HIGH LINEARITY OPTICAL TRANSMITTER
FIELD OF THE INVENTION The present invention relates to optical transmitters. More specifically, the present invention relates to optical transmitters which are particularly suited to the transmission of analog signals in fiber optic systems.
BACKGROUND OF THE INVENTION Analog optical transmission has recently attracted renewed interest owing to the inherently low-noise characteristics of fiber optic systems. With the advent of single mode systems noise levels have been greatly reduced by eliminating modal noise within the fiber. Also, recent improvements in laser diodes have led to dramatically lower noise levels. With these advances in fiber optic technology, analog transmission becomes extremely attractive in applications where the desired signals at the terminal ends are analog (for example, commercial television) . By eliminating the need for modulation conversion the terminal equipment is greatly reduced in complexity and cost. One remaining obstacle to widespread use of analog optical transmission is the intrinsic nonlinearity of typical light emitting devices. Distortion of the signal caused by nonlinearities can be a problem, particularly in multichannel systems, where a high degree of linearity is required to prevent interchannel crosstalk. Linearization of a laser diode or high bandwidth light emitting diode (LED) by direct feedback is generally not feasible due to the difficulty of controlling phase shift at the extremely high frequencies involved. However, other methods of linearization are possible. For example, a quasi- feed-forward technique by Straus and Szentesi, Electronics Letters. 13, 158 (1977), effectively linearized an LED by
using a matched LED to cancel the distortion. Linearization of an LED by predistorting the input signal was reported by Asatani and Kimura, Electronics Letters. 13, 162 (1977) and IEEE Journal of Solid-State Circuits, SC-13 , 133 (1978). Their method, using fixed predistortion, gave second and third harmonic distortion improvement of 28 dB and 31 dB respectively.
SUMMARY OF THE INVENTION The present invention utilizes a novel method of indirect (averaged) feedback to adaptively predistort the input so as to minimize the overall nonlinear distortions. Since the predistortion is adaptive and continuously acting, optimum linearity is automatically maintained as operating conditions change. Broadly stated, the present invention provides an optical transmitter capable of adaptively maintaining linear input-to-output response, comprising adjustable distortion means having a plurality of inputs and an output, adapted to receive an input signal at its first input and a plurality of control voltages at additional inputs and to vary its input-to-output system response in response to variation in said control voltages to distort said input signal and provide a resultant predistorted signal at its output; light emitter means having an input and an output, said input electrically coupled to the output of said adjustable distortion means for receiving said predistorted signal and generating an optical transmission signal through said output, whose intensity varies in response to the variation of said predistorted signal; and feedback means having an input and a plurality of outputs, optically coupled at its input to the output of said light emitter means and electrically coupled at said plurality of outputs to said plurality of inputs of said adjustable distortion means
corresponding to said plurality of feedback means outputs, for receiving said optical transmission signal and for providing said adjustable distortion means with said plurality of control voltages which are capable of varying in response to said optical transmission signal, thereby predistorting said input signal so as to adaptively linearize the overall input-to-output response of said optical transmitter. Further, the present invention provides a method for adaptively linearizing the response of an optical transmitter comprising the steps of generating an optical transmission signal corresponding to an electrical input signal; generating at least one time averaged signal, in response to said optical transmission signal, whose magnitude is indicative of the nonlinearity of the input-to output response of said optical transmitter; and altering the input-to-output response of said optical transmitter so as to minimize the magnitude of said time averaged signal, thereby adaptively linearizing the overall input-to-output response of the optical transmitter. Additionally, the present invention provides a method for adaptively linearizing the response of an optical transmitter comprising the steps of generating a predistorted input signal by distorting an input signal; generating an optical transmission signal whose intensity varies in response to variations in said predistorted input signal; generating at least one time averaged signal whose magnitude is indicative of the nonlinearity between said input signal and said optical transmission signal; and altering the degree and the kind of distortion of said input signal in response to said time averaged signal so as to minimize the magnitude of said time averaged signal, thereby adaptively linearizing the overall input-to-output response of the optical transmitter.
