US 4,825,452 AGrant
Digital FSK Demodulator
Issue Date:1989-04-25
•10 Claims
•16 Drawing Sheets
Abstract
The present invention provides a digital frequency shift keying (FSK) demodulator circuit which receives a hard-limited FSK input signal and recovers the modulating baseband signal at 300 baud or less. The FSK demodulator includes a digital phase locked loop demodulator section which receives the FSK input signal and generates a pulse width modulated output the frequency of which is a function of the input signal frequency. The output of the PLL demodulator section is provided to a quantizer which generates a set of parallel numbers. A translating means, for example a code linearizer, converts the parallel numbers into corresponding parallel outputs which are representative of the frequency of the FSK input signal. The parallel outputs are processed by a digital filter which emulates a second-order, low-pass filter. The filter processes the parallel outputs utilizing pulse density modulation. A deglitcher removes noise from the filter output to provide the demodulated data output.
Metadata
Assignee
- National Semiconductor Corporation
Inventor
- Hee Wong
Application Information
Application Number:US 021907&
Filing Date:1987-03-04
Priority Date:1987-03-04
Art Unit:263
Classifications
IPC:
H03K 708H04L 2714
Field of Search:
37532934037082;88;91;80;94;102;22;23;21126;128825.63;347 DD;347 AD9-11
Patent Drawings (16 sheets)
Description
Background of the Invention
1. Field of the Invention
The present invention relates to communications systems and, in particular, to a digital frequency shift keying (FSK) demodulator circuit which receives a hard-limited FSK input signal and recovers the modulating baseband signal at 300 baud or less.
2. Discussion of the Prior Art
The basic function of a communications system is to transmit information over a communication channel between an information source and a destination which are physically separated from one another.
Communications channels designed to handle voice transmissions (i.e., a telephone network) have inherent characteristics which make it difficult for them to transmit digital data or bit streams. Therefore, to permit the transmission of digital data over voice channels, it is necessary to convert the digital data at the transmission point to a form which is compatible with the voice channel. This is done by utilizing the digital data to modulate a carrier waveform which is within the voice frequency range, transmitting the modulated signal and then demodulating the signal at the receiver to separate the transmitted data from the modulated carrier waveform.
Data transmission systems that operate at low data transmission rates, i.e. 1200 baud or less, typically use a modulation technique known as frequency shift keying (FSK). According to this technique, the two binary states are represented by two different frequencies and are detected by using two frequency tuned sections, one tuned to each of the two frequencies. The demodulated signals are then integrated over the duration of a bit and a binary decision is made based upon the result.
Digital transmission using FSK modulation has a number of advantages. First, the implementation is similar in complexity to a conventional AM transmission system. Second, since the received signals can be amplified and limited at the receiver, a simple limiting amplifier can be used, whereas an AM system requires sophisticated automatic gain control in order to operate over a wide level range. Furthermore, FSK modulation can show a 3 or 4 dB improvement over AM systems in most noise environments, particularly at distortion threshold; that is, as the frequency shift becomes greater, the advantage of FSK over AM improves in a noisy environment. A further advantage of FSK modulation is its immunity to the affects of non-selected level variations, even when these variations are extremely rapid. Thus, FSK modulation is used almost exclusively in worldwide high frequency radio transmissions where rapid fades are a common occurrence. In the United States, FSK modulation has nearly universal application for the transmission of data at 1200 baud and below.
A common disadvantage of conventional analog FSK demodulator circuits is that they are sensitive to circuit parameter variations and are not suitable for large scale integration.
Conventional digital FSK demodulator signal processing requires A/D converters, high power consumption and very high system toggling frequencies which results in heavy expenditures on RFI shieldings.
Summary of the Invention
It is an object of the present invention to provide an all-digital FSK demodulator.
It is an object of the present invention to provide an FSK demodulator which utilizes no A/D converters.
It is an object of the present invention to provide an FSK demodulator circuit which is highly stable and integratable and has no trimming requirements.
It is a further object of the present invention to provide an FSK demodulator circuit which exhibits predictable performance.
It is a further object of the present invention to provide an FSK demodulator circuit that utilizes one system clock that determines center operating frequency.
It is a further object of the present invention to provide a phase locked loop demodulator, rather than the conventional frequency discriminator, to improve error rate performance.
