ArticlePDF Available

COMPACT DISC: System Aspects and Modulation.

Authors:

Abstract

The sound from a Compact Disc system encoded into data bits and modulated into channel bits is sent along the 'transmission channel' consisting of write laser - master disk - user disk - optical pick-up. The maximum information density on the disk is determined by the diameter d of the laser light spot on the disk and the 'number of data bits per light spot'. The effect of making d smaller is to greatly reduce the manufacturing tolerances for the player and the disk. The compromise adopted is d approximately equals 1 mu m, giving very small tolerances for objective disk tilt, disk thickness and defocusing.
.'
Compact Disc: system aspects and modulation
J.
P.
J.
Heemskerk and K. A. Schouhamer lmmink
Fig.!. The Compact Disc system, considered as a transmission system that brings sound from the
studio into the living room. The transmission channel between the encoding system (COD) at the
recording end and the decoding system (DECaD) in the player, 'transmits' the bit stream Bi to
DECOD via the write laser, the master disc (MD), the disc manufacture, the disc (D) in the player
and the optical pick-up; in the ideal case Bo is the same as Bi. The bits of Bo, as well as the clock
signal
(Cl)
for further digital operations, have to be detected from the output signalof the pick-
up unit at Q.
Philips tech. Rev. 40, 157-164, 1982, No. 6
In this article we shall deal in more detail with the
various factors that had to be weighed one against
the other in the design of the Compact Disc system.
In particular we shall discuss the EFM modulation
system ('Eight-to-Fourteen Modulation'), which helps
to produce the desired high information density on
the disc.
recording
Fig. 1 represents the complete Compact Disc system
as a 'transmission system' that brings the sound of an
orchestra into the living room. The orchestral sound
is converted at the recording end into a
bit stream
Bi,
which is recorded on the master disc. The master disc
is used as the 'pattern' for making the discs for the
user. The player in the living room derives the bit
stream Ba .:..._which in the ideal case should be iden-
Dr
J.
P.
J.
Heemskerk is with the Philips Audio Division, Eind-
hoven; Ir K. A. Schouhamer Immink is with Philips Research
Laboratories, Eindhoven.
157
tical to Bi - from the disc and reconverts it to the
orchestral sound. The system between
COD
and
DECOD
is the actual
transmission channel;
Bi and Ba
consist of 'channel bits'.
Fig. 2 shows the encoding system in more detail.
The audio signal is first converted into a stream
111
of
'audio bits' by means of pulse-code modulation. A
number of bits for 'control and display' (C&D) and
the parity bits for error correction are then added to
the bit stream
[1] [2].
This results in the 'data bit
stream'
B2.
The modulator converts this into channel
bits (Ba). The bit stream Bi is obtained by adding a
synchronization signal.
(1]
M. G. Carasso,
J.
B. H. Peek and
J.
P. Sinjou, The Compact
Disc Digital Audio system, this issue, p. 151.
(2]
H. Hoeve,
J.
Timmermans and
L.
B. Vries, Error correction
and concealment in the Compact Disc system, this issue,
p.166.
158
J.
P.
J.
HEEMSKERK and K. A. SCHOUHAMER IMMINK Philips tech. Rev. 40, No. 6
The number of data bits nthat can be stored on the
disc is given by:
where
A
is the useful area of the disc surface,
d
is the
diameter of the laser light spot on the disc and 17is the
'number of data bits per spot' (the number of data
bits that can be resolved per length dof track). A/d2
is the number of spots that can be accomodated side
by side on the disc. The information density n/A is
thus given by:
The spot diameter dis one of the most important
parameters of the channel. The modulation can give a
higher value of 17.We shall now briefly discuss some
of the aspects of the channel that determine the speci-
fication for the modulation system.
We shall consider one example here to illustrate the
way in which such tolerances affect the design: the
choice of the 'spot diameter'
d.
We define
d
as the
half-value diameter for the light intensity; we have
d
=
O.6À/NA,
(1)
where Àis the wavelength of the laser light and
NA
is
the numerical aperture of the objective. To achieve a
high information density (1) dmust be as small as
possible. The laser chosen for this system is the small
CQL10
[3],
which is inexpensive and only requires a
low voltage; the wavelength is thus fixed;
À ::::;
800nm.
This means that we must make the numerical aperture
as large as possible. With increasing
NA,
however, the
manufacturing tolerances of the player and the disc
rapidly become smaller. For example, the tolerance in
the local 'skew' of the disc (the 'disc tilt') relative to
the objective-lens axis is proportional to
NA
-3.