In one embodiment of the invention the time-averaged feedback signal is derived from spurious harmonic frequencies generated as a result of uncompensated nonlinearities. In this version the feedback means acts to suppress the unwanted harmonics, thereby automatically enhancing the system linearity. In another embodiment of the invention the time- averaged feedback signal is derived from a comparison of the output signal with the time-delayed input signal. In this version the feedback means acts to reduce the time-averaged discrepancy between the actual output and the desired output. The present invention provides an optical transmitter having input-to-output linearity substantially superior to that of the transmitters of the prior art. It also provides an optical transmitter having the novel feature of automatically maintaining its high degree of linearity over long periods of time under changing conditions. Areas of application of the present invention are very broad, encompassing those applications where optical analog transmission is desired. Of particular interest is multichannel video transmission with either AM or FM modulation. The invention might also be used effectively for digital data transmission in local area networks. In this application the digital information would be encoded by modulating an RF carrier, allowing simultaneous operation of multiple data channels with each channel using a different carrier frequency.
BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a block schematic diagram illustrating the essential features of a first embodiment of the optical transmitter of the present invention wherein linearization is accomplished by the suppression of harmonics.
TITUTE SHEEΪ
Figure 2 is a block schematic diagram showing the essential features of a second embodiment of the invention wherein linearization is accomplished by minimizing a ti e- averaged error signal. Figure 3 illustrates an embodiment of the present invention incorporating as the light emitting device a laser diode having two facets. Figure 4 illustrates an exemplary configuration of the adjustable distortion unit.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The invention and its various aspects will be best understood by reference to the Figures and the following text. Figure 1 is a block diagram schematically illustrating the essential features of one embodiment of the optical transmitter of the present invention. An input signal is applied to input line 1 of the optical transmitter. The input signal may take a variety of forms but in this embodiment it should contain at least one carrier frequency or tone so that any uncompensated nonlinearities will result in spurious harmonic frequencies. For example, the input signal may consist of one or more modulated carriers with the modulation being any of several types including amplitude modulation, frequency modulation, or phase modulation. Alternatively, the input signal may contain a steady tone superimposed on an information carrying signal of arbitrary form. The first input of adjustable distortion unit 2 is coupled to input line 1, and receives the input signal therefrom. Adjustable distortion unit 2 is for distorting the input signal in a manner which is as close as possible to the exact inverse of the distortion which will occur later in light emitting device 3. Adjustable distortion
unit 2 is further coupled at a plurality of other inputs to a corresponding plurality of outputs of processor 14 . In response to voltage levels provided on lines 15a-15e by processor 14 , adj ustable distortion unit 2 is capable of altering its output variation as a function of its input. The input of light emitting device 3 is coupled to the output of adjustable distortion unit 2 , and receives the predistorted input signal therefrom. Light emitting device 3 is for generating an optical transmission signal . Light emitting device 3 responds to the variations in the predistorted input signal by correspondingly varying the intens ity of the optical transmission signal . Light emitting device 3 may be any dev ice capable o f such operation , including a l ight emitting diode or a laser diode. The remaining components shown in Figure 1 comprise one configuration of the novel feedback means of the present invention. Light splitting device 4 is optically coupled to the output of light emitting device 3. Light splitting device 4 is for splitting off a portion of the optical transmission signal. The split portion of the signal is provided to the other elements feedback means, while the remainder of the optical transmission signal is output from the transmitter as output beam 16. The input of photodetector 5 is optically coupled to light splitting device 4 and receives the split portion of the optical transmission signal therefrom. Photodetector 5 is for- converting the split portion of the signal to an electrical feedback signal. Photodetector 5 responds to variations in the optical signal by correspondingly varying the generated electrical feedback signal. The input of amplifier 6 is coupled to the output of photodetector 5 and receives the feedback signal therefrom. Amplifier 6 is for increasing the power of the feedback
SUBSTITUTE SHEET