It is yet a further object of the present invention to provide an FSK demodulator circuit which utilizes a code linearizer to improve circuit linearity and to reduce bit-bias distortion.
It is yet a further object of the present invention to provide an FSK demodulator circuit that utilizes pulse density modulation to eliminate the need for multipliers in the filter circuits.
These and other objects of the present invention are accomplished by providing a digital frequency shift keying (FSK) demodulator circuit that conforms to Bell-103 modem specifications. The circuit includes a digital first order phase locked loop and a digital second order low-pass filter. In addition to these two major components, the circuit includes several unique circuit arrangements that allow a total digital implementation which outperforms presently-available FSK demodulators.
Thus, according to the present invention, the FSK input signal is received by a digital phase locked loop demodulator which generates a pulse width modulated output signal. The pulse width of the PWM output is a function of the input signal frequency. The output of the digital PLL demodulator is then provided to a quantizer/translator which transforms the PWM output into corresponding parallel outputs which are representative of the frequency of the FSK input signal. The parallel outputs are processed by a digital filter which emulates a second-order, low-pass filter. The filter processes the parallel outputs utilizing a pulse density modulation (PDM) technique. A deglitcher removes noise from the output of the filter to provide the demodulated data output.
Additional objects, features and advantages of the circuit of the present invention will become apparent and be appreciated by reference to the detailed description of the invention provided below, which should be considered in conjunction with the accompanying drawings.
Description of the Drawings
FIG. 1 is a diagram illustrating the proper arrangement of FIGS. 1A-1D.
FIGS. 1A-1D combine to provide a detailed schematic diagram illustrating an embodiment of an FSK demodulator circuit in accordance with the present invention.
FIG. 2 is a schematic block diagram illustrating a model first-order phase locked loop of the circuit shown in FIGS. 1A-1D.
FIG. 3 is a timing diagram illustrating the waveform of the circuit model shown in FIG. 2.
FIG. 4 is timing diagram illustrating the algorithm for the synchronous programmable divider with proportional loading which is utilized in the circuit shown in FIGS. 1A-1D.
FIG. 5 is a timing diagram illustrating the waveform of the PLL demodulator of the circuit shown in FIGS. 1A-1D.
FIG. 6 is a diagram illustrating the proper arrangement of FIGS. 6A-6F.
FIGS. 6A-6F combine to provide a schematic diagram illustrating an alternative embodiment of an FSK demodulator circuit in accordance with the present invention.
FIG. 7 is a table which summarizes the clock generator frequency assignments of the circuit shown in FIGS. 6A-6F.
FIG. 8 is a graph illustrating quantizer output coding versus input frequency.
FIG. 9 provides a series of waveforms illustrating the performance of the baseband filter of the demodulator circuit shown in FIGS. 6A-6F.
FIG. 10 is a timing diagram illustrating the waveforms of the clock generator of the circuit shown in FIGS. 6A-6F.
FIG. 11 is a timing diagram illustrating the waveforms of the output deglitcher shown in the FIGS. 6A-6F circuit.
FIG. 12 is a block diagram generally illustrating a digital FSK demodulator in accordance with the present invention.
Detailed Description of the Preferred Embodiment
Referring to FIG. 12, the present invention provides a digital FSK demodulator circuit that includes a digital phase locked loop (PLL) demodulator 100 which receives an FSK input signal 102. The PLL demodulator 100 generates a pulse width modulated serial bit stream 104 that corresponds to the input signal 102. A quantizer 106 transforms the serial bit stream 104 into an output signal 108 comprising parallel bits representative of the frequency of the FSK input signal 102. A code linearlizer 110 translates the quantizer output signal 108 into optimized corresponding signals 112, as described in greater detail below. A pulse density modulator 114 operates on the optimized signals 112 to recover the modulating baseband signal 116. A deglitcher 118 removes noise from the signal 116 to provide the demodulated data output 120.
FIGS. 1A-1D combine to provide a detailed schematic diagram of a digital FSK demodulator circuit in accordance with the present invention.
As shown in FIG. 1A, the demodulator circuit receives the following two inputs:
1. HLS High/low band select: 5 volts selects the high-band mode (H=2025 Hz/2225 Hz); 0 volts selects the low-band mode (L=1070 Hz/1270 Hz).