The
it--------------------I
I
GOD
I
I ~
S~
I
I
COD GEN
I
I
I
I I
rcr+-
A PAR
I
I
COD
I
I
I
I ,
I :
i
8
1
8
2
8
3
1
8
i
L ~
Fig. 2. The encoding system
(€OD
in fig. I). The system is highly simplified here; in practice for
example there are two audio channels for stereo recording at the input, which together supply the
bit stream
Bl
by means of PCM, and the various digital operations are controlled by a 'clock',
which is not shown. The bit stream
B,
is supplemented by parity and C&D (control and display)
bits
(B2),
modulated (Ba), and provided with synchronization signals
(Bi). MUX:
multiplexers.
Fig. 9 gives the various bit streams in more detail.
The channel
The bit stream
Bi
in fig. lis converted into a signal
at
P
that switches the light beam from the write laser
on and off. The channel should be of high enough
quality to allow the bit stream
Bi
to be reconstituted
from the read signal at
Q.
To achieve this quality all the stages in the transmis-
sion path must meet exacting requirements, from the
recording on the master disc, through the disc manu-
facture, to the actual playing of the disc. The quality
of the channel is determined by the player and the
disc: these are mass-produced and the tolerances can-
not be made unacceptably small.
tolerance for the disc thickness is proportional to
NA
-4,
and the depth of focus, which determines the
focusing tolerance, is proportional to
NA
-2.
After
considering all these factors in relation to one an-
other, we arrived at a value of 0.45 for
NA.
We thus
find a value of 1
urn
for the spot diameter
d.
The quality of the channel is evaluated by means of
an 'eye pattern', which is obtained by connecting the
point
Q
in fig. 1 to an oscilloscope synchronized with
the clock for the bit stream
Bo;
seefig. 3a. The signals
originating from different pits and lands are super-
imposed on the screen; they are strongly rounded,
---------------------~- --
Philips tech. Rev. 40, No. 6COMPACT DISC DIGITAL AUDIO: MODULATION 159
mainly because the spot diameter is not zero and the
pit walls are not vertical.
If
the transmission quality is
adequate, however, it is always possible to determine
whether the signal is positive or negative at the 'clock
times' (the dashes in fig. 3a), and hence to reconstitute
the bit stream. The lozenge pattern around a dash in
this case is called the 'eye'. Owing to channel im-
perfections the eye can become obscured; owing
Fig. 3. Eye pattern. The figures give the read signal (at
Q
in fig. 1)
on an oscilloscope synchronized with the bit clock. At the decision
times (marked by dashes) it must be possible to determine whether
the signal is above or below the decision level
(DL).
The curves
have been calculated for a) an ideal optical system,
b)
a defocusing
of 2
urn,
c) a defocusing of 2
urn
and a disc tilt of 1.2
0
The curves
give a good picture of experimental results.
to 'phase jitter' of the signal relative to the clock
an eye becomes narrower, and noise reduces its
height. The signals in fig. 3a were calculated for a
perfect optical system. Fig. 3b shows the effect of de-
focusing by 2 J.Lmand fig. 3c shows the effect of
a radial tilt of 1.2° in addition to the defocusing. In
fig. 3b a correct decision is still possible, but not in
fig. 3c.
This example also gives some idea of the exacting
r.equirements that the equipment has to meet. A more
general picture can be obtained from Table I, which
gives the manufacturing tolerances of a number of
important parameters, both for the player and for the
disc. The list is far from complete, of course.
With properly manufactured players and discs the
channel quality can still be impaired by dirt and
scratches forming on the discs during use. By its
nature the system is fairly insensitive to these
[1],
and
any errors they may introduce can nearly always be
corrected or masked
[2].
In the following we shall see
that the modulation system also helps to reduce the
sensitivity to imperfections.
Table
I.
Manufacturing tolerances.
Disc
Pit-edge positioning
±
50 nm
Pit depth 120
±
10 nm
Player Objective-lens tilt
±
0.2
0
Tracking
±
0.1
JlID
Focusing
±
0.5
JlID
BIBLIOT E
. H EK
NAT.
LAB
R.M.S. wavefront noise of read II\~eJ'¥aI?L9PJs,A (
(40 nm)
G
LOEIL.l
'\0::"
EKEN
Thickness 1.2 ±0.1 mm P051....,. S 80.000
Flatness
±
0.6
0
(at the rim c01PeÇ80nldk
g
~I)lQijQ,YEN
0.5 mm)
Bit-stream modulation
The playing time of a disc is equal to the track
length divided by the track velocity
v.
For a given disc
sizethe playing time therefore increases if we decrease
the track velocity in the system (the track velocity of
the master disc and of the user disc). However, if we
do this the channel becomes 'worse': the eye height
decreases and the system becomes more sensitive to
perturbations. There is therefore a lower limit to the
track velocity if a minimum value has been established
for the eye height because of the expected level of
noise and perturbation. We shall now show that we
can decrease this lower limit by an appropriate bit-
stream modulation.
