Adaptive Equalization Circuit for Magnetic Recording Channels Utilizing Signal Timing

The invention relates to adaptive equalization in magnetic recording channels. More particularly the invention concerns an active equalization circuit which utilizes signal timing of random signal data rather than amplitude to obtain a measure of the needed equalization signal.

The need for equalization or pulse shaping in magnetic recording channels is well recognized as being necessary to compensate for signal distortions that occur due to different head characteristics in different machines.

In the past signal equalization was provided by a fixed equalization circuit in the channel. However, at higher recording densities now in use, a single equalization setting is not likely to adequately compensate the signal across the total variation that may occur due to head, media and machine factors.

Other more advanced equalization techniques have been designed including automatic equalization loops wherein a training sequence of known pulses is provided, whereby the loop can measure the amplitude of the known pulses and determine an error signal that is used to control the amount of the equalization feedback signal necessary to compensate the recording signal. One example of such systems utilizes a so-called two frequency approach wherein the amplitude is measured at two different frequencies and the gain of the compensation circuit is stabilized by setting the gain at the ratio of the two frequencies to unity. This compensation technique results in ideal compensation at the two frequencies. However, it has been recognized that distortion occurs at other than the two frequencies sampled, and that the compensation applied may tend to leave the total pulse envelope in an over compensated or under compensated state.

Another disadvantage of this type of equalization is that it involves a training sequence and is therefore not adaptive since it cannot be used to vary random pulses of a continuing data signal. Also, since a training sequence is used, such systems are limited to periodic compensation and not continuous compensation.

Accordingly, a need exists in the art for a continuous equalization technique which can be utilized to derive an error signal from random data of a magnetic recording signal.

An object of the invention is to provide an adaptive equalization circuit which utilizes a measurement of signal timing to determine the amount of equalization applied on a continuous basis.

This object and other features of the invention are attained in an equalization circuit which utilizes a feedback compensation path. In the circuit the relative timing of random data pulses is sensed as an indication of error, the relative timing signal is provided to a logic circuit and the logic circuit output controls an equalizer amplifier to provide a variable factor of compensation in accordance with whether the pulse timing is in an over compensated or an under compensated state. The equalization is adaptive in the sense that compensation is continuously carried out with respect to the data pulses, and in the sense that random bursts of data pulses are evaluated rather than a training sequence of known pulse format.

The above features and advantages of the invention are described in a detailed description included hereinafter, taken in conjunction with the drawings wherein

FIGS. 1(a) and 1(b) represent typical graphs of under compensation and over compensation with respect to peak shift pulses, and

FIG. 1(c) represents a graph of ideal compensation.

FIG. 2 is a block diagram of the reproduction circuitry of a digital magnetic recording channel.

FIG. 3 is a block diagram where the reproduction circuitry of FIG. 2 has been enhanced according to the teaching of this invention.

FIG. 4 is a circuit embodiment used to generate gain control signals in accordance with the invention.

Referring now to the drawings, a preferred embodiment of the invention is described comprising an active symmetrical equalization circuit for magnetic recording channels capable of handling high density signals. It should be recognized that the term adaptive equalization is intended to mean compensation of random data without prior knowledge of what the data is, as opposed to other techniques which use "training sequences" or pulses of known data content to vary the amount of equalization.

FIG. 2 illustrates a block diagram of the reproduction circuitry of a digital magnetic recording channel. Read signals are picked up from the magnetic media (tape or disk) by reproduce head 21. These signals are amplified by amplifier 22 and then shaped by fixed equalizer 23. Data pulser 24 outputs a logic pulse at the time of a positive or negative peak in the signal on conductor 33. Block 25 is a variable frequency clock that phase and frequency locks to the logic pulse outputs of data pulser 24. Detector block 26 detects whether the incoming data is a one or zero using inputs from clock 25, fixed equalizer 23 and data pulser 24. The above is all well known and understood in the prior art.

Fixed equalizer 23 is designed to equalize (shape) signals from nominal reproduce heads, media and machines. In high density magnetic recording, these components can vary considerably from their ideal values. This results in under-compensated and/or over-compensated signals such as those shown in FIGS. 1(a) and 1(b). It is the purpose of this invention to measure the resulting signal distortion by measuring the variation in signal peak timing and comparing it with the detected data. This comparison is used to generate a correction signal which will increase or decrease the amount of compensation until a signal substantially like that of FIG. 1(c) is achieved. When this signal shape is achieved, the signal peak timing will also be substantially the nominal desired value.