2. HFI FSK signal input: 0-5 volts digital input level. This input is taken from the output of a hard-limiter that is placed after a channel filter. The channel filter and the hard limiter are conventional and, therefore, not described.
The Demodulated Data Output (FIG. 1D) is the only output of the demodulator.
As shown in FIG. 1A, flip-flops 10 and 12 and divide-by module 14 combine with XOR gate 16 and a state machine formed by ROM 18 and "D" flip-flip 20 to form a digital phase locked loop (PLL) demodulator. The function of the PLL demodulator is to generate a pulse-width modulated signal stream such that the width of the positive output pulse of the PLL demodulator is a function of the input frequency.
The modeling of the PLL demodulator can be described with reference to FIGS. 2 and 3.
FIG. 2 provides a simplified block diagram of the PLL demodulator and quantizer portions of the circuit shown in FIGS. 1A-1D. The hard limited FSK input HLS is provided as one input FIN to XOR gate 16. The output of XOR gate 16 is the "Select" input to the state machine 18,20 as well as the input signal to the quantizer stage of the circuit. The output of state machine 18,20 feeds .div.N divider 14 which generates the second input F.sub.OSC of XOR gate 16.
FIG. 3 shows the waveforms associated with the FIG. 2 model. Trace 1 in FIG. 3 shows the hard limited FSK input frequency F.sub.IN measured at the input of XOR gate 16. Trace 2 shows signal F.sub.OSC measured at the output of .div.N divider 14. Trace 3 shows the Select input of state machine 18,20. Trace 4 shows the output F.sub.N of state machine 18,20.
If the FSK input frequency F.sub.IN has a 0.5/0.5 duty cycle, then the following equation can be written: ##EQU1## If the circuit's quantizer (described below is designed to perform the following equation: ##EQU2##
Table 1 below provides a summary of circuit performance when the circuit parameters are as follows:
The input frequency F.sub.IN of the input signal HFI and the pulse width period "P" are related by equation (2) above.
The configuration of the above-described PLL demodulator is similar to a circuit proposed by Pasternack & Whalin, "Analysis and Synthesis of a Digital Phase-locked Loop for FM Demodulation". B.S.T.J. vol. 47, page 2207-2239; December 1968. The Pasternack/Whalin circuit uses two clocks to operate the PLL. The circuit of the present invention uses a synchronous programmable divider with a proportional loading circuit to derive the two clocks. The advantages of this latter approach are:
1. Circuits implemented with integer dividers may not be able to generate the two desired frequencies if the system clock is low, i.e., less than 2 MHz.
2. If the two clocks are generated asynchronously, jitter performance may become unacceptable due to spurious switching. Asynchronous clocking also produces hardware design difficulties in pipeline designs.
3. A programmable divider without the use of proportional loading greatly increases the jitter content in the demodulated output.
Referring back to FIG. 1A, state machine 18,20 is synchronous to the demodulator clock .phi..sub.D. The changeover between the two divisors is performed in a proportional manner. The algorithm for implementing the changeover is described in conjunction with FIGS. 3 and 4, where traces 1-4 in FIG. 3 correspond to the signals shown in FIG. 4. The circuit reads the current count "x", then computes ##EQU3##
The circuit then loads "y" counts into state machine 18,20 during the next clock interval.
The pulse width modulated signal stream from the output of XOR gate 16 is pipelined through flip-flop 12 and routed to the quantizer stage, shown in FIG. 1C, formed by counters 22 and 24, XOR gates 26 and 28, flip-flops 34 and 36, and NOR gate 32. The function of the quantizer is to transform the bit stream into parallel outputs so that the quantities can be further processed by the code linearizer, described below. Equation (5) above denotes the relationship between the output counts X.sub.c of the quantizer and the frequency of the input FSK signal F.sub.IN.
Look-up ROM 38 and latch 40 combine to form a code linearizer. The function of the code linearizer is to translate the parallel numbers from the quantizer to optimized corresponding quantities. This allows the performance of the filter circuit which follows the code linearizer to be optimized. The translating function is a simple look-up process utilizing ROM 38.