We first consider the situation without modulation.
The incoming data bit stream is an arbitrary sequence
of ones and zeros. We consider a group of 8 data bits
in which the change of bit value is fastest (fig. 4a).
Uncoded recording (1: pit; 0: land, or vice versa) then
gives the pattern of fig. 4b. This results in the
rounded-off signalof fig. 4c at Q in fig. 1; fig. 4d gives
the eye pattern. The signal in fig. 4c represents the
highest frequency Urnl) for this mode of transmis-
sion, and we have!rnl =!!d, where
ja
is the data bit
rate. The half eye height al is equal to the amplitude
Al of the highest-frequency signal.
[3)
J.
C.
J.
Finek, H.
J.
M. van der Laak and
J.
T. Schrama,
A semiconductor laser for information read-out, Philips tech.
Rev. 39, 37-47, 1980.
160
J.
P.
J.
HEEMSKERK and K. A. SCHOUHAMER IMMINK Philips tech. Rev. 40, No. 6
-i
Q
1 :
0
I
I
I
I
I
I
I
o
o
o
I
Tmin:
I :
I
I
I
I :
I
I
I
I I
I
I : : I
Aln
V"\ ~ ~
It
P:
~'C7~
14
.1
I
1
1/fml
1
1 :
I
I
I
I
I
I
01>oOOc
I I
I I
~
T=1/fd
Fig. 4. Direct recording of the data bit stream on the disc. a) Data
bit stream of the highest frequency that can occur.
b)
Direct trans-
lation of the bit stream into a pattern of pits.
c)
The corresponding
output signal (at
Q
in fig. I); its amplitude Al is found with the aid
of fig. 5. d) The eye pattern that follows from
(c).
Tmin minimum
pit or land length;fml highest frequency; Tdata bit length;/d data
bit rate. We have Tml
n
=
T;
fml
=
!/d.
The relation between the eye height and the track
velocity now follows indirectly from the 'amplitude-
frequency characteristic' of the channel; seefig. 5. In
this diagram
A
is the amplitude of the sinusoidal sig-
nal at
Q
in fig. 1 when a sinusoidal unit signalof fre-
quency
i
is presented at
P.
With the aid of Fourier
A
-t
A2~------------~
0;= '/
2
A
2
ol=A,~--------------+-------~
00
f
m
2
f
m
l
-f
Fig. 5. Amplitude-frequency characteristic of the channel. The dia-
gram gives the amplitude
A
of the sinusoidal signal at
Q
(fig. 1)
when a sinusoidal unit signal is presented at
P
as a function of the
frequency f. The transfer is 'cut off' at the frequency fe, which is
given by fe
=
(2NA/À)v. The line shown applies to an ideal optical
system; in reality Ais always somewhat lower; the cut-off frequency
is then effectively lower. The 'maximum frequencies' fmb
[ea,
the
amplitudes Al. A2 and the -'half eye heights' a}, a2 relate to the
'direct' and 'modulated' writing of the data bits on the disc; see
figs 4 and 6.
-i
OiO
1
1
I I
I I
I
1
10:0:0
11
: I
1
: : I
I I
I
o
0
I
1
'00000"1
1
1
~-------~--
L
J
I
Q
A2'~""""'7""V--.i ...
..._1_... ...
!"-"
Q2
---¥--¥--¥
Fig. 6. Eight-to-sixteen modulation. Each group of 8 data bits (a) is
translated with the aid of a dictionary into 16 channel bits
(a'),
in
such a way that the run length is equal to at least three channel bits.
b) Pattern of pits produced from the bit stream
(a').
b') pattern of
pits obtained with a different input signal.
c)
The read signal cor-
responding to (b); its amplitude is again determined from fig. 5.
d) The resultant eye pattern. The half eye height (a2) here is only
half the amplitude (A2) of the approximately sinusoidal signal of
maximum frequency
(fm2).
analysis and synthesis the output signal can be cal-
culated from
A(f)
for any input signal. The line in
the diagram represents a channel with a perfect optical
system. In the first part of this section we shall take
this for granted. The true situation will always be less
favourable. The 'cut-off frequency' is determined by
the spot diameter and the track velocity v; in the ideal
casefe
=
(2NA/À.)v.
For a given track velocity we now obtain the half
eyeheight
al
in fig. 4 directly from fig. 5: it is equal to
the amplitude
A
I
at the frequency
f
mi .
If
v,
and hence
[«; is varied, the line in fig. 5 rotates about the point 1
on the A-axis. For a given minimum value of
al,
the
figure indicates how far ie can be decreased; this
establishes the lower limit for
v.
In particular, if the
minimum value for
al
is very small, ie can be de-
creased to a value slightly above
i
mt
(=
!
id)._
Fig.
