Technical Papers

» MULTI-CHANNEL INTERFEROMETRIC DEMODULATOR

JEFF BUSH and ALLEN CEKORICH
Optiphase Inc.
7652 Haskell Ave.
Van Nuys, CA 91406
Phone: 818-782-0997
Fax: 818-782-0999

CLAY KIRKENDALL
Naval Research Laboratory
4555 Overlook Ave, SW,
Washington DC
Code 5674

Abstract

This paper outlines recent progress made by Optiphase Inc. in the development of low-cost, high performance interferometric demodulators applied to multi-channel interrogation. The paper’s focus is on the Optiphase digital demodulation concept applied first for single channel operation and then to multichannel operation. This paper also presents single and multi-channel demodulation test data taken from hardware developed for these applications.

 1. INTRODUCTION

The concept of using fiber optic sensors as an advantageous replacement of conventional sensors is one which has been in existence for 2 decades. The touted advantages are; EMI immunity, higher performance, lower cost and in the case of multi-channel sensing, passive multiplexing. To date, for the most part, the EMI immunity and the passive multiplexing have been demonstrated, while the "higher performance" and "low cost" virtues exist mainly as ideas. The lack of solutions to these two shortcoming has significantly hampered the commercialization and acceptance of many fiber optic sensing products. The generally recognized remedies to the solutions involve improvements or innovations in transducer design, deployment techniques and sensor interrogation techniques. This paper addresses a sensor interrogation approach which shows significant promise to eliminate the aforementioned problem. The sensor interrogation approach presented is applicable to many fiber sensor types, such as Interferometric, Polarimetric, Fabry Perot and possibly Fiber Gratings. Due to the brevity of the paper, and the authors desire to focus the topic on interrogation, we discuss it mainly in conjunction with interferometric sensing.

The sensor interrogation (or demodulation) approach involves direct digitization of the optical (received) signals returning from the sensor and implementation of an efficient DSP based (optical) phase measurement algorithm. The merits of such an approach are:

• It is an open-loop approach (applicability to many sensor types);
• It minimizes the amount of samples required to make accurate measurements;
• It provides an extremely high dynamic range (micro-radians to thousands of radians);
• It can be used by unskilled operators
• It can be produced in moderate volumes at a low cost

2. BACKGROUND: INTERFEROMETRIC TERMINOLOGY AND PARAMETERS

Optiphase has spent a considerable amount of time exploring the development of interferometric demodulators. It is common knowledge that there are a great many approaches for performing interferometric demodulation. It is also common knowledge that the type of electronics required for demodulation requires a high degree of specificity. Currently, most interferometric demodulators are custom developed for the fiber sensor being developed. This typically results in the demodulator comprising a significant cost of the fiber sensor. If Interferometric Fiber Sensors are to be low cost, the demodulators will also need to be low cost. The approach taken by Optiphase to accomplish such an objective is to focus on demodulator approaches which are applicable over a wide range of sensing applications. This concept assumes that volume requirements for such a device will drive prices down. As stated earlier, Optiphase believes that the most general type of interferometric demodulator will exist as an Open Loop processing device. Further we believe that if high accuracy demodulation is to be accomplished, this demodulator will have to incorporate a means for providing a modulation (or dither) signal into the interferometer which will enable an accurate (optical) phase measuring approach. The approach we have selected for producing the modulation is commonly known in the field as the Phase Generated Carrier (PGC) modulation scheme.

DEFINITION OF TERMS; PHASE GENERATED CARRIER

There are a number of ways to induce dynamic phase modulation in interferometers. An outline of such techniques is beyond the scope of this paper and we simply make the assumption that the interferometer is equipped with some means of producing the modulation. If the interferometer’s "phase modulator" is driven with a sinusoidal signal, it will produce a sinusoidal variation of the interferometer’s optical phase. The generalized expression for the received signal of the interferometer which employs this Phase Generated Carrier (PGC) is shown in equation 1.

