January, 1997
By request, this is an upload of a file I dug out of my archives. It is
basically a 'handout' intended to accompany a talk I gave to an amateur
radio club. Thus there are some references to visuals that I no longer
have. Note that the date is 1984. I have re read the document and
I've decided to leave it intact, but where appropriate add some updated
comments, which I will enclose in square brackets [[]]. To put this in
perspective, this was written shortly after Harris caved in to accommodate
the Motorola integrated circuit, but it appears to have been before the
Delco (division of General Motors) tests, in which they made it clear that
Motorola was the only system they would build radios for. This was
no surprise since Motorola was already a major parts supplier to General
Motors.
Oct 14, 1984
AM STEREO
(talk before the Fullerton Amateur Radio Club Oct 16, 1984)
by Chris Hays, KRLA
INTRODUCTION
Beginnings of AM stereo
The development actually paralleled similar developments in
television and FM stereo. In late 50's and early 60's and
involved such names as Philco, RCA, Westinghouse, and Kahn
Communications. Interestingly, Today only one of those names is
still in the race: Kahn Communications. But the FCC rejected all
these approaches and decided to encourage the growth of the FM
band, which at that time was floundering and starving.
[[The author recalls listening to experimental KAHN transmissions
as early as 1959]]
As FM continued to erode AM ratings, the screaming by AM
broadcasters became louder, and in 1975, the National AM
Stereophonic Radio Committee (NAMSRC) was formed and submitted a
report to the FCC in 1977. This report had its problems, not the
least of which was Mr. Kahn's refusal to participate in it,
citing the politics of the committee and other reasons. The FCC
then authorized Limited testing of the then five competing
systems, and asked the proponents to submit their results and
invited the public and other interests to file comments in the
normal manner. Curiously, the only comments filed by a company
who was not affiliated with one of the competing systems were
from Sony Corporation.
[[SONY in this report appears to be the first to discover the
image instability problem, also known as 'platform motion']]
The FCC announced their decision of Magnavox at the 1982 National
Association of Broadcaster's convention, and then all Hell broke
loose. Proponents and broadcasters alike screamed like stuck
pigs at the stupidity of choosing that system. The FCC placed
their tail between their legs, and went back to Washington to try
to figure out what went wrong. It turns out that the FCC had
turned the selection process over to the science and technology
committee, who, apparently had used a flawed matrix in their
evaluation.
[[The truth is, it was a pretty sound decision: Magnavox was the
only remaining proponent that had a strong consumer electronics
interest. I suspect the FCC considered this (although they never
admitted it) as important to have receivers produced. It is ironic
that one of the few decisions they ever made that was right, they
backed down on]]
The FCC decided to try again, but before they released their
second choice (which was Harris' VCPM) people were beginning to
think that the marketplace was the only place where this would be
settled without the possibility of the whole thing getting bogged
down in the courts for five years. So here we are.
[[Even so, to this day some of KAHN'S patent suits are still pending]]
The systems
The five competing systems were Belar, Magnavox, Harris, Kahn-
Hazeltine, and Motorola. Upon the marketplace decision, the
field was reduced to four, as Belar backed out, saying that their
company did not have the resources to slug it out in the
marketplace.
[[Side bar: there was a sixth system proposed by Fisher, which was
not compatible with existing receivers]]
The four remaining systems can be divided into three groups:
Mixed mode which includes Magnavox (and formerly Belar), Modified
quadrature which includes Harris and Motorola, and Independent
Sideband which is Kahn-Hazeltine.
[[Actually Motorola could also be viewed as a mixed mode system]]
Technical considerations: AM vs FM stereo
FM was a monaural medium once, but it was designed from the
beginning to be somewhat esoteric. What with channels on 200khz
centers, plus or minus 75khz deviation, FM had plenty of room to
grow. The only rigid criteria that the FCC ever laid down for
Stereo transmission, is that it had to be compatible with
existing monaural receivers.
Since the designers of FM stereo had a lot of spectrum to play
with, they decided on a supersonic stereo subchannel approach.
This way, a sum of the two channels (L+R) could be transmitted in
the baseband to keep the existing Mono radios happy. The stereo
is transmitted then as a sum and difference. The difference (L-
R) is placed on a suppressed-carrier sideband signal centered at
38khz. A 19khz pilot tone is provided for phase and frequency
reference (as well as lighting the red light). The only drawback
to this approach is that it results in a 22db degradation of the
signal-to-noise ratio over monaural. But with 75khz deviation and
the signal levels involved with broadcast, this was thought to be
a reasonable sacrifice. (FM is inherently much quieter than AM
to start with).
