DSPs go forth and multiply

Even as digital signal processors get faster, chipmakers look for new ways to increase performance, enabling new AV applications in the process. Tim Kridel reports.

As the world goes digital, and AV along with it, digital signal processors (DSPs) are becoming an increasingly powerful tool for integrators. One big draw is the ability to do more with less, compared to old-school analogue technology.

“What digital signal processing has done to the world of audio is made all of the functions of the analog tools that you’d need to tune a room or to sweeten a sound [doable] in software,” says Glenn Polly, owner of VideoSonic, a New York City-based integrator. “One of the first things we noticed with DSP processors five or six years ago is that the equipment disappeared. You basically didn’t need a rack full of outboard audio gear any more. That enabled us to be able to deliver a system without having to go back to the client and say, ‘This system would benefit from adding this extra piece of audio equipment.’”

The ability to deliver good sound on a tight budget is always a plus, but it’s even more of a competitive advantage today, when a sluggish economy has clients focusing on price more than ever. One example is an amplifier with an integrated DSP.

“We can do a great-sounding, two-zone, mono system for a restaurant or a hair salon with an amplifier and an iPod dock if the amplifier has a DSP,” Polly says. “In the economy we’re in today, it’s a lot harder to sell systems now than six months or a year ago. They want to spend the bare minimum. So if we’re able to come up with a solution that gives them quality sound yet saves them money, they’re going to understand that.”

Not surprisingly, AV vendors are adding DSPs to more of their products, including digital mixers, speaker processors and effect processors. One example is Yamaha Commercial Audio Systems, which will begin shipping a power amp with integrated DSP this summer because customers have been asking for them.

“It’s a design philosophy on the audio consultant’s part: Some people believe in having discrete components, and some believe in processing per amplifier,” says Marc Lopez, marketing manager at Yamaha Commercial Audio Systems.

One consideration is cost.

“Each amplifier costs more than having a bunch of DSP in front of a bunch of ‘dumb’ amps,” Lopez says. “But in that scenario, you’re relying on one large DSP unit or several DSP units ahead of the amps rather than a distributed DSP system. In the case of having DSP on each amp, you only lose one amp or two channels. You don’t lose clusters at a time. It depends on the design philosophy and cost.”

Some integrators say that DSP-based products are still a bit pricey, even in light of the value and savings that they can provide.

“There are processors in the low $500 [€315] area,” Polly says. “We’d like to see the price of DSP processors come down.”

Value also depends on the ability of getting them to do everything that they’re capable of. For example, although DSPs are digital at heart, a solid background in analog can be helpful in terms of maximizing all of their features.

“You need a good analog audio background and a little computer savvy to use them,” Polly says. “If you don’t know what to listen for, you won’t know what to select. If you’re aware of what all the analog audio tools did, you can make the best use of a DSP processor. I think that’s really important.”

DSP 101

All DSPs are microprocessors that are designed to crunch data in real time – in other words, without the delays that would be noticeable to end users.

“One thing we’ve done recently is support stereo acoustic echo cancellation, which is a very complex problem to solve,” says Craig Richardson, vice president and general manager of the Installed Audio division at Polycom. “You’ve got a couple of independent audio sources coming to the room, creating echoes, and you can’t do that with a single channel or mono echo canceller. The processing power for that goes up by a factor of four to support the full stereo processing. So having higher performance chips enables that kind of solution. Every time the DSP manufacturers leap forward, there’s no shortage of ideas for using those [new capabilities].”

That speed is why DSPs are widely used in a variety of devices outside of commercial and pro AV, such as cell phones and the networks that they connect to. All of those applications across all of those industries increase DSP volumes and in turn help drive down their prices.

“Everybody is using commercially available DSP processors just because of the price point,” says Richardson, whose company has been using Texas Instruments DSPs for more than 20 years.

Many consumer AV products use DSPs. That’s noteworthy because consumer is a larger market than pro, so it’s not uncommon for chipmakers to develop DSPs for consumer applications and then adapt them for the pro market.

“I think we benefit because somewhere, there is one or more reasonably high-volume applications – such as $700 AV receivers – that has some of the same needs that pro AV products have,” Richardson says. “I don’t think that Texas Instruments thinks, ‘Gee, what do the pro AV guys really need?’ Generally the big push is coming from other applications.”