P(t) = COS { b SIN(wt) - f }                                              (1)

Here we show only the portion of the received signal representing interference terms. For this expression,

f = the phase of the interferometer (includes sensed signal and offset terms);

b = the PGC modulation index, expressed in peak radians;

w = the frequency of the PGC phase modulation.

Equation 1 can be broken out into harmonic terms of the PGC "carrier" frequency.

P(t) =                  J0(b ) COS{ f }                       (dc term)                                      (2)

+                        2J1(b ) SIN (wt) SIN{f }            (1w term)

+                        2J2(b ) COS(2wt) COS{ f }       (2w term)

+                        2J3(b ) SIN (3wt) SIN{ f }         (3w term)

+                        2J4(b ) COS(4wt) COS{ f }       (4w term)

+ ...

where Jn(x) is the Bessel Function of the first kind of order n and argument x.

Equation 2 shows that the resulting break out of terms produces even harmonics of w which vary with COS f and odd harmonics vary with SIN f.

To more clearly illustrate the PGC modulation, An diagram showing such signals is shown in figure 1. Here we show a commonly used modulation index (or modulation depth) b of p radians. The plot on the bottom shows the sinusoidal modulation drive signal. The top plot shows the signal seen at the interferometers receiver given the signal phase, f is 0 radians. The center plot shows the signal when f = p/2.

GENERALIZED DIGITAL DEMODULATION

Once the PGC signals are generated, and given the quadrature relationships defined in equation 1, theobjective of the demodulator is to digitally process the quadrature terms COS f and SIN f to the effect that the term f may be measured. This generalized process is shown in figure 2.

Here the SINf and COSf are digitized. An algorithm is used to determine the magnitude of both values, and their ratio is taken. An inverse trigonometric process is performed to determine the phase within the quadrant, s. The process on the right side of figure 2 involves determining the quadrant. This is performed with simple sign checking. The "offset" phase from the quadrant determination, d is determined and added to s to determine f, which is the phase (on the unit circle). The digital approach also implements a fringe counter. This is a simple algorithm which compares prior values of d to current values of d to determine if the unit circle boundary has been crossed. When a crossing is detected, the fringe count a is incremented or decremented, depending on the direction of the crossing.

The first obvious benefit of using a digital demodulation approach is the advantage very large phase measurement ability through the implementation of the fringe counter. Other benefits will be seen in subsequent sections.

3. DIGITAL DEMODULATION, SINGLE CHANNEL

The algorithm shown in figure 2 does not incorporate the fact that the interferometer signals are in the "carrier" domain and not yet baseband quadrature phase terms. Some type of process is required to perform this conversion. In earlier years, Optiphase looked into digital demodulation approaches like that shown in figure 3, called the "mixed signal" demodulation approach. Here we used an analog front end to "beat" signals of odd and even harmonics of the modulation drive frequency (1F and 2F) to realize the baseband quadrature signals which provides inputs which are compatible with the processing approach in figure 2. In figure 3, we also show a third term which gets processed (from 4F) which is used as a reference indicator (relative to the level of the 2F term) to maintain a constant modulation depth.

Although this mixed signal demodulation approach has been shown to provide high performance, its analog circuitry makes the approach somewhat less than desirable for the following reasons.

• It adds to component cost and circuit board size and complexity;
• It requires manual tuning; and mainly;
• It is incompatible with multi-channel interrogation schemes.

These inflexibility’s were enough to encourage investigations to alternative methods.

The most obvious alternative approach is one where the ADC is placed directly behind the receiver, as shown in figure 4.

The simplicity of the block diagram speaks for itself. This is truly a much more desirable configuration (from the standpoint of electronic "block diagram" complexity).

What this design does, is transfer the complexity to the ADC and DSP in that the they are now operating in the carrier domain at much higher speeds while maintaining a high bit count for the ADC.

In the investigation performed on the all digital approach it was determined early-on that in order for such an approach to be generally effective, the following three requirements must be met.