What should a good AM stereo system do?
1. good separation
2. good frequency response
3. low distortion
4. no loss of coverage
5. low decoder cost to consumer
6. compatibility with existing radios (this is the kicker!)
AM radio is not as bad as the public perceives it. Although AM
stations are on 10khz channels, they are protected to 20khz
spacing or better depending on the frequency. The trouble really
began in the 50's with the ALL-AMERICAN five tube radio. Radios
were starting to get cheaper. The first thing they eliminated
was the tuned RF stage (one of the reasons car radios work better
than home radios on AM is that you can't eliminate the Tuned rf
stage in the car.). Without signal-frequency selectivity, the
receiver receives noise right into the mixer stage, where it is
likely to produce harmonics of itself or couple directly to the
following stages. So the noise floor deteriorates, and the only
CHEAP way to solve the problem is to narrow the IF bandpass of
the receiver causing loss of fidelity.
The final death blow to AM fidelity came with the advent of FM
stereo, and the stereo buying boom of the 60's. Receiver
manufacturers, faced with intense competition were determined to
get the price of the AM tuner to an absolute minimum: after all,
FM Stereo was 'where it was at'. As a result, even a $3000-$5000
stereo system very likely the AM section of that expensive tuner
is only a $0.39 integrated circuit doing the whole thing.
BROADCASTERS see AM STEREO as a way to get high-fidelity back
into AM receivers. (a classic case of the cart before the horse,
but we really have no choice).
[[In hindsight, it was mostly engineers that thought this: we
needed to do a better job of education]]
THE SYSTEMS
All the systems have one thing in common: they work by adding an
exciter to an existing AM transmitter. The exciter has two
outputs, one of which is L+R audio which is fed to the
transmitter's normal audio input, and the other is the RF
carrier, which replaces the signal from the transmitter's normal
crystal oscillator stage. This signal contains the necessary PM
(quadrature) information. Because these exciters are retrofitted
onto transmitters which were never designed for stereo operation,
they also must contain corrective equalizers and adjustable
delays to insure that all of the phase relationships are correct
at the modulated stage of the transmitter.
Mixed Mode
Magnavox is a mixed mode approach. The sum is modulated onto the
AM carrier in the normal manner, and the difference is phase-
modulated onto the carrier. The decoder then uses a conventional
envelope detector (diode) to extract the L+R signal, and a
quadrature demodulator (a type of product detector) to extract
the PM component. The system employs a pilot tone of 5hz,
modulated +/- 20 hz, for stereo identification purposes only.
The advantage to this system is its simplicity. Simplicity tends
to reduce the cost to the consumer. This is truly a receiver
designer's system.
The disadvantages are it has a bandwidth somewhat wider than
normal monaural AM, causing possible increase in interference to
adjacent channels (modulation index=1.0). This wide bandwidth
may also cause mistuning distortion to be higher than monaural.
However, the most serious problem is the problem of image
instability, which all of the phase-encoded systems suffer from.
IMAGE MOTION
Image motion is caused by the receiver receiving two uncorrelated
signals at the same time. This occurs in one of two ways.
Either a co-channel interfering station skips in at night, or the
station receives reflections from its own transmitted signal
bounced off the ionosphere causing a multipath problem. In the
former case, the interfering signal will appear to revolve about
the desired signal. In the latter case, the desired signal will
rotate back and forth (because the ionosphere is unstable). This
is a serious problem, as in some people it can cause a form of
motion sickness. All of the systems except Kahn-Hazeltine suffer
from this problem. More on this later.
Modified Quadrature (Harris and Motorola)
Quadrature in its pure form would be two conventional AM
transmitters on the same frequency with their carriers in phase
quadrature (90 degrees apart). This approach would yield good
stereo in a proper decoder, but has serious envelope detector
compatibility problems, as well as spectrum problems.
HARRIS: A tale of three systems
The original Harris system handled these problems by reducing the
spread between carriers from 90 degrees to 30 degrees. This
approach reduced the envelope detector distortion to a very
tolerable 4%, while allowing linear synchronous detection in
stereo. The problem was that by reducing the stereo information
by one third, the stereo coverage deteriorated seriously over
monaural.
