WO1993013612A1

WO1993013612A1 - A robust optical signal transmission system

Info

Publication number: WO1993013612A1
Application number: PCT/US1992/011215
Authority: WO
Inventors: Salim N. Jabr
Original assignee: Raychem Corp
Current assignee: Raychem Corp
Priority date: 1991-12-20
Filing date: 1992-12-18
Publication date: 1993-07-08
Anticipated expiration: 1994-06-20
Also published as: AU3421993A

Abstract

A signal transmission system having a switch that enables failed bit channels to be reallocated to bit signals in a manner that keeps this transmission system operable while repairs are made. Such failed bit channel can be reallocated to a least significant bit of data in systems in which the lack of such bit does not unduly degrade operation. A second switch can invert this allocation so that bit signals exit from this system over the same output path, regardless of this reallocation of transmission channels. A simple wavelength division multiplexer/demultiplexer is presented that utilizes the product of fiber length, spacing between adjacent wavelength division multiplexed channels and fiber dispersion to produce a desired temporal offset of bits in different channels to be achieved.

Description

A ROBUST OPTICAL SIGNAL TRANSMISSION SYSTEM

Field of the Invention

This invention relates to signal transmission systems, such as those utilized to distribute telephone and/or video signals to customers in their homes, and relates more particularly to such a system that enables defective components to be replaced without interrupting transmission of signals over such a system.

Background of the Invention

In the figures, the first digit of a reference numeral indicates the first figure in which is presented the element indicated by that reference numeral.

The use of optical fibers in long distance telecommunications has expanded dramatically in recent years. Optical fibers exhibit several advantages over the copper wires that traditionally have carried telecommunications signals. Optical fibers exhibit much greater bandwidth than copper, thereby enabling an optical fiber to carry data at a much higher rate than can be carried by a pair of copper wires. Optical fibers also exhibit much less signal attenuation than pairs of copper wires, thereby enabling optical signals to be carried over much longer distances before amplification is required. Fiber cables are also much less bulky than copper cables, thereby making the installation process easier. In addition, optical fiber systems are virtually immune to electromagnetic interference that causes static or other line noise in copper systems. Therefore, the quality of transmission through such optical fiber systems is much higher.

A basic optical fiber transmission path consists of two light- carrying optical fibers — one for transmission in a downstream direction along this transmission system and one for transmission in the opposite, upstream direction. Each fiber utilizes an optical source, such as a diode laser, at the transmission end of the fiber to inject optical signals into the fiber and utilizes an optical receiver, such as a photodiode, at the reception end of the optical fiber to receive such optical signals.

It is advantageous to transmit data in parallel along more than one transmission path to increase the rate of data transmission. For example, when several different data signals are to be transmitted, each signal can be transmitted over a parallel array of transmission paths. Examples of such a parallel array of transmission paths include: a parallel array of optical fibers; a plurality of different wavelength channels transmitted through a single optical fiber; and a combination of these two types of parallel transmission paths. Because useful information is transferred only when each complete word of data is transferred, it is advantageous for the parallel data to consist of an integral number of words of data.

Because it is important that all parallel-transmitted bits of data for each word of transmitted data arrive at the receiver at substantially the same time so that there is a clear indication as to which word each bit is a part, it is important that the transit time for each bit be nearly identical for all bit paths for each word of data. Since small differences in path length of different fibers will often introduce an unacceptable amount of skew between bits in any given word, it is advantageous to transmit each word of data as wavelength division multiplexed bits of data. In this format, each bit of parallel data is transmitted at a uniquely associated wavelength through a single optical fiber. This avoids the introduction in any word of bit skew arising by differences in lengths of fibers in a parallel bundle of optical fibers. This type of transmission is known as wavelength division multiplexing (WDM).

Optical fiber communication links are utilized in a wide variety of applications, including the transmission of binary computer data, transmission of telephone conversations and transmission of television programs. In order to avoid running separate optical fiber links for each of these types of data, it is advantageous to provide optical communication links that can carry all three of these types of data. For example, Raynet's LOC^TM-2 fiber optic system provides superior telephone service and can be upgraded to carry broadband integrated services digital network transmissions without the need for additional optical fiber. All that is needed for such an upgrade is to upgrade the electronics of the system.

Because of the wide bandwidth of optical fiber communications, many data signals are typically multiplexed into a single optical fiber. At the present time, each telephone customer in the United States is coupled by a pair of copper wires to the phone distribution network. Each of these pairs of wires for a local cluster of customers is connected to a subscriber interface unit (SIU) that multiplexes these signals for transmission within the phone distribution network. Increasingly, these multiplexed signals from the SIU are carried over optical fiber links, through other intermediate transmission centers and finally to other subscriber interface units, thereby enabling the customers to communicate with one another. When it becomes economically feasible, even these local copper pairs will be replaced by optical fibers.

In another method of multiplexing data, referred to as time division multiplexing (TDM), bits of data from a set of N different data signals are interleaved temporally to produce a serial, binary data signal. In time division multiplexing, the data can be interleaved a bit at a time so that the nth signal consists of the data in bit locations p*N + n, where n == 1, ..., N and p = 0, 1, 2, .... The data can also be interleaved in clusters of M bits from the same data signal (e.g., a word at a time for the case of M = 8) so that the pth cluster of data for the nth signal consists of the data in bit locations p*N*M + n*M + m where p = 0, 1, 2, ..., where n = 0, ..., N-l, and where m = 0,...,M-1. All of those bits of data corresponding the same bit position in data words (i.e., to a fixed value of m) are referred to herein as the "mth bit signal". The physical path of transmission of this bit signal through a signal transmission medium, such as an optical fiber or bundle of optical fibers, is referred to herein as the "mth bit channel". All of those bits of data corresponding to the same data signal (e.g., to a fixed value of n in this time division multiplexed scheme) are referred to herein as a "data channel". Each data channel consists of those words that have been multiplexed from a single data transmission.

