A Method to Improve Conventional FM Stereo
     


The author is president of Omnia Audio in Cleveland.

The FM stereo transmission system employed in worldwide broadcasting has been in place since 1961. The rules governing stereophonic performance have not been altered since the mid-1980s (in the USA) when they were modified to allow an additional 0.5 percent total modulation (maximum of 110 percent total), for every 1 percent of SCA modulation, if an SCA was being utilized. The rules governing the requirements of the FM stereo baseband signal are quite explicit, and leave little — if any — room for improvement of the stereo transmission system.

This paper will offer, in detail, a method utilizing single-sideband suppressed carrier (SSBSC) modulation of the stereophonic subcarrier in the FM multiplex baseband that is compatible with existing radio receivers. Additionally, there are multiple overall benefits to the broadcast signal, which are perceivable to the listener. They reduce multipath-induced distortion, and offer additional protection to the spectrum used for RDS, SCA signals and HD Radio content — thereby improving data robustness in the receiver.

COMPETE FOR EVERY POSSIBLE LISTENER

Radio broadcasting has a good fight on its hands. As a media transom to the public, it battles a multitude of additional delivery methods and systems like never before. Until recently, the choices for the listener’s ear were television, phonograph records, compact discs or tape. Now, with the advent of good portable audio playback devices and wireless streaming, there are many more franchises available to steal the listener away from radio. What can radio do, technically, to improve its performance so a listener has less reason to abandon it as a media outlet?

HD Radio was introduced to the marketplace within the last 10 years, but has yet to make enough of an impact to keep the casual listener. What’s needed is an improvement to the existing infrastructure, one that does not require any change or added expense to the listener. Within present-day radio listening, the FM band is the preferred choice. Recent research and development unveiled a unique means to improve the performance of FM stereo. What follows is the result of that research, along with a recommendation for FM broadcasting.

SILVER ANNIVERSARY OF FM STEREO

April 2011 marks 50 years since the Federal Communications Commission (FCC) approved stereophonic transmission for the United States. The commission, after evaluating 14 proponents, decided upon a method that was of similar design from both Zenith and General Electric.

A quick refresher course how the Zenith/GE system works, courtesy of subpart 73 from the FCC Rules and Regulations:

§ 73.322 FM stereophonic sound transmission standards.

(a) An FM broadcast station shall not use 19 kHz ±20 Hz, except as the stereophonic pilot frequency in a transmission system meeting the following parameters:

(1) The modulating signal for the main channel consists of the sum of the right and left signals.
(2) The pilot subcarrier at 19 kHz ±2 Hz, must frequency modulate the main carrier between the limits of 8 and 10 percent.
(3) One stereophonic subcarrier must be the second harmonic of the pilot subcarrier (i.e., 38 kHz) and must cross the time axis with a positive slope simultaneously with each crossing of the time axis by the pilot subcarrier. Additional stereophonic subcarriers are not precluded.
(4) Double-sideband, suppressed-carrier, amplitude modulation of the stereophonic subcarrier at 38 kHz must be used.
(5) The stereophonic subcarrier at 38 kHz must be suppressed to a level less than 1 percent modulation of the main carrier.
(6) The modulating signal for the required stereophonic subcarrier must be equal to the difference of the left and right signals.
(7) The following modulation levels apply:
____(i) When a signal exists in only one channel of a two-channel (biphonic) sound transmission, modulation of the carrier by audio components within the baseband range of 50 Hz to 15 kHz shall not exceed 45 percent and modulation of the carrier by the sum of the amplitude modulated subcarrier in the baseband range of 23 kHz to 53 kHz shall not exceed 45 percent.
____(ii) When a signal exists in only one channel of a stereophonic sound transmission having more than one stereophonic subcarrier in the baseband, the modulation of the carrier by audio components within the audio baseband range of 23 kHz to 99 kHz shall not exceed 53 percent with total modulation not to exceed 90 percent.

