Interview with Rich Maez of Boulder Amplifiers

October 15, 2009


Rich Maez of Boulder Amplifiers has agreed to an exclusive interview with the Ultra High-End Audio and Home Theater Forum to discuss his company and its cutting edge preamplifiers, amplifiers and digital products.

Welcome, Rich. Would you start by giving our readers a little background on your company – when and why it was formed, what its design philosophy and goals are, and where it sees itself positioned relative to other high-end manufacturers?

Boulder was founded 25 years ago by Jeff Nelson in Boulder, Colo. At that time, the original focus of the company was the professional market, specifically for the Southern California recording, mastering and broadcast studios. We produced numerous products for the professional industries that were (and still are) used by Skywalker Sound, Sony Studios, Westlake Studios and countless radio stations when sound quality was of the utmost importance. Some of these products eventually made it to the home consumer market, and as demand from the home audio and audiophile markets continued to grow over the years we eventually left the professional world entirely.

Our design philosophy comes from a different perspective than that of the traditional audiophile company: We do not pursue a particular “sound” or voicing for our products. We don’t choose parts based on their sound or the flavor they impart on the overall character of a product. Rather, we develop products that are as inherently neutral and distortion-free as possible. These terms are used quite a lot in marketing and slogan circles, yet other manufacturers still continue to develop circuits that are slaves to their parts instead of the other way around. We don’t listen to solder, we don’t choose a particular type of output device because of the way it sounds, we don’t select transformers or caps because they were used in old tube amps. We engineer rather than design by ear. Products are prototyped and then performance is evaluated and verified via listening by a number of people, including ourselves, dealers, distributors, friends of the company, etc. There are no “golden ears” at Boulder, as golden ears lead to listening biases that result in distortions. All circuits are balanced, input and output impedances are benign, stability and reliability are paramount, all resonances are minimized as much as possible, distortion is reduced to the lowest possible levels, no outside tweaks or additions are necessary (our gear works to the best of its ability right out of the box), and every product is designed to be able to drive any load within reason. Our equipment shouldn’t have limitations as far as what it can be connected to.

Over the years we have become the only U.S.-based manufacturer that we are aware of that is primarily a manufacturer instead of an assembly house. In addition to creating electronic designs from the ground up, we do CAD design, machine all of our own metalwork in-house, texture and finish the metal, stuff all surface mount circuit boards with our own pick-and-place machine, do all assembly and complete all electronic testing of both subassemblies and finished products before listening to each of them and releasing them for shipment. The only processes not completed within our factory walls are the anodizing of chassis metal and the manufacturing of bare or raw circuit boards. This internalizing of all manufacturing and testing gives us (and by extension, our customers) a great amount of confidence in our quality control, reliability, performance and future value. We tend to be more expensive relative to other companies because of this, but it’s necessary in order to be able to build products the way we want.

Your original preamplifier and amplifier designs were based on the 990 gain stage. Would you give us a little history of that design and why you chose it as a basis for your initial products?

The original 990 was a discretely implemented gain stage or operational amplifier (op-amp) consisting of 43 individual parts that Deane Jensen designed in the 1970s. It was used in everything from amplifiers to mixing consoles to microphone preamps. In those days, the integrated op-amp chip was new and a rather poor performer. Deane sought to create a gain stage that would far surpass the abilities of the op-amps that were available at the time. By contrast, the 990 was an extraordinary performer. Jeff Nelson was a good friend and colleague of Deane’s and wanted to use the 990 in his original designs because of its exceptionally low distortion, ability to drive a load and wide bandwidth. No other gain stage came close.

You most recent series of preamplifiers and amplifiers have been designed based on the 993 gain cell. Is the 993 your own development of the 990 design? What are its areas of improvement?

The Boulder 993 is a much more complex evolution of Deane’s original 990 – there are more than 150 parts and each takes at least 200 hours to complete from beginning to end. The design is still discretely implemented, however it has grown in physical size, power capability and complexity. The operating voltage is much higher, each is thermally potted, the housings are machined and every one is trimmed and matched to spec. It has significantly lower noise levels (harmonic, floor and thermal), a huge amount of current drive capability, and nearly bulletproof reliability. What this means to the listener is a massive amount of revealed information: textures, instrumental resonances, hall ambience, separation, even production values and editing are rendered with incredible detail while each individual image has a foundation and solidity that is almost organic in its presentation. The 993 is used in only the 2000 Series products, our current top-of-the-line gear, because of the time and labor involved in the manufacture of each piece. The 2010 Preamplifier, for example, uses 18 of the 993s.

All of your amplifiers are high output, ranging from 150 watts/channel at 8 ohms in you 860 stereo amplifier and 865 integrated amplifier all the way up to 1000 watts per channel at 8 ohms in your 2050 monoblocks. Why is there a need for such huge amounts of power?