6 gives the situation
with
modulation: an ima-
ginary 8-+ 16 modulation, which is very close to EFM,
however. Each group of 8 incoming data bits (fig. 6a)
is converted into 16 channel bits (fig. 6a'). This is
done by using a 'dictionary' that assigns unambigu-
ously but otherwise arbitrarily to each word of 8 bits a
word of 16 bits, but in such a way that the resultant
channel bit stream only produces pits and lands that
Philips tech. Rev. 40, No. 6 COMPACT DISC DIGITAL AUDIO: MODULATION 161
are at least three channel bits long (fig. 6b). On the
time scale the minimum pit and land lengths ('the
minimum run length'
Tmin)
have become
I~
times as
long as in fig. 4, but a simple calculation shows that
about as much information can nevertheless be trans-
mitted as in fig. 4 (256 combinations for 8 data bits),
because there is a greater choice of pit-edge positions
per unit length (see fig. 6b and bi); the 'channel bit
length'
Tc
has decreased by a half.
With the modulation we have managed to reduce
the highest frequency (fm2) in the signal (see fig. 6c,
left; fm2
=
kfd
=
ifml). Therefore fc and vcan be
reduced by a factor of
I~
for the case in which a very
small eye height is tolerable (see fig. 5); this represents
an increase of 50% in playing time.
The modulation also has its disadvantages. In the
first place the half eye height (a2) in this case is only
half of the amplitude (A2) of the signal at the highest
frequency (see fig. 6d). This has consequences if the
minimum eye height is not very small. For example,
the modulation becomes completely unusable if the
half eye height in fig. 5 has to remain larger than ~
(a2
>~
implies A2> 1); uncoded recording is then
still possible
(AI
=al). In the second place, the tol-
erance for time errors and for the positioning of pit
edges, together with the eye width (Tc), has decreased
by a half. In designing a system, the various factors
have to be carefully weighed against one another ..
To show qualitatively how a choice can be made, we
have plotted the half eye height infig.
7
as a function
of the 'linear information density'
(J
(the number
of incoming data bits per unit length of the track;
(J
=fd/V) for three systems:' '8~8 modulation' (i.e.
uncoded recording), 8~I6 modulation, and a system
that also has about the same information capacity (256
combinations for 8 data bits) in which, however, the
minimum run length has been increased still further,
again at the expense of eye width of course ('8~24
modulation',
Tmin
=
2T,
Tc
=
1
T).
The figure is
a
direct consequence of the reasoning above, with the as-
sumption that the cut-off frequency is 20% lower than
the ideal value (2NA/À)v, as a first rough adjustment to
what we find in practice for the function
A(f).
In qualitative terms, the 8~ 16 system has been
chosen because the nature of the noise and perturba-
tions is such that the eye can be smaller than at
A
in
. fig. 7, but becomes too small at C. An improvement is
therefore possible with 8~I6 modulation; but not
with 8~24 modulation.
For our Compact Disc system we have
(J
=
1.55
data bits/urn
(fd
=
1.94 Mb/s,
v
=
1.25 mis [1]); the
operating point would therefore be at Pin fig. 7. The
model used is however rather crude and in better
models
A,
Band C lie more to the left, so that
P
ap-
proaches C. But 8~ 16 modulation is still preferable
to 8~24 modulation, even close to
C,
since the eye
width is I~ times as large as for 8~24 modulation.
EFM is a refinement of 8~I6 modulation.
It
has
been chosen on the basis of more detailed models and
many experiments. At the eye height used, it gives a
gain of 250/0 in information density, compared with'
uncoded recording.
a
t
Fig. 7. Half eye height
a
as a function of the linear information
density
u,
for 8-+8, 8-+16 and 8-+24 modulation. These systems
are characterized by the following values for the channel bit length
Tc and the minimum run length Tmin:
8-+8: Tc
=
T,
Tmin
=
T(fig. 4),
8-+16: Tc
=
l
T,
Tmin
= ~
T(fig. 6),
8-+24: Tc
=
1
T, Tmin
=
2T,
where
T
is the data bit length. The straight lines give the relations
that follow from fig. 5:
al
=
cI(1 -
fml/fe) -+ al
=
1 - u/1.8,
a2
=
C2(1 - fm2/fe) -+ az
=
0.5(1 - u/2.7),
a3
=
cs(l -
fms/fe) -+
as
=
0.26(1 - u/3.6),
where
u
is the numerical value of the linear information density,
expressed in data bits per
urn,
The c's are the ratios of the half eye
height to the amplitude, and thefm's the maximum frequencies for
the three systems
(Cl
=
1,
C2
=
sin 30°
=
0.5,
Cs
=
sin 15°
=
0.26,
fml
=
lid,
fm2
=
lid,
fm3
=
!/d; /d
is the data bit rate). The
second set of equations follows from the first set by substituting
0.8 x
(2NA/À.)v
for fe, with
NA
=
0.45,
À.