signal and for providing the amplified feedback signal onto line 13. The inputs of filters 7 and 8 are coupled to the output of amplifier 6, and each receive the amplified feedback signal therefrom. Filters 7 and 8 represent a plurality of narrow band frequency filters which are tuned to a harmonic frequency of one or more of the carrier frequencies of the input signal on line 1. For example, filters 7 and.8 may be tuned to the second and third harmonics, respectively, of one of the carriers. Alternatively, if two or more carrier frequencies are present, filters 7 and 8 may be tuned to beat frequencies (sum or difference frequencies) of those carrier frequencies present. The number of filters is not restricted to two; a larger number may be desirable for multi-carrier signals, while just one filter may be sufficient in some cases. Filters 7 and 8 are for isolating the energies in the feedback signal which are in a very narrow frequency neighborhood around the frequencies to which they are tuned. The inputs of averaging filters 9 and 10 are coupled to the outputs of filters 7 and 8, respectively, and receive the isolated frequencies of the feedback signal therefrom. Averaging filters 9 and 10 represent a plurality of averaging filters which may be coupled to the outputs of corresponding narrow band frequency filters such as filters 7 and 8. Averaging filters 9 and 10 are for averaging, OVPΓ some period of time, the energies of the isolated frequency signals they each receive, and for providing voltage levels on lines 11 and 12, respectively, which are indicative of the harmonic or beat content of the signal on line 13. A first and a second input of processor 14 are coupled to lines 11 and 12, respectively. Processor 14 is for varying the voltages on lines 15a-15e so as to adjust the
degree and kind of distortion produced by distortion unit 2. Lines 15a-15e represent a plurality of lines coupling outputs of processor 14 to corresponding inputs of adjustable distortion unit 2. Processor 14 is normally in continuous operation and responsive to the signals on lines 11 and 12, seeking constantly to adjust the voltage levels on lines 15a-15e so as to minimize the voltage levels on lines 11 and 12. In this way the feedback adjusts the system so as to suppress those harmonic frequencies passed by filters 7 and 8. Suppressing these harmonics has the affect of linearizing the response of the optical transmitter since these harmonics are generated as a result of nonlinearities in the system. The result is that the output signal produced on output beam 16 is a faithful reproduction of the input signal on line 1. "' Figure 2 schematically illustrates an alternate embodiment of the present invention in which the time averaged feedback is derived from a comparison of the output optical transmission signal with a time delayed input signal rather than the harmonic content of the optical transmission signal. In this embodiment adjustable distortion unit 2, light emitting device 3, light splitting device 4 and photodetector 5 function as described in the previously discussed embodiment. The input of delay unit 19 is coupled to input line 1 and receives the input signal therefrom. Delay unit 19 is for delaying the input signal on line 1 by a fixed time which is chosen to closely match the total of the delay time of adjustable distortion unit 2, the delay time of light emitting device 3, the optical transmit time to photodetector 5 and the delay time of photodetector 5. Delay unit 19 may be any device capable of such operation,
SUBSTITUTE SHEl .
including a co-axial line of fixed length. Differential amplifier 17 has a first input coupled to the output of delay unit 19 and a second input coupled to the output of photodetector 5. Differential amplifier 17 is for generating an error signal whose magnitude is indicative of the discrepancy between the signals produced by photodetector 5 and the input signal, and for providing the error signal on line 18. The input of averaging filter 20 is coupled to line 18 and received the error signal therefrom. Averaging filter 20 is for generating a voltage level indicative of the magnitude of the error signal averaged over some time, and for providing the voltage level onto line 21. Processor 14 functions much the same as described in the previous discussed embodiment. The input of processor 14 is coupled to line 21. Processor 14 is for varying the voltage on lines 15a-15e so as to adjust the kind of degree and distortion by distortion unit 2. Lines 15a-15e represent a plurality of lines coupling outputs of processor 14 to corresponding inputs of adjustable distortion unit 2. Processor 14 is normally in continuous operation and responsive to the signal on line 21, seeking constantly to adjust the voltage levels on lines 15a-15e so as to minimize the voltage level on line 21. In this way the feedback adjusts the system so as to minimize the discrepancy between the signal produced by photodetector 5 and the input signal, thereby having the effect of linearizing the response of the optical transmitter. The result is that the output signal produced on output beam 16 is a faithful reproduction of the input signal on line 1. The embodiment of Figure 2 obviates the need for the input signal to possess a carrier frequency or tone, permitting the input signal to be of arbitrary form. In contrast to the embodiment of Figure 1, however, the
ET
