Knowing where distortion creeps into your signal is the first step toward removing it—and vastly improving your sound
By Paul Stamler
We all want our recordings to sound like the Big Guys’ studios Downtown. Ocean Way. Trident. Muscle Shoals. Hit Factory. We want that big sound, that clean, clear, driving, high energy, smooth, alive, and suave sound that spells professionalism and (we hope) sells records.
But too often we don’t get there. For reasons subtle and unsubtle, our productions sound wrong—thin or tubby; noisy; flat and one-dimensional. Or just plain grubby, for no apparent reason.
There’s a multitude of reasons why this is so, and I could write hundreds of pages about how to do better. (In fact, I have; that’s why this magazine’s here in the first place.) Rather than cover acres of territory, this time I’ll approach the problem from a single angle: keeping the recording chain clean, with particular emphasis on the front end.
Why the front? Come with me for a moment into the spacey world of information and distortion theory, and I’ll tell you.
Links on the chain
One of the truisms of life is that a chain is only as strong as its weakest link. That may be true for bicycles, but it’s rampantly false for anything involving information, such as photography or audio. In an information transmission or storage system, unless all the other links are perfect, a chain is weaker than its weakest link.
There are two ways to look at audio transmission (which includes storage). They correspond to viewing a cup as half full and half empty. Instead of describing how good a signal is, information theory looks at how bad it is; transmission chains are characterized by the amount of degradation they impart to a signal. One maximizes quality by minimizing degradation, thus lowering the barriers between music and listeners.
Taking this dour attitude leads to important practical insights, because it includes the precept that a signal can never get better (in terms of information content and lack of distortion) than it was at the beginning of its journey. To bring this down to a concrete level, once a signal’s been distorted you can’t undistort it, any more than you can unscramble an egg.
In the case of audio distortion, however, there’s more to the story. A perfect audio device is “linear”—that is, if you draw a graph with the X-axis showing input voltage and the Y-axis showing output voltage, the graph will be a straight line and will remain one at all expected levels and frequencies.
Unfortunately no audio device is perfectly linear—no, not even a simple resistor. Audio devices generate distortion in different ways and in varying amounts.
Most of you know that one type of distortion is called harmonic. When a nonlinear circuit (which means, in practice, any real world circuit) is fed a sine wave of a particular frequency, its output includes new, unwanted signals at integer multiples of that frequency. Thus a 400 Hz tone might generate extra frequencies (harmonics) at 800 Hz, 1200 Hz, 1600 Hz, 2000 Hz, and on up the scale. These harmonics correspond to the overtones generated by musical instruments.
Most musical tones come from signals more complex than sine waves. A clarinet, for example, produces its own harmonics at odd multiples (3x, 5x, 7x, etc.) of the fundamental frequency, giving the clarinet its typical tone. And a strummed guitar creates a whole galaxy of fundamentals and overtones as its strings resound. These combinations of fundamental and harmonics create the characteristic sound, or timbre, of each musical instrument.
Harmonic distortion changes the timbre of instruments by adding additional harmonics. This is why, for example, a highly distorted single note on an electric guitar can sound like a saxophone.
Another type of distortion is called intermodulation or IM distortion. This occurs when two different frequencies travel through a single nonlinear device. This time, instead of multiplying, the signals add and subtract, forming new distortion components at the “sum and difference” frequencies.
Confused? Let’s take an example. Say there are two sine waves mixing in an imperfect (= real world) amplifier, and that their frequencies are 8,500 Hz (call this f1) and 10,000 Hz (call this f2). When they combine, they’ll produce a “sum” frequency at f1 + f2, or 18,500 Hz; they’ll also produce a “difference” frequency at f1 – f2, or 1,500 Hz. These are known as the “first order” IM distortion products.
The main thing to notice about these new frequencies is that unlike harmonic distortion products, they don’t form integer multiples of the original frequencies and thus don’t fall into the normal overtone pattern of a musical instrument. This makes them highly audible.
On rare occasions these distortion products are fun to add deliberately; this is done with a gizmo called a ring modulator. But most of the time, they’re unwanted and do a lot of audio damage.
