Are your audio tracks not sounding quite right? There are several different types of EQ and each kind, shelving, parametric, bandpass, can help you in different ways. If you want to shape the tone in broad strokes, if things just seem too bright or too boomy, or conversely, too dull or thin: try boosting or cutting a few dB with a low or high shelf. It’s kind of like messing with the bass and treble controls on your stereo, although most shelving EQs will give you more control over which frequencies you are adjusting. If there’s a particular frequency bump that’s bugging you, try cutting it back with a parametric. And if there are entire ranges of sound that you’d like to remove, try high-pass or low-pass filters. Some combination of these will likely get you closer to what you’d like to hear. But be careful, too much EQ can degrade the sound, introducing phasey artifacts or metallic ringing. And first, try cutting a few dB rather than boosting, you’ll often get where you need to be faster, and you’ll have less chance of overdriving the sound. Remember: adding EQ can dramatically change your overall level, so be careful that the changes you make don’t cause the sound to overmodulate the track level or the master mix.
Here are a few places to start: if there’s low rumble interfering with the intelligibility of your interview, perhaps from traffic, machinery, wind, mic-handling noise or vibrations, try a high-pass filter. The name is a bit counterintuitive, because we’re adjusting the low frequencies, but remember, we’re letting the high frequencies pass (it might be clearer to call it a “low-cut” filter, but engineers made the rules and we’ll have to play by them.)
You can adjust the frequency at which the roll-off begins, and some EQs will allow you to tweak the “Q” or slope of the curve. The proper adjustments here will depend largely on the character of the sound you are trying to reduce, and that which you are trying to preserve. To pull a voice out of the murk, try starting with a high-pass set between 100-150 hz. Even a resonant male voice will maintain much of its fullness with a cut in this range, the filter isn’t whacking everything below the frequency you set, it’s just starting to roll off there, cutting more and more the lower the frequency is. Higher voices, such as many women’s, or children’s, have most of their energy starting in even higher frequencies, so one could roll off at 200-250hz without impacting the vocal quality too severely.
The offending low-frequency element of the sound could start to sound very weird at extreme settings, so try to find a balance between clarifying your sound and retaining a natural timbre.
A low-pass filter can go a long way toward reducing the effect of wind or plosive distortion from P-pops. It won’t remove them completely, but the sound can be improved by reducing the rumble that results. Instead of filtering the entire soundfile by inserting the EQ on the channel, try selecting only the offending segments and running the high-pass on the rumbles only, as a destructive pass (in ProTools, select a process from the Audiosuite menu.) If your editing program allows it, try processing a bit more than you need to before and after the target, and crossfade in and out of the processed sound.
Low-Pass filtering can be very effective in reducing tape-hiss or other high-frequency irritations. Just as with High Pass filtering, the frequency one selects isn’t a brick wall, rather the point at which the roll-off begins. One can often start rolling off high-end at 8-10 khz without destroying clarity. But be careful: the consonants so important to intelligibility have information up high, so don’t be too brutal. You’ll find Low-Pass filters much more effective at reducing hiss than cutting with a high shelf, but sometimes too severe. If the resulting sound seems a bit dull, try putting a little edge back in with a fairly narrow parametric EQ at 3-4 khz. The two filters will be fighting each other, but you can punch up the “s” and “t” sounds a bit with the parametric while getting the general high-end reduction you need. If the sound is still too dull, raise the cutoff frequency, or make the slope of the curve more gentle.
If “s” sounds are your problem, and you need to tame excessive sibilance, the best tool for the job is a “de-esser”. It’s a compressor that’s keyed to reduce the level of the sound only when a certain targeted frequency crosses the threshold. You can build one from a conventional compressor that has a side-chain input, keyed via a band-pass filter, but it’s easier to just get a de-esser, either hardware or software, if this is a frequent problem. If that’s not an option, try a parametric EQ centered at 3-5khz (it will depend on the nature of the S, probably different for every speaker).
As for any specific frequency that might need correction, play some audio, adjust the Q value of the parametric EQ up to about 7, boost 5 or 6 dB and sweep the frequency around until the offending frequency sounds especially bad. (Careful: you can damage your speakers and your ears if you do this at high volumes, or with extreme boosts on the EQ.) When you find the offending frequency this way, then start cutting the “boost” level into negative numbers. Adjust the Q value until it seems that only the frequency that you want to fix is being affected. A higher Q results in a narrower, more specific peak or trough, while lower Q levels will give a more gentle, gradual effect. Extremely high Q levels can be very effective in “notching-out” specific problems, but can create ringing at some settings, so don’t go too crazy with these controls.
Some hums and buzzes can be improved by notching the fundamental frequency (usually 60hz or 50hz, depending on the frequency of the local A/C current: 60hz in the US) and then harmonics above that, 120, 240, 480, 960, 1920, 3840, etc… although it’s tricky to do this well manually. Restoration tools that deploy a vast number of notches automatically, work better, but you can still help things a bit with a few cuts.
While it won’t make a hum disappear, sometimes a high-pass filter can make it less obvious. Start with a roll-off at about 150 hz and adjust up and down to find a balance where you get hum reduction without thinning your sound excessively. Great advances have been made in noise reduction software that can “learn” a steady-state noise and remove it from a soundfile. These processes can often be more effective than standard EQ, although artifacts resulting from these calculations can be unpleasant as well.
Phone transmissions have almost all of their energy in the range between 400hz and 3,500 hz, and so any signal above or below those frequencies is likely to be noise. Try using high and low-pass filters to eliminate rumbles down low or hiss up high. The larger problem is the signature shrillness of phone audio. In order to optimize voice intelligibility, phones are designed to hype the frequencies from about 1,000-3,000 hz, and so reducing those frequencies a bit can make phone audio less harsh. Each phone and voice will be a bit different, but try applying a parametric EQ and cutting a few dB at about 1,000-2,000 hz. Sweep the center frequency up and down until you find the best spot to reduce. But remember that phones will still sound like phones, large ranges of frequencies have been thrown away in order to have efficient transmission down a cable, so no amount of EQ will make it sound as if your subject is speaking into a studio microphone. That bright EQ peak was designed to make voices easier to understand, so don’t reduce too much, but rolling back some of those high-mids can reduce the harshness.
Reduce real world harshness!
Help Transom get new work and voices to public radio by donating now.
Wider parametrics (low Q values) can be very useful in boosting or cutting a range of frequencies: if your sound feels a bit thin in the middle, give it a couple dB at 400 hz; if it seems a little honky, pull out a few dB at 2 khz; if there’s a bass resonance from a small room, cut a little around 100 hz. Sweep the center frequency around a bit to find the right spot, and be gentle, a couple dB will do a lot.
And keep in mind that applying EQ is kind of like laying cinderblocks on a waterbed: cutting one frequency will make the surrounding frequencies seem louder, boosting one band will make the others seem to drop. Sometimes the best way to make something brighter is to cut some bass rather than adding treble. Rather than boosting several frequencies, try cutting in the area you were not going to boost, and raising the volume overall. It’s easy to start piling on multiple EQs and end up with something even murkier that what you started with, so try to be as gentle as possible, and remember that all the frequencies are interacting, so what you do in one range will affect the others as well.