FIX YOUR MIX .com » BLOG

Over the course of hundreds of interactions with clients through Fix Your Mix, both in a mixing and mastering capacity, I have noticed that there is a great disagreement out there on the practical frequencies in audio.  This is strange to me because we have such a vague lexicon for our enterprise (boomy, boxy, tinny, etc.) that you’d think we’d all latch on to terms with such defined parameters as Low, Low-Mid, High, et al.

But nevertheless, every couple months I get a client who says “I love the mix, but I’d really like to hear more bass, can you boost 10 Hz by like 5 dB?”  So for all of you loyal readers out there and as a reference for future clients, I have composed a series of articles describing the portions of the frequency spectrum.

Here is an excellent primer for discussing frequency ranges. Jay works in post-production (television, film, etc.), so his end goals are different from those of us in the music business. He also neglects to emphasize the importance of upper frequencies for imbuing a recording with presence, clarity, and professional quality.  But other than that it is an excellent breakdown of the frequency bands.  For this week though, we’ll be talking about the audible frequency spectrum at large.

The audible frequency range is generally accepted to run from 20 to 20,000 Hz.  Some people hear more, most people hear less.  However, it is important to understand that this broad frequency range is supposed to include the frequencies that the average person is physically able to hear.  For the purposes of experimentation, frequencies outside of the range can be heard, but they have to be amplified to such an extreme that they are not worth measuring.

To the left is the Fletcher-Munson Equal Loudness Curve, established in 1937.  It is probably the most cited graph in psychoacoustics (although the Robinson-Dadson Equal Loudness Curve of 1956 has been shown to be more accurate, since Fletcher-Munson is the most widely used, the following commentary will focus on that).  This graph plots sound pressure level (SPL) in phons against frequency.  The lines indicate equal apparent loudness.  That is, if you were to follow each line, from 20 to 20k, you’d see the variation in amplitude necessary to make each frequency sound equal in loudness.  For example, on the top curve, take 1000 Hz sounding at 120 phons as the baseline.  In order to hear 20 Hz at the same apparent level, you’d have to amplify it to 130 phons.  The same goes for 20k.

Another interesting phenomenon about this curve is how exaggerated the differences become at lower amplitudes.  For instance, when you look at 1000 Hz at 20 phons (the third line from the bottom), you can see that it takes almost 80 phons to sound at the same apparent level.

Now bear in mind, this is not to say that you want to go and quadruple your bass content to get a booming mix.  On the contrary, this is to say that you really shouldn’t expect to hear anything beyond a certain points in the mix.  In almost all instances of music recording, there will be frequency content below easy audibility.  The point of mixing is not necessarily to make them audible.  Sometimes these frequencies are meant to be felt rather than heard.  Other times, these frequencies don’t really add much to the mix at all—eating up large portions of the usable power spectrum and overloading your mix with unnecessary content that either will hurt fidelity due to digital encoding or broadcast algorithms, or will be cast off anyway due to physical limitations of sound reproduction systems.

Here is a graph of all the frequency ranges for common instruments and their notes as shown on a piano.  What you’ll notice is that the range for a concert bass is from ~90 Hz to ~350 Hz.  The absolute lowest note on the piano is around ~28 Hz, and that is a note that you will likely never hit.  Practically all the action in musical instruments occurs between 60 and 5000 Hz.  Allowing for formants, harmonics, and other sonic phenomena outside of the fundamental frequency of the note, it is safe to say that practically all usable and desirable sounds fall within 20-20K and that range could even reasonably be made smaller.

In next week’s article I will examine these specific limitations and discuss why the low frequencies are the most problematic.

More from Phil’s Audible Spectrum series:

• The Audible Spectrum, Part 2: Capture and Reproduction Limitations

• The Audible Spectrum, Part 3: Components
• The Audible Spectrum, Part 4: Sub-Bass Issues
• The Audible Spectrum, Part 5: Harmonics, Overtones, and Formants Primer

In 1978 the Yamaha NS-10 first hit the home audio market. The speakers were originally designed for the consumer rather than the professional sphere. The only problem was that the speakers sounded terrible and no one wanted them for that purpose. They were often described as overly bright and harsh and the frequency response was abysmal in the low end (criticisms which are founded and still exist to this day). However, despite its audiophilic shortcomings, Fate found other uses for this Little-Speaker-That-Couldn’t.

