Introduction to digital audio

Introduction to Digital Audio

Recording in the digital realm came about due to huge advances in technology and research and as a way of finding a more efficient medium. Magnetic tape and vinyl discs were prone to many types of errors such as hiss, wow and flutter, drop-outs and scratches etc. Transmissions over the airwaves also suffered from interference.

Digital recording transfers the data from acoustic sound waves into a more stable form, which is less affected by system errors. Binary code is used as the carrier of information. This is similar to Morse code in a way, but it uses 1's and 0's. More on this later.

There are three stages involved in digital recording:

Sampling
Quantizing
Coding
Before we look at each separately I will just mention now something that many of you may already know. Most consumer products that use digital audio operate at 16 bit 44.1 kHz. Keep this figure in mind as we go along and you will soon see what this means.

Sampling is done by an analogue to digital converter or an ADC. This examines minutely the sound wave and it can be thought of like taking a photograph at a moment in time and recording the data from it. You will have all seen the flip books with a picture that changes slightly on each page. The more pages you have the more realistic the moving image becomes. It's similar to video and TV. These run at 24 frames per second - so 24 different pictures a second gives us our life- like viewing without jumps and stutters.

In digital audio, however, 44 100 (forty four thousand one hundred) photographs are taken each second and the data from these is recorded in binary form. This is the 44.1 kHz mentioned earlier. It is called the sampling rate. Other rates can be used but this was set out to be the consumer norm and so far it has stayed that way.

Modern recording systems often now sample at much higher rates 96 kHz or 192 kHz.

This basically allows a bit more leeway when manipulating a sound with digital processors, and gives a better signal to noise ratio. At the end of processing it still needs to be re-sampled back to 44.1 to go on a commercial cd.

Quantizing - the word comes from physics and means ‘a definitive and fixed quantity.' If you've used it whilst recording midi you will know that it shifts notes to the nearest user set level e.g. quarter notes or eighth notes. It works in a similar way in digital audio and it is how the wave form as each of the sampling photographs are measured. The number of levels can vary, but essentially, quantizing records the data at the nearest level to where the wave form is. If you used a ruler with levels at centimetre intervals and the wave form was measured at 8.23cm it would be recorded as 8cm. obviously the more levels the more accurate the reading. With the levels set at cm there are only 10. If we set the levels to millimetres there would be 100 and our waveform would have been recorded as 8.2cm - much closer to its actual level. The more levels, the more data is needed to record something and the bigger the file size. The number of levels we have in a typical consumer cd is 65 536 and we arrive at this from having 16bits of binary data.

Coding
Binary code - This is a way of storing numbers using 0's or 1's. It works using columns and each column represents a number which is double the previous column. The first column represents 1's the second 2's then 4's then 8's and so on. Let's first take a look at our regular number system known as base 10. Taking the number 284 we know that the 4 represents "ones" and we have 4 of them, the 8 represents "tens" and we have 8 of them, and the 2 represents "hundreds" and we have two of them. When you understand what the columns mean you can work out the number. In base 10 we go from 1-9 before we move and add one to the next column. In binary code we have either 0 or 1 in each column. A 2-bit system would consist of two columns and the number possibilities are as follows:

Columns

2nd 1st number represented

0 0 = 0

0 1 = 1

1 0 = 2

1 1 = 3


So in a 2- bit system you can represent only 4 values. If you were quantizing to these levels you would not get a very accurate picture of a waveform. 16- bit audio therefore has 16 columns and the total numbers represented is 65 536.

Here is an example of a 16 bit binary number:

0 0 0 0 0 0 0 1 1 1 1 1 0 1 0 0 this represents the number 500



So why use binary when you can just say 500?

Because it's computers or circuits doing the math not humans. Using binary means numbers can be represented a lot quicker and by different systems. With just 2 options numbers can be represented as on or off or by o volts and +5 volts. And by only needing to count up to 1 large numbers can be calculated much faster. For the world of digital recordings, specifically cds, these are made using a laser. This laser is highly accurate and etches on to a disk. The laser beam is switched either on or off dependant on the code being written. That's the basics.

www.song-production.co.uk