Telos/Omnia Disputes Criticism of Studio Costs And Questions Statements by Neural Audio
Continuing a debate over the best way to accomplish radio surround sound, this is in response to a Guest Commentary authored by Neural Audio co-founder/CTO Robert Reams in the April 13 issue.
At risk of giving the Neural surround system more attention than it deserves, we must reply to the statements the company made in its guest commentary. To not do so would leave a lot of incorrect information standing.
Neural seems to be under the mistaken impression that we, Telos/Omnia, developed the MPEG surround system we are supporting. Not so.
The inventors are Fraunhofer Institute for Integrated Circuits, a German publicly owned research laboratory with 450 staff and years of experience in digital audio. Fraunhofer developed both MPEG Layer 3 (MP3) and most of MPEG AAC. FhG IIS (the German acronym for Fraunhofer) worked with Agere Systems in the United States, which absorbed many of the audio researchers formerly associated with Bell Labs. There are significant contributions from Coding Technologies, who supplied the HD codec, and Philips, the consumer electronics company.
We are not paid by them or anyone else to promote the MPEG system. If surround takes off, we expect to make and sell encoders, processors and studio equipment – but because the system would be a standard, so would any number of vendors.
If we weren’t promoting this system, who would?
Third-party testing
As a supplier to broadcasters, we do have an interest in keeping our industry healthy. We think surround is a great application for HDFM, and would be good for the future of radio broadcasting. That’s why we are working to increase the odds that our industry gets it right.
As a sole source of a proprietary product, Neural’s position is plainly different. They react defensively to our call for disclosure and testing by pointing to a few promo broadcasts as if they were somehow a significant test.
We believe disclosure and unbiased third-party comparative testing are the way to go. Remember how Ibiquity endured plenty of careful testing before the NRSC endorsed and the FCC approved its system? That process proved quite beneficial. Important flaws were caught and improvements made, such as the codec upgrade that very much enhanced audio quality.
Here are our responses to some specific Neural statements:
“The ‘mechanical downmix’ is something that has to happen in all spatial coding, including Fraunhofer, Agere, Philips and Coding Technologies.”
The Neural guys simply don’t understand, and have got this point completely wrong. The MPEG system is able to pass an original unaltered “artistic” stereo mix to stereo receivers. There is no downmixing or any other alteration of any kind to the stereo signal.
Matrix systems, including Neural’s, must impose a fixed-downmix. They also must apply phase shifts. As we learned in the ’70s, matrixed surround often plays quite weirdly on stereo or mono receivers. As would, indeed, Neural-encoded material.
“Neural has been working with MPEG members for years.”
Stefan Geyersberger, FhG IIS intellectual property licensing and business development director, replies, “We’ve been working within MPEG audio from the start. It is most irritating to read that lot of nonsense about MPEG: That MPEG knows nothing about encoding (!), that MPEG is a consumer organization, and, most absurd, that Neural was ‘talking to MPEG’ and is an MPEG member. I just checked with our current representative, who has been in all but one meeting since summer 1996. He says he knows all regular MPEG audio participants, and Neural has not been among them.”
“If you were to actually talk to MPEG, as we actually do, Neural Audio is an MPEG member, you would find that they (MPEG) don’t have much to say about encoding standards (like on the broadcast end). They only recommend technology on the consumer or decoding end.”
Neural’s claim that MPEG only cares about the consumer end really is nonsense. Of course, you must have both ends for any producer-to-consumer chain to work.
And MPEG standards are most certainly used professionally, with MPEG ISDN codecs the most obvious example.
“Only recently has it been appropriate to disclose the technology to consumer organizations (like MPEG).”
As of the time of this writing, there has been no Neural disclosure to MPEG. There were five entrants into MPEG’s surround coding testing, and Neural was not among them.
“The Neural technology has been ‘on the air’ for quite some time now and has been proven to be stable, reliable and practical.”