=
0.8 urn, V
=/d/u.
The
factor 0.8 is introduced as a rough first-order correction to the
'ideal' amplitude characteristic.
Further requirements for the modulation system
In developing the modulation system further we
still had two more requirements to take into account .
In the first place it must be possible to regenerate
the bit clock in the player from the read-out signal (the
signal at Q in fig. I). To permit this the number of pit
edges per second must be sufficiently large, and in par-
ticular the 'maximum run length'
T
max
must be as
small as possible.
The second requirement' relates to the 'low-fre-
quency content' of the read signal. This has to be as
162
J.
P.
J.
HEEMSKERK and K. A. SCHOUHAMER IMMINK Philips tech. Rev. 40, No. 6
small as possible. There are two reasons for this. In
the first place, the servosystems for track following
and focusing
[1]
are controlled by low-frequency
signals, so that low-frequency components of the in-
formation signal could interfere with the servo-
systems. The second reason is illustrated in fig. 8, in
.which the read signal is shown for a clean disc (a) and
for a disc that has been soiled, e.g. by fingermarks (b).
This causes the amplitude and average level of the
signal to fall. The fall in level causes a completely
100% 100% 50%-----
DL-~ ~ O%~-
0% 0% -50%-----
g
Fig. 8. The read-out signal for six pit edges on the disc, a) for a
clean disc,
b)
for a soiled disc, c) for a soiled disc after the low fre-
quencies have been filtered out.
DL
decision level. Because of the
soiling, both the amplitude and the signal level decrease; the deci-
sion errors that this would cause are eliminated by the filter.
wrong read-out if the signal falls below the decision
level. Errors of this type are avoided by eliminating
the low-frequency components with a filter (c), but
the use of such a filter is only permissible provided the
information signal itself contains no low-frequency
components. In the Compact Disc system the fre-
quency range from 20 kHz to 1.5 MHz is used for in-
formation transmission; the servosystems operate on
signals in the range 0-20 kHz.
\
The EFM modulation system
Fig.
9 gives a schematic general picture of the bit
streams in the encoding system. The information is
divided into 'frames'. One frame contains 6 sampling
periods, each of 32 audio bits (16 bits for each of the
two audio channels). These are divided into symbols
of 8 bits. The bit stream
Bi
thus contains 24 symbols
per frame. In
B2
eight parity symbols have been added
and one C&D symbol, resulting in 33 'data symbols'.
The modulator translates each symbol into a new
symbol of 14 bits. Added to these are three 'merging
bits', for reasons that will appear shortly. After the
addition of a synchronization symbol of 27 bits to the
frame, the bit stream Bi is obtained. Bi therefore con-
tains 33
X
17
+
27 =588 channel bits per frame.
Finally, Bi is converted into a control signal for the
write laser.
It
should be noted that in Bi '1' or '0' does
not mean 'pit' or 'land', as we assumed for simplicity
in fig. 6, but a '1' indicates a pit edge. The informa-
tion is thus completely recorded by the positions of
the pit edges; it therefore makes no difference to the
decoding system if 'pit' and 'land' are interchanged
on the disc.
Opting for the translation of series of 8 bits following the divi-
sion into symbols in the parity coding has the effect of avoiding
error propagation. This is because in the error-correction system an
entire symbol is always either 'wrong' or 'not wrong'. One channel-
bit error that occurs in the transmission spoils an entire symbol, but
- because of the correspondence between modulation symbols and
data symbols - never more than one symbol.
If
a different modul-
ation system is used, in which the data bits are not translated in
groups of 8, but in groups of 6 or 10, say, then the bit stream
B2
is
in fact first divided up into 6 or IQbit 'modulation symbols'. Al-
though one channel-bit error then spoils only one modulation sym-
bol, it usually spoils two of the original 8 bit symbols.
In EFM the data bits are translated 8at a time
into 14 channel bits, with a
Tmin
of 3 and a
Tmax
of
11 channel bits (this means at least 2 and at the most
10 successive zeros in Bi). This choice came about
more or less as follows. We have already seen that the
choice of about I! data bits for
Tmin,
with about 16
channel bits on 8data bits, is about the optimum for
the Compact Disc system
[4].
A simple calculation
shows that at least 14 channel bits are necessary for
the reproduetion of all the 256 possible symbols of 8
data bits under the conditions
Tmin
=3,
Tmax
=11
channel bits. The choice of
Tmax
was dictated by the
fact that a larger choice does not make things very
much easier, whereas a smaller choice does create far
more difficulties.
