The mighty processor. Like so many parts of the air chain it has undergone changes over the past decade; and the promised land of IP audio beckons to it. Where is the processor today and where is it going tomorrow?
Seeking answers, Radio World sought a sampling of expert manufacturer opinion. Participating were Vincent Defretin, product & project manager, Sound4; Mike Erickson, systems and support engineer, Wheatstone; Frank Foti, CEO, The Telos Alliance and founder, Omnia Audio; Peter Howarth, director of sales, special markets, AEQ; Gregory Mercier, product manager, Ecreso and HQSound, WorldCast Systems; Bob Orban, founder, Orban Labs; and Jim Wood, founder, Inovonics. Each replied individually via email; answers are excerpted below..
Peter Howarth
AEQJim Wood
InovonicsBob Orban
Orban LabsVincent Defretin
Sound4Frank Foti
The Telos AllianceMike Erickson
WheatstoneGregory Mercier
WorldCast Systems
Radio World: It seems like recent audio processor model upgrades and introductions offer nuanced improvements, some new monitoring and UI tools compared to their predecessors. Has the technology of audio processing reached a plateau of diminished returns? Put another way, is there any blood left in the “getting louder and cleaner” turnip?
Jim Wood: Whenever someone says “We’ve taken this as far as it’s possible to go,” someone else comes along with either a significant or an incremental improvement in the subject technology. Big changes come rapidly, eventually giving way to modest ones, and then, more often than not, that technology is displaced by something totally revolutionary.
Take sound recording, for example. Acoustical recording was refined until electrical processes consigned it to the junkyard. Shellac records gave way to microgroove LPs, which yielded to digital CDs, internet downloads and music collections saved in solid-state memory.
The drive for “louder” radio spawned the loudness wars of the 1960s, which sadly are still with us, much to the detriment of the listener’s enjoyment through aural fatigue factors. Certainly the historic, blatant artifacts of audio processing have been reduced or obscured, but the concept of “louder and cleaner” is a farce. Program material can always be made louder, but cleaner? No, certainly not cleaner than the unprocessed source, which is what the performer intended the listener to hear in the first place.
Mike Erickson: The marketing answer will always be yes, there is more blood. But the reality is we’re getting smarter about processing, particularly how the processor handles the differences in source material. Anyone who has been reading processing Q&As since the dawn of lossy audio has heard about the importance of linear audio. Linear audio is important, but we realize it’s not always practical. Then there’s the stark reality of source material that continues to be overmastered. Getting louder and cleaner with each new update is important for sure, but the algorithms and ideas needed to do that (especially the cleaner part) are much different than what everyone was doing 10 years ago.
At Wheatstone, our AirAura X3 processor upgrade in 2013 was the first broadcast processor to really tackle cleaning up lossy audio files to make them more presentable after aggressive processing. The algorithms we needed to do that were at the front of the processor as opposed to the back-end clipper, where much of the improvements in loudness and making the box sound “cleaner” have traditionally taken place. The same goes for our first-generation SST dynamics management and our second-generation iAGC dynamics management technology. SST and iAGC help adjust the processing based on a three-platform approach, and not just the traditional adjustments of spectral balance and amplitude. Here, the dynamic range of the source audio is also considered. These algorithms, like our approach to lossy audio management, are found in the front end of the box and that makes a huge difference.
So the short answer is, yes, there is blood left, but not in places and ways we’ve traditionally gone to maximize the turnip.
Gregory Mercier: Is there a need to increase loudness yet further? Definitely not! Is there any need to make it cleaner? In most cases, yes, although some may disagree.
Many processors, including those from WorldCast Systems, now enable you to reach very loud levels with minimum audible distortion when properly adjusted. For some formats, loudness is fine and can increase the audience but there is a point where it no longer makes sense to compete on loudness. High audio quality also includes dynamics. Our first priority should not be to always offer maximum loudness but to ensure maximum quality at each stage of the signal chain. So, while processors are a key element, the other devices through which the audio passes are also important. We need to look at the complete solution.
Bob Orban: It’s very hard to predict the future of technology. In the late 19th century some eminent scientists warned that the patent office might soon need to close because everything useful had already been invented. Despite Orban’s 40+ years in the FM processing business, innovations like our MX peak limiter and “multipath mitigator” phase corrector were developed in the last few years. While it is certainly true that “all of the easy stuff has been done,” there are certainly people working hard in our labs to significantly improve the radio experience and push the boundaries of FM processing.
Frank Foti: Ahem … Not quite so sure our latest effort known as “G-Force” fits this category of “nuanced improvements.” We basically introduced a whole new processor that loads into the existing Omnia.11 platform. End users proclaim it’s our best work yet. We introduced a novel dynamic EQ system, which further improves sonic consistency, further enhances the intermodulation-reduction methods in the dynamics processing, along with offering the Perfect Declipper as an optional purchase. This, by the way, is not an upgrade. We refer to it as a plug-in, and all for a very affordable cost.
In this fashion, an existing Omnia.11 customer does not need to shell out big dollars for an entire new box. They can add this plug-in, and basically have a new processor for pennies on the dollar.
The end result is an audible improvement on the air, as it offers additional clarity, and does not compromise loudness. Likewise, we have further improved the interfacing of Omnia.11 with the Nielsen PPM watermarking system. If you think these items are nuanced improvements, then you might also think the Cleveland Browns will win the Super Bowl!
Is there any more blood left in the “louder and cleaner” turnip? The answer is — Yes, there is! Actually, I’m working right now on some new research that deals with how the ear perceives sonic annoyance (distortion), and methods to eliminate it before it gets created. Just as Captain Kirk was always looking towards that “new frontier,” we do the same thing here at Omnia.
Vincent Defretin: Audio processing has not reached a plateau! Yes, getting louder and cleaner is the goal of each manufacturer. Just have a look at MPX power: 10 years ago it was very difficult to go higher than 8 dBr (in many cases, distortions above that were not acceptable). Then 9 dBr! And today, we reach up to 10 dBr! Of course, not all brands have reached this … Today we have such nice clipper designs, the game is moving before the clippers for more punch, more clarity and more and more sound consistency, and this is where we have a real advantage compared to our competitors! And yes, we have a lot more blood left to put into audio processing.
Peter Howarth: We have a feeling that we are at a point where the technology is adapting to the transmission channel — internet or FM/AM. All advancements so far have been related to the FM/AM transmission. Now internet is the big thing and algorithms and processing is focused on this media and trying for all it can to sound as good as FM — subjectively.
Radio World: Are the less-expensive “lite” processor models offered by many manufacturers achieving comparable quality and loudness as their more expensive high-end big brothers? What does all the extra money for the high-end boxes really buy the user?
Orban: Orban’s “lite” processor, the Optimod 5500, performs better than Orban’s top-of-the-line 8200 from 20 years ago. Nevertheless, the 5500 cannot compete with the MX limiter technology in our higher-end 8600 and 8700 processors, which offer lower distortion, greater transient impact, and up to 3 dB more high frequency power handling capacity compared to the 5500.
Howarth: Certainly not. The “extra money” gives you flexibility and processing power to customize your signature sound and to dynamically change with your program content and hours. You sound as good as your budget — more or less. Another thing is that the gaps between the extremes are really closing in. That’s the way technology goes.
Foti: It’s always about competitive audio quality. Always has, and always will be! To date, I have not heard a “lite” box that can stand up on the dial with an Omnia.9 or an Omnia.11. Any broadcaster who wants to sound the best on-the-air will achieve that performance with one of our top-of-the-line products.
Defretin: This is a real question. In some cases, both big and affordable models use the same final clipper stages, so in fact, again, everything comes from the stages before clippers. Compared to big models, there are fewer processing stages and less processing power in less expensive processors. The result is not bad, but sound consistency and sound signature are not the same thing!
Mercier: It’s hard to answer for all brands, but in general, “lite” processors provide less processing power, quality, loudness and/or customization settings. One of those criteria, or several combined, can make a significant difference. Additional features may make a processor more expensive when compared to a “lite” version but they can save the broadcaster money or time elsewhere in terms of monitoring, redundancy and reliability.
Erickson: Low-cost processors are (or should be) designed with a few things in mind. The first is that the person adjusting it may not have the time or processing know-how available to them. So the controls have to make sense and the interface cannot be overwhelming. It has to be easy to set up, put on the air and use. Wheatstone’s FM-55 is that. But it also has to actually work and be competitive. It has to have that no-nonsense, great sound. Some low-cost processors are stripped-down versions of bigger boxes that are purposely designed not to be as competitive as their bigger brothers. We thought there had to be a better way. When we designed the AM/FM 55 series, we built these new budget-priced boxes from the ground up so they could stand on their own merits at their price point. Our expensive boxes add obvious features like a second audio processing path for HD that helps users manage the HD codec, but you also get plenty of extra tools in a bigger box. If you look at Wheatstone, our bigger boxes add visual and audible tools like audio analysis displays not found in other big boxes as well as extra refinement of our iAGC dynamics and lossy audio source management systems I mentioned, plus the ability to really customize bass and stereo enhancement along with extra controls for the tweakers who really want (and maybe need) to dig in!
Wood: From a hardware standpoint, audio processing has become a relatively simple matter. The onus of squeezing program dynamics without destroying program quality has moved from FETs and op-amps to strings of numbers. It’s cheaper than ever to manipulate the audio program without regard to circuit board real estate and the cost of esoteric analog parts. Digital processing algorithms may be stacked end-to-end without introducing additional noise or, in most cases, additional distortion.
And because sources for digital signal processing chips has multiplied, and the cost of the parts plummeted, the manufacturing cost difference between a “lite” audio processor and its high-end big brother is more about how fancy it looks, and how many features want to be excluded from the lesser offering for marketing considerations. (You sure don’t want to kill sales of the high-end model!) So, yes, “baby” processors can be made just as functional as their more expensive siblings. Spending really big bucks may give the broadcaster peace of mind, thinking that expense equates to quality, but a Honda will take you from Omaha to Minneapolis as quickly, and even perhaps in greater comfort, as a Ferrari. It’s just that you may not be perceived as “competitive” by your broadcaster peers. Listeners won’t know the difference.
Radio World: Whatever happened to the SSB FM option once highly touted by Omnia and also offered in Orban Optimods?
Wood: What happened to Dolby FM, Quadraphonic FM and FMX? How about 3D television, for that matter? Ideas and proposals may have merit from a technical or a Wow! factor standpoint, but the ones that quickly lose momentum are generally a “last hurrah” in a technology that is rapidly evolving or is being replaced by the next generation. All-digital radio and the specter of the “connected car” eclipses a marginal and controversial improvement in FM’s noise and multipath artifacts.
Foti: Well, the NRSC basically decided the SSB system does not degrade the FM broadcast system. In the end, they took a neutral stance on the topic. So, the ball is actually in our court. Meaning, for an effective and optional FCC rule to be changed, it’s now on us to take it up with the FCC. It’s on our To-Do list. We are willing to lead in an area that will help FM broadcasters worldwide, even at our expense, much like we did with AES67, MP3, and many other technologies.
Defretin: We have also been doing SSB for two years. SSB is a good approach, our experience with current customers show better results in some cases and no results in others. It depends mostly on whether problems are coming from multipath, bad coverage or frequency disturbed by others… In the case of multipath, it can solve reception problems, but a new problem may occur: some receivers stay in Mono when they receive an SSB signal! The final question is: Is it better to have a few listeners in Mono and more in stereo in a better condition??? This is the choice of radios!
Orban: The careful lab work done by John Kean at NPR Labs under contract from the NRSC revealed that SSB had no advantages over conventional DSB in distortion or signal-to-noise ratio, and usually made things worse with current radios. Additionally, it revealed that some existing FM radios were completely incompatible and would blend to mono when receiving an SSB signal. The lab results suggested that anecdotes touting SSB’s alleged multipath improvements were probably caused by the radios’ blending more toward mono when receiving SSB, giving an illusory advantage. The only potential advantage of SSB could come from purpose-built radios that ignore the upper sideband, unlike all current radios. These new radios could achieve a few dB improvement in signal-to-noise ratio compared to DSB transmission.
Erickson: In our tests, the cure was worse than the disease and I think, after the initial marketing dust settled, people who used the technology have come around to seeing that there is no free lunch. It may work in some cases, on some receivers, but our tests of 100 radios from headphone radios to car stereos to boom boxes and home stereo tuners found that nearly 40 percent of radios we tested in SSB exhibited either severe blending, severe blending with high frequency roll-off, or increased distortion. This was particularly true with Sony aftermarket HD car receivers and HD receivers in GM and Toyota vehicles. In one demo I did late in 2015, the PD had heard about SSB and wanted to try it. How disappointed he was when he heard what it sounded like in his GMC. On the other hand, there is and has been a better way of reducing multipath. I showed him Wheatstone’s multipath controller, which started life in DSP form in our AP2000 processor and has been refined over the years (based on user feedback) to be sentient of user-defined stereo enhancement in all of our FM processors. The system helps reduce the annoying blend in car stereos and has no adverse effect on the behavior of receivers. While the multipath controller is available in all of our FM processors, our bigger boxes also have the ability to change our MPX processor from an oversampled clipper to a high-speed lookahead limiter. This can further reduce reception issues in extreme cases without sacrificing loudness.
Radio World: Is the day of the hardware processor, the single-application box, coming to an end, to be replaced by processing cards in the transmitter, soft processors inhabiting PCs or an app based in a section module on an IP network?
Foti: Good question. The shift has begun towards locating the processing inside a transmitter, and at some point it could become cloud-based. But, with just about all technology, there will always be those whose comfort level will be having a “box” mounted in a rack somewhere. Omnia Audio will always support that platform. In fact, we will support all platforms that our customers feel are important.
Mercier: I am not a fan of hardware integration, especially inside the transmitter. Given that, in WorldCast Systems, we have the three brands, APT, Ecreso and Audemat, it would have been easy for us to combine all our broadcasting expertise into a single device. However, our core priority in transmitter design is to ensure that the Ecreso FM range remains as simple, robust and as easy to service as possible. For small to medium-sized stations, pricing is also an issue. So, overall, moving a processing board from a standalone device to a transmitter has no significant benefits. On the other hand, software integration, when possible, can make sense both in term of design and savings.
Erickson: I think, in the United States, you’ll start to see that coming in smaller markets and with processing enthusiasts, and it will work its way up to medium and larger markets. We are flooded with all different kinds of PC-based processing right now that is very inexpensive. Some of it is quite good and might be suitable for people who want to tinker with DIY processing and who have the know-how and the PC hardware and sound cards needed to properly use this software in place of a traditional hardware box. With all that said, the vast majority of customers are still looking for that purpose-built hardware box for AM and FM and we are fortunate enough at Wheatstone to control our own destiny in that respect. Doing everything in-house, working with our processing team and our own algorithms, we can design those algorithms to work on whatever platform the market decides it wants, without having to worry about the whims of third-party designers. Right now we are focused on our hardware boxes because that’s what is in highest demand.
Howarth: This is coming more and more and is natural for the internet radio and TV stations. Again this is related to budget. For the “traditional” TX chain we don’t see all the transmitter manufacturers being interested in adding modularity to their products if they can avoid this. In essence, modularity is adding cost and that results in a higher list price. The soft processors or “apps” on the IP network may be a future product, but the reason for using dedicated hardware is because you seek reliability so for the time being, we believe that critical operations will continue relying on dedicated hardware solutions rather than risking the “blue screen — I am rebooting.”
Wood: Certainly the transportation and delivery of the broadcast signal is undergoing changes on all levels as we migrate from an analog to a digital broadcasting world. As Bob Orban taught us decades ago, modulation efficiency (loudness) initially demanded the integration of audio processing with the generation of the stereo multiplex signal. Thanks in large part to fallout from the digital revolution, we now are able to use a more “distributed” approach to the program signal. This, and the fact that multiple delivery channels (AM/FM/HD Radio/streaming) each require slightly different treatment, means that we will see more bits and pieces, here and there.
Defretin: I’m not sure if it is the end of box-processors but yes, for sure this is the end of single-application boxes for three reasons: First, customers are looking for less expensive chains, so getting a multiple-application box is the right answer. Second, today we can create versatile platforms where we can host more and more applications. Sound4 was the inventor of multi-application products in this field when we launched our PCI and later PCIe cards Year after year, we have been adding new applications (RDS, IP codec, streaming backup player, SNMP …). Third, it takes up less space in racks. This is a real demand from the market.
The market is slowly moving to new architectures. Often there is a PC inside the chain and stability depends on hardware quality and OS. Yes, some manufacturers hide a consumer PC inside their boxes and this is a shame. At Sound4, we are thinking differently: for all PCIe products, audio never passes inside the PC — the OS can crash or reboot with no effect. A great advantage is also the boot time — at Sound4, we are on air in 2.5 seconds, while some others may need up to a minute. For the new in-box products, we don’t use any PC or PCIe cards. We have created dedicated hardware made in a robust way and it will always be versatile for upcoming future applications.
Radio World: Conversely, does the direct digital multiplex give new purpose to the box right before the transmitter?
Orban: Putting the processor next to the transmitter and connecting to the transmitter via a 192 kHz direct digital MPX link makes most efficient use of the bandwidth of existing STLs, as this placement allows the STL to carry only two channels of 44.1 or 48 kHz audio as opposed to one channel of 192 kHz audio. If the STL bandwidth for 192 kHz is available, then it may prove to be more convenient to locate the processor at the studio. Because of the well-developed PC remote control applications available for current digital processors, there is little downside to locating the processor at the transmitter unless it is necessary to perform maintenance on the processor.
Defretin: It is interesting because of the full digital chain up to the exciter, but in fact, do listeners hear the difference?
Erickson: Every processing manufacturer has always stressed the importance of having the processor at the transmitter site. Being able to take advantage of new technology like digital multiplex and automatic HD time-alignment reinforces that opinion. But each station is unique, and having the processor at the transmitter is not always possible. In those cases, we work with customers to maximize what benefits they can get from their processor, even if it needs to be placed at the studio end.
Foti: Not really. Digital multiplex actually enables a lot of flexibility for processor location. Our µMPX is a perfect example, as it enables a processor to be located at a network head-end, and then distribute composite over a wide area network, while using a lower data payload.
Radio World: Most audio processing companies have decried the amount of heavy processing and clipping being applied to new music releases in post-production by record companies. Is there any evidence this practice is abating and if not, what if anything can be done to combat that with broadcast processing techniques? Are concepts like “unprocessing,” “unclipping” and other so-called distortion removal schemes effective or are they marketing ploys?
Wood: It is a sad state of affairs that the “loudness wars” migrated upstream to the source of our entertainment. The same mentality that prompted paranoid program directors to demand a louder sound now permeates the recording industry. CD “authoring” and mastering labs advertise their ability to cram more and more level into a mix, and the performers are buying into it.
With a linear companding system, level compression can be “undone” by complementary level expansion. Complementary companding was the basis for Dolby and dbx noise-reduction systems used for analog recording. But it is near-impossible to reverse audio compression and remove distortion without knowing exactly what was done in the first place. The best one can do is to introduce even more processing artifacts that may, or may not, be perceived as some form of “sonic restoration.”