This claim is quite a stretch. We don’t know of any HDFM station regularly broadcasting with the Neural scheme. The few promo broadcasts have been with jazz and classical music, not the pop and rock programming that dominate on the FM band. Jazz and classical producers usually use surround to recover room ambience, while pop and rock production techniques involve more aggressive mixes with lots of channel separation, a tougher case for matrix.
“Somehow, Telos/Omnia continues to cling to the dangerous assumption that bitstream-based technologies are compatible with the day-to-day operations of a real, live broadcast facility.”
Both of us are former CEs and we have plenty of experience with “real, live broadcast facilities.” We now head a company that prospers only if real engineers decide to use our goods within real radio stations.
We note with a bit of wistfulness that we have between us over 60 years of real radio experience. (We started young!)
And there are another dozen ex-broadcasters among the Telos/Omnia crew. (One among them laughed that Neural’s chastising us for not knowing broadcasting was akin to the Rutles, the so-called Pre-Fab Four, criticizing the Beatles.) Seriously, we do have our feet pretty firmly planted on the ground and wouldn’t be proposing anything that wasn’t truly practical.
“I would bet all of the Radio World readers would love to see a detailed cost accounting showing how there is ‘near zero incremental cost differential’ in building a discrete six-channel plant over a conventional stereo one.”
Steve has given a paper on exactly this topic at the recent NAB show titled “Studio Structures for Surround Broadcasting.” This forum is not the place to discuss this topic in detail, so we will direct those interested to www.axiaaudio.com/surround for an explanation.
Indeed, it could well cost more to use the Neural approach, as our Web article details.
“I am also sure that the folks at NPR would be fascinated in Mr. Church and Mr. Foti’s proposal for shipping a discrete six-channel bitstream though the PRSS satellite distribution system to all of the NPR member stations for ‘near zero additional cost.'”
Certainly there is no reason to hobble broadcasting forever just to accommodate a particular existing distribution system. On this logic, FM would not today have 15 kHz fidelity or stereo, since the mono telephone lines that radio networks used back when for distribution only went to 5 kHz.
Regarding our concern that the Neural system has not been the subject of the kind of careful testing that MPEG puts its proponents through, Neural stated:
“Mr. Pappas has authored a painfully detailed account of validating the performance of the Neural system. The article appeared in a recent RW supplement about the ‘Toast of the Nation’ broadcast.”
We hope Radio World’s readers recognized that the supplement was an “advertorial” insert, produced by the advertising department in cooperation with the participating vendors. Mr. Pappas and the NPR guys were right to be thrilled with having pulled off such an ambitious project.
But if there was something like a “validation test” described there, we missed it. (And citing one’s own advertising as being indicative of such a test is pushing the credibility envelope a bit far, no?)
The article described no comparisons to the original, which would have been difficult owing to the broadcasts being live remotes.
We expressed skepticism regarding Neural’s claims involving “watermarking” and remain doubtful. A previous Radio World article had quoted a rate of 16 kbps, and we reacted by observing that this was unlikely, and further said that, “Neural’s secrecy is a barrier to making a valid assessment of their quite outrageous claims.”
To which they replied:
“Neural’s claims are not outrageous to those skilled in the art.”
It was Fraunhofer’s experts who alerted us to the improbability of Neural’s claim. They are most certainly “skilled in the art” as they’ve done research in the area of watermarking for a decade, have a number of published papers to their credit, and have developed and licensed numerous watermarking technologies. The couple hundred bits per second that watermarking could reliably provide would not be particularly useful to solving the problems inherent in the matrix approach.
“Talk to the radio stations, guys; 16 kbps can produce much higher quality than a cell phone. There are many speech and data services that generate far more revenue than the promise of surround ‘someday.'”
We think surround is worth some dedicated bits in the broadcast stream. Acura is running full-page ads touting surround sound in its cars – from DVDs, not radio. In the same ads, the company is promoting satellite. Shouldn’t we broadcasters be concerned?
