Interview with Andreas Koch of Playback Designs

November 8, 2009

Andreas Koch Of Playback Designs, whose MPS-5 CD/SACD player and MPD-5 digital-to-analog converter have received universal acclaim, was kind enough to consent to an exclusive interview with the Ultra High-End Audio and Home Theater Forum.

Thank you for agreeing to be interviewed for the Ultra High-End Audio and Home Theater Forum. Most of our members are unfamiliar with your background. Would you give us a brief biographical sketch?

First of all, thank you for the opportunity to talk about some of the technologies I have been researching and developing over the past 29 years. It is my hope that these technologies continue to contribute to the ever evolving and incremental improvements in sonic performance that we can experience in our industry.

My field of experience has always been digital signal processing, and because I already pushed the limits of the technology while studying it at the University in Switzerland where I built the fastest signal processor, I quickly became involved in digital audio right at its beginning, which then seemed so fast that hardly any hardware could keep up with it. At Studer ReVox I researched the theory and built the hardware for the first fully asynchronous sample rate converter with completely arbitrary frequencies. It was 1982 and barely a 16-bit world. I recognized that 16 bit would not be good enough, so I expanded my hardware to 24 bits.

Shortly after that I joined Dolby Labs in San Francisco as its first DSP engineer. I was surrounded by all these brilliant analog engineers and together we pushed the technical envelope and built the AC-1 audio compression scheme for TV broadcast. That was just the precursor of what I added shortly afterwards, the first real time processor to implement algorithms that later evolved into the widely used AC-3 compression which as many of you know is what DVD-Video and movie theaters are based on.

Meanwhile in 1987 computer technology had gotten just fast enough to handle the most basic tasks in digital audio. It was a new frontier calling me and I decided to return to Studer to head a project where digital audio would be recorded onto computer hard disk for editing. Soon I launched one of the most complete computer audio platforms for professional recording, editing and mixing, known as Dyaxis.

What could possibly be the next technical frontier after such a gigantic project? Sony helped me answer that question: launch a new consumer audio format with more channels and vastly improved sonic performance. While developing the very first SACD mastering recorder / editor (“Sonoma”) with native DSD processing, I also had to develop a new kind of A/D and D/A converter that would be able to highlight the advantages of SACD. Price was not the objective. It was only sonic performance that mattered. Sonoma is still used today to master many SACD’s.

Following my work with Sony, I had gained so much experience and knowledge of technologies in digital audio processing that I thought I should try marketing my know-how as an independent contract engineer. I was very successful with it up until very recently when Jonathan Tinn and I teamed up and formed Playback Designs.

What led you to start Playback Designs?

Jonathan and I met when we were both heavily involved with EMM Labs. Once both of us left EMM Labs, we kept in touch as friends and started talking about forming a new company that would allow us to implement our own ideas without the limitations put on us by past or existing products or any corporate culture. The decision was made and we formed Playback Designs. I started with a clean sheet of paper and close to 30 years of experience and designed the Playback Designs MPS-5. This player represents the pinnacle of digital audio design. There has never been another player like it and is what I believe to be the true “state of the art”.

What in general is your design philosophy?

In audio design as with almost anything else, keeping it simple is almost always the hardest and most beneficial thing to do. When designing, I keep this in mind at all times. I am constantly considering how to simplify or shorten the signal path or simplify an algorithm. This allows a purer and more lifelike sound which I believe is truly the goal.

What is a “two-dimensional” DAC?

This is discussed quite in-depth on our website. Perhaps it would be best if I repeated it here:

2 Dimensional DAC Technology and Computer Audio

“Audio is represented in a y/x-axis system: the y-axis for amplitude and the x-axis for time. Mostly because of analog audio’s sensitivity problems in the y-axis, digital audio was introduced. But digital audio not only quantizes the y-axis, it does so as well on the x-axis. Sounds like we got more than we wanted – true and too bad. A typical state-of-the-art DAC converts between quantization levels in the digital y-axis and the analog y-axis and is completely transparent and open as to what happens on the x-axis (time domain). Sounds like we forgot the quantization on the x-axis.

