PxPixel
Semiconductors in Audio Part I - ProSoundNetwork.com

Semiconductors in Audio Part I

For our February Special Report: Semiconductors in Audio, Pro Sound News asked a number of leading design engineers a number of questions about modern solid state components and their application.
Author:
Publish date:

For our February Special Report: Semiconductors in Audio, Pro Sound News asked a number of leading design engineers a number of questions about modern solid state components and their application. Our dialog was with Lucas van der Mee, Chris O'Malley and Kevin Vanwulpen from Apogee Electronics, Ian Dennis from Prism Media, Bruce Jackson and David McGrath from Lake Technology, Daniel Weiss from Weiss Electronics, John Petrucelli of Crest Audio, Jim McTigue from Euphonix, Mark Brunner from Shure, Alan Kraemer from SRS Labs, Elon Coats from Peavey Electronics and George Massenburg from GML/Massenburg Designworks. A full transcript of their replies follows.

1. Has your component selection changed significantly over the past couple of years when choosing semiconductors for analog design? Have new parts brought new capabilities to your designs (faster, quieter, cheaper, more modularity requiring fewer external components, etc...)?

van der Mee: Not really, several new parts have been released for analog application, but most of them are not very applicable for pro-audio. The majority of new analog parts is more focused on lower power applications and the majority is single rail only. Efficiency is the hot topic. Which makes total sense for the ever-increasing industry of portable equipment like cell phones, mp3 players, etc. But for high quality audio applications, where dual power rails of at least +/- 15 volt are a bare minimum, they are of no use at all.

When it comes to op-amps I am pretty much using the same ones as I did five years ago, but have become better in blending the different flavors to get the end result I am after. As a matter of fact the choice has even become smaller, as one of my favorite op-amps has been discontinued. We have about 80000 of them in stock, but when were through those, it is over...I still haven't found a part that comes even close to this one, seen from a price/performance point of view.

Dennis: It is rare that an advance in component technology suddenly makes a new type of product realizable (although it does happen). The common experience is more mundane: the march of technology continually yields components which are cheaper, smaller, faster and consume less power than their forbears. In the audio equipment marketplace this translates to more 'bangs per buck' and some miniaturization, but among those products at the pinnacle of audio performance irrespective of cost, the impact of component advances is arguably less dramatic.

Component cost is usually an important consideration in audio circuit design. Certainly within my company there are product divisions concerned with speech recording and test equipment where cost is always an important factor. We also develop products for other companies, and in these cases production costs are usually capped up-front, and the capabilities and performance of the design must fit in with that. However, I think I can honestly say that in our 'music recording' product division, cost is a very minor consideration - we are far more preoccupied with performance and features within the limitations of media and form-factor. Sure, we design in the cheaper part where there is no reason not to, but in general our barriers are heat, size and the laws of physics. All of these are the enemies of low-noise, low-distortion analogue design! So in these designs the effects of component advances are more subtle.

Jackson: A good component is still a good component. Using the better quality resistors, capacitors and semiconductors still makes an audible improvement. And measurements still miss the mark when judging audio quality. You can't beat tweaking your circuit and listening to the result. That said, the general audio quality delivered by electronic components has moved up several notches over the last few years.

Even inexpensive, garden variety op amps are fast and clean. It no longer costs as much to give the customer excellent performance.

It still pays to do all the little tricks a designer learns over the years. What is great is the choice of great components from which to select. It's like a kid in a candy store surrounded by so much good stuff, you don't know where to begin.

Petrucelli: For analog designs, there have been various "building block ICs" available for some time. Probably the first widely used ICs were the early mic preamp chips from SSM. These eventually evolved into the parts offered by Analog Devices. These offered good performance, but there is usually some tradeoff between the building-block and the discrete design. You DO gain the benefit of lower parts count and easier board assembly, but you may end up with a design that's not as quiet or doesn't have the gain-range you'd like. The actual cost of the building-block and its associated circuitry is not always cheaper than the discrete approach. You're paying for the convenience of a lower parts count and the benefits to purchasing and manufacturing that result.

There are some more-complex or patent-protected circuits that DO provide benefits to both Engineering and Manufacturing from the building-block approach. Early VCAs and RMS detectors were typically built from discrete parts that needed to be well matched and thermally coupled to get the type of performance expected from a Pro-Audio piece. The ICs now available (from THAT) make the analog end of dynamics-control much easier than the good-old days

McTigue: Yes - trend is toward higher performance analog to keep pace with improved quantization. Primary issues are dynamic range, bandwidth, distortion/non-linearity, and noise.

Brunner: Yes! We have changed the analog engine of many of our wireless designs from the traditional compander that is used in virtually all analog pro audio wireless systems to a new patented design using a state-of-the-art component. We chose this component for its versatility, which when combined with our own patented Audio Reference Companding design, gives us a lot of flexibility to optimize the compression and expansion techniques inherent in all analog wireless systems. This has resulted in greatly reduced distortion, increased headroom, and a much lower the noise floor. This makes the sound reproduction comparable to that of a wired microphone

Kraemer: No, we only use basic opamps for the analog reference design for our algorithms.

