Digital FSK Demodulator

This invention relates to data demodulators and more particularly to FSK (frequency shift keyed) data demodulators.

FSK demodulators have in the past primarily employed analog processing which utilized a first analog circuit to detect the "mark" frequency and a second analog circuit to detect the "space" frequency to demodulate FSK signals.

An object of the present invention is to provide an FSK demodulator which employs digital processing techniques.

A feature of the present invention is the provision of a digital frequency shift keyed (FSK) demodulator comprising: a first source of an analog reference signal; a second source of analog FSK signals; first digital means coupled to the first source to produce an inphase digital representation of the reference signal having the form cos A, where A is an angle, and a quadrature digital representation of the reference signals having the form sin A; second digital means coupled to the second source to produce an inphase digital representation of the FSK signals having the form cos B, were B is an angle, and a quadrature digital representation of the FSK signals having the form sin B; third digital coupled to the first and second digital means to produce an inphase code signal and a quadrature code signal for estimating the value of (B- A); fourth digital means coupled to the third digital means responsive to the inphase and quadrature code signals to determine whether the value of (B- A) is increasing or decreasing, the fourth digital means having an inphase output and a quadrature output; fifth digital means coupled to the inphase and quadrature outputs to determine whether a first given code is detected more frequently than a second given code and vice versa; and averaging means coupled to the fifth digital means to average the difference between the first and second given codes to produce a high voltage for a mark and a low voltage for a space.

Above-mentioned and other features and objects of this invention will become more apparent by reference to the following description taken in conjunction with the accompanying drawing, in which:

FIG. 1 is a block diagram of the digital FSK demodulator in accordance with the principles of the present invention; and

FIG. 2 is a block diagram of a delta modulator that may be employed for the delta modulators of FIG. 1.

Referring to FIG. 1 there is illustrated therein common circuity 1 which is common to a plurality of channel circuitry 2-2n. The channel circuitry 2 illustrates the block diagram of the digital FSK demodulator of the present invention.

Common circuitry 1 includes a reference oscillator 3 generating a sinusoidal signal having a center frequency f.sub.o and a processing rate oscillator 4 to generate a clock having a frequency f.sub.p. The output of oscillator 3 is converted to digital form in delta modulator 5 to provide a digital inphase reference signal I.sub.1. The operation of modulator 5 is controlled by the clock signal of oscillator 4. The output of modulator 5 is delayed in an N-stage shift register 6 to provide a quadrature digital reference signal Q.sub.1. The operation of register 6 is also controlled by the clock of oscillator 4. The FSK input to channel circuitry 2 is a sinusoidal signal having a frequency f.sub.o - df to represent a mark or logical 1 and a frequency f.sub.o + df to represent a space or a logical 0. The input signal is converted to digital form in delta modulator 7 under control of the clock of oscillator 4. The output of modulator 7 is an inphase digital FSK signal I.sub.2 which is coupled to an N-stage shift register 8 to provide a quadrature digital FSK signal Q.sub.2. Register 8 is also under control of the clock of oscillator 4.

The circuitry of modulators 5 and 7 is illustrated in FIG. 2 to include an amplitude comparator 9 whose output is coupled to the D input of a D-type flip flop 10 under control of the clock of oscillator 4. The digital output is removed from the Q output of flip flop 10 and is also coupled through a low pass filter 11 to the minus input of comparator 9 which has its plus input coupled to the input signal, either the output of oscillator 3 or the FSK input.

The delta modulator employed herein has the property that the short-term average or low-frequency component of the output is equal to the input. That is, if the input is sin A, where the angle A is a linear function of time, than the probability that the output is a logical 1 will be (1+ sin A)/2, and the probability that the output will be a logical 0 is [1- (1+ sin A)]/2= (1- sin A)/2. This point of view allows the response of logic circuit to be predicted by statistical analysis.