Here’s a second example: given starting frequencies of 18,000 Hz and 18,100 Hz, the sum frequency (f1 + f2) will be 36,100 Hz (which you can’t hear) while the difference frequency (f1 – f2) will be 100 Hz (which you can). You can easily demonstrate this if you have a couple of oscillators handy by playing them simultaneously through a guitar amp; as you tune them to similar high frequencies, a “ghost note” appears in the bass. This high-frequency IM distortion, which creates distortion products at much lower frequencies, turns out to be crucial in making the audio chain clean or dirty. We’ll return to it in a moment.
First, let’s go back to our earlier example. An amplifier with “simple” nonlinearities (such as a gentle curve in its transfer function) might produce only the first-order distortion products we calculated. A device with a more complex curve, however, might generate distortions from multiples of the original frequencies.
So for example you might create 2f1 + f2 (17,000 Hz + 10,000 Hz, or 27,000 Hz—inaudible again) and 2f1 – f2 (17,000 – 10,000 Hz, or 7,000 Hz—definitely audible). Continue combining (2f2 – f1 = 12,500 Hz, 2f2 – 2f1 = 3,000 Hz, etc.) and you’ll wind up with an entire spectrum filled with extraneous frequencies, which will be very audible indeed.
Congratulations, you’ve just invented the fuzz box.
These will be on the exam
There are two vital theoretical points that need to be emphasized, and one practical point.
Theory Point #1: Simple, relatively linear devices produce relatively simple, first-order distortions, while complex devices often produce complex high order distortions involving all those confusing multiples we saw in the last example. In practice, we hear high-order distortion far more easily than first-order.
Theory Point #2: The situation becomes much more complex when one imperfect device follows another. The second device will not only generate distortion from the original signals (which are still, of course, present), but also from the distortion products created by the first device. In other words, it distorts the distortion as well as the signal, leaving you with hash browns.
Let’s sum up for a moment:
- All audio equipment distorts to some extent.
- A signal, once distorted, cannot be undistorted.
- Devices with simple nonlinearities create simple distortions; complex devices create complex distortions, which are more audible.
- Distortion is cumulative; two imperfect devices in succession can produce far more audible distortion than either one by itself.
Getting back to those Downtown studios—to get Downtown quality, you need to do what they do: concentrate on keeping it clean from the get-go.
Oh yes, the practical point: most of the solid-state integrated circuit chips found in low to moderately priced audio gear have complex nonlinearities, with particular vulnerability to high frequency IM distortion. This includes amplifier chips (op amps) in mixing consoles and the input stages of most recorders (including digital multitracks); more importantly, it includes the analog-to-digital and digital-to-analog converters that are integral parts of digital gear.
So how can we keep it clean? Or in gloomy information theory terms, how do we minimize the cumulative degradation in our audio chain?
Since the sources of degradation are multiple, the remedies must be as well. This is a good thing, since it gives us a chance to upgrade one piece at a time, and that’s easier on the budget. The centerpiece of the strategy is to make the front end super-clean, get it into the digital domain as quickly as possible, and avoid messing it up once it’s there.
The first target is the microphone; since that’s the place the audio signal is generated, it’s also the first chance to screw it up. I’ve reviewed a lot of microphones for Recording, and you may have noticed I’m inordinately tough on microphones with a “tizzy” or a “hashy” sound up top.
This sound, which overemphasizes sounds like sibilant “esses” or pick noise on a guitar, often comes from high-frequency intermodulation distortion in the mic. Most condenser microphones have small transistorized amplifiers inside, and these (when badly designed) are classic high-frequency IM generators.
Better mic designs don’t do this; most tubed mics don’t, for example, which is the main reason tubed mics are famous for sounding “warm.” (It’s not that they generate warmth, usually; instead they fail to generate the icy chill up top that characterizes bad solid-state mics.) Modern mics like the LEWITT LCT 940 and Telefunken Copperhead have brought good tubed designs within project studio budgets, for which I sound a loud hurrah.
