For the last thirty years, I have been asked questions on one topic at least once a month, sometimes a few times a week: Compression. What its it? How does it work? How do I control it? Do I need more or less? My response is generally dependent on the context, because in truth, compression is a LOT of different things, but they’re all based on one concept.
The word itself means “press together,” usually to attempt keeping the same amount of something in a smaller space, like 10-pounds of doggy-do in a 5-pound bag. When you press a big tub of grapes together with (hopefully clean) feet, the result is a fine Bordeaux or Chablis. When you compress the left button on your mouse, you computer will do a little trick for you. When you cover one finger with super glue and then press that finger and another together, you get…um, very uncomfortable. In the bigger world of professional applications, Architects deal with compression (along with tension and shear) every day in their designs, particularly on large buildings and bridges.
Cardiologists are very concerned with compression: systole, the compression on your blood stream as the heart contracts and diastole, the compression on your bloodstream when your heart muscle relaxes between beats.
Your mechanic is likewise concerned with compression within each cylinder of your engine as the crankshaft revolves thousands of times every minute. With each of these professional concerns, the wrong compression can be fatal, to your engine, your health or the tens of thousands of people crossing the Brooklyn Bridge every day in cars, on bikes and on foot.
To a modern radio professional, compression can mean completely different things depending on whether you’re talking about time, dynamic range or data, but the concept remains the same. (See doggy-do reference above.) Just like Architects, Doctors and Mechanics, the wrong compression can be fatal to your work. Usually, a radio pro is talking about dynamic range compression, so I’ll save that for last as it is more detailed.
When you’re dealing with an item that is going to be inserted into a larger program, like a commercial going into a network break, you have to pay particular attention to the length of time your item takes to play. When the local break is exactly 30-seconds, your commercial needs to BE exactly 30-seconds long. Any longer and it will be cut off. Any shorter and you will leave a gap in the program stream. Neither is a happy result.
You have a couple of options. If it is 5% too long (1.5-seconds on a 30-second piece) or less, you can generally use a plug-in which uses an algorithm to uniformly speed up your item without making the vocal sound like Alvin and The Chipmunks. If it’s more than 5% too long, you not only risk having Alvin as your spokesperson, you also could add some digital noise which, depending how severe, can be quite irritating. The 5% rule works well with time expansion as well, though if you correct it too much, instead of Alvin you could end up with Jabba The Hut.
If it’s more than 5% too long OR short, you’re stuck with the second option, which I prefer anyway: editing to time.
Before I start building anything, I roll through the VO and separate every phrase, losing all the breath noise, mouth clicking and swallowing that show up in just about every session, even my own. I then rebuild the VO on two mono tracks, alternating phrases on each track. This gives me the freedom to “nudge” each phrase forward or back to give the read a natural sound while keeping an eye on the time. If the VO is seriously short, you always have the option of adding effects or drops, even repeating a phrase from earlier in the piece.
This method also allows you to work an even closer kind of magic: getting the VO to match up better to a musical track. That actually deserves a much longer explanation that I’ll save for another time.
If you have ever converted a WAV file to MP3 or an MOV file to MP4, you have done
some data compression. Digital sound or video is, by definition, data compression. The very process of committing the analog world to ones and zeros compresses what was real into a data stream, so some of the information is lost in translation. Engineers who design these formats describe them as being “lossy.” How MUCH is lost makes a format more, or less “lossy.”
The WAV format is a PCM file used strictly for audio. PCM stands for Pulse Code Modulation, which means the amplitude of the analog signal is sampled regularly at uniform intervals, and each sample is quantized to the nearest value within a range of digital steps and there is zero additional compression for that file. It is as close to analog as the digital domain can get. AIFF is another kind PCM file that can also be used for other media. On average, a WAV or AIFF file that contains 1-minute of music will be about 10Mb in size.
When your storage is limited or the bandwidth of a data stream is rather narrow, you need to compress the WAV or AIF file. That 10Mb WAV becomes a 1Mb MP3, depending on the sample rate. The more samples, the less “lossy” and better sounding your file is. If you keep it up around 320Mbps, it’s gonna sound pretty decent and still save a ton of space.
DYNAMIC RANGE COMPRESSION
First, a quick explanation of dynamic range: The difference between the most quiet part of a recording and the loudest. If the quietest part is -91db and the loudest peak is -3db, the dynamic range is 88db. Easy math…right? Most classical music has a broad dynamic range. A great example is Tchaikovsky’s The 1812 Overture. From the almost whispering piccolos in the beginning to the cannons and church bells of the finale is going to be a tremendous difference. Screaming metal by a band like Megadeth will, in the main, have the least dynamic range.
Dynamic Range Compression (DRC) would mean the reduction of the dynamic range. You can boost the quiet parts or take down the loudest parts, or both. For the record, a “limiter” is not a dynamic range “compressor,” although it is often used that way to similar effect by lowering the threshold setting.
There are four main settings you’re confronted with when opening most compressors:
Threshold – the level of the incoming signal at which compression begins. This is always the first knob you should grab and adjust while audio is playing. How deeply you compress is directly affected by the threshold. Generally, it is good to start off with the threshold just below a few peaks, then start getting more aggressive.
Knee – the moment compression begins (hard=instant, soft=slowly). The knee is often confused with “attack.” The difference is the knee only deals with time. Attack deals with time AND strength.
Attack – how quickly compression builds once the knee has activated.
Release – how quickly compression lets go when the input drops below threshold.
The last two questions are, “When do I need it?” and “How much do I need?” My style of production calls for a fair amount DRC on the VO and almost none on the music. I want the vocal to “float” on top of the music, giving the final product a unified sound…something very polished. Then I usually use a very slight amount on the finished product to control those pesky peaks and put a nice shiny gloss on it.
Unfortunately, “How much," is a question only you can answer. Every radio station has DRC built into the processing chain and the amount you use on your piece will largely be determined by how much is already on your signal. A CHR station will typically have a lot. A News/Talk station, not so much. A Classical Music station will have as little as they can get away with. (The signal-to-noise ratio will determine just how quiet your program can be before it’s swamped by noise.)
Although it’s not technically correct, compressing your dynamic range sounds very much like turning on “loudness” on your receiver. Putting a strongly compressed bit of programming on TOP of a signal that’s already been compressed a lot, can have some pretty awful results. It will sound lifeless and (believe it or not) very often it will sound less loud, compared to whatever is around it. So, as you get used to the DRC your station is using, start off with just a little on a few pieces, then slowly up the ante over the next few. Once your work is “popping” from the rest of the programming, you are set.
Does that cover it? Barely!