Since the inception of stereophonic broadcasting, there has been no technical change to the infrastructure of the Zenith/GE system at all. The FCC rules are quite specific regarding the multiplex spectrum, and its interoperability as a system. After 50 years of service, the system works fairly well, but it could be better. Considering the alternatives a listener now has, it makes practical as well as good business sense to investigate improvements to the present system. It stands to reason that any means proposed must be backward-compatible with existing stereo receivers. Also, after 50 years, a nice anniversary present is in order!

TECHNICAL CHALLENGES FOR FM STEREO

The FM stereo system, as described above, has worked quite well for 50 years, but not without challenges. Most notable is multipath distortion, especially in areas of hills or mountainous terrain. Also, radio broadcasters have added incremental signals within the multiplexed spectra. Radio Data Services (RDS), as well as a 92 kHz-based SCA can additionally occupy the signal, if desired by the broadcaster. The modulation index of the FM carrier is further reduced with each and every added signal.

Increased multipath is a direct result of low modulation index within the FM carrier. As more of the channel spectrum is utilized within the multiplex signal, the modulation index of the carrier is reduced. The following condition generates the lowest modulation index: a single audio channel of a two-channel system, either Left or Right channel only at 100 percent modulation. For example, a 15 kHz tone in the left channel only will produce multiplex spectra at 15 kHz, 19 kHz (stereo pilot tone), 23 kHz and 53 kHz. Each of these signals will reduce the modulation index to its smallest level, and this increases sensitivity to multipath in the receiver. Fig. 1 on page 1 is an illustration of this.

(click thumbnail)
Fig. 1: 15 kHz, Left Channel Only
Note the 30 kHz difference in the L–R subcarrier of the two sidebands located at 23 kHz and 53 kHz. These are generated by the DSBSC process of (38 kHz minus 15 kHz) for the lower sideband, and (38 kHz plus 15 kHz) for the upper sideband. During multipath, as the multiple reflections of the FM carrier arrive at, and then become demodulated in the receiver, the time delay difference created by the carrier reflections will offset the phase of the upper and lower sidebands. During the stereo de-multiplexing process, stereo separation at these frequencies is reduced as the recovered L–R level is negatively altered due to phase shift brought on by multipath.

Bandwidth of the conventional analog FM channel is allocated for 99 kHz of spectrum use. The FM stereo system requires 53 kHz (0 Hz–53 kHz) of this available real estate. The remaining 46 kHz (53 kHz–99 kHz) is used for RDS and SCA services. Common practice requires the use of audio processing to ensure proper peak level and bandwidth control of the various signals present in the multiplex spectrum. Current generation processors are capable of creating near-theoretical multiplex signals. In these cases, there are little, if any, transmission difficulties for the signal.

However, some broadcasters choose to employ a form of processing known as composite clipping. This technique inserts a hard limiter (clipper) at the output of the stereo baseband generator, and will induce up to as much as 3 dB of clipping to the multiplex signal. These devices provide no additional filtering to remove unwanted harmonic content from the clipping process. The additional harmonics will cover the entire 46 kHz, and beyond, used for RDS and SCA services. This creates interference and distortion to those signals. These harmonics may also interfere with the digital carriers generated for HD Radio, as these carriers are set 129 kHz out from the main channel carrier.

Another known challenge for the system is the compromised signal-to-noise (SNR) level when broadcasting stereo. FM transmission noise will rise at 6 dB per octave over the channel’s passband range of 99 kHz. It has been generally accepted that FM stereo suffers a 23 dB noise penalty compared to monophonic broadcasting. This is due to the rising noise floor over the subcarrier range of 23 kHz–53 kHz, as compared to the SNR over the range of 0 Hz–15 kHz which is used for mono. Fig. 2 is an illustration of the composite baseband signal. Fig. 3 shows the 6 dB/octave noise floor slope of an FM channel, as it would appear at the output of an IF section in a receiver. Fig. 4 represents the SNR response through the composite baseband signal. This illustration superimposes the noise response across the composite spectrum. It is easy to observe the most severe noise occurs in the upper sideband of the stereo subchannel.