Accordingly, the reproduce circuitry is enhanced by the addition of circuitry to measure the signal distortion and a variable equalizer to adjust signal shapes. This arrangement is shown in FIG. 3. Blocks 21a, 22a, 23a, 24a, 25a and 26a of FIG. 3 perform substantially the same function as do blocks 21, 22, 23, 24, 25 and 26 of FIG. 2, respectively. In FIG. 3, blocks 47, 48, 49 and 69 have been added, in accordance with the invention. Block 48 is an equalizer filter designed to provide necessary additional compensation. A possible transfer function for block 48 is: ##EQU1## This is the general equation of a high pass filter designed to accentuate high frequencies since high frequency loss is a common magnetic recording problem. It should be understood that other filters may be used in this block. Block 47 is a variable gain amplifier that varies the amount of signal output to equalizer filter 48. The gain of block 47 is varied in response to the control voltage on conductor 65. The value of the control voltage is determined by the measurements of signal peak timing error which is determined by block 69.

The detailed hardware of block 69 is shown in FIG. 4. It is the purpose of this circuitry to determine the amount of additional equalization required to properly equalize (shape) the read signals so that they will be substantially like the signals of FIG. 1(c). From the wave forms of FIG. 1, it can be seen that under-compensated pulses shift away from the closest neighboring pulses, whereas over-compensated pulses shift toward the closest neighboring pulses. For the case of ideal compensation, there is substantially zero peak shift.

Based upon this recognition, the logic in Table 1 has been generated for (d, k) run length code, where d is the minimum number of zeroes in the code and k is the maximum number. In the table T is equal to d+1:

In the discussion that follows, D.sub.N is defined as the value of the bit detected at time N; D.sub.N+T is defined as the value of the bit at time N+T for any arbitrary time j or k.

In accordance with the logic of Table 1, a correction will be made at time N only if D.sub.N is equal to 1, where D.sub.N has possible values of (1, 0).

The logic of Table 1 is based on an assumption that the interference in the read pulse is reasonably symmetrical and that the read pulse is of reasonably narrow configuration. This is essentially the function of fixed read equalizer 23. Equalizer 23 is designed such that the widest pulse at its output will not be more than two code bits wider than the ideally compensated pulse. In general for an arbitrary (d, k) code the pulse width should be less than (4+2d) code bits wide. Thus, for the graphs of FIG. 1 where d=0, the pulse at the output of fixed equalizer 23 should be less than 4 code bits wide.

The logic of Table 1 can be implemented with the detailed hardware illustrated in FIG. 4. Referring to FIG. 4, block 73 is a clock correction circuit that determines if the logic data pulse on conductor 68 occurs to the left (prior in time) of the center of the clock period or to the right (later in time) of the center of the clock period. The clock wave form is on conductor 55. This circuit function is well known in the prior art and is used in phase-lock loop designs.

If the logic data pulse on conductor 54 is left shifted, a positive level is output on conductor 91. If the logic data pulse is right shifted, a positive logic level is output on conductor 92. Data is input on conductor 56 and goes to flip-flop 71. Flip-flops 71, 72, 74 and 75 are logic delays and are used to properly time align the input signals to AND gates 76, 77, 78 and 79, respectively.

The reader will understand that the arrangement of FIG. 4 is for a (d, k) code where d=0. In general for an arbitrary d, k code, each flip-flop block is replaced by d+1 flip-flops. Thus, for the case shown for d=0, each flip-flop becomes 0+1 or one flip-flop. For a code with d=1, each flip-flop would be replaced by 1+1 or two flip-flops. In all other respects, the circuitry of FIG. 4 would remain the same for all other (d, k) codes.

AND gates 76, 77, 78, 79 and OR gates 80, 81 implement the logic of Table 1. The output of gate 80 is high if overcompensation is measured. The output of gate 81 is high if under-compensation is measured. The circuitry including transistors Ql, Q2, Q3, Q4, Q5, Q6 and resistors R1, R2, R3, R4, R5, R6, R7, R8, R9, R10 form a charge pump. The electric charge (voltage) on capacitor C1 is increased if the output of gate 80 is high representing an over-compensation condition. The voltage on capacitor C1 is decreased if the output of gate 81 is high representing an under-compensation condition. The voltage on capacitor C1 appears on conductor 65. As seen in FIG. 3, this is used as the control voltage for variable gain amplifier 47. The gain of amplifier 47 is thus increased or decreased to correct the measured over or under compensation condition.

The equalization circuit described hereinbefore is believed to be significantly different from the prior art in that it is truly adaptive, with the term adaptive being used in the sense that it can adapt even in random data without prior knowledge of what that data is. This enables the system to equalize the signals being read from tape or disk defected areas that might otherwise affect data reliability. The technique of changing the equalizer in response to an error signal that relates to the relative timing of present and past detected data values is advantageous and permits the circuit to be used in a continuously adaptive mode in random data. The use of signal timing to effect changes in the equalizer response makes this invention insensitive to signal amplitude loss that occurs in media defects. Thus, the equalized signal is continuously compared with the data to determine whether the amount of equalization is correct.