, the results are as follows.

• A scan range in excess of 10 mm was achieved (This is a true scan distance. In order to scan 10 mm, the interferometer needs to produce 20 mm of modulation which double passes the target path.)
• 10 mm triangle wave scans up to 24 Hz (amplifier limited)
• 10 mm sine wave scans up to 56 Hz (amplifier limited)
• Decreased scan ranges can be performed faster. For example, a 1 mm scan can be performed at 500+ Hz for a sine wave drive.
• Polarization crosstalk caused by the modulators was negligible (holding better than –20 dB);
• Overall optical excess loss is very small, with no measurable residual AM modulation.

The configuration of the dual modulators used in the PM Michelson interferometer tested is shown in figure 8. They are shown mounted on a printed circuit board which also contains the high voltage amplifiers used to drive them. The other elements of the interferometer are on the opposite side of the board.

Scan Linearity: Measurements were also made to determine the scan linearity. For the 10 mm scans, we obtained the following results.

• Sine Wave                                             ±0.02 mm , frequency independent, up to 56 Hz

• Triangle Wave                                        up to 15%, frequency independent up to 24 Hz.

The errors measured for the for the Sine and Triangle wave modulation inputs are shown in figures 9 and 10 respectively. The high voltage amplifier driving the modulators was outfitted with a 400 Hz low-pass filter. Actual displacement for both waveforms was taken by digitizing the receiver output during the sweep and then Hilbert transforming this data to obtain instantaneous frequency vs. time information for the sweep. This frequency information was then converted back to "true" displacement data, then compared with the modulation waveforms applied to the input to the high voltage amplifier.

 

Figure 9. Sine Wave Scan Measured Errors Straight line is measured

scan data vrs. Input (left scale). Difference from from Sine fit ( right

 

Triangle Wave Nonlinearity for 3 Drive Rates

Examination of the error data from the sine wave data (figure 9) shows errors generally less than 1%. This data was taken with a 56 Hz drive. Lower frequency drive inputs produced similar errors (less than 0.02 mm).

The errors for the triangle wave inputs are somewhat larger. Here we show the error for the positive slope of the ramp signal. When the scan is in the range of 6 to 10 mm, it appears that modulator hysterisis is present in this data set. The drop out of the 26.8 Hz data above 8 mm is an artifact caused by the amplifier used and not the modulator itself.

Figure 10. Triangle Wave measured errors. Three frequencies of 6.7, 13.4 and 26.8 Hz used. High voltage amplifier configured with 400 Hz Low pass filter.

At the low end of the scan (below 1 mm), the errors are caused by the 400 Hz filter used in conjunction with the high voltage amplifier. This is evidenced by the varying errors for the three different frequencies used. We recognize that some of the errors shown for the triangle wave data are likely caused by the amplifier used to drive the complex impedance loads (of the dual PZT elements). We plan future activities to investigate the PZT drive amplifier design or waveform generation pattern with an aim to improve the scan linearity for ramp input waveforms.

2.3. All-Fiber Michelson Scanning Assembly

An all-fiber Michelson white light scanning instrument is realized when the optical configuration of figure 7 is integrated with an electronic set in figure 4. A typical scan resulting from this type of assembly is shown in figure 11. Here we used an angle connector fiber output (probe) so that it doesn’t provide a back reflection. A reflector was placed 3 mm away from this probe. The scan approach was a triangle wave drive at 10 Hz, and the optical source was an ELED at 1300 nm. Data averaging was used to reduce the scatter of the noise floor.

The resulting output shows up at the 3mm point, as expected. The –3 dB scan resolution width was measured to be slightly less than 30 um, consistent with its coherence function of the ELED. This indicates the all-fiber scanning configuration did not broaden the width of this peak through mismatched dispersion of the fibers in arms of the interferometer. Other tests performed (not shown) did show that this peak could be broadened if care is not taken to match dispersion of the fibers forming the interferometer.

2.4. MIF-01 Scanning Interferometer

The work conducted in the development of the all-fiber white-light Michelson interferometer led to the continued development of an OEM scanner product (see figure 12). This is basically a scanning assembly with the broad-band source, envelope detection, and modulator drive electronics contained within a small rugged enclosure. The fiber leads shown are intended to be connected to user developed probe and reference fibers which complete the interferometer. The unit also requires external power and will accept a wide variety of modulator control input signals.

