WHY DATA VERIFICATION IS NOT
ENOUGH!
The concept of disc testing is new to most
CD-ROM publishers. Until recently, only disc replication plants
tested the discs, and quality control was considered to be their
responsibility. Sampling output is an essential part of the
replication process, so most CD test equipment is designed for the
manufacturing environment. However, in today’s market,
developers, publishers, and distributors may find that doing their
own testing is desirable. And It is now possible to make accurate
and consistent quantitative measurements of disc quality using
low-cost desktop systems.
Traditionally, CD-ROM publishers have been
limited to testing their discs by comparing the data bit for bit
with the original source. However, data verification does not
reveal anything about the quality of the disc unless the
disc proves to be unreadable. Because of the superb error
correction capability of CD-ROMs, even a very poor disc still can
be readable. Data verification does not provide information that
indicates whether a disc is about to fail, is likely to have a
short life, or can be played on all players. In addition, this
test is only possible if the original data is available. Users of
desktop testing systems can place discs created by replicators or
on CD-Recorders in the unit and run a series of diagnostic tests
which provide feedback on disc quality. If the disc is
unsatisfactory, the cause can be determined and the problems can
then be solved. The desktop testing systems are priced in the
$2,500 - $15,000 range, as compared to starting prices around
$35,000 for commercial testing systems.
Why Test CD-ROM Discs?
Some of the key reasons to test discs are:
To verify disc quality.
Manufacturing CDs is a delicate art, and
information about how to make good discs is hard to find. Often,
this information has been kept proprietary by companies wanting to
maintain a competitive edge. Although most manufacturers produce
good quality products, some are better than others. Discs that are
made thinner than usual and physical defects such as pit
distortion from short manufacturing cycles indicate poor
manufacturing techniques and can result in unusable discs. If you
are paying for premium discs, you want to be sure the quality
measures up. In any case, no quality control program is perfect,
and problems can sometimes slip through.
To enhance quality control for CD-Recordable
users.
CD-R technology has revolutionized many aspects
of CD-ROM publishing. The low cost and rapid turnaround time make
the technology ideal for small volume publishing. Distribution on
CD-Rs is more economical than replication for quantities of less
than 2000 pieces. However, with CD-R usage, the publisher becomes
responsible for quality control, and CD-Rs are more subject to
problems than replicas. The continuous pregroove can sometimes
cause tracking problems on some players. Any variations or defects
in the dye layer can prevent proper recording. Although CD-R
technology has made great strides in the past few years, it is
fair to say that CD-Rs have a higher failure rate than replicas.
To provide confidence in product quality.
When problems occur, they can be resolved more
quickly and easily if publishers know the quality of their discs
and have a way to check them. Also, once a product is in the
field, fixing a problem is more difficult and expensive.
Publishers can avoid returns and maintain a good reputation by
tracking the quality of their discs.
To detect media and recorder compatibility
problems.
Compatibility between media, writer, and
software also is necessary to produce good discs. Certain
combinations of write strategy, record speed, and media type can
cause poor results. Determining the best media to use with a given
recorder (or vice versa) is greatly facilitated by testing disc
quality.
To test for errors that result from use of
premastering software.
Some premastering software leaves gaps between
sessions. These gaps generate uncorrectable errors. If
uncorrectable errors are found on a disc that otherwise looks
good, check to see if the errors are occurring between sessions.
The best way to avoid this problem is to write the disc in one
session ("Disc at Once").
To eliminate problems with mastering that
result from disc errors.
In the majority of glass mastering systems,
uncorrectable errors (referred to as E32; see discussion below)
automatically abort the mastering session, causing costly delays
for both the mastering house and the customer. Most replicators
prefer that you be able to demonstrate that your CD-R does not
reach this level of error.
To optimize selection of discs for archiving.
The initial quality of a disc profoundly affects
its potential lifetime. A good disc has greater tolerance for the
effects of aging than a poor one. If you are archiving either CD-Recordables
or replicas, you will want to select the highest quality discs to
archive. This is possible only if you have a way of measuring disc
quality.