(click thumbnail)
Fig. 2: Baseband Spectra

(click thumbnail)
Fig 3: FM Channel Noise Response

ALTERNATIVE APPROACH

(click thumbnail)
Fig. 4: Receiver IF Output SNR
Prior to 1973, when the phase locked loop (PLL) circuit was introduced into the consumer receiver market, FM detectors required the transmitted multiplex signal to employ double-sideband suppressed-carrier (DSBSC) to faithfully recover the stereophonic (L–R) signal. Today, there are very few, if any, pre-1973 stereo radios in use.

An alternative approach for stereo transmission would be the use of single-sideband suppressed carrier (SSBSC) as the mechanism to carry to the L–R payload. The lower sideband is chosen as it reduces the occupied spectrum from 53 kHz down to 38 kHz. In order to support the correct L+R/L–R matrixing in the receiver, the amplitude of the lower sideband is increased by 6 dB. This offers numerous benefits to the receiver:
  • • Reduction of occupied bandwidth in the L–R subchannel range increases the FM modulation index by a factor of two. This directly reduces multipath distortion.
  • • Narrows the overall FM transmission bandwidth and reduces degradation of stereo performance caused by finite bandwidth of passband filters, cavities, multiplexing systems and antennas. If adopted internationally, this further benefits broadcasters 100 kHz channel spacing used in some countries, as compared to the 200 kHz spacing used here in the USA.
  • • Creates additional and significant protection for RDS, SCA and HD Radio signals. Note: With a HD Radio power increase looming, reduction of the composite spectrum benefits conventional receivers due to less demodulated overlap of the HD Radio signal.
  • • Compatible with all existing modulation monitoring systems.
  • • Compatible with detectors in current model (post- 1973) receivers.
  • • Less harmonic content generated throughout the channel spectrum when composite clipping is employed in the transmission audio processor.

The concept of utilizing SSB modulation for the L–R payload has been written about before. A white paper on this topic, titled “A New Method of Generating FM and Television Stereo Composite Baseband Yields Improved Broadcast Performance,” was presented by William Gillman at the 1997 NAB Engineering Conference. Reviewing Mr. Gillman’s paper and subsequent testing by this author confirms his findings, and along with technological advances in the transmission firmware, makes this concept much more plausible.

TECH FINDINGS

Implementing SSB modulation of the L–R signal is relatively easy to accomplish using DSP. Fig. 5 is a spectral diagram of a 15 kHz single-channel tone using DSBSC system. Fig. 6 is the same condition, except SSBSC modulation is utilized.

Note the 6 dB increase in level of the SSB carrier in Fig. 6. This illustrates the manner in which the L+R/L–R mathematics are upheld in the receiver.

(click thumbnail)
Fig. 5: 15 kHz Tone, Single Channel, DSBSC

(click thumbnail)
Fig. 6: 15 kHz Tone, Single Channel, SSBSC

It is easy to observe the significant difference in spectrum used. The DSBSC method forces the single- channel condition of 15 kHz to exist across a broad range. The fundamental is at 15 kHz, and the two sidebands are at 23 kHz and 53 kHz respectively. The DSBSC example illustrates the fragility in faithful reproduction of stereophonic high frequencies during instances of multipath. The group delay at 15 kHz, 23 kHz and 53 kHz becomes non-linear, during multipath, and this is why stereophonic high frequencies are so fragile with respect to multipath.

Compare the spectra of Fig. 5 with that of Fig. 6. The close proximity of the 15 kHz fundamental, and the 23 kHz SSB carrier improves high-frequency robustness during multipath. Due to the closeness of these two frequencies, there is less adverse affect when multipath non-linearities are created, and high-frequency stereophonic performance is audibly improved.

Using a noise source in one channel, the same tests were performed. The results are illustrated in Figs. 7 & 8.

(click thumbnail)
Fig. 7: Noise, Single Channel, DSBSC

(click thumbnail)
Fig. 8: Noise, Single Channel, SSBSC

SSB subchannel modulation makes FM channel occupancy more efficient. Fig. 9 demonstrates carrier deviation of the RF signal using DSBSC modulation of a single channel noise source. Fig. 10 is the same test signal, except SSBSC modulation is employed. For the example shown here, the carrier frequency is 400,000 kHz, with deviation set to 4 kHz.