The reasons for the look-ups are:
1. Center alignment - the middle count of counters 22 and 24 may not equate to the center frequency of the input frequency band F.sub.IN. This is due to the fact that some parameters in the PLL demodulator section are integer values, i.e., H, L, N and etc. Exact centering is impossible unless a very high system clock frequency is utilized. This kind of offset introduces bit-bias distortion. The look-up process eliminates the offset.
2. Linearity correction--the transferred characteristic of the PLL demodulator is not linear as demonstrated by equation (5) above. The look-up process corrects this problem.
3. Frequency ratio look-up--the look-up values are expressed in the logarithmic ratio of F.sub.IN over the transmitted frequency band's geometric center. This helps to reduce the distortion of the demodulation output and can be treated as part of the linearity correction.
4. Dynamic range amplification--the PLL demodulator is designed to produce the least Inter-Symbol Interference (ISI) distortion at the demodulator output. The band-width (locking range) requirement is about three times the frequency deviation of the FSK signal. In spite of this requirement, the typical output swing produced by the FSK input signal is only 1/3 of the total dynamic range of the PLL demodulator. The look-up process "amplifies" the coding dynamic range so that the filter can operate at an optimum condition.
5. Intentional offset addition--for compensating the intrinsic offset of the filters.
The program listing "FDLR.BAS" to support these statements is provided as Appendix F at the end of this specification.
The architecture of the filter portion of the circuit shown in FIGS. 1A-1D emulates a state variable second-order low-pass filter. In order to reduce the amount of circuitry, the following techniques are introduced:
1. Processing of the signals is done serially by employing a pulse density modulation (PDM) technique.
2. Due to the use of the PDM technique, the summing element becomes a simple look-up table.
3. Due to the use of the PDM technique, the utilization of counters for performing the integrating function becomes feasible.
4. By selecting the length of the counters, a scaling function can be achieved without the use of hardware multipliers.
Referring to FIGS. 1B-C, rate generators 42 and 44 and latch 46 translate parallel outputs from latch 40 into a PDM bit stream. The dynamic range of the output is from -1 to 1, i.e.,
The instantaneous weight is determined by the running average of the pulse density of the output. Note that the output changes state only at the edges of the filter clock which operates at a very high frequency compared to the modulating base-band signal. The filter circuit after this pulse density modulator is designed to operate with this kind of signal.
Component 72 (FIG. 1B) performs the summing function of the filter. It has three inputs and two outputs. The first input (pin 12) is driven by the demodulated serial signal mentioned above. The second input (pin 11) receives its signal from the filter's first integrator, while the third input (pin 10) is connected to the output of the filter's second integrator. The signals at the last two input pins 11 and 10 have the same kind of modulation as the first input pin 12 and are the same dynamic range, i.e., -1 to 1.
The output of this portion of the circuit is the difference between the input signal and the sum of the outputs of the filters to the integrators. The dynamic range is from -3 to 3 and has the following discrete values: -3, -1, 1, and 3.
Components 48 and 50 form an 8-bit up/down counter and create the filter's first inteqrator which accumulates the output from summing device 72.
The four most significant bits of the first integrator are fed to a PDM modulator formed by components 52, 54 and 56. The selection of the 4-bit value is based solely on the tradeoff between the amount of hardware and system performance. Since this integrator has more resolution than its output, the intrinsic offset is introduced into the filter circuit. This offset is also due to the asymmetrical characteristic of the rate generator 54; i.e., there is a code "0" for "no" output and no code for a "full" output (the circuit does not produce a full output). Component 52 is a gated four-bit "D" flip-flop which is updated when rate generator 54 is at the terminal count. The connection between pin 7 of rate generator 54 and pin 9 of flip-flop 52 is for the purpose of synchronous loading.
The signal at pin 5 of latch 56 is the output of the first integrator which, as described above, is routed back to summing device 72. This output is also used to increment or decrement the filter's second integrator.
Referring to FIG. 1D, the filter's second integrator is formed by components 58, 60 and 62. In order to achieve a "Bessel" filter characteristic, the length of this counter chain is arranged to have one bit more than the first integrator. The up/down function is controlled by the first integrator. Latch 64 senses all the carry flags from the counter chain and inserts upper and lower limits to the "output swing". Again, only the four most significant bits are used to drive a PDM modulator that feeds the summing circuit 72. This modulator is formed by components 66, 68 and 70.
