|
|
Audio
Processing: A Retrospective
Frank Foti
Cutting Edge Technologies
Cleveland, Ohio, USA
One interesting facet about broadcasting is that many
engineers and program directors have long thought of audio
processing as "black magic," using "black
box" technology to accomplish apparently magical things.
Well, there is definitely real world technology that is
applied to the task, but it's the proper application
thereof-the creative aspects of the technology-that makes it
mystical. If you were to review the history of this art, it is
usually a disjointed collection of stories about what this
engineer did, or that company attempted. So, as a matter of
reference, we've tried to map out a history of audio
processing. This may be the first known "public"
assemblage. Thus, if there are items missed or events that
might be out of date, we apologize. This is not meant as a
"be all, end all," in-depth research piece, but more
as a matter of tracing the lineage of this interesting
technology. Enjoy! Our heartfelt "thanks" to Jim
Somich, of Processing Solutions for contributing this
wonderful essay!
Contrary to popular belief, audio
processing is not as old as broadcasting itself. If we
consider the introduction of the CBS Audimax I in 1959 to be
the birth of modern audio processing (and many do), then the
art is less than forty years old. Before Audimax, there were
various AGC amplifiers and peak limiters, but their primary
purpose-and the way they were marketed-was to either
"ride gain" or prevent overmodulation. It wasn't
until CBS Laboratories, working on a research project for the
CBS Radio Network, introduced the Audimax I that true audio
processing was born. Even then, the primary touted advantage
of the Audimax was "gain riding." The Audimax was a
true audio processor, but the broadcast community had yet to
understand audio processing as an enhancement rather than a
tool for technical housekeeping. CBS Laboratories sold the
products directly and offered a 30-day trial. This was one of
the most successful marketing ploys in processing history, and
is still used today.
The early Audimax was based on the
GE-6386 remote-mu-controlled tube. A derived DC control
voltage was mixed with the program audio, and set the tube
gain. To cancel this DC bias at the output, the units employed
push-pull (balanced) circuitry. The identical DC component in
each side of the push-pull circuit was therefore canceled at
the output. This assumed that the push-pull circuit was well
balanced, and there were the omni-present "thump"
suppression controls to accomplish this task.
The range of 6386-based devices in the
50s included the GE-Unilevel series and Gates Sta-Level, both
simple ungated AGC compressors. Feedback control was utilized
to develop the DC control voltage, which was further modified
by simple R-C networks to determine attack and release times.
The Gates Level Devil added simple gating and return-to-zero
functions to the Sta-Level circuit.
Before the 6386, it was common
practice to employ grid bias as the controlling element in
peak limiters and AGCs. A popular box in the late forties was
the Langevin "ProGar". The ProGar employed push-pull
6J7s driving push-pull 6V6s. The principle was the same as the
units to come later. The ProGar might be considered the first
"true" audio processor, except that it sounded quite
bad!
In the era ruled by AM, the prime
purpose of non-linear audio processing devices was to prevent
overmodulation. When an AM carrier reaches cutoff (at 100%
negative modulation), distortion increases very rapidly and
spurious harmonics are generated en masse! Every AM station,
from the smallest "peanut whistle" to the giant 50kW
powerhouse, employed a peak limiter at the transmitter to
avoid the dreaded "cutoff" at 100% negative
modulation. A popular unit in the early 50s was the GE
BA-6.This limiter suffered from thumps if the push-pull stages
were not balanced perfectly. In the mid-fifties, GE came out
with a peak limiter that-in an effort to reduce processing
artifacts-amplitude-modulated the audio onto an RF carrier,
controlled the carrier amplitude, and demodulated the audio.
Suffice to say that, until the
introduction of the Audimax, "riding gain" was a
profession in the broadcast station rather than a piece of
hardware. Anyone attempting to use a compressor to ride gain
reaped the disadvantage of severe processing artifacts such as
pumping, breathing, thumps, background noise buildup, and
clipping distortion.
Enter FM
While FM did not become a reality until the late forties,
there were no processors specifically designed for this new
medium until the 70s! Either a station used a standard peak
limiter to prevent overmodulation, or they used no processor
or limiter at all. This was not as unusual as you might think.
Average modulation levels were kept low and careful gain
riding was employed. The 75 microsecond pre-emphasis curve
played havoc with these simple processing schemes. I can still
remember muted trumpets pinning our Hewlett-Packard FM
modulation monitor in the late fifties!
