Dynamic magnetic information storage or retrieval – Automatic control of a recorder mechanism – Controlling the head
Reexamination Certificate
2001-01-12
2003-06-03
Sniezek, Andrew L. (Department: 2651)
Dynamic magnetic information storage or retrieval
Automatic control of a recorder mechanism
Controlling the head
Reexamination Certificate
active
06574067
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to servo control systems for disc drives, and particularly to repetitive learning compensators (RPCs) for disc drive servo systems.
BACKGROUND OF THE INVENTION
Accurate positioning of a read/write head relative to a disc medium is crucial to the operation of disc drive systems. To reliably read data from or write data to the disc surface, the read and write transducers of the head must be positioned precisely over the center of the track of the media surface from which data is read or on which data is written. A servo system operates the actuator arm assembly on which the head is mounted to position the head relative to the track so the read or write transducer, as the case may be, is centered on the track. Failure to accurately position the read/write head over the desired track during a read operation may result in unreliable data retrieval. Improper positioning of the read/write head during a write operation may result in the loss of the data being written, and in overwriting and destroying data on adjacent tracks.
A growing need for more data storage capacity has resulted in increased track density, thus increasing the number of tracks per radial inch (TPI) on the disc. As track density increases, the track width decreases, requiring head positioning to be even more accurate. Consequently, the servo system must be designed for highly precise head position control.
There are many control schemes used in servo systems of hard disc drives. Most disc drives employ a proportional-integral-derivative (PID) or state-feedback (SFB) controller for major control operations. To enhance the servo performance, adaptive and robust schemes are also used as feedforward loops. Examples of servo systems employing feedforward loops include model reference control, adaptive feedforward cancellation (AFC), proximate time optimal servomechanism (PTOS), iterative learning control (ILC), command shaping, repetitive learning compensators (RLC), and others.
Repetitive learning compensators (also called “repetitive learning controllers” or “RLCs”) are implemented in code to reject repetitive disturbances, such as repeatable runout (RRO). Examples of sources of RRO include disturbances associated with the spindle motor, and written-in RRO which is a special repetitive disturbance having a fundamental frequency equal to the spindle frequency.
Some RLCs employ a positive feedback of the output of a delay line to form a filter that cancels the repeatable disturbances at frequencies which are integer multiples (harmonics) of a fundamental frequency f
0
that is based on the disc rotation. More particularly, for a drive spindle motor speed of RPM, the fundamental frequency f
0
is equal to RPM/60. Typically the filter is a digital filter-in-the form of a memory buffer, the length of which is based on a ratio of a sample frequency f
s
and the fundamental frequency f
0
. Most RLCs employ a comb filter, which exhibits an infinite number of spikes or comb teeth to amplify signals at frequencies that are multiples (harmonics) of the fundamental frequency. In a feedback control system, this selective signal amplification at specified frequencies can be used to attenuate disturbances at harmonics of the fundamental frequency. Consequently, a high-gain control is achieved at these frequencies.
In hard disc drive servo controls, it is often necessary to cancel multiple harmonics of the spindle motor. Consequently, not only must the fundamental harmonic (1×) be cancelled, but also 2×, 3×, 4× and so on. The adaptive feedforward compensation (AFC) can selectively cancel the RRO of different harmonic frequencies, for example, 1×, 3×, 5× etc., but the AFC is computationally more complex and expensive to implement. The RLC, on the other hand, is relatively simple to implement.
FIG. 1
illustrates one form of repetitive learning compensator (RLC) according to the prior art. Plant 80 is driven by a servo signal r and is subject to external repetitive disturbances. In a disc drive, plant
80
is the actuator assembly that includes an actuator arm driven by a voice coil motor to position a read/write head relative to the disc medium. The read/write head reads data from the disc and returns information concerning track positioning to summing node
81
where it modifies the input servo signal. A general filter
82
contains a low pass filter, known as Q-filter, and a memory buffer. The Q-filter provides a response Q(z
−1
) to shape the ideal internal model G
r0
(z
−1
), where
G
r0
(
z
-
1
)
=
1
1
-
z
-
N
.
