The author is president of StreamS/Modulation Index LLC.
There are many more technical details to optimizing broadcast and netcast audio than simply installing an audio processor. What comes before and after the audio processor can have a huge effect on audio processing results.
First, let’s understand what audio processing is. With many new technical staff, there is little to no real understanding of this, since many with IT backgrounds, now responsible for broadcast/netcast operations, lack technical media experience. Outsourcing to third parties has become another problem. This is known as the big “digital divide.”
With more and more audience migrating to streaming from traditional radio, it is more important than ever to audio process streams correctly. Much of what is available now still sounds like the amateur hour, because the proper engineering has not been applied.
The primary goal of audio processing is to solve the biggest age-old listener complaints of volume and tonal variations, improving your listener experience.
The secondary goal may be to give your sound a “sonic signature” so you sound better than your competition, in hopes that this keeps your listeners coming back for more.
The tertiary goal is to prevent transmitters and encoders from over-modulating or being over-driven causing clipping distortion.
Before and after
Some claim audio processing is unnecessary, especially in a digital world. This couldn’t be further from the truth.
The same level and tonal variations are still present in digital media delivery. Broadcast/netcasting is not a content delivery audio link or studio-to-transmitter function. It is a “last-mile” delivery function to provide entertainment-grade audio.
You are providing a program for listener consumption that should be frictionless to consume, the same way movies are shot, corrected and produced before release for viewing. Delivering unprocessed audio to your audience is exactly what they don’t want. Although this might be tempting in theory for audio purists, be careful what you wish for here.
Before we continue, a few important things must be noted.
Like analog audio equipment implementations, digital solutions suffer from the good, the bad and the ugly. Just because something is digital doesn’t necessarily mean it is right or better. Digital audio solutions are only as good as the arithmetic behind them, and suffice it to say, many software developers did not pay attention in math class. Not all digital is the same!
The slimming down of broadcast facilities and staff reductions does not leave much time for serious vetting of products and services. Technical personnel have had no alternative but to turn into “emergency firefighters,” leaving little time for solutions research.
This leaves broadcasters open to vendors taking advantage or providing bad advice. Proceed with caution.
Over the years there have been many articles on audio processing adjustment. As co-developer of Optimod audio processing, there are many available, including a library of Optimod Processing Presets to get you started, or use as is.
Instead of another subjective audio processing discussion, here is a brief factual account of what comes before and after audio processing to make sure your audio processing system will shine.
AM, FM, HD/Streaming are audio processed in entirely different ways.
AM and FM are the most severe cases, with quality levels that can never achieve that of HD or streaming. They involve the use of pre-emphasis, which puts AM and FM at a severe disadvantage compared to digital systems such as HD Radio and streaming.
And streaming has yet another added advantage, lossless audio, provided you are using a good streaming encoder, such as StreamS HLSdirect Encoders and a content distribution network (CDN), StreamServerHLS, that supports it.
Because of the pre-emphasis requirement of AM and FM processors, these devices spend much of their time rolling off high frequencies to prevent distortion and over-modulation. This audio can never compete favorably with digital systems, unless severely turned down, which many broadcasters are not willing to do, not to mention ratings encoder abuse.
Seems like distortion rules here! Wonder why new audience is going to streaming? Wonder no more.
Best practices
First, the input audio source. This is important. No audio processing can fix bad audio sources. Both analog and digital systems must maintain a flat frequency response and have adequate headroom to prevent clipping.
Digital facilities should have a house reference media clock. All digital I/O should be synchronous as much as possible, rather than relying on sample rate converters everywhere. Not all SRC performance is the same and can cause quality issues, especially when many are used in the signal path. Streaming encoders should also be locked to this reference.
If sources are audio files from an audio playout system, the playout system audio performance should be vetted. These software players, once again, are not created equal. Player engines vary, and with the introduction of audio over Ethernet, sample rate conversion performance is now very important. Measuring this performance is tricky. Everyone assumes they are right. Many are not, affecting your sound.