(click thumbnail)
Fig. 9: FM Deviation, DSBSC, Noise, Left

(click thumbnail)
Fig. 10: FM Deviation, SSBSC, Noise, Left

Once again, it is easy to observe the reduction in utilized spectrum. The signal shown in Fig. 10 will pass through narrow cavities, combiners and mal-adjusted antennas with better stereo performance than the broader signal shown in Fig. 9.

(click thumbnail)
Fig. 11: Single Channel Only Pink Noise, SSBSC
Another benefit SSB brings to the transmission method is added spectral protection to RDS, SCA and HD Radio services. Fig. 11 is an example. Single-channel-only pink noise is used to generate the baseband signal with SSBSC modulation. Notice the extremely wide guard band that exists between 38 kHz and where the first SCA carrier would appear at 57 kHz. The reduction in cross-talk to ancillary services is exceptional!

WHAT’S NEXT?

Transmitting SSBSC modulation of the FM stereo signal can be done right now! Software exists to implement this method today. One minor item must be addressed: FCC rule 73.322, section (a), subpart (4) which states “Double-sideband, suppressed-carrier, amplitude modulation of the stereophonic subcarrier at 38 kHz must be used.” There was a time, when rule 73.322(a) (4) was required. Times have changed. Both transmission and reception firmware have improved significantly to enable a change in the rules and regulations governing the FM stereo baseband signal.

The theory, testing and findings presented here should be more than enough evidence for the FCC to consider, at the very least, a Special Temporary Authority (STA) to enable all broadcasters to implement the SSBSC transmission method. Benefit will occur immediately to those whom employ this, especially those in areas of rough terrain with significant hills, and mountains.

After 50 years of service, a modification to the rules and regulations governing FM stereo service would be a wonderful anniversary present indeed! Most importantly, the benefactors are the general public as audible annoyances will be suppressed, and in some cases eliminated. At a time when broadcasters are looking to find every possible way to enhance their customers (the listener’s) experience, this change in the rules would benefit everyone. This is total upside, with nothing downside for all.

The author would like to thank William Gillman for his exceptional paper written in 1997, which provided a point of reference in this investigation. Additionally, thanks to Steve Church for his support and enthusiasm in our joint audio processing and transmission system efforts.


Rating People: 6   Average Rating:     
Comment List:

