Big Numbers

Marketing gurus love big numbers. I’m sure some psychologist could tell you why humans gravitate to ever-higher numbers but my guess says that we’re just wired to appreciate larger values. Of course, we want faster cars, more pixels, bigger bank accounts and more free time. But on the other hand…we want less girth around our midsections, lower times for the 5 or 10K races that some of us run and fewer taxes, right?

So I’ve always been amused when I see larger and larger numbers in the audiophile marketplace. We’re all familiar with the gradual move from 44.1/48 kHz sampling rates and 16-bit word lengths to 88.2/96 kHz and 24-bit PCM specifications. The ubiquitous Pro Tools digital audio workstation blasted right past those numbers and put out it “HD” system a number of years ago. The I/O units were actually called 192s because they could run at 192 kHz. No matter that virtually no one was operating at that level because none of the plugins were able to keep up and the computer power was too anemic to run more than one at a time. It just felt good to know that is was capable of 192 kHz.

There is ample justification for higher sample rates and longer word lengths to a point…at least for PCM encoded audio. But can anyone justify a “35-bit 844 kHz” DAC and Digital Preamp? I don’t even know how they got those numbers. The product is real and made by NAD…a company whose products I have owned and enjoyed. It’s been a while since I had one of their early preamps but I always appreciated the simplicity and sonic purity of their designs.

The M51 doesn’t sound like they’ve continued with the “simple is better” philosophy of their earlier models. Here’s how they describe the operation of the unit:

“NAD’s technology uses a very powerful processing engine that operates at much higher speed and with much greater accuracy than typical Digital-to-Analogue converters. Running at 108MHz, the M51 resamples the incoming pulse code modulated (PCM) signal and converts it to a pulse width modulation signal (PWM) with a sampling rate of 844kHz. Operating in a differential mode with double precision, the M51 has state-of-the-art specifications for low noise and freedom from distortion. The extreme headroom afforded by the 35-bit architecture allows for a DSP-based volume control that does not reduce resolution. Even with 24-bit high definition signals, the output can be attenuated by 66dB (very, very quiet) before bit truncation begins.”

The above description sounds really promising, but remember it was written by the marketing department and not by the technical people at NAD. What they are saying is that they believe they can improve on the quality of the output signal but transforming the digital input stream into an completely different data representation (the Pulse Width Modulation) at a very high sampling rate AND using longer word lengths. They also put tremendous emphasis on the ability to attenuate a digital signal in the digital domain. I don’t buy any of it.

If the incoming PCM digital signal is already pristine, highly dynamic (using perhaps 20 bits of the available bits…and that would be a lot!) and properly clocked, then recasting it as a PWM signal at 844 kHz with 35-bits wouldn’t “improve” the sound at all. It might change it…but if it were already as clean as it could get, then mucking around with it in this manner wouldn’t help. It couldn’t because the dynamic range, frequency response and all of the other specifications have already eclipsed the ability of humans to perceive any additional enhancement.

What could you do to improve distilled water with some sort of filtering or modifier…nothing. Same idea.

Just think about the 35-bit specification for starters. As a computer guy (MS in Computer Science CSUN 1992), I can’t figure out how or why they thought to use 35-bits. There’s a reason why bit lengths have always been powers of 2 (the binary language of digital computers or multiples of 8. We’ve had 8, 16, 24 and now 32 bits…but 35? This is a fell good number as is 32-bits. The dynamic range of human hearing is not much greater than 130 dB SPL. AND 24-bits gives us the potential for 144 dB, so why bother. Also keep in mind that most recorded music doesn’t reach even 10-bits of dynamic range.

The conversion to 844 kHz PWM is bogus too. The knowledge that skilled designers have about PCM is very thorough and deep. Current designs like the Benchmark DAC2 reach well beyond the sonic realities of the recordings that we’re buying. So why mess about with PWM and bigger numbers? Because it gives the marketing people, the reviewers and addicted audiophile consumers something to talk about.

Bigger…at least in this case…is most definitely not better.


Mark Waldrep, aka Dr. AIX, has been producing and engineering music for over 40 years. He learned electronics as a teenager from his HAM radio father while learning to play the guitar. Mark received the first doctorate in music composition from UCLA in 1986 for a "binaural" electronic music composition. Other advanced degrees include an MS in computer science, an MFA/MA in music, BM in music and a BA in art. As an engineer and producer, Mark has worked on projects for the Rolling Stones, 311, Tool, KISS, Blink 182, Blues Traveler, Britney Spears, the San Francisco Symphony, The Dover Quartet, Willie Nelson, Paul Williams, The Allman Brothers, Bad Company and many more. Dr. Waldrep has been an innovator when it comes to multimedia and music. He created the first enhanced CDs in the 90s, the first DVD-Videos released in the U.S., the first web-connected DVD, the first DVD-Audio title, the first music Blu-ray disc and the first 3D Music Album. Additionally, he launched the first High Definition Music Download site in 2007 called iTrax.com. A frequency speaker at audio events, author of numerous articles, Dr. Waldrep is currently writing a book on the production and reproduction of high-end music called, "High-End Audio: A Practical Guide to Production and Playback". The book should be completed in the fall of 2013.