With 14 channel bits it is possible to make up 267
symbols that satisfy the run-length conditions. Since
we only require 256, we omitted 10 that would have
introduced difficulties with the 'merging' of symbols
under these conditions, and one other chosen at
random. The dictionary was compiled with the aid of
computer optimization in such a way that the transla-
tion in the player can be carried out with the simplest
possible circuit, i.e. a circuit that contains the mini-
mum of logic gates.
The merging bits are primarily intended to ensure
that the run-length conditions continue to be satisfied
when the symbols are 'merged' .
If
the run length is in
danger of becoming too short we choose 'O's for the
merging bits; if it is too long we choose a '1' for one
of them.
If
we do this we still retain a large measure of
freedom in the choice of the merging bits, and we use
this freedom to minimize the low-frequency content
of the signal. In itself, two merging bits would be.
sufficient for continuing to satisfy the run-length con-
Philips tech. Rev. 40, No. 6 COMPACT DISC DIGITAL AUDIO: MODULATION 163
are shown two data symbols of
B2
and their transla-
tion from the dictionary into channel symbols (Ba).
From the
Tmin
rule the first of the merging bits in this
case must be a zero; this position is marked
'X'.
In
ditions. A third is necessary, however, to give sufficient
freedom for effective suppression of the low-frequency
content, even though it means a loss of
6070
of the
information density on the disc. The merging bits
-t
1frame
t
6 sampling periods
32 bits per sampling period
" l
,
4
symbols of
8
bits
,,
24 audio symbols
1
+1C&D
-Bporiiy
33 data symbols
,
"
"
"
"
"
"
+14 (EFM)
!
i
3 (merging) ;
i
~~~~~~~~~~~~~~-1717channelbffs: per symbol:
,,
,
33
x
17 channel bits
+
27
sync bits
588 channel bits
sync
2xTmax
D
_j
,--
or
.,
L___
Fig. 9. Bit streams in the encoding system (fig. 2). The information is divided into frames; the
figure gives one frame of the successive bit streams. There are six sampling periods forone frame,
each sampling period giving 32 bits (16 for each of the two audio channels). These 32 bits are
divided to make four symbols in the 'audio bit stream' BI. In the 'data bit stream'
B2
eight parity
and one C&D symbols have been added to the 24 audio symbols. To scatter possible errors, the
symbols of different frames in BI are interleaved, so that the audio signals in one frame of
B2
originate from different frames in BI. The modulation translates the eight data bits of a symbol of
B2
into fourteen channel bits, to which three 'merging bits' are added
(Bs).
The frames are
marked with a synchronization signalof the form illustrated (bottom right); the final result is the
'channel bit stream' (BI) used for writing on the.master disc, in such a way that each' l' indicates a
pit edge
(D).
contain no audio information, and they are removed
from the bit stream in the demodulator. /
Fig. 10 illustrates, finally, how the merging bits are
determined. Our measure of the low-frequency con-
tent is the 'digital sum value' (DSV); this is the dif-
ference between the totals of pit and land lengths ac-
cumulated from the beginning of the disc. At the top
the two following positions the choice is free; these
are marked
'M'.
The three possible choices
XMM
=,
000,010 and 001 would give rise to the patterns of pits
as illustrated, and to the indicated waveform of the
[4]
A more detailed discussion is given in K. A. Immink, Modu-
lation systems for digital audio discs with optical readout,
Proc. IEEE Int. Conf. on Acoustics, speech and signal pro-
cessing, Atlanta 1981, pp. 587-589.
164 COMPACT DISC DIGITAL AUDIO: MODULATION Philips tech. Rev. 40, No. 6
DSV, on the assumption that the DSV was equal to 0
at the beginning. The system now opts for the merging
combination that makes the DSV at the end of the
second symbol as small as possible, i.e. 000 in this
case.
If
the initial value had been
-3,
the merging
combination 001 would have been chosen.
o
1
100
0 0 1
channel bits
'r:-:=-=""""':-:::-:~'"""';-;-;"="'+'O::-:O::-:I;-:O::-:O::-:I;-;O::-:O'-:I;-:O::-:O;-:';-:O'"'Od:M:-:-
(8
3)
XMM=OOO
XMM=010·
XMM=001
/., /., !I
010
).. .1'.,.'.,
.>; " : 001
,.:.,. .' ,. .'.:»
r
{ ........I
I
I
DSV
t
000
Fig.
10.
Strategy for minimizing the digital sum value (DSV). After
translation of the data bits into channel bits, the symbols are
merged together by means of three extra bits in such a way that the
rim-length conditions continue to be satisfied and the DSV remains
as small as possible. The first run-length rule (at least two zeros one
after the other) requires a zero at the first position in the case illus-
trated here, while the choice remains free for the second and third
positions. In this case there are thus three merging alternatives: 000,
010 and 001. These alternatives give the patterns of pits shown in
the diagram and the illustrated DSV waveform. The system chooses
the alternative that gives the lowest value of DSV at the end of the
next symbol. The system looks 'one symbol ahead'; strategies for
looking further ahead are also possible in principle.