3. ALL-FIBER AUTOCORRELATOR

A variation of the All-Fiber Michelson white light interferometer is realized when the probe (fiber) is located external to the interferometer. In this case, the light returning from the sample is sent to a scanning interferometer and then processed. We call this the autocorrelator configuration. This configuration has the advantage of using a probe of arbitrary length without having to match the length of a probe fiber to the length of a reference fiber.

3.1. All-Fiber Autocorrelator, Simple Configuration

The block diagram of the all-fiber autocorrelator is shown in figure 13 in its most generic form. Here we incorporate the same type of fiber modulator scanning mechanism. A disadvantage to this approach when compared to the Michelson approach is that it has a larger optical loss resulting from the extra coupler. This coupler could however be replaced (at an additional expense) with a circulator to improve the throughput power.

If a similar envelope detection system is employed, the response of the autocorrelator shown with a rear surface mirror as the sample is shown in figure 14. Here we assume the probe fiber has no reflection. Note that the autocorrelator response is double sided, resulting from the two reflections, the weak front surface and the strong rear surface of the mirror (different from the response of the Michelson approach shown in figure 5).

 

3.1. Enhanced Configuration, All-Fiber Autocorrelator using reference laser

A collaborative effort between Optiphase, Inc. and Eastman Kodak has been focused on investigating the effectiveness of an all-fiber autocorrelator to make precision displacement measurements. Eastman Kodak has significant experience using bulk optic scanners for precision applications and recognizes a potential cost savings by using an all-fiber configuration, such as that shown in figure 15.

This configuration uses an additional coherent optical source which co-propagates with the broadband light inside the scanning interferometer. WDM’s or other appropriate filters are used to inject and separate out the returns from the broadband and coherent sources. The detected fringe crossings from the coherent source are used to determine the exact displacement of the scan at all points in the sweep.

The bulk optic scanning approach used by Eastman Kodak is capable of very high resolution as it implements a helium-neon laser as the coherent (reference) source. Although the all-fiber approach could be configured with a helium-neon reference source, it is much more desirable to use a semiconductor source, which has the desirable traits of being compact, rugged and an inherent long lifetime.

Given this interest, we selected a stable DFB laser for the coherent source. Wavelength stability measurements are shown in figure 16. It is evident that this source is not as stable as a helium-neon source, however it proves to be suitable for most precision measurement applications. An all-fiber white-light autocorrelator using this reference source was built to be evaluated against the Eastman Kodak bulk optic scanner

Figure 16. DFB Wavelength Stability Test

3.2. Comparison of All-Fiber Autocorrelator with Bulk Optic Autocorrelator

The fiber system was tested at Eastman Kodak labs. Both systems used the same broadband source which was a 1.3 um ELED, with 60 nm spectral width and approximately 50 uW of output power. Additionally, Kodak used the same processing electronics and software on both systems, which incorporate a Kodak developed technique to obtain sub-micron displacement accuracy.

A typical scan comparing the two systems is shown in figure 17. The two plots show oscilloscope traces of the receiver outputs for both systems with a repetitive scan input waveform. The box on each plot shows the range of a scan cycle (start to finish). Note on the bottom plot, the arrows point to anomalous data likely caused by an imperfect reflector in the fiber system. Since this additional reflection did not overlap the other reflections it did not cause an error in the processing technique. A comparison of various sample tests (reduced data) using the two systems is shown below.

Comparison of Bulk Optic Scanner and All Fiber Scanner

Thickness in microns

All measurements from both systems agree to within 0.1 um, indicating good convergence, and generally a successful all-fiber implementation.

4. CONCLUSION

We have demonstrated the feasibility and general capability of all-fiber scanning white-light interferometers. Advantages of such instrument configurations include:

• Lower manufacturing cost;
• Smaller and more robust packaging;
• Increased scan rates;
• Accuracy comparable to bulk optic systems;
• Longer lifetime;
• Less maintenance (lower life cycle costs).

Future plans include commercialization of this technology.