Requirement #1: The Number of Samples per PGC Carrier must be Low

goal: Less than 16 samples per PGC cycle where PGC frequencies can be as high as 200 KHz

Requirement #2: The Sampling Approach Must also be Compatible with Multi-Channel Sensing

goal: Compatibility with TDM multiplexed PGC sensor arrays (this includes arrays with MZ or Michelson interferometer sensors, Fabry Perot sensors and Polarimetric Sensors)

Requirement #3: Requirement #1 must not compromise Requirement #2 (or visa-versa)

The results of a multi-year, multi-program development effort (sponsors include Optiphase, DOE, NRL, and the US ARMY MICOM) have resulted in a rather elegant evolved approach. What we found was that there were a wide range of signal sampling strategies which could be used, and that it was actually possible to accomplish the demodulation with as little as 4 samples per PGC carrier cycle. It turns out that if the desired measurement results are to be performed accurately, slightly higher sample counts would be necessary. In general what we found is summarized in Table 1.

Based on the findings in Table, 1, Optiphase developed its first-article single-channel, All-Digital Demodulator. The objective of this development was to keep it simple and low cost. The key characteristics of this demodulator are as follows.

• Uses Low Cost Integer DSP (TMS320C50) @ 40 MIPS; 12 samples per PGC Cycle
• Provides Modulation Drive Signal, 1 KHz to 100 KHz range, user selectable;
• Firmware provides automatic modulation level adjustment (maintains high accuracy);
• Outputs a 32 bit data word (+/- 0.6 urad to +/- 12,000 radians);
• 20 Bit DAC used for Analog Output. User selectable for operation anywhere over the 32 bit range of the digital output.
• Digital or Analog Output Data rate is equal to PGC cycle rate.
• Self Noise £ 4 urad/rt-Hz (current version, 1 urad/rt-Hz is achievable)
• Slew Rate Limit = pi radians per PGC cycle time.
• Distortion < -50 dB (signals less than 20% of the PGC carrier rate)
• Optical Phase Measurement Accuracy is better than 0.3%.
• Current design fits on 8" X 4" circuit board.

The current version single channel demodulator uses electronic components conducive to low cost manufacturing at moderate volumes. Additionally, when requirements for high volume production present themselves, PLD and/or ASIC design approaches will be applicable to meet the lower cost demands.

Some Test data depicting the performance of the single channel demodulator is shown in figures 5 and 6. Figure 5 shows the Self Noise performance where the interferometer is operating under optimum conditions (shot noise is much less than demodulator noise, and the receiver output level fills the demodulators input ADC by 90% or more). Figure 6 shows the dynamic performance for various levels of ac input signals, ranging from small to large levels. In figure 6, the PGC carrier frequency was 60 KHz.

4. DIGITAL DEMODULATION, MULTI-CHANNEL

In the previous section, we discussed that the ideal design approach for the digital demodulator would be such that it could made applicable to the processing of multiple sensor channels. The criteria for this to occur involves using a time domain multiplexed PGC interrogation approach. We have found that a wide range of sensor types can be made to fit this criteria. Figure 7 shows a few of these approaches.

In figure 7, the top two sensor configurations are mismatch-path interferometers. In the case of intrinsic polarimeter, we assume birefringent fibers used for the sensing elements, which in effect causes a mismatch path interferometer. In all of these cases, the PGC signal is generated by frequency shifting the optical source and then pulsing it.

For sensors with long path mismatches (typically the Michelson or Mach Zehnder configurations), the optical source is typically a CW laser which is frequency modulated by cavity tuning, or in the case of a laser diode, injection current modulation can be used. For sensors with short path mismatches, such as the Fabry Perot and polarimetric sensors, larger wavelength excursions are required for the PGC modulation. Here the typical optical sources are CW broad spectral sources such as SLD’s or Er Fiber fluorescent devices where tuning occurs external to the source, usually with Acousto-Optical Tunable Filters or low finesse tunable filters.

The optical pulse forming in any of these sensor systems is typically created using Integrated optic type switches or in-line Bragg Cells.

INVESTIGATION WITH NRL ACOUSTIC SENSOR ARRAY

In Late 1995, NRL took an interest in the Digital Demodulator approach for their multi-sensor acoustic system applications. The traditional approach used by NRL for multi-channel interferometric demodulation was to use a single analog (differentiate and cross multiply) circuit demodulator per sensor channel. What NRL had found regarding such an approach is that as the channel count increases, the circuit cost, size and power consumption increases linearly with the channel count.