embodiment of Figure 2 requires careful gain control and critical timing to produce a usable feedback signal. In addition, differential amplifier 17 must be capable of tracking the highest frequencies present in the output signal. Thus the embodiment of Figure 2 has the advantage in applications where the input signal must be of general form, whereas the embodiment of Figure 1 is more easily implemented in high frequency applications where the input signal either possesses a carrier or when a tone can be added. Figure 3 illustrates an embodiment of the present invention incorporating as the light emitting device a laser diode having two facets. Laser diode 3* of Figure 3 may replace both light emitting device 3 and light splitting device 4 of Figures 1 and 2. In such an embodiment laser diode 3 has its input coupled to the output of adjustable distortion unit 2 and receives the predistorted input signal therefrom. Further a first output of laser diode 3' is optically coupled to the photodetector, and a second output produces the optical transmission signal as output beam 16 of Figures 1 and 2. Laser diode 31 responds to variations in the predistorted input signal by correspondingly varying the intensity of its output signals. The use of this variation is advantageous when the light beams emitted from the two facets of the laser diode are closely matched in their characteristics. If this is not the case the configurations of Figure 1 and Figure 2 are preferred. It is common practice to package laser diodes with a monitor photodiode built into the enclosure so as to collect light from the rear facet of the laser diode. This built in monitor photodiode may be used according to Figure 3 in those applications where the response of the photodiode is fast enough. Normally, such photodiodes are not designed to have a very fast response time since they are intended only
SUBSTITUTE SHEET
to monitor average power levels. In high frequency applications a laser diode package may be constructed in a similar manner, substituting a higher speed photodetector. Figure 4 illustrates an exemplary configuration of adjustable distortion unit 2 of Figures 1 and 2. This adjustable distortion unit comprises transistor 22 having its base coupled to input line 1, its collector coupled a first end of resistor 25, and its emitter coupled to emitter circuit 23. More specifically the emitter of transistor 22 is coupled via line 24 to a first end of resistor 28 and also a first end of a plurality of series resistor diode pairs 24-33. The second end of resistor 25 is coupled to a voltage source. The second end of resistor 29 is coupled to ground and the second end of each series resistor diode pair is coupled to a corresponding output of processor 14. For any given input level on line 1 the gain of the amplifier is given approximately by the ratio of the collector resistance 25 to the total resistance of the emitter circuit 23. Consider now the operation of emitter circuit 23, as the voltage on line 24 rises above the voltage on adjustable distortion unit input line 15a, switching diode 34 becomes forward biased, effectively adding the resistance of resistor 35 in parallel with the emitter resistance of resistor 29. The total resistance of the emitter circuit 23 thus decreases, thereby increasing the gain of the amplifier. Similarly, diodes 36, 37, 38 and 39 turn on as the voltage on line 24 passes the preselected voltage levels imposed on lines 15b, 15c, 15c and 15e. Since the diodes turn on gradually rather than abruptly, by choosing appropriate voltage levels the overall response can be made smooth and of the desired shape.
As diagramed in Figure 4 , the response of the predistortion unit is superlinear, with the gain increasing
SUBSTITUTE SHEET
as the input signal level increases. The response may also be made εublinear over some range of input levels by reversing the polarities Of some of the diodes. In any embodiment of the present invention processor 14 must minimize the voltage level of its input by altering the plurality of voltage levels at its outputs. Processor 14 accomplishes this by executing an optimization algorithm. Any algorithm which can minimize a function of several variables under changing conditions may be used. The function to be minimized may be changing its form slowly with time and may also change abruptly under certain conditions (for example if one of the carriers in a multi- channel system is suddenly turned off) . It is therefore desirable to employ an algorithm which constantly searches the immediate neighborhood of the current minimum. The search must encompass only the immediate neighborhood of the current minimum since each step away from the minimum degrades the linearity of the system slightly. It is therefore necessary to carefully adjust the size of the search region. If the search region is too large the fluctuation in system linearity will be unacceptably large. If the search region is too small the system will not adapt rapidly to changing conditions. During the initial adjustment of the predistortion unit the system may become trapped in a local minimum some distance from the true global minimum. In order to avoid this, an optimization technique that allows tunneling or hill-climbing for escaping from local minima (e.g. a simulated annealing optimization algorithm) may be used. Such a technique is described by Savin AIP Conference Proceedings 151, Neural Networks for Computing, Snowbird UT (1986). The Boltz an machine and the Cauchy machine are appropriate algorithms for the initial adjustment. The previous descriptions of the exemplary embodiments
are provided to enable any person skilled in the art to make or use the present invention. Numerous embodiments which are clearly within the scope and spirit of the invention will be readily apparent to those skilled in the art. Thus, the present invention is not intended to be limited other than by the appended claims. I CLAIM: