Digital is Obsolete!
by Fenris Wulf, KDVS Studio Tech
In the course of recording music for 15 years, I've used a lot of digital technology that didn't work as advertised, or didn't work at all. It cost me a lot of time, money, and frustration. When I moved from digital recording to analog recording, it was a revelation. This article is a summary of everything I wish I had known 15 years ago.
Various companies make software instruments that emulate synthesizers, organs, pianos, or other instruments. They don't have the same texture as the real thing, and they sound even less real in a full mix. Save your money and buy an old analog synth, or an old electric piano, or plug an old organ into a DIY rotating speaker. You’ll be amazed at the difference.
There are software samplers that can play back massive multi-sampled orchestral instruments that are streamed from the hard drive and stored in the computer’s RAM. They can sound "real" in a bland way, if you spend hours programming them. But they can't duplicate the expressiveness of a real musician on a real instrument. You're not actually playing an instrument; you're triggering a recording of someone else playing a single note, divorced from any musical context. The emotional feedback between player and instrument is missing.
The best way to use a sampler is for non-realistic sounds, and the best kind of sampler is an old one that isn't too hi-fi, like the 8-bit Ensoniq Mirage that was used on the classic Public Enemy records. A Mellotron tape-based sampler is even better.
A keyboard controller uses MIDI to control a software instrument or a sound module. Some have built-in sounds, some don't.
Some of the mid-priced models cut costs by using a limited number of velocity steps, 20 or 30 instead of 127. The velocity response is jerky and unpredictable; a small variation in velocity produces a large variation in volume. These controllers are very frustrating to play.
Even the best controllers cannot reproduce the smooth, intuitive response of a real piano. A good pianist can make a real piano "speak" in a way that no digital piano can duplicate. A student who learns on a digital piano will learn to play the right notes, but he won't learn to play with emotion.
For less than the cost of a fully-weighted digital piano, you can buy a used upright piano in good condition. (A full-size upright is preferable to a spinet.) It will have to be tuned and regulated by a qualified technician. If you want it to be in tune with other instruments, use standard tuning and not stretch tuning. For recording, you can remove the front panel for a brighter sound. An upright has a more resonant sound and lighter action than a grand, and is preferable for many styles of music.
Another type of MIDI controller is an electronic drum set. Unfortunately, the designers have never solved the basic problem of accurately reading the velocity of the drumstick as is strikes the head. They use "sensors" underneath the heads which are nothing more than cheap contact microphones, and a "brain" that translates the amplitude of the sound into MIDI velocity. It doesn't work very well, because it senses the effect instead of the cause. The head has "hot spots," the velocity response is jerky and unpredictable, and it maxes out around mezzo-forte and doesn't get any louder. A student should never practice on an electronic drum set; it will give him bad habits and rob his playing of any nuance or feel.
Electronic drums are not silent and you won't be able to play them in your apartment, not even a ground-floor apartment with concrete floors. For the price of a full-size electronic set, you can buy a very nice acoustic set and have money left over to soundproof your garage or rent a practice space.
A MIDI sequencer records the MIDI data, allows you to edit the notes and expression, and plays it back. Most DAW (Digital Audio Workstation) programs include MIDI sequencing.
DAW’s have perfect MIDI timing -- as long as you stay inside the DAW and use software instruments. As soon as you go outside the computer and connect the DAW to a controller or sound module, the MIDI timing goes to hell.
Every DAW in existence, PC or Mac, has a certain amount of "jitter" or random timing error. The jitter ranges from 5 milliseconds, which is almost inaudible, to 20 milliseconds, which is enough to destroy the feel of the music. Worst of all, it’s completely unpredictable; the amount of jitter might change every time you open the program. If you use an external controller and an external sound module, the total amount of jitter is doubled. Some MIDI interfaces have "time-stamping" that is supposed to solve the problem; it doesn't.
It's unbelievable, but it's true. Modern computers, with their multi-gigahertz processors, can't even deliver accurate MIDI timing.
If you want jitter under 2 milliseconds, you have to find an Atari ST, a computer made in the 1980's, with a compatible MIDI interface and timecode reader. The MIDI is built into the motherboard and it's not fighting with other processes for CPU cycles. Quite a few serious MIDI people are still using these computers.
A British company called Expert Sleepers makes plug-ins that convert MIDI to control voltages, which are sent to the audio outputs. By using a CV-to-MIDI converter, you can obtain sample-accurate MIDI timing.
DIGITAL AUDIO WORKSTATIONS
DAW's are advertised as a "studio in a box" that replaces a roomful of equipment. Almost everyone is using them, from bedroom hobbyists to multi-million-dollar studios.