This filter has a "Bessel" low-pass frequency characteristic; the three dB cutoff frequency is at 155 Hz.
The output signal of the filter block, i.e., pin 14 of counter 58, may contain glitches due to noisy input signals or for other reasons. The glitches normally are of a high frequency nature and scatter around the data transitions. This happens to be a problem common to any FSK demodulator circuit. A deglitcher circuit formed by components 74 and 76 is designed to eliminate this problem by introducing hysteresis.
The deglitcher circuit functions, basically, as a programmable counter such that when the output count crosses the mid-count, the counter will jump to either the maximum or the minimum count.
FIGS. 6A-6D show a preferred embodiment of a digital FSK demodulator circuit in accordance with the present invention. The embodiment shown in FIGS. 6A-6F differs from the embodiment shown in FIGS. 1A-1D as follows:
1. The code linearizer portion of the circuit is omitted in the FIGS. 6A-6F embodiment to reduce system cost. Center alignment and intentional offset addition are achieved by adding a base number to the quantizer. Furthermore, the linearity correction, the frequency ratio look-up and the dynamic range amplification processes are not performed in the FIG. 6A-6F circuit. The degradation and system performance due to the elimination of the linearizer portion of the circuit is compensated by applying a higher clock frequency (two times) to the quantizer portion of the circuit.
2. The clocking signal connected to the demodulator output latch 140 is gated-off in the FIG. 6A-6F circuit when the incoming frequency is above 2360 Hz (high band) or 1430 Hz (low band). These high frequency components are due mainly to noisy edges embedded in the "hard-limited" incoming FSK signal under low input S/N ratio conditions. Due to the presence of these noisy edges, the quantizer portion of the circuit will malfunction and erroneous data will be produced at the output of a latch. Therefore, a 4-input NOR gate 130 is added at the output of counter 122 to perform the "gate-off" function.
3. The second-order low-pass filter portion of the FIGS. 6A-6F circuit is constructed with a multiplexing adder scheme. The processing of signals is also done serially by employing the PDM technique as described in conjunction with the FIGS. 1A-1D circuit. The filter parameters remain unchanged, while the circuit complexity is reduced. A 5-phase clock generator is also implemented for controlling the multiplexing cycles.
The PLL demodulator portion of the circuit shown in FIGS. 6A-6D is the same as the demodulator block shown in the FIGS. 1A-1D embodiment. A program listing "FDPD.SRC", which is provided at the end of the specification as Appendix A, provides the coding statements for the state machine 118, 120 of the demodulator of the FIGS. 6A-6F circuit.
The clock generator frequency assignments for the demodulator clock .phi..sub.D, the filter clock .phi..sub.F and the pulse width quantizer clock Q.sub.Q of the FIGS. 6A-6F embodiment are summarized in FIG. 7.
As shown in FIG. 6B, due to the elimination of the code linearizer, the loading circuits of the two counters 122 and 124 are modified to start with a base count before each quantization process. The base count produces offset in the transfer characteristic of the quantizer such that a "center alignment" and an "intentional offset addition" can be achieved. The base count is "8" when the circuit operates in the high band mode and "3" in the low band mode to eliminate bit bias distortion.
To retain the performance of the FIGS. 1A-1D embodiment, the clocking frequency .phi..sub.Q is increased from 614.4 KHz to 1.2288 MHz in the high band mode and from 307.2 KHz to 614.4 KHz in the low band mode. In addition, the counter length is increased from 7 to 8 bits.
The outputs D.sub.0 -D.sub.3 of counter 124 and the outputs D.sub.4 -D.sub.7 of counter 122 are then fed to an 8-bit D-type data latch 140 (FIG. 6E). This latch is clocked at twice the carrier rate, as shown in FIG. 5.
In FIG. 5, Trace 1 shows the input FSK analog waveform measured before the hard limiter in the low band mode at 300 baud. Trace 2 shows the phase detector output measured at the output of XOR gate 116. Trace 3 shows the demodulated output measured at the output of latch 140.
The binary value of the output of latch 140 depicts the frequency reading of the incoming FSK signal. A plot of the transfer characteristic is provided in FIG. 8. It should be noted that the curves are shown in FIG. 8 without the effect of the base counts. The plot follows equation (5) above. The source listing of the plotting program "FQF.SRC" is provided in Appendix B at the end of this specification.
