Errors

(pp 133-138 of text)

There are a multitude of digital audio formats. Different formats have become the subject of heated discussions as to which format is 'superior'. These format wars are fueled by manufacturers claiming that 'theirs is better'. Often the discussion is centered around which format is less prone to error and the methods used to correct errors.

Error correction in digital audio works in a similar fashion to spelling checkers in some word processors. Some word processors fix spelling mistakes as you type. For instance, you may type the word 'yis'...the spelling checker finds the error and changes the word to 'yes'. It does this by looking at the y and s and after comparing 'yis' to a data bank of words assumes it was supposed to be 'yes'.

Let us take a brief look at what causes errors .

CAUSES OF ERRORS

As we have learned, errors increase as the level decreases. As hard as we strive to eliminate errors, we have seen that errors are as common as viruses found in a healthy body.

One of the remarkable benefits of digital audio technology is that errors can be fixed. It is a good thing since we have seen many areas in the digital stream where errors can be introduced. Jitter, droop, aliasing, and more occur throughout the encoding and decoding process.

The medium that the data is stored on can also introduce error. Defects or even a particle of dust can cause the loss of thousands of bits.

The computer industry standard allows for an error rate of 10^-13 , that is less than one uncorrectable error in 10^-12 (one trillion) bits. The audio industry is a little more forgiving. DAT assumes a data error rate of 10^-5 (at 48kHz) ..about 2 errors every second.

Tape based mediums can be affected by dust, scratches, fingerprints, smoke, stretched tape, impure oxide, and more.

Optical data devices can be affected by pit asymmetry, bubbles, defects and coating defects.

The most severe type of error is drop out. Dropout is caused by:

1.a defect in the medium during manufacturing,

2.a defect introduced during use.

A defect causes misrepresentation consisting of erroneous data, or loss of data. Loss of data can often be heard as clicks and pops caused by the DAC jumping to new amplitudes.

While an error in the LSB may go undetected, an error in the MSB would affect one half of the signal amplitude and would be quite noticeable. (A MSB error of .01% -1 part in 10,000 - negates 2 LSB).

KINDS OF ERRORS

There are two basic kinds of errors.

1. Random-bit. Errors that have no relation to each other occur singly and are easily corrected. Where individual bits are corrupted.

2. Burst error. Where a sequence of bits are corrupted - e.g. a long scratch in a CD .A large, sustained error caused by noise spikes, connector problems, transmission problems in the connecting cables, defects, dust, etc. Burst error usually results in data and redundant data loss. Correction is difficult.

Burst error is a serious phenomenon. So much so that systems are measured by how they react to burst errors. The maximum number of continuous error bits that can be corrected is a measure of the quality of a system.

Also, a system must be able to correct random and burst errors simultaneously.

Drop Out on Tape format.

Dust, while it may be barely perceptible to the eye, can create havoc as it moves to different locations on the tape. The dust can also come between the tape and head resulting in spacing loss.

Manufactured drop out usually occurs at the very first or end of the tape, where the tape was cut.

(DAT machines - 1/2 of each channel is recorded by each head. 1 head records rt. ch even sounds- 1 head records left. ch odd sound.)

Drop Out on Optical Drive

During the manufacturing process, defects could be created which cause drop out.

Dust and external particles can be cleaned off of optical media like with C.D., although if a scratch occurs, any resulting error will be permanent. That is why a C.D. cleaning kit recommends wiping with a soft cloth across the C.D. instead of around the perimeter. A scratch across the C.D. will cause small errors in multiple sections rather than one big error in one section.

Other Causes

The cables used can cause error. A cable can reduce high frequencies, reduce bandwidth, and add delay.

Cross-talk, imperfectly erased signals, and mechanical malfunction...

Impedance mismatch, RF interference, improper shielding and grounding can cause radio frequency interference. Especially in copper cabling. Fiber optic cable is preferred since it can not be affected by RF.

How frequently do bit errors occur?

If you look at bit error rate in terms of media, some electronic systems experience more errors than others. For instance, optical disk has a higher error rate than magnetic disk. Magnetic tape has a higher error rate than magnetic disk. Fiber optic communications cable and semiconductor memory have a low error rate.

One way to measure bit error rate is in terms of the number of bit errors divided by the total bits transferred. Using magnetic disk as an example, if you were to look at bits coming right off a magnetic disk you would see, on the average, about 1 bit error for every billion bits transferred. If you were to look at bits coming right off an optical disk, you would see a much higher rate of errors, about 1 bit error in every 100,000 bits transferred. Without some form of error detection and correction, most storage devices would be too unreliable to be useful.

Another way to measure bit error rate is in terms of how many bit errors occur in a unit of time. Again, using magnetic disk as an example, if you transfer a million bits per second, on the average, you'd have a bit error every thousand seconds, or every 16.6 minutes. If you transfer a billion bits per second, on the average, you'd have a bit error every second. Currently, some disk drives transfer at the rate of 40 million bits per second, so that a bit error occurs every 25 seconds.

As a general rule, in all electronic devices, transfer rates are increasing as time goes on. Also, errors will occur more frequently in any kind of storage device because manufacturers are squeezing more bits into smaller and smaller spaces. As speed and density increase, error correction becomes more of a necessity.

Data Integrity Measure

As manufacturers squeeze more bits into smaller spaces, measuring the error rate and correction becomes very critical. If a medium has a lower error rate (tape), the digital stream can be broken into larger blocks (frames). High error rates (optical) call for shorter blocks.

There are three ways that errors are measured in digital systems:

1. BER (Bit Error Rate), is one method of testing the integrity of data. BER is the number of bits divided by the total number of bits received. The problem with BER is that it counts large and small burst errors equally and not the distribution of the errors.

2. BLER (Block Error Rate) measures the number of blocks of data per second that have one occurrence of un corrected data.

3. BERL ( Burst Error Length) , counts the number of consecutive blocks in error.

(BLER and BERL determine the rate of error per second. BER determines the total count.).

Optical media has a high error rate of 1 bit out of every 100,000.

Magnetic disk has a lower error rate of 1 bit out of every billion.