EQ, compression, and effects are provided by "plug-ins." The most popular plug-ins are emulations of vintage analog gear. In A/B comparisons with real analog gear, including cheap analog gear, the analog wins every time. It has more texture and complexity, it has a wider range of useful settings, and it requires less tweaking to sound good. (To do this test properly, you have to use an analog console and listen in the context of a full mix. Looping the outboard back into the DAW degrades the sound and makes the difference less audible.)
Recording and mixing in a DAW leads to an unsatisfying sound. The engineer can choose from hundreds of different plug-ins, but none of them sound very good. He ends up over-processing the sound in a futile attempt to make it more "analog." By providing too many options and encouraging musicians to rely on editing instead of playing it right, a DAW makes the recording process slower instead of faster.
Serious DAW users are reluctant to abandon their system, after spending thousands of dollars on it. So they spend more money to improve the sound and workflow. First they buy the control surface with motorized faders, then the latest digital converters, then the analog summing box to "warm up" the sound, then the rack of vintage preamps and outboard gear, then the special interface that integrates the DAW with an analog tape machine. They can easily spend ten times the price of a good analog console that has faders, summing, preamps, EQ's, and tape monitoring built in.
Most DAW users spend endless hours dealing with computer problems. This is because computers have five basic flaws:
1. Overcomplexity. Most software is designed by dozens of people, no one person understands the whole, new features are constantly added to stay ahead of competitors, and bug fixes are a never-ending process. Operating systems are designed by hundreds of people, they have to support a multitude of devices and programs, and they represent a compromise between many conflicting requirements.
2. Lack of modularity. Your entire studio, comprising dozens of hardware devices and software programs, runs on a single machine. If one component malfunctions or has a conflict with another component, the system doesn't work at all, or the sound has random glitches. Tracking down the problem can take hours, and you might have to re-install everything from scratch. In some cases, the only solution is to buy a whole new system and pray that it works.
3. Rapid obsolescence. Standards and protocols constantly change. New software doesn't work with old hardware and vice versa. Software A can't exchange files with Software B or even a different version of Software A. Computer hardware is not designed to be repaired, and when it breaks down you have to throw it away. Hardware is frequently "orphaned" when the manufacturer stops updating the drivers. If you update your operating system or your DAW, some of your plug-ins will stop working and you'll have to pay for an update. Depending on your system, you might have to re-purchase all your plug-ins every few years.
4. Malware. A never-ending barrage of malware forces users to install anti-virus programs that work in a similar way to an airport checkpoint and slow the system to a crawl. Rampant piracy forces software companies to use copy protection that behaves like malware and crashes the computer.
5. Inefficiency. A general-purpose CPU is the wrong kind of processor for audio. It's not designed for real-time processing, it can't multi-task properly, and it wastes about 80% of the processing cycles. Certain plug-ins or combinations of plug-ins cause CPU spikes: the audio becomes choppy and you have to restart the program. Some plug-ins run on proprietary DSP cards: these cards put a heavy load on the PCI bus and cause CPU spikes on about 20% of all systems.
In addition to CPU spikes, DAW's suffer from drop-outs, clicks, crackles, random bursts of full-scale white noise, plug-ins that suddenly output full-scale DC, corrupted audio files, corrupted session files, and random "poofs" where the DAW suddenly closes without saving the session. Your DAW might work perfectly (until you install some new piece of software and upset the delicate balance of the computer), or it might be totally unusable. There is no way to predict how a given configuration will behave, because there are too many variables.
The "industry standard" DAW runs the entire system on proprietary DSP cards; but most users max out the cards and supplement them with "native" plug-ins that run on the CPU. Despite its much higher price, this system is no more reliable than native DAW's.
Another problem with DAW's is latency, the slight delay as the computer processes incoming data. Controller latency makes the faders feel slow or mushy, audio latency affects what the musicians hear in the headphones, and plug-in latency introduces random amounts of delay and changes the sound of your mix (most DAW's have plug-in delay compensation).
Analog equipment has none of these problems. You can connect anything to anything else; it's all compatible. The basic standards and electronic principles were established before WWII. You can use a microphone from 1930 with a preamp from 2010. A console is usually designed by one person, it consists of a few simple circuits repeated many times, and bug fixes are rare. It doesn't get viruses and it doesn't need copy protection. Professional tape machines and consoles are modular; if one channel develops a problem, you can quickly swap in a spare card. You can obtain a schematic for almost any piece of equipment, and you don't need a college degree to understand it. If something doesn't work right, you can modify it. If something breaks, you can fix it.