The clock signal to latch 140 may contain noisy edges due to the low S/N ratio at the input of the demodulator. These noisy edges are undesirable because multiple start/stop strobes are generated into the quantizer, thereby causing erroneous data at the output of latch 140. The solution is to gate-off the clock signal applied to latch 140 when the binary value of the counter output is below a specific count. The circuit formed by NOR gate 130 is designed for this purpose. For convenience, the number "16" is selected as the threshold count. As shown in FIG. 8, the corresponding frequency readings are 2360 Hz for the high band mode and 1430 Hz for the low band mode. These frequencies are beyond the maximum deviation of the incoming FSK signal frequency F.sub.IN.
Referring to FIGS. 6C-6E, a second-order, low-pass filter is formed by data latch 140, ROM decoder 142, 8-bit bit adder 144 and 146, data latches 148, 150, 152, 154 and 156 and a demultiplexing circuit formed by flip-flops 158, 160 and 162.
The architecture of this filter emulates a second order "state variable" type of low-pass filter. It includes two integrator latches 150 and 152 in the feed-back loop. The -3 dB cut-off frequency is at 155 Hz with a Bessel type of roll-off characteristic (including the pole introduced by the PLL Demodulator).
An 8-bit adder 144 and 146 is the core of the filter circuit. It is multiplexed to perform five functions within a period of 8.138 .mu.Sec (1/122.88 KHz). Each function occupies a cycle which is 1/5 (1.628 .mu.Sec) of the total period. The cycle timing is arbitrated by a five-phase clock generator described below. The five function-cycles are as follows:
Cycle 0: to perform the 1st integrating function;
Cycle 1: to convert the 1st integrator parallel output data to a (PDM) signal;
Cycle 2: to perform the 2nd integrating function;
Cycle 3: to convert the 2nd integrator parallel output data to a PDM signal;
Cycle 4: to convert the demodulated parallel output data from latch 140 to a PDM signal.
The five function-cycles can be divided into two types of cycles, the integrating cycles and the PDM conversion cycles.
The integrating cycle is a straightforward process. The result generated in each integrating cycle is the algebraic sum of the previous contents of the corresponding integrator and the PDM signal(s).
Cycle 0: The 1st Integrator cycle
Latch 150 is enabled during this cycle. The previous contents stored in its output registers are applied to one set of inputs of the adder circuit. The other set of inputs of the adder circuit are connected to the PDM outputs via ROM decoder 142. The PDM signals are categorized as follows:
Pin 5 of flip-flip 158--demodulated PDM signal
Pin 9 of flip-flop 160--1st integrator PDM output
Pin 5 of flip-flop 162--2nd integrator PDM output
During this cycle, the output of decoder 142 sums these three signals as described on line #240 of the program listing "FADR.SRC" which is provided in Appendix C at the end of this specification. The result has one of the following values:
To achieve a 7-bit equivalent adder circuit out of an 8-bit configuration, a scaling multiplier of "2x" is utilized.
Cycle 2: the 2nd Integrator cycle
This cycle is similar to Cycle 0. Integrator 152 is enabled and serves as the storage element. The output of decoder 142 only decides the value of the PDM signal of the 1st integrator, which has only two discrete values, +1 and -1. The selection between Cycle 0 or Cycle 2 in the decoder circuit is controlled by the signal applied to pin 13 of decoder 142.
The PDM conversion cycles are achieved by means of an "adding" technique. The result is promising due to the fact that the adder circuit accumulates every signal parallel input data in its registers until overflow. Upon overflow, an amount equal to the carry weighting is subtracted from the accumulator (no data is being lost). On the other hand, conventional rate-multipliers (such as the CD4089 rate-multiplier) perform a one-to-one translation on the input codes; the memory function relies on the counter states inside the multiplier for waveform continuity. For input signals that change more rapidly than the repetition rate of the internal counter, the output of the multiplier cannot accurately reproduce the input signal, thereby resulting in spurious distortion.
The "adder" approach basically performs a Sigma-Delta modulator function. The adder accumulates the input data until overflow. Upon overflow, an amount equal to the carry weighting is subtracted from the accumulator (the adder's basic function) and a logical "1" is generated at the carry output. The carry output then serves as the PDM output.