When this strategy is applied, the noise in the servo-
band frequencies
«
20 kHz) is suppressed by about
10 dB. In principle better results can be obtained,
within the agreed standard for the Compact Disc sys-
tem, by looking more than one symbol ahead, since
minimization of the DSV in the short term does not
always contribute to longer-term minimization. This
is not yet done in the present equipment.
Summary. The Compact Disc system can be considered as a trans-
mission system that brings sound from the studio into the living
room. The sound encoded into data bits and modulated into chan-
nel bits is sent along the 'transmission channel' consisting of write
laser - master disc - user disc - optical pick-up. The maximum
information density on the disc is determined by the diameter dof
the laser light spot on the disc and the 'number of data bits per light
spot'. The effect of making
d
smaller is to greatly reduce the manu-
facturing tolerances for the player and the disc. The compromise
adopted is
d'"
1
urn,
giving very small tolerances for objective and
disc tilt, disc thickness and defocusing. The basic idea of the
modulation is that, while maintaining the minimum length for 'pit'
and 'land' (the 'minimum run length') required for satisfactory
transmission, the information density can be increased by increas-
ing the number of possible positions per unit length for pit edges
(the bit density). Because of clock regeneration there is also a maxi-
mum run length, and the low-frequency content ofthe transmission
channel must be kept as low as possible. With the EFM modulation
system used each 'symbol' of eight data bits is converted into 14
channel bits with a minimum run length of 3 and a maximum run
length of 11 bits, plus three merging bits, chosen such 'that, when
the symbols are merged together, the run-length conditions con-
tinue to be satisfied and the low-frequency content is kept to the
minimum.
Philips tech. Rev. 40,1982, No. 6
This prototype player, which will be put on the market later, will
display 'information for the listener' such as title, composer,
'track number' and playing time of the piece of music. The dif-
ferent sections of the music on the disc can also be played in the
order selected by the user - the numbers on the far right.
165
... To understand how these signalprocessing methods arose, we review a few basic facts about the physical process underlying digital magnetic recording. (Readers interested in the corresponding background on optical recording may refer to [25], [84], [102,Ch. 2], and [163]. ...
... Another well-known application of this method is that of the Eight-to-Fourteen Modulation (EFM) code, a rate code which is implemented in the compact audio disc [96], [84], [109]. A collection of 256 codewords is drawn from the set of lengthwords that satisfy the constraint. ...
... We can choose 267 kinds of codes which complies with the limitation of (2,10) run-length from the 16384 candidate codes. Last we can 256 kinds of channel code which correspond the 8 bit source data [8] . ...
... The parameter comparison between RLL(2,12;8,15) modulation and several common modulation is shown in Table 7. This (8,15) modulation code choose small d and large k, aim to achieve the top limit capacity on the modulation rate and enhance the density rate DR. As a result, the performance improved 6.67% than the EFMPlus appilied in DVD [17] . ...
Article
Full-text available
In digital storage system, the aim of channel coding is to improve efficiency and reliability of the channel. In order to improve the capacity and quality for optical storage system, the study on the modulation code is significant. A new RLL (2, 12; 8, 15) run-length limited code is presented. The construction method of the code is discussed and the procedures in encoding and decoding of the code are also given. The major characteristic parameters of different run-length limited codes are compared; the decoder is implemented based on FPGA. The RLL(2, 12; 8, 15) code with high modulation rate, namely high encoding efficient, is fit for high density optical storage systems. The study of modulation code for high-density optical discs with independent intellectual property rights is also of great significance.
... The resulting bit rate (for every three original audio bits an extra bit is added) is § x 1.41 = 1.83 Mbit s-1 . Eight bits of c&D information are added per frame and the whole bit stream is now modulated according to the rules of the EFM modulation (see Heemskerk andImmink 1982, Ogawa andImmink 1982). In EFM code blocks of eight data bits are transformed into seventeen channel bits with special properties that make them suitable for the optical channel. ...
Chapter
Full-text available
The modulation and coding format to be used has to be chosen in such a way that it matches the special requirements of the optical channel. The main characteristics of the channel are bandwidth limitation and asymmetry of the recording process.
... However, when RS coded data are used with a block decodable line code, the input size of the line code word can be selected to m atch the symbol size of the error control code. This avoids error extension due to the line code since it is restricted in only one n -b it symbol of the RS code[43]. The disadvantage of this approach is th at the maximum rate of the line code th a t ...