DAW’s are good for one thing: editing bands who can’t play. That job can be done by the cheapest software running on an old computer. All the other functions can be done better, faster, and cheaper by analog gear.
LIES, DAMNED LIES, AND SPECIFICATIONS
The digital mavens point out that digital has extremely low noise and distortion and ruler-flat frequency response, and therefore it must sound better. But it's not that simple.
Simple measurements of noise, distortion, frequency response, and timebase errors (wow, flutter, and jitter) are misleading and can make cheaply built equipment look superior to quality equipment. Phase response is more important than frequency response. Distortion spectrum is more important than percentage. Correlated noise is worse than uncorrelated. High-frequency timebase errors are worse than low-frequency.
All digital converters have some level of aliasing, quantization noise, ringing in conventional filters, pre-ringing in FIR filters, phase shift, and high-frequency jitter. These artifacts are extremely unpleasant even at low levels, and designers strive to reduce them below the level of audibility. A digital converter can sound transparent at first; but downstream compression and EQ boosts these artifacts by as much as 20 dB.
Analog artifacts are more euphonic. Tubes, transformers, and tape saturate gradually and handle transients in a graceful way, instead of clipping like a transistor or a digital converter. This provides up to 20 dB more headroom and better signal-to-noise than "specs" would indicate. The engineer can vary the amount of saturation to change the dynamic and harmonic characteristics of each instrument; engineers have been "abusing" the equipment in this way since the 1960's. Getting the right sound in tracking is faster and more effective than trying to fix it later with plug-ins.
Tape saturation is the "softest" and resembles the behavior of the human ear at high decibel levels. A digital recording is closer to the sound coming from the mic preamps -- but an analog recording is closer to the sound in the room!
Even tape flutter is less objectionable than digital jitter. Flutter is seldom audible; most instruments have an inherent vibrato that masks it.
Tape hiss first became a problem in the 1970's, when studios started using tape machines with more and narrower tracks. When digital recording was introduced in the late '70's, it was welcomed by classical engineers and some rock engineers for its low noise. But with modern high-output tape formulations, noise is no longer an issue. The biggest source of noise is the room itself.
WHAT IS FIDELITY?
The most primitive link in the recording chain is the loudspeakers. The last major innovation was the introduction of electrostatic speakers in the 1950's. Every loudspeaker has resonances, phase and frequency anomalies, and distortion at high and low frequencies that would be totally unacceptable in any other piece of equipment. These problems are exacerbated by room acoustics, even in the most well-designed room. Not even the most expensive loudspeakers can reproduce the three-dimensional soundfield of an orchestra in a concert hall, or the sound of a drum set or a Marshall stack in person. They are limited by the laws of physics.
To overcome the deficiencies of loudspeakers, audio engineers strive for a sort of exaggerated clarity. Studio acoustics are treated to randomize sound reflections and eliminate unwanted resonances. Microphones are placed much closer than listening distance, and recorded to separate tracks. Unwanted or conflicting frequencies are filtered out. Almost every instrument is altered with EQ and compression to sound more present. Artificial reverb is added to create the illusion of size and distance.
Even purist classical engineers use microphone selection and placement to alter the sound, and spot mics to bring up individual instruments or sections. Many classical engineers prefer vintage tube microphones for their natural sound, even though on paper they have far more noise and distortion than modern transistor microphones.
"Fidelity" is almost completely irrelevant to modern recording methods. Pure sounds don't exist in nature; every musical instrument is a sound-altering device, and the recording studio is an extension of that process. An audio engineer's job is to manipulate the sound until he arrives at something that his brain accepts as "real." Analog equipment makes that job much easier.
A lot of recording equipment is aimed at the semi-pro market. But broadcast equipment is built to higher standards, right? Not any more.
KDVS paid $3000 for a digital profanity delay, not realizing that the manufacturer had switched to surface-mount technology. When it stopped working, it couldn't be repaired and replacement boards were no longer available. So we bought a used one on Ebay for $460. We bought a dozen "professional" CD players for $600 apiece; they turned out to be horribly unreliable and there was nothing better on the market. So we switched to a $200 model that was easily replaced. We use vintage broadcast consoles, because the newer models are too flimsy to stand up to daily use. We use a 1980's broadcast compressor that sounds better than the newer digital units (in fact, it sounds so good that I'm planning to buy one for the recording studio), and a 1970's transmitter with a grounded-grid design that is stable and easy to repair. Our turntables are a 30-year-old design that outperforms anything else in sound and reliability (unfortunately, the company recently shut down its entire turntable division because most DJ's have switched to digital technology). We have a very extensive record collection, and the vinyl records don't skip as much as the CD's do. We have no plans to digitize our library, because it would limit the diversity of our playlist.
