Tweaking Levels
Voice Processing | Level Headed
by Jeff Towne

I don't like your tone, young man...

Back in the (not so) old days of recording to analog tape, it was standard practice to always include set-up tones at the head of the tape, often several tones at multiple frequencies, to allow proper alignment of the playback machine and gain-setting on the mixer or dub recorder. The most crucial of those tones was a 1 kHz sine wave, set to indicate reference level of "0dB". This tone is still quite valuable even in this digital age, as a guide for the still common analog stages in uplinking and/or broadcast.

The bad news is that there are not absolute rules for the tones, recommended values range from -18dB to -12dB, but the important concept is that the level of your tone should indicate the average volume of your program. The proper ratio of average level to peak level is subject to much debate, especially as average levels in the commercial music and radio world creep upward every day, as new compression and limiting tools allow record producers and broadcast engineers to squeeze program material upward more and more. Pop music is increasingly "crunched" and radio stations are engaging in loudness wars, resulting in a sound that seems to keep meters pinned within a few dB of full-scale at all times.

But this constant quest for more volume often results in audio that sounds lifeless, unnatural and distorted, and is rarely appropriate for the documentary-style programs we usually discuss here. But it is worth keeping in mind that the other extreme: wide dynamic range with very quiet and loud elements rarely sounds good on radio or internet broadcast either, the quiet elements often lost in road noise or mired in streaming audio's noise floor. Try to engineer your audio to have a consistent apparent volume. Your meters can help you with this, indeed your levels should bounce around your reference level, but you really need to rely on your ears too, some sounds seem louder or softer at identical meter-readings.

Why bother with a tone at all? Whether you are mailing CDs to radio stations, using the satellite system, or FTPing programs to PRX for stations to download, before that program hits the air, there will almost always be at least one analog stage: a station's control room mixer or the satellite modulator for instance, and it will always be more likely that your program will air at the optimal volume (good and loud without distortion, with proper left-right balance) if there's a some way for the users to check the input gain trims or mixer fader levels on their recorders or broadcast mixers. Giving a reference tone that accurately reflects the average level of your program will help stations broadcast you carefully-engineered production at an optimal level. There are valid concerns that including tones with your program material will increase the probability of an inattentive board operator broadcasting those tones, but careful labeling and communication with the engineering staff can reduce that possibility.

If you distribute through the Public Radio Exchange, it's actually not advisable to include a tone as part of your soundfile. As stations download the digital files and in most cases load them directly into a computer for airing, there are not many opportunities for the levels to change, making the tones less helpful. While it could be handy to have a setup tone to align broadcast mixer levels, that benefit is outweighed by the potential for the tones to be aired accidentally, or for the head of your program to be damaged if the station is required to trim tone off of the soundfile. This model will likely hold true for the PRSS Content Depot, but check their producer's FAQ for up-to-date guidelines.

If you are distributing through the Public Radio Satellite System, you must give the uplink a setup tone to align the input optimally to the satellite modulator, and those modulators will clip if peaks get beyond 12dB over average. There are guidelines here at the PRSS website.

What's the Reference?

Even if you are not including a tone on your final soundfile, it's good practice to mix to a reference, pick an average level and mix against it, with peaks extending no more than 10-12dB above the average.

What level should your reference tone be, whether you include it on your media or just refer to it with your FTP-ed soundfile? The simplest answer is -12dB. There are good arguments to be made for other settings, but if you are engineering your master to have peaks near full-scale (0dB) 12 dB of dynamic range between average and maximum levels is a good-sounding, but still safe range for radio or Internet streaming. You can format your recording with tone at -15 dB with no peaks exceeding -3dB and maintain that same 12dB dynamic range, and PRSS and some other uplinks actually prefer that arrangement, especially on DAT tapes, because it allows cleaner interfacing of professional DAT machines and the modulators. But it's simpler to put tone at -12dB if only because it's physically impossible for your peaks to extend higher than 0dB, keeping you within the safe range.

As for regulating your program level so that your reference tone is actually representative of the real volume of your program, that's a murky and subjective art. With better compressors and limiters, it's possible to push the average levels up very high relative to peak. You'll have to use your ears and your meters to determine where that is. Broadcast chains often apply severe limiting and compression, but there's no consensus about whether audio sounds better "pre-compressed" or left with a greater dynamic range before being subjected to the indignities of a broadcast processing chain. Keep in mind that your program might have ancillary uses now or in the future, so make it sound as good as you can, so that its potential eventual uses on the Internet or on CD or whatever new formats come our way are not compromised.

There's nothing wrong with being average

Old-style VU-meters, common on analog gear, especially tape recorders, were actually very good at indicating "average" levels, although not so good at indicating peaks. So if there's a way to employ VU-style metering as you mix, that additional information can be very helpful.

One wonderful tool is the free VU-meter plug in (for almost all plug-in formats) from PSP Audioware. You can set the reference level to wherever you want, and the digital needles faithfully replicate the ballistics of the old mechanical meters, and are very good for indicating if your average levels are too high or low. Set your reference tone to display at 0vu, and have your program create good healthy bounces on the meters, not in the red all the time, but traveling over there occasionally.