Because wavelength division multiplexing produces parallel transmission of bits from each of the N different data signals, whereas time division multiplexing produces serial data, the data transmission rate is usually greater for wavelength division multiplexing. However, because optical fibers are typically dispersive, wavelength division multiplexing has the disadvantage that the signal transmission velocities at the different wavelengths are generally unequal. Therefore, those bits that are transmitted concurrently will not typically be received concurrently — i.e., there will be bit skew between the different wavelength signals.

The article by Loeb, et al entitled High-Speed Data Transmission on an Optical Fiber Using a Byte-Wide WDM System, Journal of Lightwave Technology, Vol. 6, No. 8, August 1988, shows the surprising result that, because of this skew, the maximum aggregate bit rate is the same for a serial data link as for a bytewide, wavelength division multiplexing (WDM) system. This maximum aggregate bit rate occurs when the most delayed pulse in one word just overlaps the most advanced pulse in the succeeding word. This shows that, for WDM systems that have been pushed to the maximum rate allowable in view of such skew, such a WDM system exhibits the same data transmission rate as a serial system. For such a situation, increased maximum aggregate bit rate by WDM transmission requires that bit skew be at least partially corrected.

U.S. patent 4,677,618 entitled Method And Apparatus For Deslewing WDM Data Transmitted Through A Dispersive Medium issued to Lee C. Haas, et al on June 30, 1987 provides a device in which bit skew is substantially eliminated. At the receiving end of the WDM device, the wavelength division multiplexed signal is demultiplexed to produce a sequence of M-bit words. Because of the skew, the bits in each word do not arrive concurrently. To recover such concurrency, the data stream for each bit position is subjected to an adjustable delay sufficient to eliminate the skew between these bits. The amount of skew needed for each bit is determined from two pieces of data: a measured skew between a pair of reference bits; and the known frequency-related dispersion characteristics of the transmission medium.

Parallel data transmission systems, such as wavelength division multiplexing (WDM) systems and data transmission over parallel fibers, suffer from an increased amount of down time, because a separate optical source (e.g., a photodiode) is required for each wavelength channel. For an N-bit wide WDM system the rate at which one of these diodes fails is, on the average, N times the rate for a serial data system, thereby producing an N-fold increase in down time. It would therefore be advantageous to reduce or even eliminate the amount of system down time produced by failures of these sources.

Summary of the Invention with Objects

In accordance with the illustrated preferred embodiment, a multichannel signal transmission system is presented having a reduced amount of down time. Each channel consists of a signal transmission medium, an associated signal source for transmitting digital data through that medium in a downstream direction and a receiver for receiving signals transmitted through that transmission medium. Such transmission medium for a given channel can be, for example, a separate optical waveguide carrying data only for that channel, can be a particular frequency of light in a wavelength division multiplexed transmission of multiple bits over a single optical waveguide or can be a combination of these parallel transmission channels.

Because incorrect data will be transmitted if any of these signal sources, transmission channels or signal receivers malfunctions, such a system will generally need to be repaired when any of these signal sources, transmission media or receivers fails. The channel that contains this failed source, transmission medium or receiver will be referred to herein as the "failed bit channel". This improved multichannel transmission system includes a mechanism that enables reassignment of signal sources to channels in such a manner that, once a malfunction of one of these sources or receivers is detected, a reassignment is made in a manner that makes this system operable while that signal source or receiver is repaired or replaced. A first switch is located in the signal path upstream from said signal sources to reallocate incoming signals to these signal sources. A second switch is located in the signal path downstream from said receivers to reallocate signals produced by the receivers. This second switch inverts the allocation so that, when the failed bit channel is repaired, the data exiting this signal distribution system will be unaffected by this reallocation of signals to these channels.

The inclusion of these switches not only enables a failed bit channel to be assigned to an ignorable bit signal, it also enables 8! (approximately 40,000) reassignments of bit channels to bit signals for purposes of addressing subscribers and interdiction.

In a particular class of embodiments, the number N of channels is equal to an integral number times the word length of the data so that one or more words of data are transmitted in parallel along these channels. Each of these channels carries data for a single bit position and is referred to herein as a "bit channel". The reallocation of signals to channels is selected to allocate the failed bit channel to the least significant bit (LSB) of incoming data. This is a suitable choice when an error in the least significant bit does not significantly impact device operation. This is true for the case of transmission of video signals, because errors in the least significant bit do not significantly degrade the displayed video picture. Indeed, if device operation is not unduly disturbed by faulty data in any of the k least significant bits, then this failed bit channel can be allocated to carry any one of the k least significant bits.

Preferably, each of these two switches is an NxN matrix switch that can make any of the N! possible reaUocations of incoming signals to channels of the multichannel distribution system, because this enables any pair of channels to be interchanged for purposes of carrying data. This enables the simplest possible reallocation to be implemented — namely, the failed bit channel and the selected one of the k least significant bit channels are functionally interchanged by these two switches. Thus, if the pth channel fails, this pth channel can be interchanged with any channel carrying data that can be deleted without undue degradation of data transmission. Which channels satisfy this criterion is very dependent on the data being transmitted. The above- indicated example of the least significant bit of data in transmissions of numerical data is only one example of this type of data. If a less versatile switch is utilized, then it will typically be necessary to reassign more than two channels.