Then there is the audio file format. It cannot be stressed enough that MP3 is dead! It is over 30 years old and a primitive, inferior audio codec by comparison to what is available today.
However, since storage has now become cheap, there is absolutely no reason to use lossy coded audio for playout. FLAC or PCM WAV should be used. There is no need to worry about double coding for HD or streaming audio encoders.
FLAC is probably preferred, since metadata tagging is an enforced standard unlike PCM WAV. Contrary to popular belief, PCM WAV can be tagged, although there are support and standards limitations.
This brings us to audio music sources, the most popular being the CD rip.
Here, once again, not all CD ripping software and CD/DVD drives are created equal. Unlike data CDs or DVDs, audio CDs do not have the same kind of error correction, so it is important to get the right stuff to rip CDs.
But this is just the tip of the iceberg now. With so many releases of the same songs on multiple CDs, the audio quality is now all over the road, even from the same record companies. This requires careful audio knowledge vetting if you are to achieve the best audio quality. We often assist users with procurement of their audio libraries. We’ve seen and heard it all here!
If you are running talk formats with voice and telephone source audio, modern HDvoice VoIP systems should be employed. This can achieve 8 kHz audio with a rather stunning improvement to your callers. POTS and low-rent VoIP providers are to be avoided.
Even new VoIP telephones with a USB headset jack can be used with an audio interface such as StreamS IOdigi2X or a USB headset adaptor for minimal cost, and the telephone acts as the controller. VoIP telephones do not require hybrids.
Tradeoffs
All this said, you may be ready to apply an audio input to the audio processor. What comes next is also important in achieving your processing goals.
The peak-limited output of an audio processor has strict requirements to maintain peak control for maximum processing efficiency.
The signal path following the processing must maintain this peak control to achieve maximum loudness without overmodulation or overload. Otherwise you will need more processing to compensate for this loss of peak control to maintain competitive loudness. This fragile processed signal requires careful distribution and high-performance transmitters and streaming encoders. And yet, again, none of these devices are created equal. It’s all in the details.
The peak-limited signal requires extremely low-frequency response to prevent low-frequency square wave tilt. If tilted, levels must be reduced, costing loudness. The peak-limited signal requires a high-frequency response that does not overshoot on properly bandlimited signals. If overshoot or ringing is present, levels must be reduced, once again, costing loudness.
Keeping the signal path in the digital domain helps preserve these requirements.
Beware of insufficient AC coupling of analog stages, or high-pass filters associated with digital receivers. Many developers are not aware of these strict requirements. And when passing FM-DMPX 192 kHz signals, stereo separation must be maintained.
Not all streaming encoders are created equal. What happens on the other side of the input meter, inside the encoder, is not at all what it may seem.
First, depending upon the coding bit rate and how much energy is actually removed from the signal, significant peak overshoot happens. If the encoder operates in fixed-point arithmetic, this can be an internal clipping disaster, and nothing indicates this on the input level meter. StreamS Encoders use commercial floating-point arithmetic to alleviate this problem.
What happens on the other side of the encoder at the decoder is even more interesting (and sets us up for a forthcoming article in depth). Simply put, the decoder should also operate in floating-point arithmetic, to recover the peak overshoot without distortion. This problem also exists with coded audio music files. To accommodate fixed-point systems, it is probably best to reduce encoder audio drive levels to at least –3 dBTP (True Peak).
And yes, different browsers sound different. This is not audio crackpottery. It is fact. They use different audio engines with different audio implementations. What you don’t know will hurt you.
We will discuss computer and browser audio another time. This will present some surprising performance results, some good, some not so good.
One would think that after being at it for so many years, computer operating system suppliers have had enough time to work all of this out. Sad, but “I guess you just don’t want to rush into these things …”
For the ultimate article on this topic, the “gold book” Maintaining Audio Quality in the Broadcast and Netcast Facility is a comprehensive guide to getting this all right. An updated edition will be available soon.