In order to accurately reproduce exactly what the artist, producer and engineer involved in the creation of a recording wanted you to hear, an amplifier must have absolute control of each of the drivers in a loudspeaker. As soon as the amplifier loses control of the drivers, you’ve lost the subtle meanings that the artists spent so much time and effort trying to get across to you. Not only is absolute grip important, but large amounts of peak power as well. The more accurate the upstream components are, the more an amplifier needs to be able to respond to the dynamic and transient needs of the incoming signal (dynamics in = dynamics out). Even at low listening levels, power output capability and headroom is important when reproducing both macro- and microdynamics. Quiet passages, loud passages and contrasts between the two must all be reproduced as accurately as possible without strain in order to make the music involving. Because not every loudspeaker is a simple load, we want to be sure we can control and drive just about any loudspeaker that our gear may get connected to.

To what use are microprocessors put in your amplifier designs?

There is a microprocessor in every product that I can think of off the top of my head. Amplifiers have microprocessors to control turn on (to keep massive inrush current draw in the big ones from popping a breaker in your house), protection circuits and housekeeping; preamps have them to enable the selection of different inputs…and, well, all user interface functions, really; the same goes for the 865 Integrated Amplifier, and the digital products have them in a number of places for all of the above reasons and more. Even the 2008 phono stage has a microprocessor to enable the user interface to function properly as well as remember settings for each input. No product in our line uses them anywhere in or near the signal path, however.

Audiophiles have traditionally associated microprocessors with problems such as noise, RF, distortion, etc. Micros are actually not the source of the problem – poor implementation is. Bad grounding, improper isolation, poor shielding, poor board layout, etc. can make all the difference in the world and negate any potential benefits.

All of your components, including your digital components, rely on balanced rather than single-ended connections. Are there any advantages to balanced connections other than reduction of hum in long runs?

Most definitely! In addition to noise cancellation in cable runs, you double your signal while leaving the noise floor at the same level, thus significantly improving your signal-to-noise ratio. Also, if implemented properly, balanced circuitry can also eliminate noise within the product itself by taking advantage of common mode rejection within the electronics circuits. This is true of not only hum, but also of other forms of noise as well. If you’re trying to render a recording as it was originally intended and with no flavor or colorations imparted, there’s no reason not to use balanced circuitry throughout the system.

Speaking of digital components, your 2020 Advance DAC employs what you refer to as Upandoversampling. What is the difference between upsampling and oversampling, and what is Upandoversampling.

OK, this is going to be greatly simplified, so keep in mind that there is a lot that isn’t going to be covered here. Upsampling is a conversion process by which the original data rate is increased, whereas oversampling is a form of inserting additional samples between the original data samples to raise the sample rate to the DAC to increase the signal rate. Simple upsampling by itself will have absolutely no effect on sound quality – you can’t create or retrieve information that didn’t exist to begin with. Oversampling can improve sound quality by spreading quantization noise over a greater spectrum and thus lowering it in the signal data band.

So why do we upsample at all? Because we want to feed the absolute maximum data rate to the DACs at all times. A 44.1 kHz signal is automatically increased to a combined rate of 705.6 kHz. In addition to the benefit of lowered quantization noise offered by oversampling, we also benefit from the ability to use much better filters now that our data rate is that much higher because of upsampling. Upandoversampling is simply using both processes in tandem so that we can gain advantages elsewhere.

What are the advantages of using 32-bit interpolation when processing a 24-bit signal?

We’re going to be simple here again. When doing any sort of signal processing, you’re doing math. If you’ll remember from grade school, any time you multiply two numbers, you get a larger number. A 24-bit word length means a 24-digit number. To do any sort of digital processing, you’ll need to apply math to that 24-bit number. If you simply used a 24-bit interpolation scheme, you would end up with a number much longer than 24-bits and have to cut off (or truncate) the end of the number in order to keep it at 24-bits and thus reduce its accuracy after the decimal. If we increase the bit rate to 32, we can use a number that is much more accurate (and less musically distorted) because there is less truncation.

Which Burr-Brown DACs do you use and why is using five per channel advantageous.

The 2020 and the 1012 use the BB 1704K. We’ve evaluated the 1794K and didn’t like the measured or the audible performance. Using five DACs per channel has a number of advantages, but the biggest is a much better signal-to-noise ratio by way of summing the multiple outputs of all five while keeping the noise floor at the same level. After I/V (current to voltage) conversion, the voltage output is much greater than if we were to use one or a pair of DACs.

Would you explain to us how the “time advance” feature in the 2020 is able to function as a balance control? Is it superior to the analog balance control found in your preamplifiers?