The existence of a failed bit channel can be detected by observing the occurrence of a string of more than Q successive zero bits at one of the receivers at a time during which nonzero bits are being received on at least one other bit channel. Q is selected to be large enough that the probability of such a string occurring without that bit channel having failed is negligible. If there is no correlation between values of a given bit in successive words, then the probability of Q successive bits being zero in a bit stream through a functional bit channel is 2-Q. For such an occurrence to occur no more than once a year on the average, this requires that Q be on the order of or greater than 50. In response to this detected failure of a bit channel, signals are transmitted to both switches, directing that they reallocate channels. If errors persist at the detector side even after the bit channels have been reallocated to allocate the failed bit channel to the least significant bit signal, then the failure is due to some source other than the failure of the signal source or the signal detector for the failed channel. For example, the failure could occur because an optical fiber carrying several wavelength division multiplexed bit signals has broken. Additional detection circuitry is therefore included to pinpoint the actual source of the failure.

Multiple data channels can be carried over this system: by including a sufficient number of parallel bit channels that more than one data channel can be carried in parallel; and/or by time division multiplexing these data channels. For example, either or both of these techniques can be used to transmit many cable television channels over a single optical fiber.

In order to ensure that a word transmitted in the pth data channel is not incorrectly detected as being in some other data channel, at least one of these data channels transmits only a special synchronism word that is recognized by special circuitry to assist in maintaining synchronism between the transmission and reception of data words. A particularly simple synchronism scheme involves transmitting only the NULL word (i.e., the word containing only zeroes) over two adjacent data channels tacked on to a string of data channels. These two data channels can be considered to be the last two data channels. This scheme is simple because the occurrence of the NULL word of length N is inexpensively detected by an N-input OR gate. An equally simple embodiment involves transmitting only the MAXIMUM word (i.e., the word containing only ones) because the MAXIMUM word of length N can be inexpensively detected by an N-bit AND gate.

For the particular case in which these different data signals correspond to data for P different television channels, it is advantageous to separate temporally those bits corresponding to channels that are actually being received by televisions in a particular household. By maximizing this temporal separation between R channels that are actually being received, the electronic circuitry utilized to receive and utilize such data will have a maximal amount of time for successive words of received data to settle before being utilized. This enables use of less expensive electronics in the signal transmission and reception systems. This technique can be applied to transmission of pay-per-view televisions because the television station will be aware of which channels are being transmitted to its customers. Ideally, in each sequence of P words, the transmitted words for these R channels are spaced from one another by at least P mod R positions in this sequence.

This separation of actually received data signals can be approximated by sequencing the channels such that all channels carrying the same type of programming are clustered together temporally within the transmission. For example, the first Ni channels can be allocated to the Ni news channels, the N2 child programming channels can be allocated to the next N2 channels, etc. Under the assumption that each household will not have more than one television tuned in to the same class of programming, such allocation of channels will generally temporally space apart the actually received channels.

Because conventional optical fiber (i.e., germanium doped silica fiber) exhibits low absorption in the wavelength ranges from 1280 to 1360 nm and from 1480 to 1580 urn, the optical data signals are preferably transmitted within one or both of these two ranges. Preferably the wavelength differences between any two of the wavelengths utilized for wavelength division multiplexing is at least one nanometer so that such wavelength multiplexed data can be demultiplexed by inexpensive multiplexers. In particular, this minimal spacing enables the use of inexpensive demultiplexers utilizing filters to separate these multiplexed signals. Each bit channel has a wavelength spread that is much less than the spacing between channels so that at an output end of these channels the bit channels do not overlap significantly in wavelength. Typically, a wavelength spread of less than 0.1 times the wavelength spacing between channels avoids significant signal crossover between adjacent bit channels.

In some embodiments of this invention, it is advantageous to include an optical signal multiplexer at an input end of the signal transmission system to convert serial data into parallel data and/or to include an optical signal demultiplexer at an output end of the signal transmission system to convert the parallel data transmitted through this system into serial optical output data. A particularly simple embodiment for this multiplexer or demultiplexer utilizes an optical fiber having a known dispersion distribution and known length. The wavelength spacing between each pair of adjacent wavelength channels is selected such that the resulting bit skew between adjacent channels equals the period of a bit. The choice of cladding of this optical fiber can be selected to produce a time of transmission through this fiber that is an approximately linear function of wavelength. For such an embodiment, the wavelength differences between adjacent wavelength channels will be substantially equal. This device can be used as either a multiplexer or as a demultiplexer. Li the multiplexer embodiment, the allocation of bit channels to wavelengths is in the reverse order as in the demultiplexer embodiment.

Brief Description of the Drawings

Figure 1A illustrates a transmitter portion of an optical signal embodiment of the robust signal transmission system.

Figure IB illustrates a receiver portion of an optical signal embodiment of the robust signal transmission system.

Figure 1C illustrates an alternate embodiment of the receiver portion of robust signal transmission system, adapted to receive concurrently three different signals.

Figure 2 illustrates a suitable SYNCH generator for use in the circuit of Figures 1A and IB for a system utihzing a particular type of synchronization signal.

Figure 3 illustrates an alternate embodiment of the SYNCH generator of Figure 2. Figure 4 illustrates a circuit suitable for detecting whether a bit channel has failed.

Figure 5 is an alternate embodiment of a circuit suitable for detecting whether a bit channel has failed.

Figure 6 illustrates a new, inexpensive optical signal demultiplexer that not only functions as a wavelength division demultiplexer, but in addition converts parallel data to serial data.

Figure 7 is a timing diagram for signals in Figures 1A - lC.