6 thoughts on “Big Numbers

  • Gerald Pratt

    Mark. Regarding your comments on NAD’s M-51:
    Perhaps NAD selected 844 KHz so those of us with good hearing could hear the 422 KHz audio content. Of course there isn’t anything like that in the audio signal and whatever is there is captured with 96 KHz anyway.
    35 Bit D/A and A/D converters? Good luck. We don’t really get 24 bit depth even now do we? Yes the converters are built 24 bit, but I suspect those last bits are just toggling around as they sense the temperature change and thermocouple effects in the metal junctions. We don’t want or need more than 24 bits. Oh, and I will bet those digital files are going to be large at 844 KHz sampling.
    This is the kind of crap that will damage the acceptance of real audio (what do you call it? HD, Super HD but not SACD, I always thought it should be called High Resolution since the additional bits give you more resolution. Gerald Pratt

  • Andrea

    I understand that it’s hard work coming up with a new blog entry every day, but a little research and thought might have been in order. There are two reviews — one on Soundstage and the other in HI-fi News — that could help you understand the numbers. This DAC uses a 7-bit PWM architecture running at 844 kHz, which requires a clock speed of 108 MHz. Would you get all bothered about a DAC that up samples your 96 kHz recordings to 6.144 MHz? If you read the spec sheets on the DAC chips in most pro and consumer components, that’s what you’re likely to find in order to make the multibit sigma delta converters work. I believe that the Benchmark DAC you reference uses an ESS Sabre chip, which uses an architecture similar to delta sigma. Unless you are using a true ladder DAC — like MSB –, the bit depth and sampling rate are going to have to change from the 16/44.1 or 24/96 PCM storage format. As for the 35-bit processing, even a computer science textbook from 1992 could tell you that more bits yields higher precision. If you are converting audio data to non-integer sample rates, precise calculations are a good thing. You are right that the errors from using fewer bits would be buried in the noise floor, but audiophiles tend to get all bent out of shape when they know that there’s been a loss of digital resolution. I’ve heard the M51. It’s nice, but nothing extraordinary.

    • Admin

      Andrea, thanks for you comment. I appreciate the clarification and I do know that all incoming digital signals are transformed in the chips. Gerald’s assertion about the 422 kHz hearing range is off the mark. What bothers me about the NAD piece is the use of these big numbers just for the sake of using big numbers. Audiophiles should be given real world listening specifications…saying something is 35-bits (other than for internal processing and maintenance of the LSB when any signal processing is applied) is stretching things when the dynamic range of most recordings I look at are at most 10-12 bits.

      As I said, I’ve always like NAD equipment and have no doubt that the unit performs quite well…but I wish they would leave the techno babble out of the promotional effort.

  • Greg Stidsen

    I feel like a comment directly from NAD is appropriate here. Andrea has the technology description essentially correct. We have a white paper available on our website that gives a complete explanation of the technology we use which was co-developed with Zetex and is now owned by CSR, both semiconductor companies based in the UK. We also decry the ‘numbers race’ as it doesn’t really tell the whole story, but try NOT publishing the numbers! The first question we get from audiophiles is how many bits and what is the sampling rate. So we are really trying to respond to our customers (and critics) who want to know this stuff. It is more accurate to say that the Direct Digital technology we use has a minimum word length of 35 bits anywhere in the processing. We use the extra bit depth to allow many preamp functions to take place in the digital domain, like volume control and EQ filters. With a 24 bit signal we can attenuate the signal by 66dB without truncating any bits, and since the signal path and circuit condition is exactly the same for all volume settings, the volume control is completely transparent – this cannot be said for variable resistors in the analog domain. As pointed out, most/all DACs do not actually deliver the resolution promised as these are usually theoretical capabilities. If you look at the measured results for the M51 (and M2 and C 390DD, and D 7050 all using this technology) you’ll see that we have perfect linearity to -120dB and still useful decoding to -135dB with 24 bit signals. This means that we attain about 22 bits of resolution which is world class. There are only a couple of other DACs on the market that can do this and they all cost a lot more. Quite frankly, to address the core complaint of the article, we would rather have audiophiles rely on measured specifications and listening rather than play the numbers game. But everybody wants to know the numbers. How about a rant on power ratings, my pet peeve?!

    • Admin

      Greg, thanks for the comment. You’re quite right that using longer word length internally provides the necessary resolution to accomplish volume, EQ and other DSP processes. This has been one of the reasons that I’ve avoided ProTools for all these years…everything was processed at 24 bits, which results in truncation in many cases (Sonic Solutions on the other hand uses 56 bits). Still the output is 24-bits, which is more than enough to handle real world dynamics.

      I’m a big fan of Benchmark equipment and know that their DAC2 delivers 132 dB of dynamic range. We’re past the point of worrying about reproducing actual dynamics…all we need are recordings that capture real world dynamics.

      Numbers do matter but my take on the NAD M51 marketing blurb was that they introduced a whole new set of numbers that audiophiles haven’t see or “heard” and that can only lead to more confusion.

      I remain a fan of NAD gear. Thanks again.

  • Yves Smolders

    Hello Mark,

    You are comparing apples to oranges. The M51 is not a PCM based D/A converter. the PCM signal is converted to a Pulse Width Modulated signal – that is why you see the completely different numbers, which are not powers of 2.

    the processed stream is closer to SACD (DSD) than it is to PCM at this point.

    Gerald, do some background reading, this is not a 35-bit PCM D/A converter.


Leave a Reply

Your email address will not be published. Required fields are marked *