As New Wave, punk, and other lo-fi genres began to take hold on the world, a DIY spirit took over and smaller, cheaper recording studios were created that catered to a clientele who didn’t necessarily place a premium on fidelity. Near-field monitoring became the fashionable choice for these studios because it minimized the effect of listening environment on the sound of a mix. This allowed bedrooms, basements, strip-malls and other ostensibly acoustically unsound venues to become mixing environments.

In these situations the NS-10s weaknesses became strengths. Their lack of low-end capability meant that room nodes (standing waves in a listening environment which cause certain frequencies to be accentuated because of the geometry of the room) weren’t much of an issue since these acoustic phenomena are largely confined to the lower frequencies. Furthermore, their use with cheaper, lower output amplifiers (as was common in these smaller studios) meant that the program output was lower. These volume levels are generally agreed to be the NS-10s’ most accurate operating range. And of course the price, as a previously undesirable commodity, was just right for small studios.

Over the course of the 1980s, the NS-10 became a mainstay of the recording studio and their ubiquity, coupled with the fact that their poor sonic characteristics generally do not incite the individual characteristics of a listening environment, meant that the NS-10 could become a fairly universal reference. By and large, NS-10s were thought to sound reasonably similar in every listening environment. Thus, most mixing decisions are themselves adequately portable.

However, the NS-10 is only as useful as you are familiar with its sonic characteristics. A +7 dB peak at around 1500 Hz contributes to the audibility of some mid-range sounds such as the human voice and acoustic guitar. Operating without this knowledge may result in a weak vocal or acoustic in the mix when you take your songs to other environs.

It is also very difficult to judge a mix’s low-lows on NS-10s. The speaker simply was not designed to reproduce those frequencies. If you aren’t aware of this, then you may find yourself pumping in a ton of low-end just so that the sub frequencies are audible, but if you took it to the club, you’d probably blow out the speakers with all that 808!

It is now agreed in most professional circles that NS-10s are an excellent reference at low volume levels and for gross judgments that do not invoke the sub-frequencies. Armed with this knowledge, you’ll have a better understanding of how to use this omnipresent piece of gear and knowing how to properly use a tool is the most important part of the audio world.

There are some instances when a limited amount of knowledge can do a great deal of harm. For instance, you might know that a bit of sun is good for you. If you are not fully versed in the effects of sun exposure to the skin, you might be wondering what those strange, asymmetrical spots are that keep popping up all over your body. Get those checked out; seriously I worry about you sometimes…

Other times, a basic understanding of something might be helpful the most of the time. Take Euclidean geometry for example. If you aren’t an astrophysicist or a nuclear scientist, pretty much everything you need to know falls into Euclidean space.

But there are also times when the common sense understanding of something gets you by enough so that you don’t realize all the other times that it is absolutely wrong and leads you astray. This is the case with our friend the decibel.

I was working on a record a while back with producer/engineer extraordinaire Paul Kolderie (Radiohead, Pixies, Mighty Mighty Bosstones) and he mentioned something in passing that really caught my attention. I can’t really recall what the situation was, but we were setting up a session and he said to me “I can’t stand it when people ask me to change something by half a dB. A dB is the lowest possible change you can perceive, so saying half a dB is meaningless.”

Many nights I woke abruptly from sleep in a cold sweat tormented by what he had said. Something sounded so right and yet so wrong about that. I mean, if I told you to change something by half a dB twice—both equally insignificant changes by his definition—I would get a change of full dB, and therefore a significant change. Using some simple extrapolation, you can’t keep considering fractional changes in decibels as insignificant, because surely enough they add up.