On the other hand, just where are the “many speech and data services” that “generate far more revenue”? Those seem to us to be more “someday” than a here-and-now enhancement of our main program service.
Anyway, it turns out that we don’t need 16 kbps. Following the MPEG meeting April 18-22 in Korea, it has emerged that performance is very good down to 6 kbps. Isn’t providing a high-quality compatible surround service worth that?
“Yes, there is secrecy. All broadcast and receiver manufacturers/OEMs work with Neural under strict non-disclosure agreements. With an NDA in place there is no barrier to disclosure and assessment.”
So, Neural is saying it will explain how their system works only to those who are not allowed to talk about it. How convenient! An opportunity for scrutiny must be afforded to those who actually are motivated and allowed to scrutinize. We’re talking about the future of broadcasting here!
“Telos/Omnia has received countless invitations to visit the Neural facility and experience Neural’s surround technology. They apparently refuse. And: Telos/Omnia has had a long-standing invitation to come to Neural to enjoy surround and salmon.”
(click thumbnail)Fig. 1 The test setup.
(click thumbnail)Fig. 2 Roxy Music ‘More Than This’ ITU downmix.
(click thumbnail)Roxy Music ‘More Than This’ Neural 5225 Lt/Rt output.
Were Neural’s salmon to give us as much indigestion as their claims, we might better pass! Seriously, what do they think – that an in-person pitch would wipe away our doubts and satisfy our call for disclosure and testing? Sorry, but a fishy meeting is hardly the point.
“Stop wasting time. Get into the arena and make your system function in the real world. Compete. Prove that it works in a real (from mic to antenna) broadcast environment.”
About this, we agree. At NAB2005 in Las Vegas, we demonstrated a complete, practical chain from studio to consumer receiver. We also had live broadcasts to a prototype Bose Cabin Surround audio system installed in an Acura MDX SUV.
Yes, it took some time, but we are ready now to begin real on-air broadcasts. And we, indeed, look very much forward to competing.
Yet again in their RW commentary Neural spilled a lot of words, but didn’t offer any useful information as to how the system works. They also did not dispute the superior performance of the MPEG system, because they know they’d lose that argument.
We have a Harris Neustar 5225 encoder/decoder which uses Neural’s surround sound technology in our lab and have been listening to its effect on various kinds of source audio. We can say with confidence that you and your PD are not going to like what this box produces on your station’s air.
In particular, there is quite a lot of general muddying and softening of impact, particularly in the bass. There is a pronounced change to the stereo image, with voices and instruments warbling and wandering around quite strangely at times. There is a “fake stereo” effect on mono material such as announcer voice.
We’ve also been making some measurements. See Fig. 1 for the test set-up. We ran the surround version of Roxy Music’s “More Than This” though a simple fixed downmix according to the ITU-R BS-775-1 recommendation (Fig. 2). Then we ran it through the 5225 (Fig. 3).
Notice the significant change in the L-R in the Neural downmix. There is a large 90-degree phase shift, which would sound like a hole in the center of the soundstage. The increased L-R level will also exaggerate multipath effects on the analog FM and cause the station dynamics processor to inflict more clipping distortion.
With matrix systems, there is always the risk of L-R manipulations causing problems. We learned this lesson in the 1970s and again a few years ago as stations struggled with “Q-Sound” recordings. We are convinced that Neural and other matrix systems will be unacceptable to stations with regard to both surround separation and stereo/mono compatibility.
The MPEG alternative is just the thing for the HDFM application, offering excellent performance and perfect compatibility. Neural says that it would be impractical to upgrade radio studios to discrete surround. We counter that modern networking techniques offer a simple, elegant, low-cost solution.
And we say again, radio broadcasting needs better than re-hashed matrix schemes to effectively compete in the digital era.
RW welcomes other points of view to radioworld@imaspub.com.