Cycle 4: the conversion cycle for the incoming data
The input data is fed to inputs of latch 140 while the storage element is latch 148.
Cycle 1: the conversion cycle for the 1st integrator
Latch 150 is the input port, latch 154 is the storage element.
Cycle 3: the conversion cycle for the 2nd integrator
Latch 152 is the input port; latch 156 is the storage element.
The PDM output signals generated in these cycles are multiplexed at the carry pin, i.e., pin 9, of adder 144. The demultiplexing circuit 158, 160, 162 is included to retrieve the signals.
There are several distinct properties of this kind of filter:
1. The input signal to the filter can be either in PDM format or in parallel data format.
2. The PDM output signal or parallel output data are available simultaneously from the filter.
3. In the event of a parallel-in and parallel-out configuration, the output data is oversampled compared to the input data. The interpolating characteristic implies that the filter output is almost free from the sampling alias frequencies (sampling frequency>>base band signal). In this application, the filter input is sampled at twice the carrier rate, i.e., about 2 to 4 KHz, while the output sampling rate is 614.4 KHz. This can be seen in the series of waveforms provided in FIG. 9 wherein the input signal is a 300 baud FSK data sequence and waveforms "A" illustrate the demodulator output with an 8-bit DAC included and waveforms "B" illustrate the filter output with an 8-bit DAC/latch included.
In spite of these properties, a variety of circuits, such as sine-wave generators, modulators, etc., can be built without the use of digital-to-analog converters (use PDM output) or continuous anti-alias filters (parallel outputs).
As shown in FIG. 6D, the five filter cycles are arbitrated by a filter clock generator consisting of divide-by module 164, invertor 166, ROM decoder 168 and NOR gates 170, 172, 174, 176 and 178. Divide-by module 164 divides the input clocking frequency (614.4 KHz) by 5 and its outputs are fed to ROM decoder 168. The outputs from decoder 168 consist of five enable signals, E0-E4, and three combined enable signals, E2E3, E0E2, and E0E1. The five enable signals E0-E4 are also used for generating the five phase clocking signals .phi..sub.0 -.phi..sub.4. The timing relationship is shown in FIG. 10. The program listing "FCLK.SRC", provided in Appendix D at the end of this specification, provides the coding statement for ROM decoder 168.
As shown in FIG. 6F, D-type flip-flop 180 is used to latch the most significant bit of the parallel output data from the filter circuit during Cycle 2 (the 2nd integrating cycle). The output of this flip-flop 180 is then used to drive an output deglitcher circuit formed by programmable counter 182 and flip-flop 184.
The output signal from the filter may contain glitches due to noisy input signals or other reasons. The glitches normally are of high frequency nature and scatter around the data transitions. This is a problem common to any demodulator circuit. The output deglitcher formed by programmable counter 182 and flip-flop 184 is designed to eliminate this problem by introducing hysteresis.
As illustrated in FIG. 11, the output deglitcher basically functions as a programmable counter such that when the output count crosses the mid-count, the counter will jump to either the maximum (15) or the minimum (0) count. Note that the threshold does not change (mid-count is a fixed number) and hysteresis is achieved by changing the state of the counter. The performance of this deglitcher has been found to be better than conventional Schmitt trigger circuits as far as Intersymbol Interference (ISI) output jitter is concerned. The coding algorithm for counter 182 is provided in Appendix D.
The output of the deglitcher circuit is available at pin 11 of flip-flop 184. An invertor 186 is added to provide the proper polarity at the demodulator output.
Conventional commercially-available demodulators 5-20% jitter readings.
When all bit jitter components are allowed to interact, under no external noise conditions, the circuit shown in FIGS. 6A-6F produces 2% and 3% jitter readings in the high-band mode and the low-band mode, respectively, with negligible bit bias distortion.
The filter performance and the eye-diagrams are shown in FIG. 9. The eye-diagram for the low-band mode shows unsymmetrical excursion on the two peaks; this is due to the absence of a "linearizer" circuit, which is excluded from the FIG. 1A-1D embodiment because of cost considerations. The spectrum measurement indicates the noise floor is almost 60 dB below the signal main peak, which is more than adequate for this kind of application.