So what exactly is a dB and what change in dBs is significant to our ear and in our mix? Well, without getting overly scientific about it and also restricting the question to audio applications (sorry electrical engineers), a decibel is a convenient unit of measure that expresses very large changes in magnitude against a reference level in a concise manner. Concision was important back in the days of hand calculation.

When they were busy wiring up the world for telephone usage, Bell Laboratories thought it’d be really swell if they could measure the amount of degradation in audio level over a mile of telephone cable. They did the calculations but soon found that expressing the quantities in conventional terms meant using insanely large and unwieldy numbers. So they decided to use a logarithmic function to bring the numbers to more manageable figures for simple calculation. Logarithms of numbers are useful because they have some of the same arithmetic applications as regular integers (for example, you can add two logarithms with the same base just like adding to regular numbers). The unit they came up with became known as a Bell in honor of the company and Mr. Alexander Graham Bell. So a decibel is actually 1/10 of a Bell.

So why do we talk about tenths of something? After all we don’t regularly deal in decimeters or decigrams. Well in the mid 1800s, some very clever psychophysicists began studying something called Just Noticeable Differences (JND) in sensation. A JND is the smallest incremental change in a sensation that is perceptible to the average person. This could be the JND in touch as measured in PSI or the JND in sight as measured in lumens. Someone discovered that a tenth of a Bell roughly correlated to the smallest detectable change in a sound to the human ear. As such, the decibel became a very important measurement in audio because it was simple to express changes that actually meant something with regard to common perception.

It is important to note that JNDs relate to the AVERAGE person. As such, musicians and audio professionals are often able to detect much more minute changes in audio level.

When studying JNDs, another useful but perhaps counterintuitive aspect of the decibel arose—a doubling of volume roughly correlated in a change of +/- 10 dB. This is useful but strange in that the arithmetic is skewed—you ’d expect a doubling in the perceived volume of something that sounds at +2 dB to be +4 dB. But then again, what is a doubling of something that measures 0 dB? This exposes some of the fundamental limitations in the simple definition of the decibel—human perception complicates the simple calculations.

Such problems spurred further investigation into situational applications of JNDs and Signal Detection Theory was born. In basic terms, the object of Signal Detection Theory is to figure out what extra factors go in to our perception of a sound and how it compares against “noise” or unrelated signals. For instance, does a +1 dB change to a signal still sound like an increase of 1 JND if the sound is played over white noise? What about if the original signal is 100 Hz sine wave? What about 30 KHz?  What if the original signal is a voice played over a country band?  Or a metal band?

It was discovered that the JND of a signal changes based on frequency range and initial level. A JND is around 1 dB for soft sounds at frequencies in the low and mid range—the frequencies we perceive most readily. Really loud sounds can have a JND of 1/3 to 1/2 dB. Really soft sounds on the edge of audibility might have JNDs of a couple dB.

Furthermore, other things can color sounds in such a way that you can take the same sound, add something to it and suddenly the JND might be more or less than a dB. Perceptual Encoding Theorists look for factors outside the Critical Band of Frequency for a sound (the frequency or frequencies that define a sound) that would alter our perception of it. For instance, adding a slight reverb in some cases might cause the JND to rise (meaning you need to turn the signal up more to get a perceivable change) or adding a harmonic exciter in most cases would cause the JND to lower (meaning you wouldn’t need to turn the signal up as much to get a perceivable change). This is because new nerve endings are being excited and these cause our minds to perceive the sound in a different way than we had previously.

As you can see, the decibel is not quite as simple as its common sense understanding in the audio world. So when you need to make something appear twice as loud, you know what to do. When somebody tells you to make their vocals 20 dB louder, you know that that is laughably extreme (for the most part) and you should adjust your corrections appropriately. When someone asks you to turn something down by 1/3 of a dB, you know that it is really only going to be detectable if that sound is already pretty loud.

When it comes to audio, there are two types of compression and both are widely misunderstood, sometimes even by audio engineers. To briefly sum it up:

Data compression is used to reduced the size of computer files. Sound compression is used to affect the apparent loudness, energy level, or impact of sounds.

This post is Part 1 of 2 from Data Compression vs. Sound Compression. Today I’ll be explaining data compression and its two different subcategories. To learn about sound compression, stay tuned for Part 2.

Two Types of Data Compression

As written above, data compression (also known as file compression) means the size of the original audio file gets reduced. Depending on the type of data compression, though, sound quality may also be reduced.

When sending final mixdowns of a Fix Your Mix project, clients receive a .zip file and an MP3. Both of these files utilize data compression, however .zip format is lossless (temporarily compressed) while .mp3 format is lossy (permanently compressed). Both are necessary for different applications.

Lossless Data Compression
(Common file format extensions include .zip, .rar, and .sit)

Lossless data compression is temporary, which means that sound quality is not reduced. Once the file is decompressed (”decoded”), it goes back to its original file format and file size. Popular lossless data compressors include WinRAR, WinZip, and Stuffit Expander, as well as good old fashioned operating systems including Windows XP and Mac OS X.

Pros:

• Highest sound quality possible.
• Allows you to compress multiple files into a single file. This can be useful for internet transfers since web browsers do not allow you to download entire folders at once.

Cons:

• Files cannot be played back directly by audio players: they must first be decompressed (”extracted”) to their original format by the operating system. This may take up to several minutes depending on the speed of your computer.
• File sizes usually not as small as lossy formats (e.g. MP3)

Common Uses:

• Sending to a CD replication plant.
• Sending  to a mastering engineer.
• Sending to a video production company for sync licensing to film or video.

Lossy Data Compression
(Common file format extensions include .mp3, .m4a, and .wma)

Lossy data compression is permanent, meaning sound quality is reduced. Popular lossy audio encoders include iTunes, which uses a proprietary codec, and LameBrain, which uses the LAME codec. Most DAW programs will export directly to lossy formats, however this option costs extra for ProTools.

Pros:

• Files can be played directly by audio players. (Files are decompressed by the audio player itself rather than the operating system.)
• Plays on iPods and other portable audio playback devices.
• Smaller file sizes than lossless formats.

Cons:

• Lower sound quality.

Common Uses:

• Posting to Myspace, Facebook, etc.
• Posting on a website or blog.
• Email attachments.
• Playback via iPods, cell phones and other portable audio playback devices.
  

The MP3 format consists of data compression and data compression only. I’ve heard several old school engineers mess this one up so allow me to reiterate:

Myth: MP3 encoders compress both the sound (like a compressor/limiter would do) and the data (to reduce file size).

Truth: The MP3 format is entirely, 100% data compression. No sound compression is involved.

In the future, I’ll be writing about sound compression in depth and the quality of MP3s.

Additional notes:

• WAV and AIF are both lossless file formats, however no data compression is involved so they are high in file size but may be played back instantaneously like an MP3.
• This section of this post regarding lossy data compression deals specifically with audio formats, but there are lossy compression formats for images and video as well, such as a .jpg file.
• The algorithms which comprise data compression formats are known as “codecs.” Some of these codecs excel in certain applications and not in others. Some have no advantages at all and were developed solely for branding purposes. For an example of the latter, there is no technical advantage to compressing audio to .wma format–it was developed so that Microsoft could force people to listen to music in their Windows Media Player.

Whether you’ve researched production and engineering in magazines (we recommend Tape Op) or on the web, you’re well aware of the ongoing debate between the virtues of analog and digital recording. Eventually, Phil and I will discuss the merits and limitations of both, but for now I will define the two terms in order to lay the foundation for future articles including next week’s Producer Speak: “Bit Depth, Bit Rate, and Sample Rate.”

Digital audio relies on a series of points (called samples) and works similarly to film. A reel of film is comprised of a series of still photos which, when projected at high speed, gives the illusion of fluid motion to the naked eye. Your brain “connects the dots” from one image to the next. Digital audio works like film in that sound is captured via a series of samples (which could be thought of as snapshots of sound pressure levels). These dots are then connected to form waveforms:

The scale is very different, however. Though film could theoretically run at an unlimited rate of frames (images) per second, we only need to capture and play back about 30 frames per second to give the appearance of realism. A CD, on the other hand, plays back at a rate of over 44 thousand samples per second.

Analog audio does not rely on samples at all. Analog is so called because when sound is captured to an analog medium, the waveform that is created is analogous with the sound wave being captured. This means that an analog audio signal has a higher potential for quality, although analog signal decreases in fidelity (quality of exactness) each time it is copied or transferred, whereas a digital signal will retain its quality no matter how many times it is copied.

Things that are analog: reel-to-reel tape, cassettes, microphones, preamps (not including built-in analog-to-digital converters).

Things that are digital: Protools, CDs, DAT, MP3s, WAVs, DVDs, anything on a computer.

Analog to Digital converters such as the Digidesign 192 bridge the two formats together. These are sometimes called A2D or simply “converters” when also referring to Digital to Analog conversion devices.

Next week’s Producer Speak: “Bit Depth, Bit Rate, and Sample Rate.”

You’ve probably heard someone say, “don’t buy [insert hard drive brand]… mine crashed on me and I lost everything.” You may have heard that LaCie drives do not fail.

But you’re still not sure, so you do a little research.

You check out some reviews online, do a search on Gearslutz.com… maybe you’ll go to Guitar Center and have Joe Shred* tell you what he likes to use, next thing you know you’re leaving confused, half-naked with a bunch of $3,000 Monster cables…

Or you ask me and I’ll tell you to pick whatever has cool-looking lights on the front. And I’ll insist that you buy three of them.

But we’ll get back to that in a minute. As far as deliverables** go, LaCie is the industry standard. Like Apple and Pro Tools, many people won’t take you seriously if you’re using anything but LaCie. While I was working for Avatar in NYC, we used exclusively LaCie drives for data storage and deliverables (unless requested otherwise).

So I used to believe they didn’t crash too. That is, until my D2 Quadra crashed after 3 weeks for no apparent reason. At that point I decided to do some research of my own:

That’s right, the hard disk itself isn’t made by LaCie at all. Turns out, this is true for most external hard drive manufacturers.

You can get flashy. Avastor and Glyph drives use more expensive components and are often considered the best by those in the know. Personally, I put spinners on my Glyph 050Q:

But even Glyph uses Seagate disks*** (they have found Seagate to be the most reliable). In other words:

No matter how much money you spend, your drive will fail. You will lose data. And you will be pissed.

So what does all of this mean? Buy anything and pray?

Yes! …well no, but when you’re picking out an external hard drive you can pretty much remove reliability from the equation (barring any widely-reported glitches). I went with Glyph for three reasons:

1. Excellent warranty and replacement policy
2. High-quality components, including the bridging chip (bridging is supposedly the second most common source of external hard drive failures , although I could find no official study to confirm this)
3. Good tech support

The moral of the story is this: the only real way to be safe is to have at least three copies of everything, one of which should be in another location to account for physical damage or theft. I back up sessions to several hard drives as well as data DVDs, which I mail out of state bi-weekly. That means even if Fix Your Mix HQ gets nuked, your session is in Atlanta somewhere…

To be fair:

• LaCie customer service was very good to me and tech support was moderately prompt. I would not hesitate to use a LaCie in the future. I’m just saying that I also wouldn’t hesitate to use anything else.
• My Glyph 050Q fan was clicking and whirring within a week, but tech support told me a temporary fix (stick a paper clip in between the fan blades during boot up) and sent me a replacement fan, free of charge.

Have your own hard drive horror story? Share it in the comments section.

*Okay, okay. Not all Guitar Center sales reps are evil, soul-sucking capitalist pigs. But you know who you are.

**A deliverable is exactly what it sounds like: an item, product or artifact which must be created and then delivered as part of an obligation. In the audio industry that may mean hard drives, data DVDs, CD masters, session recall notes, et al.

***(Source)

Now listening to:

“Living Without” by Slim Thug Presents Boss Hogg Outlawz