That standard 1hHz sine wave is used as a reference because the physical nature of that tone, a "pure" wave with no complicating harmonics, is a good representation of the average level of the much more complex sounds of voice, musical instruments and ambiences. Notice how a -12dB 1kHz sinewave tone appears as less than half of full-scale on the ProTools peak-reading meters (the vertical green bars) but show, by design, exactly "0vu" on the "Vintage Meter" plug-in, which is better-suited to showing average levels. Well-engineered voices and music placed in those tracks should show peaks near the top of the vertical green meters, while registering on the VU-style meters in the middle of the range, with only occasional needle-travel into the red. If your program causes VU-style meters to stay in the red at all times, you've probably over-compressed, and your audio will sound strident and squeezed. Keep your levels as high as possible, but still leave room to let them breathe. Voices especially sound most natural with some contour of loud and soft.

Tanned and Toned

A couple of tips about making a tone: some computer audio editors have this functionality built-in. If yours doesn't, there's a free program for Macintosh (OS9) from AudioEase. And there are numerous shareware programs for Windows, easily found with a quick Google search. Whatever program you use to make a tone, choose a sine wave, at 1,000 Hz (1 kHz) and make 30 seconds of tone. This next step is not always obvious, but it's very important: put some silence at the end of your tone. Even if you are assembling a CDR with discrete tracks, CD players "unmute" in unpredictable ways, and a small blip of tone could end up aired at the start of your program material if the previous track of tone goes right up to the end. It's much better to add 15 seconds of silence to the end of your tone, so the uplink can air your tone in the time set-aside for that very thing, or a local station can check your levels on an audition channel, then let the CDR track straight-through from tone to your program without having to re-cue. Your program will start with a clean head and there's no risk of catching a little tone as your program starts. I have a standard 1kHz tone soundfile formatted this way that I just leave on my hard drive, and always use as track 1 when burning CDRs.

To be safe, when preparing soundiles for CDR burning or upload, as you are bouncing your final mix, or doing final prep in a dedicated two-track editor, it's good practice to leave at least a half-second of silence at the head, again to compensate for the unpredictable un-mute behavior of CD players. If you have sound right at the beginning of a CD track, it could get clipped slightly by CD players that are slower to un-mute when given a "play" command. Even when delivering via FTP rather than CDR, it's safer to leave that small silence at the head, as those soundfiles might end up transferred to CDR at a local station for ease of use.

Mind the Gap

It's important to understand, and possibly compensate for, any inter-track "gap" created by your CD burning program. Ideally, you should use "disc-at-once" mode and set gaps to zero, so that tracks flow smoothly from one to another with no space between tracks. But some CD burning programs do not allow you to do this, especially if they only do "track-at-once" burning, and automatically insert gaps between tracks on an audio CD. This is not a huge problem, if your program insists on inserting a gap between tracks, simply reduce the silence by the length of the gap so that your program follows at exactly 15 seconds after the tone ends. You may have control over the length of the gap, if not, the default is usually exactly two seconds. If that's the case, simply make a soundfile with 30 seconds of tone and 13 seconds of silence, that way the following track (your program) will still start at exactly 15 seconds after the tone ends: 13 seconds of recorded silence + 2 seconds of inter-track gap.

Truth in Labeling

Of course, there are many ways to setup a DAT or CDR for uplink or broadcast, and the most important thing is that the end-users understand. The most effective way to do that is to format your media exactly the same way all the time, and to LABEL EVERYTHING CLEARLY!!! It's very common in the real world for there to be a substitute board-op or uplink engineer who has never dealt with your program before, so be obvious about every aspect of your tape or CD.

As the producer of Echoes, I send 10 CDs to an uplink every week, and they are carefully labeled with the following information: program number or title, satellite ID, uplink date and time, track descriptions and times, and a redundant instruction about what to do. They look like this:

Of course yours doesn't have to look just like mine, the most important thing is that the station or uplink that is using your program understands how to play it. There are many ways you could format and label your media, just be clear, be simple, be consistent, and don't be afraid to be redundant. DO NOT make it difficult for the user: don't put your tone at the end, requiring them to jump around and re-cue; don't put any audio after your program, requiring them to hit stop or make sure auto-cue is on; don't format your program in multiple tracks unless it is important to be modular (that users might want to air only certain segments of the program) otherwise your uplink or broadcast might be impaired by an accidental setting of the playback machine to stop after only one track. If your program must span multiple tracks, SAY SO on the label, perhaps in several places.

Silence is Golden

I always put a small file of silence at the end of CDRs, after the program track, so that it's easy for the user (and me) to skip to the end of the program track and check the end for timing or outcue. If the program plays-through onto that last track, it's only silence, it doesn't matter.

Again, the most important element is communication, make sure your users understand what you are doing, and ask for input from them, the folks who are actually pressing the buttons often have some good ideas about how to make it go smoother.

And if you have an opportunity, listen to your program on the air, or over the Internet, and pay attention to how the levels are working. Does it seem to have a consistent volume? Is it softer or louder than the programs that surround it? Is it starting cleanly? Then make adjustments, using your meters, but more importantly your ears, to adjust actual and apparent levels to the optimum settings. For more on how to do that, read Gregg McVicar's column on processing here>>.

--jeff towne

Related Articles