Coats: We have not had major changes in analog parts. The analog parts we use have improved in performance and in price, but many of the former specialty audio ICs have been discontinued since a large part of the audio industry has shifted to digital. Affordable voltage controlled amplifiers, voltage controlled oscillators, analog delays, and voltage controlled filters are difficult to find if not impossible. Many have been discontinued. The parts that are available do indeed require fewer auxiliary components and are smaller. Smaller footprints help in many ways, including lower noise and lower costs.

Massenburg: Not really. Discrete analog parts certainly aren't getting a lot better. in fact, some specific designs (thinking about the Motorola H-05, H-55, a so-called high-voltage RF transistor) have been disappearing over the years. [During a recent microphone designer panel] someone mentioned that there are custom shops that'll build anything. But that's difficult for the small shop to float.

Weiss: All can be answered with YES. Fortunately there are semiconductor companies out there who care for the higher quality audio market. And some companies are competing against others, so there is a steady flow of new chips suited for pro audio.


2. Does the same hold true for digital designs, including converters?

van der Mee: No, on the converter front some real magic has happened the last couple of years. Sigma Delta converters became a lot better and replaced pretty much any other type converter in audio. For a while it was the major advantage in price that forced us to use them, Sigma Delta converters were a lot cheaper. Especially on the DA side, I wasn't that much of a fan of the new type and preferred ladder style DAC's. It was the introduction of the multi-bit sigma delta converters that turned me over. It made a major difference in sound quality and performance. At this point I would not consider anything else anymore.

On the front of Sample Rate Conversion some giant steps have been made as well. The chips now available are very good, with pretty minimal losses. They are that good that some companies think they can get away with not revealing they are using it to solve their jitter problems...an approach I very much disapprove of, since it is a bit of a cheat and not a loss-less process. But it says something about the quality. That is why we have been implementing these in our products for the first time in our history, but they only kick in when there is no choice. When a "problem" needs to be fixed. In any other situation a high quality PLL is the choice. That sounds a lot better...even with the slightly higher jitter.

O'Malley and Vanwulpen: As far as the digital side goes, the current line of products require fewer parts. In some instances, our legacy products might have a few CPLD's, an FPGA, DSP, microcontrollers, and a lot of glue logic (LSI parts). Most of this can now be condensed into 1 FPGA coupled with a small cheap microcontroller at a fraction of the cost. The flexibility of the FPGA also speeds up development time.

The boxes designed by our competitors generally have a similar setup. In our designs, the flexibility of the FPGA allows for routing and formatting all kinds of digital audio standards as well as interfacing with other IC's, (often ones that have been chosen based on, for example, a particular sonic criteria). In the end a big part of the digital design becomes software, thereby ensuring fewer hardware mistakes and quicker debug time.

Dennis: Most of our designs have both digital and analogue elements, so I suppose the most marked effect is that as digital processing has got smaller and less power hungry, we have spare space and power for MORE... either more processing or more analogue!

Jackson: A manufacturer's IC converter can sound very different from different designers. Some designers choose to use the reference designs provided by converter manufacturers. Often these reference designs are done for economy. Throwing more components at the problem can make an audible improvement. For example D to A's inherently put out lots of high frequencies, way above our human hearing thresholds. The general wisdom is they don't matter because you can't hear them. The reality is that power amplifiers can't deal with these super high frequencies mixed in with our precious audio and end up with subtle distortions and colorations due to the fact that the semiconductors can't move quick enough to trace the clean audio mixed with the high frequencies. Adding a few more components can eliminate these troublesome high frequencies.

One nice advancement is the line between professional gear and consumer is pretty blurry. Converter specs have come a long way...and both pro and consumer benefit.

Petrucelli: For digital designs, I think that the latter part of my earlier statement holds. More and more capabilities are being integrated into the digital ICs of today. Hi-performance 20 or 24 bit A/D converters would be out of the reach of most designers (and ultimately end-users) if they had to be built discretely.

Kraemer: We have moved from Cirrus converters to AKM for lower power and higher performance. We truly feel that these are the best available and have heard more of that from within the industry.

Coats: I think that this is where most of the gains have been made. The performance is up and the price is down. Much more is integrated into chip sets, requiring far less support circuitry. Unfortunately, they seem to have a short life span as chips are being updated or phased out. We have had a few problems in this area.

Massenburg: Although it's possible to build a slightly better converter front-end (inputs stage & modulator) in discrete, what you can buy on silicon has gotten better and better. And I think you could say it's happening faster than any kind of discrete improvements. And, yes, they're faster and quieter and more stable. And undoubtedly cheaper.

Weiss: Yes


3. The availability of new analog and digital building blocks would seem to imply that circuit design is simpler than ever for good results, do you agree?

van der Mee: Yes, it is much easier now on the low end. To make a basic converter, simple amplifier, DVD player, it does not require a lot of skills. It becomes more and more a job of just reading the datasheets and application notes by the manufacturer. Resulting in a very competitive and cheap low end market. However, for high-end applications the struggle is almost the same. If not more difficult. To differentiate, more is expected and to be really groundbreaking you have to offer something special. So more in depth knowledge, more experience, more "vision" is required. Equipment has to be more reliable, then was expected in the past: when it was "ok" for an esoteric product to fail once in a while...these days are over. Our current product line is a good example of this new philosophy. It is full of innovations, completely redesigned from the bottom up, extremely efficient and reliable. We have only one repair tech nowadays and he kills his time by soldering cables...