Those adventuresome engineers who
placed the pre-emphasis network before the peak limiter
avoided this problem, but created enormous holes in their
program material due to the broadband peak limiting. Fairchild
Recording Equipment Co. introduced the "Conax" in
the late fifties to address this problem. Nothing more than a
pre-emphasized dual band clipper, the results were preferable
to wideband pre-emphasized limiting, but hardly optimal.
The Audimax Reigns
The CBS Audimax was "king of the hill" from the
late fifties through the sixties. The Audimax I was followed
quickly by the II and III, which were improved versions. A
gating circuit was added in the early sixties, which minimized
noise buildup during pauses. A key to the excellent
performance of the Audimax was its use of a gain-platform.
This was nothing more than a dual time-constant release
circuit, but it permitted a much slower time constant unless
the program audio fell outside the gain control
"window." Later, in the sixties, CBS introduced a
solid-state version of the Audimax and added the Volumax, a
pre-emphasized FM Peak Limiter, in which the gain control
element was simply biased diodes. At the end of the
decade-long reign of the Audimax, there were the new 1-rack
unit models, and CBS had added units for AM as well.
The DAP and Mike Dorrough
The reign of the Audimax abruptly came to an end in 1971
with Mike Dorrough's introduction of the DAP, the Discriminate
Audio Processor. Even though we take multiband processing for
granted today, it was a unique idea in the early 70s. Altec-Lansing
had marketed a two-band compressor in the fifties, but it was
not aimed toward broadcasters at all. I knew Mike during those
early years, and I can tell you unequivocally that the
processing world did not welcome Mike and his new idea with
open arms. The Luddites stood firmly behind their broadband
gated processors. None of the big boys would consider
marketing the DAP ,and Mike wound up building them on his
kitchen table in Burbank, California, and selling them
door-to-door. The rest, as they say, is history. Mike Dorrough
made one of the most important contributions to audio
processing. He believed in his then radical idea, and put his
money where his mouth was. The DAP made Mike Dorrough famous.
The DAP ruled the processing world.
The DAP was a three band compressor
and peak limiter with rather gentle characteristics.
Individual parameters-attack, release, band splits, etc.-were
not user-adjustable. A PWM gain control scheme was used,
similar to those still used by some manufacturers today. In an
era before low-cost, high-performance VCAs, this was a solid
engineering design. Final peaks were caught by a
"spongy" clipper across the output stage, which
consisted of several diodes in series.
Mike instinctively knew that sharp
band splits would affect audio quality. The DAPs had very
gentle slopes, on the order of 6dB/octave. Phase aligning
filters were still in the laboratory stage. Any decent
engineer around during the DAP phase knew that you could
radically alter the sound of a DAP with a "greenie"
and a little patience. There were dozens of trim pots inside
to play with.
The OptimodŽ
FM was coming on strong in the mid-seventies and the DAPs
did not directly address the 75 microsecond pre-emphasis
problem. Sure, you could get a DAP with internal pre-emphasis
and let the three band compressors control your high
frequencies, but what about filter overshoot?
FM Stereo requires very sharp low-pass
filters in the audio path to avoid interference to the stereo
pilot at 19 kHz. As the FM stereo art began to mature,
observant engineers noticed that tightly-controlled audio
became significantly looser after passing through these
filters. The culprit, as we all know today, is group delay:
the propagation of different frequencies through the filter at
different speeds. Indeed, the filters would add peaks of as
much as 6-8dB to the heretofore tightly controlled audio. Some
engineers, in an attempt to be louder without overmodulating,
removed the filters entirely from their stereo generators. As
you can imagine, the severe interference with the pilot was a
poor trade-off indeed. Fortunately, modern CDs with their
superior high frequency response (compared with carts and
records) were still years away.
Robert Orban theorized that, by
combining all processing functions with a stereo generator, he
could better control these overshoots. The Orban 8000,
introduced in 1975, proved him right. His 15kHz low pass
filters were non-linear filters without significant overshoot.
The box sounded good. The 8000 employed two bands of
processing, thereby reducing intermodulation substantially.
Later, the 8100 continued the approach pioneered by the 8000
and went on to become the most popular FM processor thus far
in history!
Other Contributions
No brief history of audio processing can ignore the
contributions of several others. While they may not have been
as precedent-setting as CBS, Dorrough or Orban, they are
important contributors nonetheless. These companies include
CRL, UREI, Pacific Recorders, Inovonics and Texar.
CRL pioneered the "phase
rotator," which kept the highest peaks of the audio in
the same direction. This allowed AM stations to overmodulate
on positive peaks and attain additional loudness-that is, if
their transmitter was capable of producing this additional
power in a linear fashion. A similar strategy was employed in
the UREI BL-40 Modulimiter and the Pacific Recorders
Multi-Limiter. Pacific also made a three-band compressor using
optical gain control, called the Multi-Max.