Description of the Preferred Embodiment

Figure 1A illustrates a transmitter portion of an optical signal embodiment of the robust signal transmission system and Figure IB illustrates a receiver portion of this robust signal transmission system. This particular signal transmission system is designed to receive one hundred electrical, wavelengthdivision-multiplexed, bit-serial, analog signals and convert them into one hundred demultiplexed, optical, word-serial, digital signals. Therefore, a multichannel, analog receiver 11 and one hundred analog-to-digital (A/D) converters 12 are included to receive these signals and convert them into eight hundred digital signals. Eight multiplexers (MUXs) 13 are each coupled to an associated output bit output of these one hundred A/D converters such that the kth of these multiplexers multiplexes the kth least significant bits from these one hundred different signals. In other embodiments, the number of multiplexers 13 can differ from eight. Indeed, an embodiment with nine multiplexers carries 9-bit data which is sufficient for video quality transmission (i.e., the signal to noise ratio is 60 dB). Eight-bit data exhibits only a distribution level of quality (i.e., the signal to noise ratio is only 45-50).

The timing of the multiplexers is controlled in response to a clock signal from a 1 GHz clock 14. Because each multiplexer needs to time domain multiplex one hundred different digital signals, each multiplexer operates one hundred times faster than each A/D converter. Therefore a divide-by-100 counter 15 is included between clock 14 and A/D converters 12 to provide to each A/D converter 12 a CONVERT signal in response to which analog data is sampled and converted to digital data.

Each of the eight time-multiplexed signals from multiplexers 13 is transmitted to a first switch, such as a first 8x8 matrix switch 16, that can reallocate each of these signals among a set of eight bit channels 17. This 8x8 matrix switch can implement any of the 8! possible reaUocations of bit signals to the 8 different bit channels. Each bit channel 17 includes: a signal transmission medium, such as electrical connections 18,19 and 110 and optical fibers 111 - 113 shown in Figures 1A and IB; a signal source, such as laser diode 114; and a receiver, such as the combination of a photodiode 115 and associated ampUfier 116, that can operate at the 1 GHz data transmission rate. Elements 18,111,114 function as a paraUel array of input signal paths and elements 19, 110, 113, 115, 116, 132 function as a paraUel array of output signal paths. In this embodiment, which utiUzes wavelength division multiplexing, a wavelength division multiplexer 117 and a wavelength division demultiplexer 118 are included to enable the eight bits of paraUel data in the input signal paths to be carried as 8 paraUel wavelength channels in single optical fiber 112. Multiplexer 117 can be a Nxl fused biconic coupler.

Each of these bit channels is coupled by electrical connections 110 to a switch, such as a second 8x8 matrix switch 119, which inverts the aUocation of data to bit channels such that, when the failed bit channel is repaired, the data exiting this signal distribution system wiU be unaffected by this reaUocation of signals to these channels. For example, if the kth data bit channel faUs, then 8x8 matrix switch 16 is activated to direct the kth data bit signal through the bit channel originaUy utilized to transmit the least significant bit of data and to direct the least significant bit signal to the fatted bit channel (i.e., the kth bit channel). The second 8x8 matrix switch 119 inverts the reallocation of signals so that each of these signals exits from switch 119 over the same electrical wire 120 as if there had been no reaUocations by these two switches.

In some embodiments, switch 119 is omitted. In such embodiments, switch 16 is unable to reaUocate bit channels to bit signals to avoid down time upon failure of a bit channel. Instead, switch 16 is utilized to code the transmitted signal either for purposes of scrambling or interdiction. Indeed, even in embodiments having the second switch, both switches can be operated to implement scrambling and interdiction as weU as to avoid down time for repairs.

Elements 115-128 can be included in a decoder located in the premises of the end user or nearby in equipment shared by several users. Electrical wires 120 connect switch 119 to a D/A converter 121 that converts the digital data back to analog data. Each D/A converter 121 includes at its input a latch that holds onto the digital input data until a subsequent digital input is strobed into this latch. A SYNCH generator 122 extracts a SYNCH signal from the digital data signals on electrical wires 120.

This SYNCH signal is passed through an adjustable delay circuit/tuner 123 that introduces a selected delay of between O and 100 nanoseconds into the SYNCH signal before it is applied to a trigger input of D/A converter 121. The amount of delay is selected to synchronize the SYNCH signal with the arrival of each word of data for the channel being received by the end user. Because the output signal from divide-by- 100 counter 15 strobes successive sets of 100 words into A/D converters 12 at 100 nanosecond intervals, this 100 nanosecond range of variable delay is sufficient to ensure that the clocking of D/A converter 121 can be synchronized with the data channel to be received. This is advantageous because no switches are required to select which channel is to be received. Instead, selection of the channel to be received is achieved by synchronizing the 10 MHz SYNCH signal with the reception of bits from this channel so that only these bits of data are strobed into D/A converter 121.

In order to achieve such synchronism, time-multiplexed data channels 99 and 100 transmit only the NULL word (i.e., the word having only zeroes). This pair of successive NULL words functions as a synchronization signal that indicates the end of the 100 channels of data. This choice of synchronization signal is advantageous because it can be detected by a particularly simple SYNCH generator 122 consisting of: an 8-input NOR gate 21, a 2-bit shift register 22 and a 2- input NAND gate 23, connected as in Figure 2. The output signal from NOR gate 21 is high only if the electrical input wires 120 to D/A converter 121 are aU zero. The 2-bit shift register 22 is responsive to the clock signal from 1 GHz clock 14 so that each successive word is tested to see if it is the NULL word. The signal on output 24 of AND gate 23 is high only if two successive words are NULL words. Therefore, SYNCH generator 122 produces a 1 bit pulse each time the 100th word is detected by detectors 115. Adjustable delay/tuner 123 introduces a delay of c+ 1 nanoseconds if the user selects channel c for reception. This delays die recovered clock signal such that D/A converter converts data bits for channel c.