Defretin: On paper, it sounds really interesting. Today’s processors which do this do a sound expansion in a detected clipping zone and to be efficient, they need a long latency … During shows, manufacturers demonstrate this in operating mode or in bypass. They never compensate levels to reach the same peaks as the input, so everybody is impressed … Do a test: Compare with the exact peak level, and improvements will be hearable in 5 percent of the cases! So yes, for us it is more for marketing purposes than for reaching a real improvement, and on our side, we have some other ways of waking up such clipped sources.
Erickson: Unfortunately, it is not abating. The challenge of dealing with overmastered material is why at Wheatstone have spent so much time developing and refining a front-end-like iAGC that recognizes this type of material and reacts appropriately. Any type of clipping restoration or repairing of overmastered audio is best done on a cut-by-cut basis when you’re ingesting audio into your playout system. We can fix a lot more today than we could 20 years ago, but developing good standards at the station will always give you an edge in the end. I’ll run into situations where levels are all over the place from song to song and spot to spot. Even the best AGCs are not going to sound smooth when you have a 20 dB-level difference between cuts. Setting standards for ingesting audio at proper levels, using proper sources and adding clip restoration when needed should be paramount. It’s the foundation to your signature sound!
Foti: Hypercompression still exists, and it’s random, based on the artist and producer. But, it seems the only processing company that has done anything about this is Omnia. We are the only ones to offer “Undo” and “Perfect Declipper,” which restores audio quality to source audio at the input of the audio processor.
The term “marketing ploy” regarding this topic only seems to emanate out of our competitors. Since they don’t have any tech like this, they have to resort to FUD (Fear Uncertainty Doubt) as a way to discredit our innovative efforts. The proof is in the details. All a user has to do is switch the Perfect Declipper on, and you can hear the improvement in audio quality. It’s as simple as that.
Orban: Declipping is a complicated issue. Information is 100 percent lost in flat-topped areas and cannot be recovered. A flat-topped waveform is a mathematical “singularity.” Declippers must make educated guesses about what’s missing based on interpolation from material surrounding the clipped samples. The interpolation must use a model of the clipping process. But many waveforms that look they have been hard-clipped have, in fact, been peak-limited by more complex limiting processes with sidechains and memory. Each limiter manufacturer has a proprietary way of computing the sidechain. For competitive reasons, these are seldom made public. Even if the sidechain is public knowledge, if the compression ratio is infinite, it is still impossible to deduce what the limiter’s input was.
Declippers can increase punch on transients by increasing peak levels by guessing what the missing waveform is. But this is not the same as cancelling IM distortion. Distortion cancellation depends on having a precise, invertible model of the peak limiting process. This is usually impossible.
Because declipping is a nonlinear process, it can make its own IM distortion that adds to any IM distortion present in the original track. The better the original peak limiting algorithm, the more likely it is that declipping will add IM distortion, not cancel it. If simple peak clipping was used on a given track, then declippers can help. But sometimes they make things worse. Therefore, the proper place for a declipper is in the production studio, so that human ears can determine if the declipper is helping or adding another layer of distortion. It is unwise to put it in the on-air processing chain and just “let it rip.” Moreover, in the broadcast processing chain, declipped waveforms force the on-air processor’s peak limiter to work harder. So use declippers with care!
Howarth: This is the fashion and the purist discussion, I suppose. Honestly, if someone spent a fortune on post-production in a recording studio, it is probably because this is the sound that they believe their product should have. If subjectively this sounds “crap” then, is it correct to try to alter that sound in for example an FM transmission? Is your signature sound of your station more important than the artist/composer’s piece of work? In any event this is probably a discussion that we should not have as manufacturers. It’s like shooting yourself in the foot. If the practice is subsiding, I am sure we will find other ways of treating the signal than trying to subjectively restore something that many hours was spent on creating.
Your thoughts? What questions would you want audio processing makers to answer? Email [email protected]with “Letter to the Editor: Processing” in the subject line.