In an effort to identify more efficient means of demodulating multi-channel sensor systems, an effort was funded by NRL to determine the applicability and performance capabilities of the Optiphase digital demodulator.

 

The objective of the work effort performed was to modify the design of the single channel demodulator design to accept TDM sampled PGC signals from a 64 channel interferometric sensor array. To keep development costs low, we elected to use the same DSP as in the single channel demodulator, and allow for user selectability of any 8 of the 64 channels to be demodulated. The design changes required to do this were as follows.

• Configure sampling and demodulation for 6 samples per PGC cycle, and a 20 KHz PGC.
• Modify the input section of the demodulator to accept the TDM data. The data sequence caused by the time delay interrogation was S1I1, S1I2, S1I3, .... S6I63, S6I64 over the course of a PGC modulation cycle, where SN represents the sample number N and IM represents interferometric sensor number M.
• Modify the demodulator to operate without active Depth of Modulation Control. When source modulation is used to develop the PGC signal through mismatch path interferometers, each different OPD for each interferometer will create different modulation depths, if they are different, controlling the modulation depth for one sensor will create a different modulation for sensors with a different OPD. Multi-channel systems will require a passive modulation depth correction, which is software or firmware based.
• Modify the processor to process the data in the TDM input format and add registers to hold fringe and phase values for each sensor.
• Modify the Output to provide a digital stream of data (32 bit words, each sensor) in the TDM  format as well as two (user selected) analog output channels (20 bit DAC used).

The resulting design changes ended up in the layout shown in figure 8 below. This design was implemented on a single printed circuit board measuring 8 by 10 inches and delivered to NRL for evaluation in an existing acoustic sensing system operating in their laboratory.

PERFORMANCE TESTING, MULTI-CHANNEL INTERFEROMETRIC DEMODULATOR

Sensor system integration and performance testing was conducted by Mr. Clay Kirkendall of the Naval Research Laboratory. Tests were made on a 64 channel interferometric sensing array at NRL. To perform the measurements, the multi-channel digital demodulator was inserted into NRL at a point directly after their receiver/ADC module. Measurements of dynamic signals, gain and phase response, and distortion were made. The results were nearly ideal, showing the capability of a sizable improvement over the performance of current technology analog demodulators.

Figure 9 shows the results of the dynamic signal data. The response to extremely large signals shown in the top left. As with the single channel demodulator, the distortion energy is spread over many harmonics. As the signal levels decrease, so does the spectral content of the distortion as shown in figure 9.

In figure 10, we show a typical channel gain and phase response when running the PGC at 20 Khz.

In figure 11, we show test data for total harmonic distortion. Here testing was performed at 250 Hz, and the signal level was varied from 70 mrad RMS (-23 dB re 1 Radian) to 22 radians RMS (+27 dB re 1 Radian). It should be noted that the distortion measurement for the input level at 70 mrad is artificially high as the noise present during the measurement represented the reference floor. Distortion measured for the multichannel demodulator was typically £ -50 dB for all channels. The data shown is typical for all channels.

In figure 11, we additionally show the distortion measured for an analog demodulator (data here also provided by NRL) . Depending on the level of the input signal, these values are 10 to 20 dB larger than the digital demodulator.

5. CONCLUSION

A highly efficient and accurate approach for interferometric demodulation has been presented. Due to the nature of its design topology, it has application with both single and multi-channel fiber optic sensing systems. Optiphase has recently begun selling prototype single channel demodulators and expects to release an instrument based product in CY 1998. Development is continuing on the multi-channel demodulator design with the objective to address a channel count of up to 256.

6. ACKNOWLEDGMENTS

The authors wish to acknowledge John Bostick and Cory Keitz of Optiphase and Dr. Anthony Dandridge and Gary Cogdell of the Naval Research Laboratory. Without their dedicated efforts, the progress made and reported on in this paper would not be possible.