Composite Clippers
Most FM stations could now be substantially louder than
before. Filter overshoots were now a thing of the past, and
integrated processors made it easy to set up the air chain for
maximum performance and minimum artifacts. Those aggressive
broadcasters seeking even greater loudness than their
competition began to experiment with composite clippers: hard
clipping of the stereo baseband signal. The early units
clipped everything, including the stereo pilot, and were thus
found to be in violation of the FCC rules. In 1982, Eric Small
of Modulation Sciences introduced and patented the CP-803
composite clipper that did not clip the pilot signal. It was
an immediate hit and is still used to this day by gonzo
broadcasters.
Check out the audio processing chains
in most stations today, and you will still find products from
the golden age of audio processor design: the 1970s. It is a
tribute to these companies that their 20-year-old designs are
still on the air and performing very well.
The Next Phase
In the 80s, the Audio Prism by Glen Clark's Texar company
was a unique four band audio processor using optical gain
control and a "hold" circuit to reduce
"hunting" in the individual bands. It was a hit and
is still used by many stations as a pre-processor. As its name
implies, it tended to bring out detail in music that had been
lost by other boxes.
Another contributor was Greg Ogonowski,
who had formed the company Gregg Labs. An excellent design
engineer in the analog domain, Greg is credited with
developing a technique to utilize multiband clippers that were
coupled with a low pass filter in each band. This concept
reduced clipper-induced IM distortion and allowed for greater
perceived loudness. In the late 80's Greg was known for a
custom FM processor that not only utilized this multiband
clipper idea, but had a "killer" bass EQ. Some of
the concepts of this custom processor was later employed in
the digital Optimod. Greg should also be given credit for
developing a method for reducing or eliminating
modulator/demodulator overshoots in STLs, exciters, and
modulation monitors.
Another innovative product was Frank
Foti's Vigilante, which was a modified Aphex Dominator. Still
on the air in many major market stations today, the Vigilante
was a force to contend with in the world of competitive audio
processing. This product launched Cutting Edge, a company
behind some new concepts that had been developed at Z-100 of
New York, when Frank was still working "in the
trenches" at a radio station. Using a completely
different approach to pre-emphasized limiting, the Vigilante
was able to generate considerable loudness, yet retain an
openness in the presence and high frequencies that had not
been heard before.
Unity 2000i
Designed around a concept known as "digitally-sampled
analog," the Unity 2000i found a wealth of popularity
worldwide. It was the first complete processor to incorporate
a composite clipper within the system. It was also one of the
first processors to allow the user to have complete remote
control of the system using a computer as the interface.
Expanding upon the concepts developed in the Vigilante, the
Unity 2000i was able to take advantage of the best features of
both digital and analog processing. As Frank Foti has since
stated, it was a great platform to use as the precursor, or
stepping stone, to an all digital processor.
Enter DSP
Digital Signal Processing (DSP) is the high-speed
real-time manipulation of digital data. It was understood as
early as the late 80s that if audio processing could be
performed in the digital domain, many advantages could be
realized. The technology was not in place at that time to make
these dreams a reality. Beginning in the early 90s, however,
inexpensive, high-speed DSP chips became available and made
the promise of digital audio processing a fact of life.
Early DSP processors included the
Audio Animation Paragon and the Valley multiband units.
Gentner even took a stab at it, after they acquired the Texar
firm in 1988. None of these efforts achieved fame to any great
degree. The old analog boxes still sounded better and louder.
The Orban Optimod 8200 was the first serious DSP design for
broadcast and achieved instant favor due in part to the Orban
reputation. In the mid-nineties, CRL introduced their own DSP-based
processor, which exhibited even better performance.
Until the Omnia.fm, all DSP-based
processors exhibited artifacts that made them more of a
curiosity than a serious processing tool. What Frank Foti
discovered was that a modern DSP processor did not have to be
a clone of old analog designs. The analog designs did what
they did because that was all they had to work with. A fresh
look, a new approach was necessary.
Omnia audio processors are the
culmination of not only all of the DSP knowledge currently
extant, but, in a peculiar way, the culmination of all of the
processing knowledge that has come before-the ProGars,
UniLevels, Audimaxes, DAPs and Optimods-but elevated to levels
heretofore unforeseen. The bar has truly been raised by Omnia.
It would be tempting to say that broadcast audio processing
has now reached perfection. It has not. But it will take
considerable ingenuity and innovation to better the technology
of the Omnia-6.
Top
|
|