In the present system, the peak level of the L-R DSB signal is -6 dB with respect to that of the main channel. In this revised system, its peak level would be increased by 6 dB to maintain existing demodulated L-R levels for proper stereo decoding, correct? By doing so, it would appear that, in the presence of composite clipping, the L-R would be clipped 6 dB harder than present, resulting in an audibly dirtier signal. Or am I missing something? What do real-world tests, with aggressive processing w/composite clipping sound like?
By RadioEngr on 11/30/2010
Sorry in advance if my reply will be seen rather off topic. Lately I was interested to find out if my simple solution to demodulate a DSB-SC signal was finally known by some others after I did it in 1979. Lately the only references I was able to hear of, are about "Costas Loop" (1950) which is rather much more complex than mine. What has attracted me in your article is that the first application of my technique is to easily recover the L-R signal without the need of the pilot tone (19 KHz). Truth be said, I used my idea 30 years ago when I needed to build a private FM link. The FM signal was modulated just with a DSB-SC at 38 KHz that holds the audio signal (voice). So an ordinary FM stereo receiver sees the low part of the basement signal (L+R) as empty. And since there is no pilot tone (19 KHz), it also cannot recover the voice modulated by the carrier 38 KHz as DSB-SC, while my receiver could. For instance, I built the DSB-SC detector as an MS thesis but since I was rather poor I couldn't continue my high studies and returned home. After a few years, I let my idea be published in "Wireless World magazine, UK" though I am not sure it was for 38 KHz or the IF 455 KHz (or perhaps both, my memory is short now). Then I forgot the whole thing since I had to design products far from communications thinking that surely some other engineers in the world will have someday the same idea that came to my mind. That is why I am surprised that no one seems to have it yet! Indeed while I was working on the thesis my teacher told me that if necessary the DSB-SC could be demodulated simply by filtering one side to get SSB-SC! Finally to please your curiosity, my detector is based on a conventional PLL circuit altered in a very simple way (hence no added cost and no I-Q signals... etc) to let it be synchronized with the missing carrier (though it still work the same if the carrier is present at any level) and therefore recover the modulating signal as any o
By Kerim on 12/18/2010
Now all you need is an FM stereo receiver that would not hear the noise from the vacant sideband. Increased power for the remaining sideband would also be nice. What would the noise level be compared to monophonic reception? The current noise level is 23 dB higher than monophonic as I remember.
By James Johnson on 10/15/2010
This is logical. Since the "AM" without carrier is modulated into the FM signal there is no extra noise added if you one sideband. If an "AM" station d one sideband then the noise received by a DSB "AM" receiver on the vacant sideband would be added to the audio. This is a good idea for an FM stereo station.
By James Johnson on 10/17/2010
In regards to the question about perceived problems with pre 1973 receivers, there really isn't a problem. The phase locked loop was introduced into consumer receivers in 1973, and we wanted to make sure earlier model radios would handle the SSB signal without difficulty. FM-Stereo receivers do employ the same decoding methods.
By Frank Foti on 11/8/2010
Regarding the question about composite clipping: The level of L-R in the DSBSC system is the same in the SSBSC system. The LSB is reduced 6dB in the DSBSC, yet the overall level is made up by the USB. There will be no change in the amount of clipping of the L-R signal in the SSB method. If you were to look closely at the composite signal, on a scope, and switch between DSB and SSB, with single channel only, modulation, you'll see the peak level is the same. The only difference is the appearance of the 'carrier' in the SSB signal. It will look 'less dense' due to the removal of the USB signal.
By Frank Foti on 11/30/2010
Please can you explain what the perceived problem with some pre 1973 receivers is exactly. As far as I remember they use much the same decoding process and just generate the 38kHz carrier using different methods.
By Brian G on 11/7/2010
If you could make the receiver detect that it was receiving the new LSB only type transmission it could ignore the frequencies in the upper sideband completely (or partially) so you wouldn't get any noise from them and this would give a big improvement in received S/N ratio. Then you just need some way of telling compatible receivers to change from the normal 53kHz LPF to a 38kHz LPF on the decoder input and you have automatic lower noise FM stereo the receiving the new LSB only transmitters. I was thinking maybe the transmitter you could amplitude modulate the 19kHz pilot tone slightly (maybe 10%) with a low frequency sine wave (maybe 10Hz) to indicate to receivers that they were LSB only. If you don't make any change to receivers there would be no great benefit to end users apart from the fact that the broadcast signal would be narrower. A possible small benefit is that the narrower signal might mean slightly lower distortion in receivers with narrow or poor quality IF filters. It also occurs to me that even with normal DSB stereo transmissions there might be some gain to be had by making a receiver that only uses the lower sideband (and boosts it's level accordingly) since the noise level in that sideband is less than in the upper sideband. One might be able to improve the S/N of the decoded stereo by a few dB.
By Brian G on 11/9/2010