O'Malley and Vanwulpen: Digital circuit design is much more based in software now. We for sure see this trend continuing as many new interface formats (firewire, for example) are very software intensive. If you look at IC's supporting those interfaces they often consist of a high performance CPU with some IO and some specific blocks. Nevertheless, the bulk of the work is handled in software.

On the other hand, the analog guys are more important and in many ways, circuit design has become easier. One could argue it is becoming more and more of an art, as all these things become more and more soft this also implies higher clock rates and much more noise across the board. Minimizing noise for example is an art in and of itself.

Dennis: There is a school of thought that with analogue 'less is more', but for me a really great analogue design which covers all the bases nicely is, by necessity, complex and a balance or priorities. Improved integration of analogue components means that circuit functions that used to be implemented discretely can now be bought 'off-the-shelf', allowing, again, more space and power for other elements to be more complex. Op amps are interesting in this respect. Modern devices are so good in almost all important areas that it is increasingly hard to argue for a discrete solution. But there is still no 'perfect' op amp - different ones are good in different jobs and different situations.

Jackson: As IC's become more and more integrated, more functionality is built into each building block, reducing the need for more and more external components. Converter Dynamic range performance, unheard of just a few years ago, now rivals the thermal noise through the very best resistors. Designers have to sharpen their chops in keeping the analog signal path performance up in the 120dB dynamic range world to match the 120dB performance of the new slew of converters.

Manufacturers are now including decent op amps in their chips. A few years ago, if you wanted real performance, you would never trust the op amps a converter manufacturer would use inside their chip to turn a current into a voltage....now they almost can't help but integrate fast, clean amplifiers right along with the rest of the goodies inside the chunk of silicon. Softening the need for external current voltage converter op amps delivers much improved sound quality for all but the very premium designs...again saving the customer money, yet still delivering great sound.

Petrucelli: I agree, up to a point. If all it took to design a good-sounding product was to take some building blocks and put them on a circuit board with some pots and switches, then we wouldn't be reading all those complaints on the web about some crappy sounding console somebody had to use at a gig. Of course, there's no guarantee that the manufacturer used those blocks in that particular unit, but you get the idea. The available blocks can make a given design easier to implement, but the sound of a given piece of audio gear involves not just the blocks, but the overall system design of the unit. When consoles are involved, you're looking at many individual channels being summed together. You can have the greatest channel circuitry in the world, but if the rest of the design is not properly implemented, then it's all for naught.

McTigue: Integration of functions is the "mega-trend' in silicon. The other is 'more for less' - more channels, performance, features, control, etc. Design is now more complex because of the increasing need to provide reconfigurability and new features during the life of a product. Digital design is driven by the ability to to add value and the ability to provide a system approach. Everyone can get the same silicon - it is how that silicon allows a company to differentiate itself in the marketplace that is critical to the hardware design.

The trend for this integration of components also manifests itself in the continuing improvement in features that appear in the "prosumer" markets. The high end continues to be driven by providing innovation.

Converters fall into the mixed-signal arena - here there has been very slow progress in performance improvement at the high end - where are the real 24 bit -384 kHz converters. There has been much progress in improving the density in the real 20 bit ~ 120 dynamic range. Big issues - getting professional level analog domain signals into and out of 5 Volt devices, time delay of the decimation and interpolation filters, non-integer related delays with regard to the sample rate. The current generation of Sample Rate converters have shown significant improvements

Brunner: While I wouldn't say circuit design is "simple," I would definitely say that Shure's designs have gotten more elegant and robust as a result of newly available components. That being said, it is all in how you implement these components. Like anything, you can have great building blocks, but if you don't show any ingenuity in how you put them together, you can still end up with an ugly house.

Kraemer: No, good design (especially in the digital domain) still requires adherence to rigorous design practices. Physics never changes so buses and signal lines need to be properly terminated and worst case timing analysis still needs to be performed. Power must be distributed properly and the design must meet regulatory requirements. So while design might be simpler from a high level conceptual point of view, it still requires an excellent circuit designer to implement correctly. People have been lulled into sloppy design by the relative simplicity of the building blocks.

Coats: That is true if you conform to the types of design that the blocks were engineered to fit, but design is still design, and it takes work to achieve your goals.

Massenburg: Well, obviously not. The systems integrations challenges (handling the input analog well, stabilizing clocks and building and distributing good power supplies) are still daunting.

Weiss: Partially, yes. It depends of the level of performance one would like to achieve. Most "building block" chips have to be compromised because of the "one size fits all" approach, which is required to get a decent number of chips sold. If a manufacturer goes the "no compromise" route, those building blocks are in most cases not suited.

Click here to read Part II of this story