Streaming Media

Are you receiving the most up-to-date streaming news? Subscribe to one or more of our newsletters to make sure you are!
Streaming Media on Facebook Streaming Media on Twitter Streaming Media on LinkedIn Streaming Media on Google+ Streaming Media on YouTube

Sure-Fire Tips for Encoding High-Quality, Low-Bandwidth Audio, Part 1
In the first of a two-part tutorial, contributor Dee McVicker runs though some helpful recording and encoding tips to deliver high-quality audio to the dial-up and mobile phone network crowd.
Learn more about the companies mentioned in this article in the Sourcebook:

More Recording Tricks

Tip 6: Use shielded cables. Unshielded cables used to hook up the microphone to a mixer can introduce noise and RF interference to the recording. Also, keep cables away from power cords, which can also introduce noise.

Tip 7: Pay the extra dough for a good audio capture card (check out cards by Digidesign, Turtle Beach and Midiman). Look for high bit width (also referred to as bit depth) and sample rate (16- or 24-bit, 48kHz sample), plus optical or wired AES/EBU/SPDIF input/output. Also, look for wide flexibility in sample rates supported. A common problem on some "consumer" sound cards is that all audio coming into the card is automatically upsampled to the default audio setting of card’s internal mixer. This doesn’t bode well for audio that was initially captured at a lower frequency than the card’s default setting, especially if you’re going to be downsampling again during the editing process.

Tip 8: When recording outside of the studio, reduce background noise and remove any audio source that will be drowned out by louder content. Background sounds serve no purpose but to overwork the codec. In much the same way there are dominant objects in your field of view, so too are there dominant sounds that drown out other sounds. Think of pigeons cooing at a park. Their cooing is the dominant sound before people arrive at the park, but lose importance once people nearby start talking, for example, if the pigeons’ cooing isn't important to your recording, shoo away the pigeons before recording the conversation. The extra sound will only overwork the codec. Some background noises aren't as easy to get rid of, so use directional microphones up close to the audio source to minimize background sounds. Unless background sounds are at the low end or high end of the spectrum, where they can be filtered out, they’re hard to remove from the track once recorded.

Tip 9: Set audio input levels in the audio chain to not exceed 0dB. Digital audio cannot tolerate hot levels like analog can. All audio equipment has a limit to how loud it can reproduce audio without distorting it. Analog gear (tape recorders) tends to go into distortion slowly. A tape recorder might have .1% distortion at 0dB and 1% distortion at +6 dB and 10% at +10dB. A digital recorder will go into gross distortion at any level above 0dB. This is because a digital recorder is calibrated to full scale, and 0dB is full scale. In fact, some digital recorders will refer to this level as 0dBf, where the "f" means full scale. It is OK to have a brief peak every now and then go above 0dB. This is because the ear is insensitive to gross distortions that last less than 30- to 50-milliseconds. But you’ll know you’ve exceeded the input levels if you hear a "crackling" noise.

Tidbit: Err on the conservative side when setting your input levels and let normalization take care of the rest. Normalization is included in most audio software whereby the application calculates how much to adjust frequency levels before distortion is introduced.

Tip 10: Apply equalization to the audio BEFORE normalizing it. Normalizing the file should be the last thing you do. If you add EQ on top of normalization, distortion will result. If your equipment’s normalization option lets you specify a percentage, try setting normalization to 95 percent of maximum to reduce the possibility of peaks choking the codec and causing distortion. If your normalization option doesn’t have a percentage setting, try turning down the volume control in your equipment a tad after normalizing.

Tip 11: Use an audio processor (available in a PCI card or separate unit) when broadcasting live audio. (Orban and Midiman make audio processors ranging in price between $1,000 and $5,000.) Audio processors, which are also referred to as compressors or limiters, provide instantaneous peak-to-average control of audio levels during sudden spikes in volume, such as a crowd cheering or a car backfiring during a speech. These sudden spikes in volume can overpower the system. Preprocessing instantaneously pulls the overshoot levels down, and raises the quieter levels up so they’re audible. These products use compression, limiting and sometimes clipping techniques. As a general rule of thumb, it’s preferable to compress first and then limit, although both will reduce the transient dynamics of the material. Clipping, which lobs off any audio overshoot past 0dB input, should be used sparingly. Clipping can produce wideband modulation distortion, which can stress perceptual codecs and force them to waste valuable bits encoding clipper distortion rather than more desirable program material.

Tidbit: A good audio processor will apply envelope shaping to the waveform using compression and limiting and then soft clip (meaning, gradually rolling a sound off) if need be. This should avoid the affects of hard digital clipping. Hard clipping is a more abrupt way to remove overshoot audio; it results in more of a square wave compared to soft clipping, which is a gentler roll-off. Clipping by its nature is a square-wave component with a lot of high frequency audio energy, which the codec reproduces as distortion. If hard clipping is left unchecked, the codec will spend more of the available bit rate to reproduce distortion than the music it’s supposed to reproduce.

Tip 12: Set the audio processor’s noise gate threshold control so that it will suppress the noise floor during pauses in speech or music. Start at the lowest setting and ratchet up to find the ideal setting that will not bring up the noise, yet leave the desired content alone, during pauses.

Tip 13: Set the audio processor compression ratio between moderate and extreme (2:1 and 10:1 range), erring on the moderate side for music. The resulting audio should be loud enough to mask artifacts that can occur during encoding.

Tip 14: If the audio clip sounds "muddy," try boosting the midrange frequencies. Boost midrange frequencies using the midrange EQ knob on a mixer (or graphic equalizer). Codecs discard some high frequency content for low bit rate targets. To compensate and trick the ear into perceiving there is more dynamic range in the content, boost the midrange frequencies (2 to 3kHz). If your mixer doesn’t have a midrange EQ knob, experiment by turning down the EQ knobs for the high and low bands and raising the volume a notch or two.

Tip 15: If you’re capturing audio, use the "DC offset" (also known as "Centering the Wave") function on your editor. You won’t hear DC offset in the original file, but not removing this low frequency noise resulting from sound card grounding problems can add a low rumbling sound to the encoded file. DC offset varies from sound card to sound card. Some editors will automatically compensate for it, others require that you determine the offset needed. To find out what the offset setting should be, record silence and check the waveform window with your software editor. If you see a flat line above or below 0 axis, that’s the DC offset. You’ll want to set your editor’s DC offset function to correct this.

Next Week: Editing Tips, Encoding Tricks, and More …