1/3 The main computational load required to implement this system is the linear-phase FIR lowpass filters needed to realize the Weaver modulator that would generate the SSB component. The L+R would also have to be delayed by the group delay of these filters. If SSB operation is required down to 0.15 Hz (which would be required to preserve the peak level of a highly processed audio waveform), then these FIR filters would be very long (they would be required to have a transition region between passband and stopband of 0.15 Hz) and would probably need to be computed by fast convolution. This implies a long coding delay, which would cause problems if talent needed to monitor off-air through headphones. (Our processors offer an "ultra-low-latency" structure that can be put on-air to address this.) I have examined the mathematics, and it turns out that one could also do a vestigial sideband realization where the upper sideband was allowed to have frequency components up to a few hundred Hz. This would substantially ease the filter requirements and would also result in less coding delay. It is possible to arrange the VSB such that the peak modulation of the composite baseband signal is still correctly controlled; the sum of the gains of corresponding upper and lower sidebands must be unity and these sideband must be-in phase, which will occur in a phase-linear realization. (Conventional DSB FM stereo is a limiting case of this, where mirror-image sidebands of 0.5 gain appear throughout the stereo subchannel spectrum.)
By Bob Orban on 11/9/2010
2/3 One other important thing to consider is how certain receivers might react to this waveform. I know that Sony in its most advanced receivers uses the quadrature component of the stereo subchannel to estimate channel noise. This noise estimate is used to control a variable-blend-to-mono algorithm in sub-bands. This exploits the fact that in conventional FM stereo, the quadrature component of the stereo subchannel is zero, so material appearing in quadrature can assumed to be due mainly to transmission channel degradation, including multipath and noise. I do not know if any other brands of receivers use this information, but if they do, it's important to know this because the SSB modulation will result in a quadrature component that is equal in magnitude to the in-phase component when demodulated with a conventional double-sideband synchronous detector. If the receiver misinterprets this, it could result in a premature blend to mono. Broadcasters need to know if implementing the new system will force a significant number of the listeners' radios into blending, which can also include HF rolloff. In addition, I believe that before transmissions using this technique are allowed to proceed, laboratory measurements of co-channel RF protection ratios should be made both under ideal conditions and with multipath. While the results may well favor the SSB technique, it is important to know this with certainty before this proposal is placed before the regulatory authority because this would affect the interference environment in the FM band. Once co-channel RF protection ratios have been measured and determined not to create worse co-channel interference than the existing system, I don't see any reason why the system can't be put on the air for tests.
By Bob Orban on 11/9/2010
Sure makes technical sense and offers several advantages to both listeners and broadcasters as well as conserve spectrum space.
By Anonymous on 10/16/2010
AS long as the new system would RETAIN STEREO in all existing FM Stereo tuners, then this is a no-brainer. Unlike C-Quam AM stereo which has no large S/N ratio penalty for stereo, FM stereo does have a huge S/N ratio noise penalty around 20dB or so for stereo, hence some stations have gone to mono FM to maintain range. I say, "let's try it out" to verify compatibility in real-world receivers, both old and new tuners and after a year of testing, make a ruling! While you're at it, let's put a minimum receiver standards ruling for AM tuners when the FM standard is d.
By Anonymous on 10/15/2010
The only problem here is that every stereo generator in existence today would have to be modified or replaced...Any guesses as to what company will be making these 'replacement stereo generators' ????
By Anonymous on 10/15/2010
Interesting, but the real solution to the present problems of analog FM is to accelerate the deployment of digital FM (HD) increasing power levels of digital while reducing analog power, giving listeners an incentive to upgrade to the superior quality and content of HD Radio. When analog is shut down, that will allow HD to broadcast at a higher bit rate and the highest possible power. It's time for broadcast industry to take a "big picture" view of the situation and understand that unless that accelerate digital radio the result will be few people listening to radio at all, as they switch to other delivery methods for content.
By Anonymous on 10/14/2010
3/3 I am in favor of any changes that will in fact benefit the listener. But when one proposes changing FCC and ITU-R rules, regulations, and standards, it is necessary to supply a great deal of data. The proposed system must be completely described technically (like, for example, the ATSC digital television system and *not* like the HD Radio system with its black-box codec) and must be thoroughly tested to determine receiver compatibility and co-channel RF protection ratios. (In my opinion, based on the RF spectra pictures in Mr. Foti's article, adjacent channel RF protection ratios should not be a problem because of the system's lower bandwidth.) I think that this is an excellent project for the NRSC and I would certainly be interested being involved with a working group to investigate this. I believe that NPR Labs has the necessary equipment, particularly the multipath simulator and other aspects of the RF path instrumentation.
By Anonymous on 11/9/2010

Post your comment

Your Name:  Required
Your Mail:       Your email will not be published.
Your Site:    

max. 800 characters


Posts are reviewed before publication, typically the next business morning. Radio World encourages multiple viewpoints, though a post will be blocked if it contains abusive language, or is repetitive or spam. Thank you for commenting!