Measuring Disc Quality
The Red Book specifies more than 50 parameters
that are used to define disc characteristics. To be certain of the
quality of a disc, tests on all parameters must be run. The
easiest way to measure disc quality, however, is to measure error
rates. This technique is effective because any serious problems on
the disc will affect the error rates. A meaningful quantitative
measure of disc quality can be obtained using a desktop disc
testing system to measure the quantity and severity of errors.
Errors are not necessarily physical features on
the disc, but are a measure of how well the total system (disc
plus player) works. In fact, all discs generate errors on playback
-even the best discs produce thousands of errors. The errors are
partly due to the fact that playing a CD is a difficult and
complicated process. The width of the pits in a disc is smaller
than the wavelength of the light used to detect them, so CD
players are operating at the limit of physical laws. Playing a
disc is not wholly deterministic, but rather is a statistical
process. Therefore, the results will not necessarily be the same
each time.
Moreover, the objective (the lens that picks up
the laser beam that reads the disc) must stay focused within a
range of less than 4 microns while the disc is moving both
horizontally and vertically. The pickup must follow the spiral
track of pits as it moves with an accuracy of less than one micron
(millionth of a meter). Several highly tuned and sophisticated
servo (control) systems are used to maintain focus, follow the
track, control the spindle speed, and handle timing issues related
to reading the data. These servos are very sensitive and work only
within a certain range. Thus, some of the errors occur as a result
of the narrow tolerances within which the equipment must function.
Two primary features of the disc itself that can
cause errors are poor pit geometry and physical defects. Pit
geometry refers to the depth, width, length, and wall slope of the
physical pits created in the disc. Although CD-R discs do not have
pits, the recording process produces areas on the disc that
function like pits and are subject to imperfections that cause
errors. Physical defects include pinholes, black spots, bubbles,
and scratches. Poor pit geometry and physical defects can make it
difficult or impossible for the servo mechanisms to read the data
properly.
A determination can often be made as to whether
problems are caused by pit geometry or local defects. A burst of
large errors confined to a small part of the disc is most likely
caused by some kind of local defect. If many large errors are
found over the whole disc (or a large part of it), then the
problem is most likely poor pit geometry.
It is extremely difficult to measure pit
geometry directly. An easier approach is to play a disc and look
at the signals produced by the pickup. By observing the pattern
(referred to as an “eye pattern”) created by the playback
signal, it is possible to measure features such as pit depth and
length.
CD-ROM Error Correction
As a way of compensating for the physical
limitation of discs and players, as well as for physical flaws in
the discs, a variety of error correction schemes are incorporated
into each CD-ROM. These schemes are described below.
Cross Interleave Reed-Solomon Code (CIRC)
All CDs incorporate an error detection and
correction scheme known as Cross Interleave Reed-Solomon Code
(CIRC). It would be impossible to make a usable CD without this
error correction, since many errors are generated playing even the
best discs under ideal conditions. CIRC is a powerful error
detection code that can detect and completely correct all errors
on a reasonably good disc. It relies on two principles for its
operation. One is redundancy. Extra parity bytes are added to the
data stream to facilitate error detection and correction. These
extra bytes reduce the available capacity of the CD by about 25%.
The other principle is interleaving. The data is
not recorded on the disc in its natural order as it would be on
magnetic tape. The data are organized into blocks of 24 bytes
(called “symbols” in CIRC terminology.) These are the blocks
referred to in the Block Error Rate measure. Four parity bytes are
added to each block of 24 symbols (bytes), making the block 32
bytes long. The data symbols belonging to one block are then
distributed over a fairly large area of the disc by
“interleaving” them with symbols from other data blocks. The
24 symbols of one data block end up distributed over 109 data
blocks. The advantage of this technique is that physical defects
on the disc do not eliminate complete data blocks, but instead,
parts of many blocks. These partially bad blocks can then be
reconstructed using the parity information.
CIRC error correction is done in two stages
referred to as C1 and C2, with deinterleaving of the data taking
place between the stages. The C1 stage is used to recover from
random errors caused by noise in the signal; the C2 stage is used
to recover from larger errors caused by physical defects such as
scratches and dirt. The error correction chip typically can
correct two bad symbols per block in the first stage, and two bad
symbols per block in the second stage. Some chips can correct four
bad symbols in the second stage in some cases.