We spent $4000 on a digital remote unit for sports games; in seven years it was obsolete and we had to buy a new one. Our old analog remote unit works fine, but we can't use it anymore since the phone companies switched from analog lines to digital lines, with low-bitrate encoding that makes everything sound like it's under a blanket.
Our broadcast chain is all-analog, except for the profanity delay and the studio-to-transmitter microwave link. (Unfortunately, telephone companies no longer offer broadcast-quality analog phone lines.) We keep our sound as clean as possible. Commercial stations use heavy compression, limiting, clipping, EQ, and overmodulation to "hype" their sound and make it as loud as possible, oblivious to the fact that it causes listeners to tune out.
Many radio and television stations have gone all-digital: they use digital consoles, digital playback systems, and digital processors. The engineers don't like it, but it's dictated by automation and networking requirements. Most of this equipment will be junk in 10 years, while the analog equipment it replaced will still be going strong. Our recording console came from CNN, we bought it for peanuts, and we'll still be using it in 30 years.
Even the sound of Hollywood movies has suffered since they moved to digital technology. Sound effects captured on a digital recorder instead of analog tape require a lot more processing to sound "real," and digital mixing leads to a cluttered and fatiguing sound.
LIVE SOUND EQUIPMENT
Digital technology still doesn't cut it for live sound. Automatic feedback suppression and speaker protection aren't very effective. Products that incorporate DSP are more likely to break down. Digital consoles lack the sound quality, ergonomics, and reliability of a good analog console, and they have latency that causes phase problems. Super-efficient Class D amplifiers are OK for car audio or lightweight powered speakers, but have inferior sound and reliability (at least the low-end models; I haven't tried the expensive ones). Digital snakes can send 64 channels over a single network cable, but analog snakes have more redundancy and can survive being run over by a forklift.
Live sound engineers work quickly and they need their equipment to be reliable, functional, and bulletproof. Given a choice, they prefer modular analog consoles and analog outboard gear. Many engineers build their own speakers and microphone snakes, because off-the-shelf equipment is either unsuitable or too expensive.
The ubiquity of digital live sound consoles is driven by the mistaken belief that audiences want to hear an exact re-creation of the band's record. A digital console provides elaborate automation and can even run the same plug-ins that were used in the studio. People have forgotten that a live performance is supposed to be spontaneous and unique.
The biggest improvement since the 1970's is in loudspeaker technology. We have high-powered amplifiers, high-output speakers, constant-directivity horns, mathematically calculated speaker arrays, and in-ear monitoring systems that eliminate feedback. If skull-crushing volume is your thing, you can have it. But audiophiles who want to enjoy the music are still using tube power amps and even vintage speakers.
LIVING IN THE PAST
In recent decades, there has been a drastic decline in the quality of all kinds of manufactured goods. They’re designed to be cheap and disposable. They sell for one-third the price, but last one-fifth as long. In the long run they’re not a bargain.
Consumer audio has gotten steadily worse in quality, with MP3's representing an absolute low point, worse than mass-produced cassettes. Practically every boom box, car stereo, and home stereo has horrible-sounding "hyped" speakers. Bass and treble controls have been replaced with digital presets labeled "rock" or "jazz." Some stereos don't even play CD's properly, chopping off the beginning of each track. Record companies are in the business of selling fake music that nobody is willing to pay for. Even good music is heavily compressed in mastering until it's unlistenable.
Computers are no exception. The faster CPU's get, the lazier programmers get. Software becomes so bloated and inefficient that your new computer ends up being slower instead of faster. A 25-year-old computer running DOS can do basic tasks like word processing faster than a modern computer. DOS apps were written in machine code, so they bypassed the operating system and addressed the hardware directly. A modern computer can’t keep up with a fast typist; there is a noticeable lag as the keyboard strokes are processed by the operating system.
That's why I choose to opt out of the system. I ride a Schwinn Continental made in 1968, because the quality of the construction is better than any bicycle made today. It doesn’t pop spokes and it doesn’t get cracks in the frame. I drive a Ford truck of the same vintage, because the last time I bought a new car, it had to be junked after 7 years. I use a computer that I got out of a dumpster, running a 10-year-old operating system and a 15-year-old word processor. I'd rather play games on an Atari 2600 than the latest Xbox. I listen to vinyl records and I will never buy an Ipod.
Appropriately enough, my computer froze up twice while I was writing this article, and I had to copy the screen with a pen and paper so I didn't lose my work. Maybe I should get a typewriter.
In my next article, I'll talk about how to put together an analog recording studio on a small budget, and how to choose equipment that will last for decades.