3 Digital Audio

3 Digital Audio

• 1.
  Digital Audio Introduction In thelate 1960’s Dr. Thomas Stockham began to experiment with digital tape recording using analogue to digital converters. By 1975 most professional audio studios began to start using digital tape recording. [1] [2] Before digital audio was created, all recording, editing and storing music was done from analogue audio. This experiment will help me discover why digital has taken over from analogue. By using Adobe Audition I can experiment with digital audio in more depth and get an understanding about aspects of digital audio such as its file size and sound quality. It was important for me to research to see how digital audio is created from an analogue signal, as well as how digital audio can be converted back to an analogue signal. Experiment Analogue to Digital Conversion (ADC) An analogue to digital converter is a device which can convert an analogue input into digital form. This is done by converting the analogue voltage into a digital number known as binary. [3] The waveform below is an example of an analogue signal. Analogue signals are continuously variable unlike a digital signal. http://micromachine.stanford.edu/~hopcroft/Research/resonator_images/sin_mov1.gif By using Adobe Audition I can create a waveform similar to the image above. To convert the analogue wave to a digital wave, it needs to have sample points taken. These are done by measuring the amplitude of the waveform at regular intervals every second. The number of samples taken each second is known as the sample rate. Sample Rate The diagram below shows a waveform I created in audition.
• 2.
  The dots onthe waveform are the samples being taken. The wave I created had the sample rate of 44100Hz which means 44100 samples are taken each second. This sample rate is that of a standard CD. The reason for this is that the sample rate of a CD has to be larger than about 40 kHz. This is because it needs to be double the maximum analogue frequency which is around 20 kHz. This theory was found out by Harry Nyquist and is therefore known as the Nyquist sampling theorem [4] By halving the sample rate of my waveform in Audition I can see this halves the number of sample points taken. To see what this does to the sound quality, I opened up a sample in Audition and changed the sample rate each time and wrote what happened in the table below. Sample Rate (Hz) Effect on test sample 48000 Sounds the same 44100 Original sample rate, so no change 22050 The sample sounds more muffled 11025 Very muffled, losing high frequencies 6000 Very poor sound, no high frequencies at all This confirmed my assumption that the lower the sample rate the poorer the sound quality. Changing the sample rate to as low as 6000Hz made a huge difference as it sounded like there were no high frequencies at all. An interesting discovery was when increasing the sample rate the quality stays the same. This is because there’s no longer the original analogue wave to sample from and increasing the number of the samples on the digital wave just keeps the wave the same. Bit Depth The values that are taken from the sample rate are stored as binary numbers. Binary is a code read by computers and is made up of 0’s and 1’s. [5] More numbers can be created by having more than one bit. Having 2 bit’s will allow 4 possible outcomes 00, 01, 10, 11. The number of bits determine bit depth, the larger the number of bits the bigger the bit depth. Bit depth rounds each sample point taken to a binary number. Therefore a higher bit depth produces a more accurate wave. The following diagram shows the bit depth of a waveform. You can see that the bit depth is 2 bit as there are 4 possible values. Each sample taken has to be rounded to the nearest binary number. This digital waveform is a very bad conversion. The bit depth needs to be much higher to get a more accurate representation of the analogue wave. http://www.pgmusic.com/images/wave6.gif The results of this poor conversion would be quantization noise. “Due to this rounding off, the raw value mis-states the actual signal by a slight amount; the error introduced by the digitization is called quantization error, and is sometimes referred to as quantization noise” [6]
• 3.
  I did anexperiment to find the difference in sound quality by lowering the bit depth and to see if I could hear quantization noise for myself. I took the same piece of audio I used for the sample rate experiment and change its bit depth. Bit Depth Effect on test sample 32 No difference 16 Original bit depth, so no change 8 Fuzzy 4 Fuzzy – Lots of hissing noise As in the previous experiment increasing the variable (in this case bit depth) had no effect on the sound quality; this is because the wave itself is not being changed. Lowering the bit depth had a much more noticeable adverse effect on the sound quality than changing the sample rate did. A big hissing noise drowned out the audio sample, and as I found out from my research this was quantization noise. One thing that can help cover quantization noise is dithering. This is done by adding random noise to the digital signal. [7] File Size When saving a file it is obvious that to get the best sound quality you need a high sample rate and a high bit depth but saving this would result in a large file size. To see how sample rate and bit depth effect file size I saved an audio file in Adobe Audition and each time saved it at a different bit depth or sample rate. Below is my results table. Sample Rate (Hz) Bit Depth File Size 44100 16 1.24MB 44100 8 636KB 44100 32 2.48MB 22050 16 636KB 11025 16 318KB 6000 8 86.5KB From this experiment I found that halving the sample rate halves the file size. This is also the case when halving the bit depth. Choosing which sample rate and bit depth depends on what I would want to do with the file. When saving an audio mix down it would be best to save the highest quality available but if I had a quick mix down I wanted to attach in an email or send over the internet it would be best to save it as a smaller file. Digital to Analogue Conversion (DAC) Getting digital audio back to analogue works in the opposite way converting analogue to digital did. A digital to analogue converter takes digital samples and outputs them as analogue. The nyquist theorem states that as long as the sampling rate is more than double the highest frequency in the original audio sample then the output will be an exact of the original signal [4] Conclusion This experiment has helped me gain a better understanding of why digital audio is now more commonly used the analogue. I found how an analogue signal is converted to digital audio and how sample rate and bit depth affects the accuracy. What I learnt was very useful if I wish to work as a studio engineer or producer as it taught me about storing data. Larger files have a higher sound quality but take up more memory, so it's a compromise I would have to bare in mind in the future. The experiment has also given me an idea of what sample rate and bit depth to save my audio as and which sound quality is appropriate for different uses.
• 4.
  References 1. About.com, The History of Digital Music [Online] Available at: http://mp3.about.com/gi/dynamic/offsite.htm? zi=1/XJ/Ya&sdn=mp3&zu=http://www.aes.org/aeshc/docs/audio.history.timeline.html [Accessed 1 December 2008]. 2. University of San Diego, Thomas Stockham and Digital Audio Recording [Online] Available at: http://history.sandiego.edu/gen/recording/stockham.html [Accessed 1 December 2008]. 3. Hardware secrets, How Analog-to-Digital Converter (ADC) Works [Online] Available at: http://www.hardwaresecrets.com/article/317 [Accessed 1 December 2008]. 4. Columbia University, Explanation of 44.1 kHz CD sampling rate Available at: http://www.cs.columbia.edu/~hgs/audio/44.1.html [Accessed 1 December 2008]. 5. CIO-Midmarket, What is binary? Available at: http://searchcio-midmarket.techtarget.com/sDefinition/0,,sid183_gci211661,00.html [Accessed 1 December 2008]. 6. University of Chicago, Noise, Dynamic Range and Bit Depth in Digital SLRs Available at: http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/#quanterror [Accessed 1 December 2008]. 7. Ear Level, Dither Available at: http://www.earlevel.com/Digital%20Audio/Dither.html [Accessed 1 December 2008].