Figure 3 iUustrates an alternate embodiment of SYNCH generator 122 for use in conjunction with a synchronization signal (SYNCH) consisting of a pair of MAXIMUM words (i.e., a word consisting of aU ones) in a row. This embodiment utilizes an AND gate 31 in place of the NOR gate of the embodiment of Figure 2.

The analog output signal from D/A converter 121 passes through a low pass filter 124 that smooths the step-function type of output from D/A converter 121. Because this embodiment is adapted for use in transmission of video signals to end users, the filtered signal from low pass filter 124 is utilized by a modulator 125 to modulate a carrier signal which is appKed to channel 3 or 4 of a television set 126. In other embodiments, the 8x8 switches 16 and 119 can be replaced by a less expensive, less complicated switch. For example, a switch that cyclicly permutes the channels by a selectable number k, is able to reaUocate the channels such that any failed bit channel k is aUocated to the least significant bit channel 0. Such a switch operates such that the pth bit signal is directed through the (p-k) MOD 8 bit channel 17, where k is the number of the failed bit channel and p = 0,...,7. However, it is advantageous to utilize 8x8 matrix switches 16 and 119 because of the added flexibility that these switches provide. For example, such switches enable implementation of a reaUocation of bit channels such that the zeroth and the kth channels are interchanged and the other channels are unaffected. This is a simpler reallocation and therefore has less potential for disrupting transmission. When the failed bit channel is repaired, the original aUocation of bit signals to bit channels can be reestablished, but such is not necessary because the properly operating system wiU operate equally well in any of the 8! different aUocations by switches 16 and 119.

A failed bit channel detector 127 is included at the end of each bit channel 17 to enable the occurrence of a failed bit channel to be detected. Each of these failed bit channel detectors is connected to a controUer 131 that controls 8x8 switches 16 and 119 to automate reaUocation of channels such that the failed bit channel is reaUocated to the least significant bit signal. This controUer can be located in either the receiver section presented in Figure IB or can be included in the transmitter section presented in Figure 1A, as is the case for this particular embodiment.

In this embodiment, the signal from failed bit channel detector 127 to controller 131 is transmitted through optical fiber 112 from the receiver section to the transmitter section. In other embodiments, this signal from the failed bit channel detector 127 can be carried over an optical fiber dedicated to the transmission of these system control signals. This signal is detected by a dedicated detector (not shown) which transmits an electrical to controller 131 carrying the information from the failed bit signal detector 127. Similarly, the signal from controUer 131 to switch 119 is most conveniently transmitted through optical fiber 112 so that controUer 131 is coupled to one of transmitters 114 or to a dedicated photodiode (not shown) which converts the electrical signal from controUer 131 into an optical signal carried through fiber 112. A dedicated detector (not shown) can be used to convert this optical signal into an electrical signal applied to switch 119.

A particularly simple embodiment of a failed bit channel detector is shown in Figure 4. This detector consists of a Q-input NOR gate 41 having each input connected to a different output of a Q-bit shift register 42. An alternate embodiment of the failed bit channel detector is iUustrated in Figure 5. This embodiment consists of a counter 51 which has a reset input connected to a line 120 and has an output on which an ALARM signal is produced if Q successive data bits on line 120 are aU zero. This counter can take the form of a Q-state countdown counter that is clocked at each cycle of clock 14. An ALARM signal goes high whenever counter 51 counts down to zero. This occurs whenever Q successive zero bits occur on the channel being monitored by this failed bit channel detector. This counter can also take the form of a count-up counter that resets to zero whenever a high bit is detected on the reset input. A particularly simple embodiment of this has its output connected to the most significant bit of this counter and requires that Q = 21 for some integer q.

Q is selected to be large enough that the probability of such a string occurring without that bit channel having failed is negligible. If there is no correlation between values of a given bit in successive words, then the probably of Q successive bits being zero in a bit stream through a functional bit channel is 2^Q. For such an occurrence to occur no more than once a year on the average, this requires that Q be on the order of or greater than 50.

In response to the detection that a bit channel has failed, a signal s transmitted from failed bit channel detector 127 to a controUer 131 that is utilized to determine the allocation of bit channels to bit signals produced by 8x8 switches 16 and 119. For the optical embodiment of Figures 1A and IB, such detectors 127 produce an optical signal that is transmitted upstream through optical fiber 112 to an optical detector 128 which transmits an electrical signal indicating which bit channel has failed. This system is successful for those cases in which the data transmission failure has occurred by failure of a laser diode 114 or a photodiode 115, but it is not successful when the data transmission failure occurred because of the failure of the optical fiber 112 over which this failure signal is transmitted upstream to detector 128.

Therefore, adjacent to each laser diode 114 is a photodetector 129 that detects light from its associated photodiode. These photodetectors 129 are connected to a second failed bit channel detector 130 that, in response to signals from photodetectors 129, detects which of laser diodes 114 has failed. In general, laser diodes exhibit some leakage of light through the back of the laser so that photodetectors 129 can be responsive to such light leaking from the back of a laser diode. In addition, breakage of fiber 112 will prevent signals from bit channel detectors 127 from reaching 8x8 switch 16 so that some mechanism is also required for informing controller 131 of such a failure. Because a broken fiber will reflect a portion of light transmitted down that fiber, detector 128 can also be used to detect such reflected light, thereby detecting that fiber 112 is broken.