DSPs aren’t the only components that can perform many of these tasks. Another type is field-programmable gate arrays (FPGAs), which the AV vendor can buy off the shelf from a chipmaker and then program to perform certain tasks. One downside of FPGAs is cost.

“Generally speaking, those are relatively more expensive to build and program than using an off-the-shelf DSP,” says Yamaha’s Lopez.

But FPGAs have their place. For example, they can work alongside a DSP to perform certain tasks in order to reduce the DSP’s workload or to handle specialized functions that the DSP can’t.

The core problem

In nearly every industry that uses DSPs, the trend is toward multicore chips. Here’s why: As applications become more demanding, each generation of DSPs must run faster in order to keep up. But simply cranking up the megahertz eventually starts to become inefficient and even problematic because it increases both the amount of electrical power required and the heat produced.

“In pro audio, power consumption is a concern, even for non-battery-powered applications,” says Sujata Neidig, product manager for Symphony audio DSP products at Freescale Semiconductor.

Heat is a problem because if it reaches a certain point, the AV vendor will have to add a fan to the product. That can put the product at a competitive disadvantage if it increases its cost or if it’s loud enough that some integrators or end users opt for a rival’s quieter product. A fan also has moving parts, which eventually will fail.

“Our customers don’t want to have to have fans in the boxes,” Neidig says.

Heat also increases power usage.

“When the temperature increases, the current goes up, as well,” says John Scott, applications engineer for Symphony audio DSP products at Freescale. “So it’s good to have the DSPs running as cool as possible.”

All of that creates challenges for DSP vendors and for AV vendors that make their own DSPs, such as Yamaha.

“It’s a juggling act,” Lopez says. “The challenge is to create more processing power with less power consumption and therefore less heat.”

A multicore design gets around those problems by dividing the DSP into multiple processor or “cores.” For example, instead of building a single-core DSP running at 3 GHz, a chipmaker could opt for a DSP with two cores, each running at 1.5 GHz. The net result is still 3 GHz worth of performance, but with less power and heat because each core is running at only 1.5 GHz.

More than the core

As more chipmakers go with multicore designs, they need to find new ways to make their products stand out from the pack. One common way is performance, which isn’t determined solely by the number and speed of the cores.

For example, performance also depends on the amount of memory available to the cores and the speed of the connection between each core and the memory. One analogy is a PC whose fast processor is hamstrung by a limited amount of memory.

“It doesn’t do any good to speed up the core if the memory can’t keep up,” says Freescale’s Neidig.

Another performance factor is the way that the vendor divvies up the workload. For example, one core could run decoders such as Dolby Digital Plus, with another doing post processing on those streams.

All of those factors are important to AV vendors because the vast majority buy their DSPs from chipmakers such as Analog Devices, Freescale and Texas Instruments. One of the few exceptions is Yamaha, which has its own semiconductor division.

“There are certain things that our engineers can specify to the chip manufacturing side,” Lopez says. “That’s a very strong advantage, that level of communication and specification.”

A major advantage of buying DSPs off the shelf is that the AV vendor doesn’t have the cost of developing them, and those savings put the vendor in a better position to price its products competitively yet profitably. The AV vendor also can use those savings to fund development of features – such as room correction – that it layers on top of the same basic DSP that the chipmaker is supplying to some of the vendor’s competitors.

Besides the DSP itself, chipmakers also typically offer their customers a library of software for basic functions such as decoders. The end result is that by outsourcing the DSP and some of the basic software, AV vendors can reduce time to market – a major plus in both the pro and consumer markets.

“In order to keep up, they need products that allow them to have quick time to market and ease of development,” Neidig says. “All of that applies to the pro audio world.”

Scaling up

The overlap between pro and consumer benefits chipmakers that target both markets, as well as their AV vendor customers. One reason is that the consumer market is much larger, so if a DSP can be sold into both the pro and consumer markets – as is sometimes the case – consumer’s volumes help drive down costs faster than if the DSP went only into the pro space.

“One of the advantages is economies of scale,” says Freescale’s Scott. “The costs will go down the more DSPs we can sell.”

Scale also is a factor that AV vendors have to consider when deciding whether to make their own DSPs.

“Most audio manufacturers don’t have the scale of business to have that capability,” says Yamaha’s Lopez.


Tim Kridel

Tim Kridel is a free-lance writer who covers technology and telecom for a variety of publications and analyst firms. He’s based in the U.S. city of Columbia, Mo., and can be reached at inavate@imlgroup.co.uk

Article Categories