Harris' second try attempted to cope with the coverage issue by
adding a rather complex companding scheme and going to a
variable-angle system. The transmission quadrature angle (??!)
would vary depending on the compatibility problem, and a sliding
pilot tone (25-55hz) would tell the receiver what was being
transmitted. This in effect creates a variable modulation index
scheme. The problems are that the decoder becomes very complex
and expensive, that sliding tone is very vulnerable to
interference 'hits' and that it only solves the coverage problems
for low to moderate amounts of stereo information, and now those
coverage problems also appear in the monaural side of the signal.
To solve at least part of these problems, Harris' third attempt
locked the pilot at 55hz, arguing that the system most of the
time would be full quadrature, and that the receiver need not
track the modulation. (the receiver would 'see' an apparent loss
of loudness on heavy stereo modulation.
The final modification was announced at the 1984 NAB convention,
and shocked the broadcasters. Harris had decided to move their
pilot tone down to Motorola's 25hz so that their system would get
through the motorola Decoder chip. Most engineers feel that this
move effectively killed the synchronous detector forever.
[[This description was garnered from the Harris literature,
and came out just as convoluted. Harris is really just linear
QUAM (Quadrature-Amplitude Modulation) of the variety you would
get out of a standard balanced modulator scheme. The variants
simply played with the modulation index]]
MOTOROLA
The Motorola system solves the envelope detector compatibility
problem, but trades it for a complex stereo decode algorithm
which can be optimized for noise or distortion, but not both at
the same time. The Motorola system generates its stereo by
starting with a true quadrature signal and removes the inphase
term by running it through a hard limiter. This leaves only the
quadrature term of the stereo signal. This is then used as the
P.M. component, and L+R is transmitted as envelope in the normal
manner. The problem with this approach is that the compatibility
problem is not eliminated, it is merely shifted to the stereo
(PM) channel, since the 'I' term of the original quadrature
signal is lost. Motorola gets around this by using a 'distortion
corrector' in their decoder. This 'corrector' compares envelope
with an 'I' demodulator, and develops an error signal, which
attempts to correct (via an inverse modulator) the Q term. The
easiest way to think of this is to realize that, for the special
case of monaural transmission, envelope is equal to I, so there
is no error and therefore no correction. As the stereo signal
increases, so will the difference between I and envelope, and
thus the correction signal will increase. The system gets into
noise trouble for two reasons: First, the correction function
turns out to be the reciprocal of a cosine function. Since
cosines vary from zero to one, the reciprocal function can vary
from one to infinity! So the gain of the loop is approaching
infinity as the stereo modulation approaches 100%. Since this is
impractical, the decoder is designed to 'give up' somewhere
between 70 and 75 percent stereo modulation. This prevents noise
rush-up in stereo, but any information above 70% will suffer
distortion in stereo (mono will be ok). To solve this problem,
Motorola recently persuaded all the manufacturers of AM
broadcast audio processors to add L-R protection to processors
used for the Motorola system, so that the corrector will not be
pushed past its limit. The second noise problem has only
recently come to light, and details are sketchy, but it appears
that when the noise floor deteriorates to about 24db, the
corrector begins to expand the noise component. This condition
is particularly acute in the presence of impulse noise (such as
noise coming from poorly shielded auto ignition systems, leaky
power line insulators, neon and florescent lights, and wheel
static). My theory is that the aggravated noise is due to the
fact that envelope detectors and I demodulators have very
different behavior patterns in the presence of this type of
noise. Because of this, a spurious correction signal is
developed at the output of the comparitor. Since this signal is
basically garbage, the Q term becomes inverse modulated with
garbage, which is placed into the L-R. Worse yet, the signal may
be nonlinearized by the reciprocal function circuitry (though
whether this actually occurs or not is not clear, to my ear,
SOMETHING is going on there). Motorola is currently denying that
the problem exists, just as they deny that image motion is a
problem. Film at 11.