In those cases where only a few channels are ever received by an end user at any given time, it is advantageous for the receiver section of the data transmission system to be able to receive aU of these channels concurrently. In the embodiment of Figure IB, the delay/tuner is responsive to user input that identifies which channels are to be received. For example, in a household having cable television, the cable television channel selector is adapted to produce an electronic signal which is applied to delay/tuner 123 to control the amount of delay introduced into the SYNCH signal. In an embodiment adapted to receive more than one different channel at a time and forward each received channel to the television for which that channel is selected, the delay/tuner will, in response to the SYNCH signal, produce an output signal that is a time multiplexed combination of three delayed SYNCH signals, each having the delay required to synchronize that delayed SYNCH signal with the arrival of each word of data for that channel.

Unfortunately, if two channels are received on two time division multiplexed channels that are adjacent or nearly adjacent temporaUy, then an expensive D/A converter may be required to provide the D/A conversion speed needed in such an embodiment. However, if these actuaUy received channels are sufficiently separated within the time multiplexed transmission of these signals, then less expensive components can be utilized in the A/D converters 12 and D/A converters 121. For example, for the case of pay-per-view television signal transmission to households, it is unUkely that more than three channels wttl be received concurrently because it is very unlikely that a household has more than three televisions and it is even more unlikely that more than three different channels wiU be received concurrently even if there are more than three televisions in that household. Therefore, a switch 133 can be included between analog receiver 11 and A/D converters 12 to reaUocate these three different actuaUy received signals to channels that are approximately thirty three timemultiplexed channels apart. In general, if R channels are to be maximaUy separated in a sequence of P different channels, then aU separations between channels should be at least P mod R channels.

To achieve optimized temporal separation of these actuaUy received channels, a 100x100 matrix switch 133 can be included to enable the video channels to be aUocated to time multiplexed channels such that the actuaUy received channels are maximaUy separated. Since the data channels transmitted over a given optical fiber 112 are typicaUy received by a local cluster of homes, this aUocation must be made in a manner that maximizes separation between actuaUy received channels for each household, subject to the constraint that such optimization is for aU of the households in this cluster.

Since the cost of such a switch can be more than the savings from the use of inexpensive components in the A/D converters 12 and D/A converter 121, a simpler, less expensive mechanism for temporaUy separating these channels is advantageous. In one particular approach, all television programs of the same subject matter type are transmitted in a single contiguous cluster of transmitted channels. For example, all news programs are in one cluster, aU sports programs are in another cluster aU music programs are in a third cluster, et cetera. This clustering is based on the assumption that any given household will not be receiving two different programs from the same cluster. If this assumption is violated, then at least one of the D/A converters will have less than its desired settling time, so that the signals for that channel will be somewhat degraded.

Figure 1C is an alternate embodiment of the receiver section of Figure IB, adapted to receive up to three different channels concurrently. In this embodiment, delay tuner 122 produces three different delayed SYNCH signals, each being applied to an associated one of three D/A converters 121. The output of each of these D/A converters 121 is directed through an associated low pass filter 124 to an associated modulator 125 which generates the signal for its associated television set 126. The channel selector for each television set controls the amount of delay introduced into the delayed SYNCH signal for that television set. The amount of delay is such that the delayed SYNCH signal for that television set arrives concurrently with the time division multiplexed data for the selected data channel. In response to the SYNCH signal, for a given television set, data is strobed from lines 120 into the D/A converter for that television set. Each of these D/A converters has available substantiaUy the full 100 nanoseconds between reception of successive bits of data for the data at the D/A converter's input to settle before D/A conversion need be initiated. Therefore, inexpensive D/A converters can be utilized in this embodiment. For conventional optical fibers 111, 112, and 113, the wavelength range of transmission is preferably 1280 to 1360 nm or 1480 to 1580 nm because conventional optical fiber (i.e., germanium doped silica fiber) exhibits low absorption in these two ranges. For conventional cladding on such fibers, the range from 1280 to 1360 nm exhibits less dispersion than the range from 1480 to 1580 nm. Therefore, it is generaUy advantageous to limit the wavelength range of transmission to the 1280 to 1360 nm range. However, in "dispersion shifted" optical fibers, the cladding is doped in a manner that alters the dispersion transmission in the optical fiber such that the range from 1480 to 1580 nm exhibits less dispersion than the range from 1280 to 1360 nm. Such dispersion shifted fiber can therefore be utilized to transmit data within the 1480 to 1580 nm range.

The individual wavelengths generated by laser diodes 114 are spaced 4 nm apart for ease of demultiplexing. The actual wavelengths used in the preferred embodiment are 1525 nm, 1529 nm, etc. Because of this separation between wavelength channels, demultiplexer 118 can be a grating/fiber device or a FabryPerot device, which have each been demonstrated to have a 2 nm resolution over this transmission range. Such demultiplexers are relatively expensive, but are commercially available.

Figure 6 ttlustrates a new wavelength division demultiplexer that is not only less expensive than the commerciaUy available demultiplexers, it also converts the paraUel data into serial optical data. This demultiplexer utilizes the dispersion of optical fiber 112 to introduce relative delays between data bits such that a set of bits transmitted in paraUel by photodiodes 14 are received as a serial sequence of data. The bit signals are assigned to wavelength channels such that the kth bit signal is transmitted over that wavelength channel exhibiting the kth most transmission time through the optical fiber, whereby the bit signals experience relative delays through this fiber such that these bits emerge from the exit end of the optical fiber sequentially in order from the most significant bit to the least significant bit. This demultiplexer can be utilized in place of elements 19, 113, 115, 116 and 118 of Figure IB.

For a 1 km length fiber 112 having a dispersion D of 15.4 picoseconds/nmkm, pulses injected into fiber 112 concurrently and separated in wavelength by 4 nm from one another will exhibit a 61.6 picosecond separation between successive pulses at the exit end of fiber 112. Because the length of fiber 112 can vary from its nominal length of 1 km by a significant amount, fiber 112 is coupled to a fiber 61 of length that has been adjusted to compensate for any difference between the actual and nominal lengths of fiber 112. This ensures that the desired temporal spacing between pulses within the demultiplexer is achieved.