Clean highs aren’t limited to tubes, though; examples of solid-state mics I’ve tried that lacked tizz included the wonderful (but expensive) Gefell M930 and the wonderful (and cheap!) Oktava MK-012. Good dynamic mics also qualify, of course—my much mentioned Electro-Voice RE15 and beyerdynamic M 260 are utterly unfizzy and very, very clean sounding. This isn’t completely a matter of flat frequency response either; the beyerdynamic M 88 has a rising, bright response but with no hash or spit at the top.
We all know that mic choice is crucial, but our strategy gives us a focus when auditioning: as Dr. Seuss might say, “Beware the Tizz!” A good torture test is a strummed acoustic guitar; the mic that maximizes pick noise and fret buzz is the one that will stimulate distortion all the way down the chain. That’s the one to avoid.
Port of entry
The first place the microphone’s signal lands is at the mic preamp, so that’s the next opportunity to degrade it. (Actually it’s not—the mic cable is. But that’s another story.) Any imperfections added by the mic preamp will be magnified downstream, and no magic box can remove them.
Many of us record through our consoles, which (if you think about it) is usually a silly idea. Let’s face it, no console is perfect, and after the mic preamp section the signal passes through (typically) half a dozen amplifiers before it leaves the mixer and heads for the recorder.
And for what? Most of the time we’re sending one mic to one track, so everything after the mic preamp section (panpots, group submasters, master busses, etc.) is unnecessary. Gang, you can usually get much cleaner results—even on very expensive consoles—if you take the signal from your mic channel insert point straight to the recorder.
And most of us don’t have very expensive consoles, which cost as much as a fancy sports car or a NY condo. Budget consoles use cheap’n’cheerful op amps in their downstream sections, which add that crispy-crunchy sound so typical of too many project studio recordings. So at the very least, make the most of what you have: take the signal straight from the mic preamp and save yourself many levels of distortion.
That’s a good first step, but it still won’t get us Downtown. One of the characteristics of the Downtown sound is a combination of purity and dynamics that can only be described as “powerful sounding.”
I’m here to tell you a secret: the big studios don’t get that sound by adding something to their signal. It comes because of what they don’t add to their signal: the crunch and grunge of cheap preamps. And the way they get that sound is by using mic preamps that amplify without grunge: Neves, Manleys, that sort of thing. We can’t afford that quality level. Or can we?
Most of the time, project studio people record one or two tracks at a time, layering instruments one by one into a well-plotted whole. There are sound reasons not to do this on occasion (you lose the interplay between musicians, for one thing), but willy-nilly, that’s how most of us need to work.
Making a virtue of necessity, we can greatly improve the overall quality of our recordings by purchasing two channels of high-grade non-grungy mic preamp, and using them for most of our tracks. This has become possible in recent years with the introduction of Downtown quality mic preamps at fairly reasonable prices.
At the forefront of these are the Grace Design m101, Great River, and the Millennia Media HV3 and its offspring. Their sonic performance is exemplary. They add no grunge—none at all—which means your signal is less vulnerable to downstream corruption.
Indeed, the un-crunchy sound of these pres takes some getting used to; at first it sounds like you’re lacking highs, until you realize that what’s missing isn’t the treble part of the spectrum, but the high-frequency distortion present in most other mic preamps. (And you can always add highs with the EQ controls, which don’t increase the grunge level.)
In general, good all-tube preamps offer the best sonic performance to my ears, but the solid-state Grace and Millennia boxes are reasonably affordable. I do not mean solid-state preamps with tube overload generators, which are erroneously called “tube preamps”; these add the same level of grunge as any cheap solid-state preamp, then use an overdriven tube to add mud. Pfui.
I love the Peavey VMP-2, a squeaky clean all-tube preamp, and it was my pre to beat for many years, as clean as it was affordable… but it’s not available any more, and the abovementioned brands most certainly are.
How do I listen to a mic preamp? After the usual obvious tests (frequency response, noise level, etc.), a crucial listening test comes when I audition acoustic guitar, kick drum, and cymbals. (Not all at once.)
On acoustic guitar I listen for excessive pick noise and fret buzzes again. These are good warning flags for high frequency intermodulation distortion.