KAHN-HAZELTINE
Kahn-Hazeltine is an independent sideband system. Normal AM
signals have an upper and a lower sideband. The Kahn system
takes advantage of this redundancy by transmitting the left
channel on the lower sideband, and the right channel on the upper
sideband. This sounds simple enough, but there is the problem of
compatibility with envelope detection. A single sideband wave
has a similar problem to quadrature, in that substantial
distortion will result if you ask an envelope detector to decode
it. Consequently, the Kahn system is not SSB, but might
better be called compatible sideband. As the stereo information
is increased and the compatibility problem develops, a second
order sideband is added to the stereo sideband in a carefully
controlled manner. This 'extra' sideband appears to the envelope
detector as if it were the missing sideband (not exactly, but
pretty close). It is important to note that there are really two
ways to do independent sideband. One of the ways would be 'pure'
and require a filter-technology receiver similar to those in use
in suppressed carrier work. This approach is not practical for
High-Fidelity reproduction due to the impossibility of designing
filters that are flat to within 50 hz of the intermediate
frequency of the receiver, and then drop off like a cliff! Even
if such a filter could be designed, its cost would be prohibitive
for a consumer-grade product. As a result, Kahn-Hazeltine is a
PM-optimized approach, designed to be received with a decoder
similar to the Magnavox decoder, but with an audio phase rotation
of 90 degrees in the L-R path (in actual practice, L+R is shifted
45 degrees one way, and L-R 45 degrees the other way. The
decoder also contains a technique which allows the correction
sidebands to be partially cancelled, such that the stereo
distortion can be held to less than one percent. Note that
cross-channel (left into right, right into left) intermodulation
products, which some detractors have said occurs in this system,
will not occur in this system. This myth started because you can
create a system which has this problem by attempting to decode a
PM optimized system with filters or vice versa. The only way
that this system may produce this is if there are nonlinearities
in the envelope path.
[[Improvements made in the STR-84 exciter made the system completely
compatible with the decoder, and gave up some of the sideband tuning
helpers, which were becoming obsolete. Truthfully, the sideband tuning
approach st6. fworks surprisingly well, although the sharper IF filters
in modern radios coupled with the pre-emphasis used by broadcasters makes
the process more difficult than it once was.]]
But isn't this alot of trouble to go to? Why not just use the
Magnavox approach? The payoff is that this system is the only
system that is not sensitive to sideband PHASE. The lack of
phase sensitivity is best shown on a vector diagram. It can be
shown that rotating the L-R audio 90 degrees, will cause the
phase term to drop out. The stereo information is then contained
in the relative AMPLITUDE of the upper and lower sidebands, so it
behaves as a true AM system. This is very important on the
standard broadcast band where signals may skip after sunset. All
of the phase encoded systems have image instability or rotation
in the presence of skywave interference from their own
transmitters (such as would be routinely encountered in the
station's secondary service area, and, in some cases, in the
primary area as well). This problem has been known to produce a
type of motion sickness in some persons, particularly if the
person is using earphones or is otherwise isolated from other
sounds. Some people don't believe this. This is partly because
it is a phenomenon that is difficult to demonstrate repeatably,
and it may be that not all persons get the physiological side
effects. The other type of problem occurs when a co-channel
interfering signal skips onto a local station. In this case,
since the interfering signal is uncorrelated in frequency by what
ever the error is in the two stations carriers, the interference
will appear to revolve about (on headphones) or flip back and
forth at the beat rate. This phenomenon is at the very least
amusing, and at the worst very distracting and annoying.
[[My observations of this phenomena have been extensive, and the most
startling thing is that if there is mild co channel interference, the
signal will develop a 'chopped up' sound that makes the signal
unlistenable on a stereo receiver. A mono receiver sounds fine.
Although the Motorola decoder is supposed to switch to mono when
this type of thing occurs, in practice it doesn't work well and
either won't switch, or will switch in and out of stereo, creating
what I have come to refer to as the 'whiplash' effect as the image
snaps form wobbly stereo to mono and back.]]
THE MARKETPLACE TODAY
As of this date, the war of attrition seems to have narrowed down
to a two-way horse race (actually a war) between Kahn-Hazeltine
and Motorola. After this description, you might be surprised at
Motorola being there. So are alot of us. The answer is in a
uniquely American syndrome: He who posseseth the greatest bucks
prevaleth. Currently, Motorola has a pluality (that's different
from a majority) of stations on the air. Their problem is that
most of these stations are in the small-to-medium market areas.
In the major market areas (where managers still listen to their
engineers), Kahn-Hazeltine is clearly the winner. In the Los
Angeles, counting stations which actually show up in the Arbitron
ratings, Magnavox has one, Harris has two, Motorola three, and
Kahn six! The pattern is similar in other major markets.
Motorola has persuaded several receiver manufacturers to build
receivers with their chip (which works with only their system).
Sony and Sansui have introduced a line of receivers that work
with all four of the systems. If you contemplate buying AM stereo
equipment, be sure that it will work with all four systems.
[[Sony built some beautiful receivers. They were stopped at the
border as Motorola exercised their political muscle, claiming they
infringed on their patents: Patents that were in dispute any way.
The real reason was that they were multi-system radios, which received
their competitors system.]]
Chris Hays, January, 199d-Si/PRE>