Differences in the lengths of fibers 113 can also introduce spurious relative delays between signals passing through each of these fibers. In the signal path of each of these fibers is a delay element 132 that compensates for such differences. In this embodiment in which such delay element is attached to fibers 113, each of these delay elements consists of a fiber section of length selected to eliminate these unwanted relative delays. Alternatively, these delay elements can be elsewhere in this circuit, such as in the electrical connections 110 to introduce electronicaUy a compensating delay in each of these paths to eliminate unwanted relative delays.

Fiber 61 includes a plurality of optical taps 62 that are bent in a sufficiently smaU loop (on the order of 3 mm diameter bend for a 9 micron diameter fiber core) that a significant fraction (on the order of 10%) of the Ught at that point exits from each bent section. Adjacent to each bent section is an associated photodetector, such as photodiode 63, which receives light from that bent section. In the optical path between each bent section and its associated photodetector is a filter that passes only a single associated one of the wavelengths of light transmitted through fiber 112. For this particular embodiment having eight different colors of Ught at the wavelengths 1525 nm, 1529 nm, 1533 nm, ..., 1553 nm, each of these wavelengths has an associated filter 64 that passes only Ught of that wavelength, whereby each photodetector 63 receives data from only a single bit signal. Each of the detectors 63 is 5 connected to an associated sample and hold circuit 65 which is triggered by the delayed SYNCH signal from SYNCH generator 122, after suitable delay by delay/tuner 12. Thus, aU of the bit signals are sampled concurrently in response to the SYNCH signal.

10 In an alternate embodiment, filters 64 can be eliminated by appropriate timing of the delayed SYNCH signal from the Delay/timer 123. In this embodiment, the sample and hold circuits 65 are strobed sequentiaUy at times selected to coincide with the arrival of a data bit in its associated bit channel. Because the product DLS was selected to

15 produce a desired equal temporal separation between pulses in adjacent bit channels, detectors 63 are sequentiaUy strobed at times separated by this temporal separation between pulses, whereby each bit channel is sampled only by its associated detector 63 at the time when a pulse in that bit channel can arrive. This desired temporal separation is in

20 general selected to be equal to a pulsewidth P of these pulses. Because only a single detector is sampled at any given time, there is no need to use filters to block Ught from other bit channels from reaching those detectors that are not associated with that bit channel.

^"25 Because the length of fiber 112 can vary from customer to customer and because the amount of temporal separation required between successive timemultiplexed channels can differ from embodiment to embodiment, a section 66 of optical fiber of adjustable length is included in the bit channels to enable a selected channel path

30 length to be achieved so that a selected temporal separation between adjacent time division multiplexed bit channels to be achieved at an output end of the demultiplexer.

It should be noted that this demultiplexer can also function as a 35 multiplexer that converts wavelength division multiplexed, serial data into paraUel data. This is achieved by: (1) allocating channels such that the kth serial bit is transmitted through the wavelength channel having the kth longest transit time and by selecting the product DLS of the fiber dispersion D, the total length L of fibers 112 and 66 and the wavelength separation S between adjacent wavelength channels such that these bit signals emerge in paraUel from the exit end of the optical fiber. In embodiments in which the dispersion D varies over the length of optical fiber 112, the factor DL in the product DLS above is equal to the spatial integral over the length of this fiber of the dispersion.

Claims

What is claimed is:

1. A signal transmission system for transmitting an integral number N of bit signals, said signal transmission system comprising:

a set of N paraUel bit channels each having a transmitter for transmitting an associated bit signal through that channel; and

a first switch having N inputs and also having N outputs each of which is coupled to an input end of a uniquely associated one of these N bit channels;

said first switch being controUably able to aUocate these N bit channels to the N bit signals such that each bit channel transmits one of these N bit signals and any selected one of the bit channels can be aUocated to transmit a preselected one of the bit signals.

2. A signal transmission system as in claim 1 wherein said paraUel bit channels comprises:

N input signal paths;

a first optical fiber, in which each bit signal is carried over a uniquely associated wavelength division bit channel;

N output signal paths;

a wavelength division multiplexer coupled between an output end of the input signal paths and an input end of the optical fiber; and

a wavelength division demultiplexer coupled between an output end of the optical fiber and an input end of the output signal paths. l 3. A signal transmission system as in claim 2 wherein:

2

3 said first optical fiber is selected from the set of dispersion shifted

4 optical fibers and nondispersion shifted optical fibers; and

5

6 said wavelength division multiplexer produces optical signals

7 within a range selected from the set consisting of 1280-1360

8 nanometers and 1480-1580 nanometers.

9 l 4. A signal transmission system as in claim 2 wherein:

2

3 said wavelength division bit channels are separated by at least 1

4 nanometer and each has a wavelength spread much less than 1

5 nanometer; and

6

7 said demultiplexer includes a plurality of spectral filters to

8 separate the wavelength division multiplexed channels from one

9 another.

1 5. A signal transmission system as in claim 2 wherein said wavelength division demultiplexer comprises:

a section of optical fiber between said optical fiber and said output paths, said section of optical fiber having a length selected

6 such that the sum of its length and the length of the first optical fiber is equal to a fixed value L;

this length L is selected in conjunction with a wavelength spacing S between adjacent wavelength channels and a dispersion D of the first optical fiber and this section of optical fiber such that pulses introduced in paraUel at an input end of the first optical fiber into adjacent wavelength channels arrive at an output end of this section of optical fiber with a spacing such that pulses in these channels are not overlapping. 6. A signal transmission system as in claim 5 wherein DLS is substantiaUy equal to a width P of digital bits in these transmitted signals.