So are cymbals. They should have their natural sheen at the top without sounding hashy or distorted; with brushes you should hear the individual strands hitting the cymbal, while with sticks the tap should be audible along with the ring. The sound on kick drum should be taut with visceral impact, rather than loose and flabby. (Make sure the drum itself sounds tight, of course.)
In general, with all EQ bypassed, the least bright preamp is the one I’ll pick. It’s usually the one that’s adding the least grunge of its own, and stimulating the least downstream.
(Incidentally, I think you’re better off auditioning mic preamps on recordings, not live; the added distortion of the converters—see next section—serves as a magnifying glass for any flaws in the preamp.)
As Fred Forssell, formerly of Millennia Media, has reminded me, another classic torture test is to take your key ring from your pocket and jingle the keys in front of a good condenser mic, adjusting the controls to ensure that nothing is overloading. On a good setup you should be able to hear the individual jingles as the keys strike one another; a bad one will mush all the sounds together into a wash of white noise. Believe me, this stressor can separate sheep from goats in a hurry.
Going from the analog to the digital domain is probably the hardest task in the audio world, and the one with the most potential for trouble. Converters can err in three major ways: analog problems, nonlinearity, and jitter.
Analog problems arise from the amplifiers that buffer the converter chips from the outside world at both input and output. Not long ago, these would be medium quality devices such as TL080-series op amps, and they tend to add crunch to the sound. Compounding the problem, many A/D converter chips constitute a difficult load, making the output stages of the buffer amps work harder than they like.
Nonlinearity in this context refers to a converter’s inaccurate response when processing low-level signals, such as the point at which a waveform crosses the zero line on its way from positive to negative. Capturing reverb tails is another tough job; early digital recordings sounded remarkably dry as a consequence of poor low-level performance.
A/D converters have greatly improved in linearity; it used to be that any signal under about –70 dBFS was reproduced with something like 100% distortion. The advent of single-bit high-speed converters and dither has greatly alleviated the problem, but it’s still not completely solved—inexpensive A/D chips can still generate surprising levels of grunge when fed high-frequency signals.
Recording-stage jitter is a variation of the clock that tells the A/D converter when to take its samples. It’s a whole ’nother article; suffice to say that recording stage jitter grunges up the sound, and it can’t be removed once it’s there. Better converter systems have lower jitter, and thanks to companies like Apogee, Antelope Audio, and others, acceptable levels of jitter are found in more and more affordable converters every year.
And that brings us to how we solve these problems. I hate to say it, but this is one of those situations that can be fixed by throwing money at it in the form of upgrades and modernization. Newer, better (and cheaper!) interfaces sound way better than their predecessors, and most of the audible improvement comes in the elimination of high-frequency IM. Staying current is costly, but this is one area where every passing year raises the bar at all price points.
However, the data on small, relatively inexpensive digital recorders or DAWs can never sound as good as what’s on the big multitrack recorders used Downtown, right?
The hell it can’t. One of the saving graces of digital is that audio quality is essentially independent of the storage medium. Whether the audio is stored on a massive networked audio storage system or a laptop from the corner store makes no difference in its audible quality. Indeed, you could encode the numbers in strokes on clay tablets and, given an appropriate retrieval mechanism, never know the difference.
This is crucial: the quality level of an unmanipulated digital recording is essentially a function of the quality of the A/D and D/A converters. Period. (The ‘essentially’ is a weasel word to cover possible increases in jitter created by bad transmission systems, but they’re irrelevant to our argument for now. Besides, a good D/A converter gets rid of added jitter—everything that’s not recorded in at the start.)
This gives us our wedge
Until now, high-quality converters have been pretty much limited to high ticket devices: professional multitracks and standalone conversion systems. But over the last few years, many products have appeared that offer much better performance than was previously available to project studio users, at more reasonable prices.
A growing number of companies offer high-quality 2-channel converters for prices that fall as fast as quality rises. I haven’t tried all of them by any means, but if the performance is as good as I suspect based on the units I have tried, devices like these offer a way to get true professional performance in a project studio, two tracks at a time.
Again, this is making a virtue of the way most of us work. If we usually record one or two tracks at a time, we don’t need more than a couple of channels of ultra-high-quality A/D conversion. A standalone converter or interface, connecting to our computer, can offer very high quality audio without breaking the budget.