Layered Error Detection Code and Error
Correction Code
CD-ROMs include an extra layer of error
detection and correction, usually called Layered Error Detection
Code and Error Correction Code (for brevity’s sake, typically
referred to as Layered ECC) which works similarly to CIRC by
adding parity information to each data sector. The extra error
correction capability can correct errors that are not correctable
by the CIRC because it adds additional parity bytes and additional
scrambling of the data.
Re-try
If a CD-ROM drive encounters an error which is
uncorrectable even by the layered ECC, it will try again to read
the sector. This is the failure mode most familiar to the CD-ROM
user. Obviously, having to re-try slows down access to the data,
but sometimes a second try is successful. If the drive cannot
recover the data within a certain number of retries, the disc is
unusable.
Analysis of Errors
Types of Errors
Disc testing equipment documents the errors that
are being detected ( and usually corrected) by CIRC. The errors
reported during disc testing are classified into a variety of
categories. An error of type E11 means that one bad symbol was
corrected in the C1 stage of CIRC. E21 means two bad symbols were
corrected in the C1 stage; E31 means that three or more errors
were found at the C1 stage. If there are 3 or more errors found at
the C1 stage, the block is uncorrectable at the C1 stage, and is
passed to the C2 stage. Because of the deinterleaving of the data
between the stages, those three (or more) bad symbols are now in
separate blocks, and so can be corrected by the C2 stage.
Similarly, E12 means that one bad symbol was
corrected in the C2 stage, and E22 means that two bad symbols were
corrected in the C2 stage. E32 means that three or more bad
symbols were found in one block at the C2 stage, and therefore
this error is not correctable.
Block Error Rate (BLER)
Block Error Rate (BLER) is defined as the number
of data blocks per second that have any bad symbols. BLER is the
most general and useful measurement of the quality of a disc. The
Red Book specification (IEC 908) calls for a maximum BLER of 220
per second averaged over ten seconds. Discs with higher BLERs are
likely to produce uncorectable errors. Presently, the best discs
have average BLERs below 10. A low BLER shows that the system as a
whole is performing well, and that the pit geometry is good.
Relying on the BLER alone is not advisable,
however. Although the BLER provides information on the number of
bad blocks per second, it does not indicate the severity of the
errors. In principle, a disc with an average BLER of five can be
unusable, if all the those errors are uncorrectable! The error
codes described above provide details that indicate the severity
of the errors and distinguish between correctable and
uncorrectable errors.
Interpretation of Error Data
Small errors (one or two bad symbols per block)
are generally considered noise. They have a random characteristic
of occurrence, and can be caused by dust, small scratches, etc.
Large errors of more than three symbols per block are considered
burst errors, and are generally due to some kind of physical
defect in the disc.
Testing equipment indicates, either on-screen or
via a printed report, the location and type of errors on the disc.
The most important aspect of testing is simply determining whether
there is a problem. From this starting point, the user then can
begin the process of determining the source-e.g., bad media,
incompatibility issues, or premastering software.
Standards for CD-ROM Error Rates
All CD-ROMs must meet Philips/Sony Yellow Book
specifications. Only two specifications are indicated for error
rates: BLER must not exceed 220, and no E22 or E32 errors may be
present. Beyond that , it is up to each supplier to decide on disc
specifications. CD-ROM users cannot tolerate any corruption of the
data, so the CD-ROM community has established some de facto
standards. Most CD-ROM manufactureres and publishers agree that
CD-ROMs should have an average BLER of less than 50, and a peak
BLER of less than 100.
Conclusion
The CD-ROM publishing community is changing
rapidly because of advances in CD-R technology, and responsibility
for testing is now shifting from manufacturers to publishers.
CD-ROM developers, publishers, and distributors can now monitor
the quality of their product using simple, affordable desk-top
test equipment which provides a quantitative measure of disc
quality.
Top
Home
Copyright © 1995 Clover
Systems