The Advance function doesn’t really act as a balance control, but as a timing control for each channel. A balance control lowers the level of one channel relative to the other in order to compensate for a difference in level from channel to channel at a fixed or stationary listening position. The Advance function changes (advances) the timing of the data from one channel to compensate for changes in the listening position. In other words, if you’re not sitting in the sweet spot, the Advance function will let you adjust the timing of the output from one channel or the other so that both channels arrive at your ears simultaneously, even though one speaker may be farther away from you. When used in combination with a balance control, the Advance function lets you move and place the sweet spot without sonic drawbacks.

The 2020 Advance appears to actually contain three separate chassis. Would you describe the function of each and explain how this separation of function benefits the sound of the component?

The 2020 is actually made up of four separate chassis! There are the left analog channel; the right analog channel; the logic, user interface and digital section; and one chassis that contains three separate power supplies. By isolating each of these sections, we keep noise such as hum, RF, or channel-to-channel interference to an absolute minimum. To the listener, this means a giant step up in resolution and a big reduction in distortion.

Given the increasing use of computers to playback high-resolution audio files, do you have any plans to add a USB or Firewire input to either your 1012 DAC preamplifier or 2020 Advance DAC.

It’s something we’ve been evaluating. USB is a poor digital transmission type but very popular because it’s so simple to implement – high school kids can do it. Some companies have done a lot of work to make it better, but it’s still not great. Firewire isn’t likely because it would have to be a proprietary interface that would exist only among our own products since there’s no common standard. There are also a number of other options that we’re looking at for computer-based audio playback.

Our position on digital interfaces is that if they can be properly implemented and made as an upgrade, we’ll do it. We don’t want to do anything that would mean any sort of compromise or sub-par performance, however. Anything we do is always the very best available.

Would you describe the unique features of your 1021 disc player?

Oh, my. This could go on for quite a while.

OK, well, the very first thing that everyone notices is the 6.5-inch display. It shows the artist, album, track title, elapsed and remaining time of the current track, as well as a number of other options during normal playback. In addition, you can scroll through the track listing and select another while the unit playing.

A very basic diagram of the path that the digital data follows would be disc mechanism ® hardware buffer ® DSP ® software buffer ®DAC section. The data is never processed as a serial data stream with an embedded clock signal (S/PDIF or AES) – everything is processed in packetized form to eliminate the chance for jitter to infect the signal at any point. Clocking is applied only at the point of digital-to-analog conversion.

There is a complete computer on board to control a number of different things within the player. All digital data transfer, software, electronic control, mechanical control and user interface are controlled here. This host computer also works in tandem with our DSP section to do some pretty amazing processing and implement our unique digital filter. Our filter uses an algorithm that is phase accurate (both pre- and post-impulse ringing are addressed, not the currently fashionable method of addressing only pre-impulse response) and is nearly perfectly frequency response accurate. All upsampling, oversampling and our volume control is implemented in DSP as well – and word length truncation is never an issue because of a new way of performing mathematical functions.

The analog section is also completely new. The DACs we use are not the usual Burr-Brown chips, but new ones that allows us to use our own digital filter instead of resorting to the internal filter within the chips themselves (convenient but not terribly good on the bench or to your ears). These DACs have the benefit of being down -8 dB in comparison to the Burr Browns when analyzing high-frequency distortion. This made for an amazing amount of new detail that we had never heard before, even in comparison to the 2020 Advance D/A. A six-pole Bessel filter is then used for analog filtering. The output section is driven by Boulder 983s (a board-implemented, half-discrete version of the 993 gain module) and the 1021 is capable of driving hundreds of feet of interconnect cable.

There are a number of other technical features in the 1021, but really, all of the techie info in the world wouldn’t matter if the performance of the player weren’t up to par and listening were a letdown. So far we’ve had a number of other manufacturers tell us that the 1021 is by far the best-sounding player they’ve ever heard (one of them going so far as to say it was better than a $150k analog rig at the 2008 Rocky Mountain AudioFest) and we’re happy about that.

Is there anything else you would like to tell us about your company or products, including any new products which may be on the horizon?

Well, the next project we’re working on (it’s scheduled for release the last week of October) is the 1008 Phono Preamplifier. The 1008 will be a single-chassis unit that will use 983 gain stages and feature the multiple EQ curves we first featured in the 2008 five or six years ago as standard. For years we’ve had people tell us how much they love the 2008 but the price is simply out of their league. The 1008 will come in at US$12,000, or just a fraction over 1/3 the price of a 2008 for a huge chunk of the performance.

After the 1008 we have a number of options open to us: an 800 Series source, a new 3000 Series (yes, we get requests) or updates to older products. We’ll have battles to determine where we go once we’ve gotten past the 1008 and it’s shipping routinely.

Thank you for joining us today.