How do I listen to a converter? My favorite torture test is to play a mandolin. (And the way I play, the torture isn’t limited to the converter.) The quick transients of the doubled high pitched strings are great stressors, which stimulate the production of high-frequency intermodulation, and there are no low frequencies to mask the distortion products.
Playing mandolin into a low-quality converter, such as those found on first-generation digital multitracks, you hear a halo of fuzz riding up and down with the level of the instrument, sounding remarkably like modulation distortion on a bad analog tape recorder. Rattling your key ring is a good test here too; listen for the system that can cleanly resolve the key clinks.
It’s important to do these torture tests at both high levels and low; some converters add more fuzz to low-level signals.
I’ve described a strategy aimed at getting the best signals possible on tape, which in the glass-half-empty world of information theory means the least corrupted. That’s the biggest part of the battle, but there’s one more hurdle: mixdown.
Here we run into a dilemma. The folks Downtown almost always mix in the analog domain using high-grade external processing and a board that occupies most of the room and costs like blazes.
The best of these consoles add much less grunge to the sound than a typical small, budget-priced board. Many use discrete solid-state circuitry, while those that incorporate op amps use higher-grade units than are found in most project studio boards. (Unless, of course, you modify your board, which is another story.) Boards like that, we can’t afford.
So I’ll go out on a limb and say that the best strategy for the project studio recordist is probably to stay digital. Let’s face it: low-cost boards have achieved miracles in terms of features and bang for the buck, but perfect they ain’t.
The Windows- and Mac-oriented software providing computerized mixdown capability has become very sophisticated, and it’s now easier to use as well. Some of the best algorithms from external digital processors are available as plug-ins for these editing systems, providing all the performance of the hardware units without the added degradation of additional A/D and D/A conversions. And many of the hardware processors now include digital ins and outs for direct connection to a digital mixing system.
Is this the ultimate mixing setup? Perhaps not; a really good analog mixdown studio can probably ace it out for ultimate quality, particularly when fed by something like a 24-track analog recorder. But I wouldn’t bet the farm on that being the case every single time; digital mixing has become very good, and it’s certainly a whale of a lot better than suffering the additional degradation of converting back to analog, running through an imperfect board, then converting to digital again for the final product.
Provided, of course, you treat your digital signal right… which primarily means keeping your hands off it as much as possible. Every manipulation of a digital signal that changes the waveform holds the possibility of degradation if the dithering algorithm isn’t perfect. Resist the temptation to “normalize” tracks before mixing them; the extra level manipulation adds a bit of rounding-off error that (like all distortion) is cumulative.
When possible, combine steps; if level changes and EQ are both necessary, do them in a single operation (most EQ plug-ins will let you). Don’t do extra operations just because you can.
How can we who live in the real world upgrade our quality? Given the constraints of daily life—most of us need to spend our money paying the mortgage, feeding the kids, and putting gas in the car—the improvements must be incremental.
For most people, the biggest bang for the buck will come when you upgrade the mic preamp; the difference between a $750–1200 unit and one that effectively costs a tenth of that will floor you.
I’d tackle the microphones next, with a good non-tizzy condenser or two and a couple of good dynamics. Working downstream, I’d upgrade the A/D converters, either a couple of channels at a time or across the board.
Last but definitely not least, I’d look seriously at the most modern DAW software available today; flexibility grows with each new edition, and as code improves and computers allow more power to crunch the numbers, the quality upgrade for most users will be significant. Combined with high quality gear upstream for tracking, these offer an awful lot of performance at a most reasonable price.
So is it possible for a project studio to sound like Downtown? Maybe, maybe not; there are limitations imposed by small rooms that are hard to overcome (another world of articles).
But I’ve demonstrated, I think, that by careful choices of upstream equipment it’s possible to approach the electronic quality of a full-scale commercial studio—a couple of tracks at a time. You’ll need to plan everything carefully, keeping an eye (ear) on all factors that could possibly degrade the audio.
And of course, if you don’t put the right mic in the right place on the right instrument, the best electronics in the world won’t save you. But that’s just being professional and using your ears—which is what Downtown is really about.