To tell the truth, this subject is peripheral to CD recording and I've been reluctant to add it. But questions come up so often that it seems to be necessary - and for many, the performance of the sound card in the PC is essential to getting good results on CD-R.
First, I shall neglect some of the factors important to users of computer audio for other purposes. The primary interest for CD-R is the ability of the sound circuitry to encode PCM - Pulse Code Modulation - from analogue. There is a special case in which the sound card serves a purely digital function - S/PDIF; it will be dealt with at the end. In addition, the focus here is on CD-DA, Compact Disc - Digital Audio.
In PCM, a collection of computer bytes corresponds to a single, specific point on an analogue waveform. The frequency with which those samples is taken is the sample rate, usually designated in samples per second but written (imprecisely) as KHz (KiloHertz). Of course, one issue is whether there are one or two channels of audio involved - monaural or stereo sound. Another is how many bits are used for each sample collected; the usual values are 8 or 16. If the encoding is set up to collect 8,192 samples per second (8 KHz) from a monaural signal and to encode at 8 bits per sample, the resulting PCM stream takes 8 K bytes per second (bps). If samples are collected to be recorded in redbook format at 44.1 KHz, stereo and 16 bits, the PCM stream takes 176.4 Kbps. That rate is labelled 1x for audio. (Because of error correction, the corresponding data rate is 150 Kbps.)
One key function of the sound card is to convert the analogue input signal it receives into digital data. The quality of the result depends essentially on two factors: the noise floor and the accuracy of ADC. Noise floor means the noise introduced by the card itself and the leads which provide input when there is no signal. It determines a level of hiss due to the card as opposed to what may be received from a tape or other input. For comparison, a very good cassette tape can deliver about 60 db SNR - Signal to Noise Ratio. That means that its noise floor is 60 db below its maximum signal. Because of quantization (the resolution capability of a digital signal) the SNR for 16 bits such as from a CD is about 96 db. A sound card which puts its quality into gaming and MIDI may deliver something like 50 db - in other words, it may provide more noise than a good, Dolby B cassette. A less expensive sound card which focusses on WAV quality may deliver 75 db. Where those cards typically cost less than $100, for $200-500 or so, one can buy still higher quality, up to about 82 db. Beyond that, the experts use outboard converters to avoid the electrical noise inside the computer's case.
The input to a sound card is typically an electrical analogue of a sound wave. The card's job is to convert that signal to a digital one of the proper number of bits and at the proper rate. The conversion process is more or less accurate depending on how well the card is designed and built. Not surprisingly, the cards with low noise floor also have high-quality ADC. High-end cards may encode to 20, 24 or more bits. That allows substantial editing without losing quality before the bit depth is reduced to 16 for writing to CD-R. In general, mastering s/w will create redbook audio from 8- or 16-bit samples, but not from others.
Similar to but worse than bit depth, sample rate is under your control and must be chosen with care. The highest frequency which can be captured in digital form is half the sample rate: 44.1 K samples per second cannot record more than 22.05 KHz audio. However, there are significant effects at lower frequencies, so it is desirable to work at a higher sampling rate when practical. Most DAT recording is done at 48 KHz; 16 and 32 KHz are common choices for lower-fidelity sources; professional mixing is often done at 96 KHz. However, those rates are not easy to convert to 44.1 for CD-R, so they will require processing outside the mastering software before they can be burned to CD-DA. Like adjusting for non-standard bit depth, that is a job for a WAV editor. Because the file size and the difficulty of editing increase with sample rate, processing low-fi sources at 22.05 KHz monaural is much more efficient than the same process in redbook - 44.1 KHz stereo. Fortunately, most mastering s/w will convert that to redbook without bothering you (though not on the fly).
Rather than repeat the information elsewhere in this primer, I will simply summarize it. A PCM file may be raw or wrapped in an envelope with a specific format name. Raw files typically have extensions PCM, RAW or SND. A raw file carries no information except the stream of bytes. If you put data on that stream to specify details such as sample rate, bit depth and mono/stereo, then you must use an envelope or wrapper such as WAV, AU or AIFF. That envelope permits other information (e.g., copyright) but does not require it. Considering only the WAV file (it's all I know), the wrapper can also hold compressed files, but those are not PCM. ADPCM and MPEG Layer 3 are typical compressed formats.
For convenience and for mixing, the sound card will accept several different inputs. Typically, one comes from CD-ROM, one from Line In (an external connector) and one from Microphone (usually monaural). There may be additional CD-ROM connectors on the card, but unless there are separate sliders for them on the mixer, they probably are all shorted together and simply provide different types of plugs. You should not connect two signal sources to a single input of your sound card. It probably won't sound good, but it definitely will not load the sources correctly. If you want to connect both your writer and your reader to a single card, you can find a card with distinct inputs (separate sliders), install a switch for the two inputs or use a cable from an analogue output such as headphone to Line In.
In order to hear a signal from your sound card, you must enable it on your mixer. However, what you hear is not necessarily what you will record. In order to record that signal, you must enable it (and set the record level) on that same mixer. On the standard Windows mixer, you go to Options, Properties, Recording in order to reach the controls you need. The next time you boot, your settings are lost and you must go there again! Other mixers make the controls more accessible and sticky, but they are not usually available except with their corresponding sound cards. An Ensoniq mixer is available only with a Creative Ensoniq card, for example.
S/PDIF designates a direct digital connection between a device such as a MiniDisc player or a CD-ROM and your computer. It is usually implemented in a high-end sound card. Note that not all outputs of CD-ROM drives which say digital or look digital will mate correctly with an S/PDIF card. Like DAE, S/PDIF avoids DAC and ADC altogether and allows you to remain in the digital domain. In addition, many sound cards implementing S/PDIF also convert from 48 to 44.1 K samples per second on command. Another advantage is that an S/PDIF input allows you to connect that second CD-ROM safely. But by now, we have gone beyond the scope of a primer. (And anyone who suggests that that means it has exhausted everything I know on the subject has been cheating!)
As usual, I make no comments on equipment I do not know. I have replaced three sound cards since getting into CD-R: a Creative AWE 32, an AWE 64 and a Yamaha. (The last meant replacing the motherboard on which it was located.) In all cases, I found both the noise floor and the quality of ADC insufficient; the Yamaha also proved almost impossible to control with the standard mixer. My current systems use an Ensoniq PCIAudio (no longer available) and its first cousin, the Creative Ensoniq PCIAudio. The cards and the mixers are different (and not interchangable), but both deliver high quality at low cost. I have not tried the Creative Live, a Turtle Beach, Card D, or any of the dozens or hundreds of alternatives.
E-mail me at cdrecording@mrichter.com
Return to Mike's home page