This oversight forced us to treat digital audio signals as if they were analog: use special cables, use all kinds of mechanical devices for our CD players, power conditioners for digital audio etc. Looks like we just shifted the original problem from the y- axis to the x-axis, but the issues are still the same. Instead of interference or crosstalk we now call it clock jitter.

Almost all DACs available today deal with the y-axis only and rely on external devices for the x-axis, such as complicated master/slave clock arrangements or external sync clock generators. At best these devices are band-aids on a wide open wound deep inside the DAC. They help, but do not resolve the problem at the source. We need a 2-dimensional DAC that not only works on the y-axis, but also on the x-axis. With this we can separate the digital world completely from the analog one and render any digital cable, transmission format, storage media and application completely irrelevant to the final sonic performance. The only analog problem that we still have then is the separation of the power supplies for digital and analog.

The DAC inside the Playback Designs product line does exactly that: clock jitter from incoming digital audio signals can be described as an analog signal that gets mixed together with a quantized digital signal (our ideal and constant sample rate clock). So before any processing can happen we need to bring these 2 components into the same domain: The Playback Designs system quantizes the clock jitter into a digital signal, where it then can be subtracted from the original sample rate while the latter is converted to analog at the same time. Of the course, the DAC also works independently in the y-axis by using a set of unique algorithms in a completely discrete architecture (not even a single Op-Amp is used).

Tests have shown that the DAC inside the Playback Designs product line can be fed by any digital source including a PC, an inexpensive Discman, a DVD player, or high-end CD transport and none of them seem to make a difference on the sonic performance of the analog output signal. Ultimately this means that as long as you are sending our DAC truthful complete bits the source does not make a difference. We believe if you own a home computer, you already own a music server that cannot be sonically bettered!”

Would you explain what DFAS (Playback Designs Frequency Arrival System) is, how it differs from other jitter reduction strategies, and how it not just reduces jitter but completely eliminates it?

Jitter is real and it exists – some of it is good, some bad. But rather than trying to tell the difference in your incoming signal and trying to filter it, I approach it from a different angle.

If you tell your friend to follow you in your car, he will inevitably regulate the speed of his car up and down to stay in sight of your car. No matter how constant you drive your car your friend will vary his speed in small amounts. That is how a PLL (phase locked loop) works and is commonly used to generate an audio clock in a DAC.

Imagine you tell your friend where you intend to go and what time exactly you expect him to be there. That knowledge enables your friend to travel completely detached and independently from you by calculating a constant speed that brings him to the destination at the exact time. Life really gets easy when we communicate and can trust each other, we just need to create a similar environment in audio where that is possible and that is what I did with PDFAS.

This is a brief analogy of what I am doing and although I realize I did not answer your question with real technical detail, the reason is simple. This proprietary knowledge is one of many things that separate us from our competitors. If I share this knowledge here, other companies might implement my discoveries in their products and we might lose some of our uniqueness.

In your judgment, at what level does jitter become audible, and what should audiophiles listen for to determine whether jitter is present in their systems?

There are so many different kinds of jitter and many people have already tried to characterize them – and that is fine, but there isn’t a simple criteria at which point you can determine that it is audible or not. Even if you eliminate the jitter in your DAC you can still hear jitter from the A/D converter that was used in the recording, because it was embedded in the digital files (worst of all) with no chance of fixing it later. Also, for various reasons it might be desirable to still have some jitter in the DAC, completely uncorrelated to the audio signal. Where do you draw the line? Everybody does it differently.

I take it you do not rely on third-party DAC chips in your components. Would you explain how you implement digital to analog conversion?

While not revealing any trade secrets I still use the same very basic concepts used in most chip solutions, but by using discrete components and implementing all the algorithms myself I can control every single step of the process. For instance, in one algorithm I found that the commonly used 32 bits or 56 bits were not sufficient. I used an insanely higher number of bits just because I could hear a difference. Discrete steps also allow me to introduce novel algorithms that address some of the issues of common signal processing. That is where the sky is the limit and I can provide improvements mostly in the form of software upgrades, all thanks to a flexible and discrete platform.

The MPS-5 plays back both CDs and SACDs. Obviously they employ different digital architectures – PCM and DSD. Is the digital to analog conversion handled by two separate circuits or do you, for example, convert DSD to PCM prior to processing?