It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention. It is intended that the following claims define the invention and that circuits within the scope of these claims and their equivalents be covered thereby. ##SPC1##
Claims
What is claimed is:
1. A digital frequency shift keying (FSK) demodulator which receives an FSK input signal and recovers a modulating baseband signal, the FSK demodulator comprising, (a) means for receiving the FSK input signal and generating a corresponding pulse width modulated serial bit stream; (b) means for transforming the pulse width modulated serial bit stream into an output signal comprising parallel bits representative of the frequency of the FSK input signal; and (c) means for processing the output signal utilizing pulse density modulation to provide the recovered modulating baseband signal.
2. A digital FSK demodulator as in claim 1 wherein the means for receiving comprises a digital phase locked loop.
3. A digital FSK demodulator as in claim 2 wherein the means for transforming includes a code linearizer comprising a look-up table for improving linearity and reducing bit-bias distortion in the recovered modulating baseband signal.
4. A digital frequency shift keying (FSK) demodulator which receives an FSK input signal and recovers a modulating base and signal, the FSK demodulator comprising, (a) a digital phase locked loop for receiving the FSK input signal and generating a corresponding pulse width modulated serial bit stream; (b) means for transforming the pulse width modulated serial bit stream into an output signal comprising parallel bits representative of the frequency of the FSK input signal; and (c) means for processing the output signal utilizing pulse density modulation to provide the recovered modulating baseband signal, the processing means comprising means for translating the output signal into a pulse density modulated serial bit stream; means for summing the output signal of the translating means, the output signal of said first integrator and the output signal of said second integrator; a first integrator which accumulates the output signal of the summing means; and a second integrator which receives an input signal from the first integrator and generates the recovered modulating baseband signal.
5. A digital FSK demodulator as in claim 4 and further including means for removing noise from the recovered modulating baseband signal.
6. A method of recovering a modulating baseband signal from an FSK modulated signal, the method comprising: (a) processing the FSK signal to generate a pulse width modulated serial bit stream the pulse width of which is a function of the frequency of the FSK signal; (b) quantizing the pulse width modulated serial bit stream to generate a corresponding output signal comprising parallel bits representative of the frequency of the FSK signal; and (c) filtering the output signal utilizing pulse density modulation to generate the recovered modulating baseband signal.
7. A method of recovering modulating baseband signal from a frequency shift keying (FSK) modulated signal, the method comprising: (A) processing the FSK signal to generate a pulse width modulated serial bit stream then pulse width of which is a function of the frequency of the FSK signal; (b) quantizing the pulse width modulated serial bit stream to generate a corresponding output signal comprising parallel bits representative of the frequency of the FSK signal; (c) translating the output signal into a pulse density modulated serial bit stream; (d) summing the serial bit stream the output signal of a first integrator and the output signal of a second integrator; (e) accumulating the sum generated in step (d) to produce the first integrator output signal; and (f) accumulating the first integrator output signal in the second integrator to produce the second integrator output signal.
8. A method as in claim 7 including the further step of removing noise from the recovered modulating baseband signal.
9. A digital frequency shift keying (FSK) demodulator which receives an FSK input signal and recovers a modulating baseband signal, the FSK demodulator comprising, (a) a digital phase locked loop demodulator means which receives the FSK input signal and generates a pulse width modulated serial bit stream corresponding to the input signal; (b) a quantizer which transforms the serial bit stream into an output signal comprising parallel bits representative of the frequency of the FSK input signal; (c) a code linearizer which translates the outputs signal from the quantizer into optimized corresponding signals; and (d) means for pulse density modulating the optimized corresponding signals to recover the modulating baseband signal.
10. A digital frequency shift keying (FSK) demodulator which receives an FSK input signal and recovers a modulating baseband signal, the FSK demodulator comprising: (a) a digital phase locked loop demodulator means which receives the FSK input signal and generates a pulse width modulated serial bit stream corresponding to the input signal; (b) a quantizer which transforms the serial bit stream into an output signal comprising parallel bits representative of the frequency of the FSK input signal; (c) a code linearizer which translates the output signal from the quantizer into optimized corresponding signals; (d) means for pulse density modulating the optimized corresponding signals to recovering the modulating baseband signal utilizing a plurality of cycles; and (e) a clock generator for arbitrating the cycles of the pulse density modulating means.