Debunking the two most common myths about digital sound.

This is a good time to be a digital video user. Now that Hollywood studios are embracing the technology, the idea that it's less worthy than old-fashioned film is disappearing. Digital audio types aren't as lucky: At least once a week, someone pops up in an Internet newsgroup or audiophile magazine to claim that because real-world sounds are analog, digital systems simply can't record them as well as analog ones.

I'm not saying that a particular model of analog recorder won't sound better than some specific digital one. But when these nontechnical types say that digital recording is inherently flawed and analog isn't--well, that just riles the engineer in me. So I'd like to take apart the two most common myths about digital sound and reveal what's really going on. It doesn't take much math or science to understand; and once you do, you'll be able to specify and use your own equipment better. Besides, maybe you'll meet one of these analog aficionados some day and you can set them straight.

The Myth of Steps One claim is: Because digital breaks signals down into discrete steps, voltages that fall between the steps won't be captured properly. Analog recording is continuous, so it can reproduce any voltage accurately: This is half true--digital recording has steppiness (called quantization error). But analog has an equivalent problem: the electronic noise of the recording medium. Both vary with system quality. Once you level the playing field--compare a first-class digital recorder with an equivalent analog mastering deck--both recording methods are equally inaccurate. You can prove this if you check with precise test equipment. But you can understand the principle using only your eyes and ears if you lower the analog and digital quality by the same amount.

Figure 1--A cheap digital signal--four bits' worth, to be precise. The steps are tilted because of the analog circuits feeding the oscilloscope.

Figure 1 shows the digital side of the experiment. I created a low-quality 100Hz 4-bit sinewave and took a photo of it in an oscilloscope. You can see how it's quantized to 16 steps, the resolution of a 4-bit system. If you want to hear the quantization error (reconverted to 16 bits for playback), go to the Reader's Corner at www.dv.com and download 4BIT.WAV.

Analog recording doesn't have quan- tization, but it has random tape noise. To compare it with our low-quality 4-bit digital recording, we'll need an equivalent amount of noise. Each digital audio bit is worth 6dB (decibels) of dynamic range-- 16-bit CDs have a 96dB range between the loudest and softest signal, because 16x6=96. So the analog noise equivalent to 4-bit digital would be 24dB below the sinewave's level. You can hear them mixed together in the NOISE24.WAV file at www.dv.com. It's a digital file at 16 bits so the additional quantization error is insignificant.

Figure 2--Digital and analog waveforms have similar problems when you examine them closely.

Compare the two files. Figure 2 has digital on the left and analog on the right. The analog noise makes the signal jump up and down so much that it's impossible to tell where the original sinewave was. In fact, it jumps exactly as much as the quantization errors in the digital signal. If you listen to both files, you'll hear the sinewave is equally obscured in each.

What's true at four bits or -24dB noise is also true at 16 bits and -96dB noise. The big difference is that even -80dB noise is hard to achieve in analog recording, while digital recorders with more than 120dB dynamic range are used every day. Your system probably uses 16 bits. Here's how to make the most of them:

• Record at the highest possible level that doesn't cause distortion. There are tips for calibrating recording levels in the June '98 Audio Solutions on www.dv.com.

• If you can't record at an optimum level, use a normalize function before equalizing or mixing to avoid subsequent damage to the sound. But it's better to record it right in the first place.

• Keep every element at its full level until the final mix. If you process a sound file to lower its level, you're throwing away bits.

The Myth of Highs Digital recording needs a fraction more than two samples for the highest frequency wave it can carry. That's why CDs work at 44.1kHz sampling--to reproduce a 20kHz bandwidth. This led some audiophiles to assert, CD-quality digital doesn't have enough data points to reproduce a complex signal--or even most squarewaves--at 20kHz. But a 20Hz-to-20kHz analog system can do it with no trouble.

The reality is that no system can reproduce any complex wave at its upper frequency limit. What makes the wave complex is its harmonics, which have much higher frequencies than the base signal. If an analog or digital recording system cuts off at 20kHz, these harmonics are eliminated. All that's left is a sinewave.

Figure 3--If you squeeze a 1kHz squarewave through a 1.5kHz filter, it stops being square.

I could demonstrate this at 20kHz, but only your pet bat would hear the result. So we'll look at the example of a squarewave at 1kHz in a system whose upper frequency limit is 1.5kHz. Figure 3 shows that squarewave on the left (1K_SQR.WAV at www.dv.com). To simulate the frequency limit, I applied a filter at 1.5kHz. The right side shows the result--the curve you see is our original squarewave, after the filter. Try it for yourself. In fact, try it with a complex waveform such as that of a trumpet or violin, playing B5 (about 1kHz) and a filter as high as 2kHz. If you've got an oscilloscope and good analog equipment, you can try the experiment at 20kHz. It's physics, and has nothing to do with specific frequencies or whether the system is analog or digital.

The myth probably originated because early digital systems had sloppy internal filters, which affected sounds below their cutoff points. Modern professional gear uses oversampling at many times the nominal rate to avoid this problem. Most editing systems aren't that sophisticated, but a few tricks can help you get the best high-frequency results:

• Bypass your soundcard's analog input filters by using an external DAT recorder or digital converter and an S/PDIF interface, or FireWire from your video deck. Or have a professional studio do the digitizing, and give you the files on CD-ROM or removable storage.

• When grabbing sound from a CD, use a direct transfer from your audio editor's CD-ROM drive instead of recording through the analog inputs. You can do this with QuickTime on the Mac, or with a shareware ripper in Windows.

• If you're releasing multimedia at 22.050kHz sample rate, record it at 44.1kHz to push the filter as high as possible. Then convert the file in software.

Digital audio isn't perfect. It's subject to manufacturing compromises as well as its own physical limitations, and can be difficult to set up properly. But exactly the same thing can be said about analog. To pretend otherwise is to myth the point.

Jay Rose (jrose@dv.com) spends most of his time creating tracks for network projects at the Digital Playroom (www.dplay.com).