The memory buffer has a length N equal to the number of servo sectors on a track on the disc medium. More particularly, the sample frequency fs is the fundamental frequency f
0
multiplied by the number n of servo sectors on the disc track, fs=n·f
0
, so N=n.
Feedback controller
84
provides a response C(z
−1
). In the control system shown in
FIG. 1
, the input signal, as modified by the signal from the head in plant
80
, is applied to filter
82
and is summed with the output from filter
82
at summing node
83
for input to feedback controller
84
. In other forms of RLCs, the output of summing node
83
is coupled to the input to filter
82
to form a loop with filter
82
, so that the filter input comprises the modified input signal summed with the filter output. In either case, filter
82
samples a modified input signal over a sampling period T
s
based on the number of embedded servo sectors in a track. The shaped internal model of filter
82
is
G
r
(
z
-
1
)
=
1
1
-
Q
(
z
-
1
)
z
-
N
.
In practice, the Q-filter in block
82
is used to restrict the bandwidth of the RLC because the maximal frequency of interest is far lower than the Nyquest frequency f
n
, which is half of the sampling frequency f
s
.
In
FIG. 1
, filter
82
is placed before feedback controller
84
. The RLC may also implemented as shown in
FIG. 2
with summing node
83
after the feedback controller
84
. The internal models of the RLC of
FIGS. 1 and 2
are similar.
Plant
80
is subject to external repetitive disturbances known as torque level disturbance denoted by d
T
and positional level denoted by dp. The RLC has a filter transfer function of H(z
−1
) determined by analyzing the RLC stability condition using the plant model. Considering the torque level repetitive disturbance d
T
and the positional level repetitive disturbance dp, the feedback controller
84
can be designed independently of the RLC. The RLC is employed to enhance the control performance of rejecting repetitive disturbances.
The ability of the RLC to reject disturbances is characterized by a disturbance rejection transfer function G
dy
(z
−1
) and a sensitivity function S(z
−1
). These are given by
G
dy
(
z
-
1
)
=
y
d
r
=
P
(
z
-
1
)
1
+
C
(
z
-
1
)
P
(
z
-
1
)
=
P
(
z
-
1
)
S
(
z
-
1
)
=
T
(
z
-
1
)
C
(
z
-
1
)
=
1
-
S
(
z
-
1
)
C
(
z
-
1
)
,
[
1
]
and
S
(
z
-
1
)
=
y
d
p
=
1
1
+
C
(
z
-
1
)
P
(
z
-
1
)
=
1
-
T
(
z
-
1
)
,
[
2
]
where T(z
−1
) is the complementary sensitivity function. In a nominal closed loop system,
G
cn
(
z
-
1
)
=
C
(
z
-
1
)
P
n
(
z
-
1
)
1
+
C
(
z
-
1
)
P
n
(
z
-
1
)
=
T
n
(
z
-
1
)
=
1
-
S
n
(
z
-
1
)
.
[
3
]
Consequently, the output, y, of the filter shown in
FIG. 2
can be represented as
y
=
Pd
r
+
d
p
1
+
CP
+
G
r
P
,
[
4
]
where &ggr; is the RLC gain. Therefore,
G
r
(
z
-
1
)
=
γ
Q
(
z
-
1
)
z
-
N
1
-
Q
(
z
-
1
)
z
-
N
H
(
z
-
1
)
.
[
5
]
The stability condition of the RLC is modeled in
FIG. 3
applying the small gain theorem with |Q(z
−1
)z
−N
|≦1, where ∀&ohgr;≦2&pgr;f
n
. Consequently, RLC stability condition may be expressed as
|1
−&ggr;G
dy
(
z
−1
)H(
z
−1
|·|Q
(
z
−1
)<1, [6]
w
Chen YangQuan
Ding MingZhong
Ooi KianKeong
Tan LeeLing
Xiu LinCheng
LandOfFree
Optimally designed parsimonious repetitive learning... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Optimally designed parsimonious repetitive learning..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Optimally designed parsimonious repetitive learning... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3161848