This is really a funny question because taking a DSD signal and converting it to PCM prior to processing is probably the worst thing you could do and I realize that there are a number of companies currently doing this very thing. I would liken it to taking a high resolution photograph and sizing it down. When you view the sized down version, it still looks great, just smaller. Now try enlarging it to its original size and all of a sudden the picture becomes extremely blurry. It is the same as converting DSD to PCM and then feeding it to a D/A which always has a stage of some form of DSD before converting the signal to analog. Very simply, the most direct route between two points is a straight line. Why would you want to take a different route if it was unnecessary?

If I am reading your website correctly, you “oversample” the data stream 128 times during processing. What is oversampling and how is it beneficial?

Yes, Playback Designs’ products take all data, whether DSD or PCM, and oversample it to 128 times the fundamental frequency (44.1 or 48kHz). If you oversample a digital signal by an infinite amount it automatically becomes analog. You want to oversample your digital signal as high as possible in order to get as close as possible to “analog” before the DAC actually outputs analog. There is nothing new about this, every commercially available converter chip does this internally to some degree.

In addition to oversampling, do the MPS-5 and MPD-5 upsample? If not, why not? If so, what are the sonic advantages of upsampling?

Although there are academic differences to be argued, there really is no difference between oversampling and upsampling. To me, it all means the same thing.

I understand that the MPS-5 can be expanded to a full six-channel setup or a four-channel system with center channel mixdown to the front left/right channels. Does this mean that in its six-channel configuration the MPS-5 can play back multi-channel SACDs? How do you implement center channel mixdown with the four channel option?

The MPS-5 can be programmed to read the multichannel area of SACD’s. Therefore, it can generate 6 channels of audio. Its internal DAC converts the front left and right channels and each additional MPD-5 that you use would convert the next pair until all 6 channels are converted. In the case of a 4-channel only setup, the MPS-5 can be programmed to mix the center channel into the front left and right channels. The LFE channel usually contains very similar information and it can be derived from either the front left or right channel directly. The mix down process in the MPS-5 is done in the DSD domain without detour to PCM. The one additional MPD-5 required then converts the surround left and right channels.

Both the MPS-5 player and the MPD-5 digital to analog converter have AES, S/PDIF, Toslink, USB and PlayLink inputs. Several Questions: first, the USB input is limited to 16-bit 44.1kHz/48kHz signals. Is that a limitation of the USB interface? What USB to AES or S/PDIF converter would you recommend for computer playback of high resolution digital files, including playback from a laptop? Finally, what is PlayLink and do you have any plans for implementing it in the near future?

The USB input is primarily targeted for applications where CD libraries are stored on personal computers and played directly via the USB connection. Of course, the USB interface in general can implement much higher data rates. The MPS-5 can be upgraded once the corresponding implementation becomes available.

The conversion between digital formats such as USB to S/PDIF or others should be irrelevant as long as you don’t lose any data bits. It is the converter’s responsibility to remember the x-dimension as explained earlier in question #4. But since most converters do not address the x-dimension, I can see that your question can be important to some people.

Frankly speaking, we feel the best all around way to connect a computer to one of our DACs, especially for high resolution file playback, is with a sound card like a Lynx AES 16. Although I do not know what one can use for a laptop, there might be a similar product that would work equally as well. The AES connection is excellent and long cable runs are used quite often with great success. The same cannot be said for USB.

Playlink is already implemented and used to link the MPS-5 to its companion MPD-5. The format is flexible enough to expand to all kinds of other applications in the future.

What special lengths have you gone to assure that the analog output stage in the MPS-5 and MPD-5 meets the same high quality standards as the digital to analog conversion stage?

The analog output stage used in the Playback Designs converter is designed with the same exact goals and principle as its digital front end: keep it simple. We use the highest quality, best sounding parts available, along with the shortest possible signal path. The design breaks quite a few rules that have been assumed so naturally by many other designers. By always looking at new ways of simplifying our design, we seem to also simplify the sound.

In closing, is there anything else you would like to tell us about your company and products?

On the very opening page of our website, we state: “Playback Designs imagines, creates and manufactures the highest performance jitter free digital playback systems available for the most discerning of listeners.” We truly believe in this and will continue to live up to this statement in the future.

Thank you again for taking time to speak with us today.