The digital FSK signal and the digital reference signal are each delayed by their associated N-stage shift registers 8 and 6, respectively. The number of stages N of the shift registers 6 and 8 is chosen to be f.sub.p /(4f.sub. o) to make the delay equal to a 90.degree. phase shift. As pointed out hereinabove, the signal before delay is called "inphase" and is identified by I and the signal after delay is called "quadrature" and is identified by Q. Thus, if the digital reference signal I is cos A, the digital reference signal Q is sin A, where A= 2.pi. f.sub.o t, where t is equal to time. Likewise, the I and Q digital FSK signals are cos B and sin B, respectively, where B equals (1.+-. m)A, where m is equal to the modulation index df/f.sub.o.

A pair of bits (I and Q) at any given time may be interpreted as an estimate of the angle A or B of the signal, coded as shown in TABLE I.

If I.sub.1, Q.sub.1 is an estimate of the angle A and I.sub.2 (Q.sub.2) is an estimate of the angle B, then a 16.times. 2 read only memory 12 connected to modulators 5 and 7 and registers 6 and 8 as illustrated and coded as shown in TABLE II can be used to form a code I.sub.0, Q.sub.0 estimating the difference of the angles B and A, namely, (B-A). The difference angle (B-A) will be increasing if the input FSK frequency is f.sub.o + df (space) and decreasing if the input FSK frequency is f.sub.o - df (mark).

To detect whether (B-A) is increasing or decreasing, that is, whether the frequency difference (B-A)/t is positive or negative, the signals I.sub.0 Q.sub.0 are each delayed by an M-stage shift register, such as registers 13 and 14, respectively, whose operation is controlled by the clock from oscillator 4. The difference angle after delay (B-A) is subtracted from the difference angle before delay (B'-A') using a second 16.times. 2 ROM 15 coded like ROM 12 as shown in TABLE II. Logic gate means in the form of gates 16 and 17 detect whether ROM 15 produces a +90.degree. code (0 1), more frequently then a -90.degree. or 270.degree. code (1 0). If this occurs, then a space is detected. However, if gates 16 and 17 detect the code (1 0) more frequently then the (0 1) code then a mark is detected. A low pass filter including resistors 18 and 19 and capacitor 20 averages the difference between these two codes, producing a high voltage for mark and a low voltage for space.

The statistical analysis of the circuit operation is briefly as follows. The probabilities of a logical 1 for the reference I and Q signals are: ##EQU1## and the probabilities of a logical 0 are: ##EQU2##

The same equations, but using B instead of A apply to the input FSK I and Q signals. The probabilities for the two-bit angle codes for A and B are products of the above probabilities for the single bits, thus: ##EQU3##

The probabilities for the two bit-codes out of the first ROM 12, representing the difference angle (B-A), follow directly from the ROM code TABLE II.

these probabilities have been computed from the above probability equations (3) to be: ##EQU4##

The same equations apply for both (B-A) and (B'-A'). Again, from the ROM code TABLE II, the probabilities P.sub.+ for the (01) code, and P.sub.- for the (10) code out of the second ROM 15, can be written:

these probabilities were computed from the equations (5) to be: ##EQU5## The difference between these two probabilities is:

if we average the probability P.sub.+ for having a (01) code and the probability (1- P.sub.-) for not having a (10) code, a similar result is obtained: ##EQU6##

If m is the modulation index df/f.sub.o, and d is the delay of the M-bit shift registers expressed in radians of the phase relative to the reference A, the output result may be restated as:

(+for space, - for mark). The block diagram of the FIG. 1 shows the logic polarities reversed to obtain logic high for mark.

By selecting M- f.sub.p /(4 d f), sin m d= 1, thus maximizing the output amplitude. The output low pass filter converts the output probability to a signal amplitude.

While I have described above the principles of my invention in connection with specific apparatus it is to be clearly understood that this description is made only by way of example and not as a limitation to the scope of my invention as set forth in the objects thereof and in the accompanying claims.