7. A signal transmission system as in claim 1 wherein said paraUel bit channels comprise:

N input signal paths;

N output signal paths; and

N optical fibers, each uniquely coupling an output end of one of the input signal paths to an input end of one of the output signal paths.

8-. A signal transmission system as in claim 1 wherein said transmission system further comprises:

means for detecting whether a bit channel has fatted and, if a channel has failed, then determining which bit channel has failed.

9. A signal transmission system as in claim 8 wherein said first switch, in response to a signal from said means for detecting whether a bit channel has failed, reaUocates bit channels to bit signals.

10. A signal transmission system as in claim 9 wherein said means for detecting whether a bit channel has failed includes:

a controUer that determines which channel has failed and transmits to said first switch a signal in response to which the first switch reaUocates bit channels such that a least significant bit signal is aUocated to a failed bit channel. 1 11. A signal transmission system as in claim 8 wherein said means for

2 detecting whether a bit channel has failed comprises:

3

4 for each channel, means for detecting if a sequence of P

5 successive zero bits have been received, thereby indicating that

6 such channel has failed.

7 l 12. A signal transmission system as in claim 8 further comprising: 2

3 a second switch, coupled to an output end of these N bit channels,

4 that controUably allocates these N bit channels to N outputs of this

5 second switch in a manner such that each bit signal is directed to a

6 fixed choice of the N outputs for all aUocations by these first and

7 second switches, whereby this second switch implements a

8 reaUocation of bit channels to bit signals in a manner inverse to

9 the reaUocation by the first switch.

10

1 13. A signal transmission system as in claim 1 wherein said first

2 switch is an NxN matrix switch that can produce any of the N! possible

3 reaUocations of bit signals to bit channels.

4 l 14. A signal transmission system as in claim 13 further comprising a

. 2 second NxN matrix switch, having N outputs and also having N inputs

3 each of which is coupled to an output end of a uniquely associated one

^" 4 of these N bit channels;

5

6 this second matrix switch being operable such that when the first

7 matrix switch is activated to reallocate channels, this second

8 matrix switch also reaUocates channels in an inverse manner to

9 the first switch such that a bit signal applied on any given one of ιo the inputs to the first switch exits from a uniquely associated one ii of the outputs of the second matrix switch, regardless of the

12 reallocation of channels by these two switches.

13 A signal transmission system as in claim 1 further comprising:

an integral number M of N-bit digital signal sources, each of these signal sources having N different outputs, each of these outputs transmitting a different one of N bit signals comprising the output of this digital signal source;

means, coupled to the N outputs of each of these M signal sources and coupled to the N inputs of the first switch, for time multiplexing these M digital signals to produce an N-bit time division multiplexed signal transmitted through said first switch.

16. A signal transmission system as in claim 15 wherein a set of Q signal sources, that provide Q adjacent time multiplexed words in the time multiplexed data signal, all produce a temporally constant pattem of equal bits, whereby these Q sources produce a contiguous sequence of Q equal synchronism words selected from the set consisting of the MAXIMUM word and the NULL word, that can be utilized to synchronize data transmission and reception, said system further comprising:

a receiver connected to an output end of the N bit channels for receiving bit signals over each of these bit channels;

a SYNCH generator, responsive to reception of Q successive synchronism words, for producing a SYNCH signal; and

a delay/tuner, connected to the SYNCH generator and responsive to a user input as to which of the time multiplexed data signals is to be received, for producing a delayed SYNCH signal;

said receiver being responsive to this delayed SYNCH signal to synchronize the receiver operation with data transmission, the delay being such that the receiver receives only data transmitted over the user selected time multiplexed data channel.

17. A signal transmission system as in claim 15 wherein said digital signal sources provide digitized video signals and wherein these digitized video signals are categorized as to type of programming carried over that channel and are allocated to time multiplexed data channels such that all signals of a given type are transmitted over adjacent time multiplexed data channels.

i 18. A wavelength division multiplexer/demultiplexer for use with

2 wavelength multiplexed digital data of bit width P and wavelength

3 spacing S between adjacent wavelength division multiplexed channels,

4 said multiplexer/demultiplexer comprising:

5

6 an optical fiber having a spatial integral I, over its length, of its

7 dispersion selected to be greater than or equal to the ratio P/S.

8 i 19. A wavelength division multiplexer/demultiplexer as in claim 18

2 wherein P/S is substantiaUy equal to I.

3

1 20. A wavelength division multiplexer/demultiplexer for use with wavelength multiplexed N bit digital data of bit width P and wavelength spacing Sk between the kth and (k+ l)st wavelength division multiplexed channels, for k = 1,...,N-1, said multiplexer/demultiplexer comprising:

a first optical fiber having a spatial integral I, over its length, of its dispersion;

a receiver, having a second optical fiber, having an input end attached to an output end of the first optical fiber, said second optical fiber having a set of N local regions from which light escapes from this optical fiber, these regions being spaced at distances d from the input end of this second optical fiber;

the sum of I*Sk + d /V being greater than P, where Vk is the velocity of transmission of signals in the kth wavelength division multiplexed channel.

21. A wavelength division multiplexer/demultiplexer as in claim 20 wherein I*Sk + dk/Vk is substantiaUy equal to P. 22. A wavelength division multiplexer/demultiplexer as in claim 20 further comprising:

adjacent to each of these local regions from which light escapes, an optical filter that transmits light for only one of the wavelength multiplexed channels, each of these filters transmitting a different one of the wavelengths of the wavelength multiplexed channels, whereby these filters separate the wavelength multiplexed signals.