Thesis
Channel coding is an important consideration influencing the design of a communications system. In particular, error control coding is used to detect and/or correct errors and line coding to modify the characteristics of the transmitted signal to suit other constraints of the channel, such as restricted frequency response. This thesis explores aspects of channel coding for such constrained channels with emphasis given to error control coding. Specifically, the hrst chapter of this thesis presents a general overview of channel coding, presents the organisation of the thesis and details the main contributions. The second chapter gives an overview of the principles of error control coding and line coding and explains a few terms that are connnonly used in the remainder of the thesis. One kind of constrained channel investigated here is the binary asymmetric error channel, where error transitions from one to zero occur with different probability than from zero to one. Error correcting codes for this channel and their properties are investigated in the third chapter. The fourth chapter introduces disparity control coding, and proposes a new error control coding structure that satisfies disparity constraints for both binary asymmetric and symmetric error channels. Run length limited channels are the subject of the hfth chapter. A new coding structure is proposed that offers advantages in performance over the one conventionally used for error control in such channels. The sixth chapter introduces peak power constraints present in multi-carrier systems. Codes that can be used limit the peak to average power ratio of such systems are presented and the application of the coding structure of the fifth chapter is also discussed. The final chapter brings the thesis to a conclusion by summarising the main results and proposing areas where further work may be fruitful.
... For the EFM look-up table a number of 256 combinations is required. There are 10 combinations rejected from the reason of keeping the DSV (Digital Sum Value) [5] as close to zero as possible, and another one randomly chosen. The EFM code is a blockdecodable code constructed by computer optimization so that the simplest implementation results. ...
Article
Full-text available
This paper proposes a method of construction for data translation codes using a look-up table (LUT) dictionary. These codes are used for binary RLL constraint channels. It is studied the case of constraints for one symbol of the binary channel. Based on back-tracking algorithm, it is developed an iterative technique to obtain the channel sequences (code words) under the (d,k) constraints from the source sequences (information words). As applications, two examples are given: the EFM code used in data translation for the optical channel from the Compact Disc system and the RLL (2,7) & (1,8) codes used in playback/recording systems for magnetic storage. However, the proposed method could be extended to any binary RLL channel with constraints for one symbol.
Chapter
Linear Sequential CircuitsConvolutional Codes and EncodersDescription in the D-Transform DomainConvolutional Encoder RepresentationsConvolutional Codes in Systematic FormGeneral Structure of Finite Impulse Response and Infinite Impulse Response FSSMsState Transfer Function Matrix: Calculation of the Transfer FunctionRelationship between the Systematic and the Non-Systematic FormsDistance Properties of Convolutional CodesMinimum Free Distance of a Convolutional CodeMaximum Likelihood DetectionDecoding of Convolutional Codes: The Viterbi AlgorithmExtended and Modified State DiagramError Probability Analysis for Convolutional CodesHard and Soft DecisionsPunctured Convolutional Codes and Rate-Compatible SchemesBibliography and ReferencesProblems
Article
Full-text available
The principles of optical fiber communication and optical data storage have evolved over the past decade into industry standards for transmission, distribution, storage, and archival of digital audio, multimedia, and computer data. The fundamentals of these sophisticated systems are traditionally taught in the lecture hall, and laboratory exposure to the hardware is at the device rather than the system level. In our educational approach, after the student has been introduced to the device basics, the compact disc audio player is used as a tool to teach the fundamentals of optical communication and storage. The ubiquitous compact disc audio system, from analog input, through optical storage and distribution, to audio reproduction, provides an excellent model of a complete real world optical transmission and storage system. The laboratory time is divided into three segments: Introduction to Electro- Optic Devices, Optical Storage, and Optical Communication. During the introduction, students learn the basics of emitters, detectors, and optical fiber. The optical data storage and retrieval function of the compact disc player is investigated in the second segment. Students are introduced to optical information storage techniques, information density limits, the optical pick-up, laser diodes, photodiode arrays, eye patterns, and the tracking and focusing sub-systems. In the third segment, the compact disc system is used as a model of an entire optical communication system. Students are introduced to sampling and quantization, channel encoding techniques, modulation, high-speed transmitters and receivers, demodulation, decoding, error correction, digital signal processing, and digital-to-analog conversion.
Conference Paper
The principal feature of a (d, k) (or other finite-type constraints) code produced with the sliding-block code algorithm is that the coded sequences can be decoded by examining a limited number of consecutive symbols without relying on external state information. As an immediate consequence, these codes have a limited amount of error propagation. The length of the decoding window is an important design parameter as it affects both the amount of error propagation and decoding hardware. The paper describes the construction of rate m/n (d, k) codes that can be decoded with a sliding-block decoder of window length at most two n-tuples. We furnish sufficient conditions for the construction of such codes. A lower bound to the code size is given. The theory is elucidated by examples of (d,k) codes that require only part of the decoding window
ResearchGate has not been able to resolve any references for this publication.