My Bits Are So Much Better Than Your Bits

A feature expressing the observations and opinions of Fred Ampel of Technology Visions.

If you even take a casual glance through our industry trade magazines' product advertising and reviews, you can't help but notice the ever rising bar being placed in front of how many bits are enough in the digital audio universe.

Not only is the "acceptable" number rising, (who would dare use anything with only 16 bit capability in 2009), but it seems as if a new form of aristocracy has been created, using an Orwellian model of some are more equal than others, with a touch of class warfare blended in for that spicy edge.

Before we go too far down this marketing fog obscured path, let's be sure we all have the same map.

The recognized definitions for what a bit is and what bit rate means are below¹.

• bit Abbr. b Abbreviation for binary unit or binary digit. 1. The smallest amount of digital information. A bit can store or represent only two states, 0 and 1. [The original term binary unit was coined by John Tukey of Bell Laboratories to represent the basic unit of information as a message representing one of two states.]
• bit rate The rate or frequency at which bits appear in a bit stream. Applied to digital audio, bit rate (kbits/sec/channel) equals the sampling rate (kHz) times the number of bits per sample. The data bit rate for a CD, for example, is 1.41M bits per second (44.1 kHz x 16 bits per sample x 2 channels)

These are NOT suggestions open to debate. They are what they are in electronic terms, parameters, and structural operations. Just because you might want to, you cannot change the composition of digital information streams to accommodate your specific marketing, specmanship or promotional needs. (Well ... I guess you can since there's a lot of it going around). But from a scientific and engineering perspective you can't!

However, spend a little time thumbing through the current crop of products and you will find some remarkable stretching of the structural frameworks and accepted operating rules going on.

First you will see the Orwellian part, where mine is bigger than yours is the only rule that counts. The logic here seems to be that if number x is good, than number x plus a few bits MUST be better. Maybe, but maybe not. It largely depends on where those extra bits are coming from. If those bits are coming from a truly higher sampling rate and the use of a longer word length then the effect might be as advertised. If they are coming from clever manipulation of the specs of components or creative use of absolute best case parameters, then the claimed audible improvement is likely to be only in the mind of the marketing writer.

Let's remember that EVERY digital audio system which deals with what were once analog signals is unquestionably affected by processes which have numerically defined results that cannot be altered to meet marketing or better because I want it to be goals.

The first, is quantization, which in case you forgot (or never cared in the first place) is the process performed by the A to D converter that changes the continuously variable amplitude of the analog waveform into one of the finite 'steps' or discrete levels defined by the sampling rate and word length of the digital signal.

The second, which is inherent in the quantization process, is some degree of quantization error, which results from the digitization. This error is the difference between the actual value of the analog signal at the sampling instant and the nearest quantization value. Therefore, in general, the longer the word length, the less the error, because there are more step sizes to choose the closest value from.²

In most cases this error is perceived as noise, or distortion, or both. There are ways to mask it (i.e. noise shaping, oversampling to move the noise out of the desired bandwidth, etc.) and reduce its annoyance value, but it cannot be removed.

These two essential parts of the digital audio process produce specific results. The error component is unavoidable, as is the inaccuracy of the process to some degree (very small to very large). Both can be minimized to tiny percentages but they do not go away.

Another absolute in this domain is dynamic range. This is nominally accepted to be 6 times the number of bits - so a 16 bit system has a dynamic range of 96dB, 20 bits produces 120dB and so forth.

However, this performance is dependent on very careful attention to input levels to keep the system close to zero level where its advertised performance will occur. If the input signal is too low then the artifacts will become more apparent and more annoying. It's much like the old analog gain structure stuff. Want max signal to noise? You have to work at it. Give two people the same rig where one pays attention to this and one doesn't- you will certainly hear the difference.

So let's say you have the coolest most bling loaded brutally expensive product from the Truly Mammoth Gigantically Impressive & Powerful Digital Audio Company.

According to its literature there is nothing it cannot do, and it's doing it all with super-maximized, mega one bit, eight times oversampled, dual layer ultra precision 64 bit silicon, producing an incredible 512 bit precision output.

That has got to be better that the 128, 32 or 24 bit stuff the less well-heeled are forced to use. You have discovered the class warfare portion of this game. Here, we operate under the largely valid assumption that most purchasers and users are not expert digital audio engineers and don't have the time or inclination to run bench tests or proof of performance tests.

So a little dithering of specs and a bit of truth shaping is applied to make sure that the highest possible, if not quite real world, numbers are produced in those marketing categories that seem to matter - chiefly the number of bits the thing supposedly uses. Throw in a bunch of digital buzz words, and a few techy terms and away you go.

Now, let's be real with each other. If you were told that the black box has 512 bit precision would you check? Could you check? No, I didn't think so.

But if you took a moment and let the adrenalin rush pass, you might want to think about this for a second. If the best anyone else in the whole world can do is 128 bit, where do these guys get their parts from to make the claims they do? Are they actually operating a chip foundry and hand producing all of their own little black pieces of plastic to go on their platinum plated, gold washed, silver vapor deposited, Teflon ( it makes the audio faster because it's slippery) coated circuit boards. Not likely!

If, on the other hand, they are actually buying parts from someone, then their ADCs and DACs and so forth are most probably not any better than anyone else who buys the same part. After all, there are only a limited number of suppliers out there and our little industry does not produce enough revenue to support too many ultra-custom pieces. It's pretty easy to track the parts back to their home. If you did that for the above product you would be devastated to find that it's the same stuff used in that $399 processor your friend spec'd for his next install. Is this adrenalin rush as much fun as one you had when you bought into the magic hype of digital marketing?

Are my bits better than your bits? Only if my bits come from a verifiably better source with real world specifications that allow its performance to be higher than something else.

About the Author

Frederick J. Ampel has been involved in the professional audiovisual industry for more than 35 years. His career has included professional live sound reinforcement, broadcast audio production, studio design and installation, and sound, AV, and video system design, installation, equipment development and sales. From 1983 through 1992 he was founding editor/editorial director of Sound & Video Contractor (S&VC). He was also editorial director for RE/P. In the fall of 1991 he founded Technology Visions, a consultancy specializing in marketing and technology assessment, business development, market research and development, import/export development, product development and planning the audio and video industries worldwide.

Ampel holds a Master of Science Degree in Engineering from Boston University and a Bachelor of Arts in English/Music from Long Island University. He has been published in the proceedings of the IoA, AES, NSCA, InfoComm and numerous journals and publications. He currently writes feature material for Residential Systems for which he is also principal editorial consultant, and contributes to LiveSound International. He was the principal editor for the 1997 edition of CEDIA's Home Theater Manual. Ampel served on the AES' SC10 committee on Computer Control of Audio Systems and both the NSCA and CEDIA education committees. He also chaired CEDIA's Certification Task Force. He has conducted classes, workshops, and papers sessions for NSCA, CEDIA, InfoComm and AES, as well as the IoA and SCIF in the UK, and several trade events in Hong Kong, Singapore and Australia. He is an active member of AES, ASA, CABA, CEDIA, IoA and NSCA.

About the Sound Vision Series

This article features the views and opinions of its author and does not necessarily reflect the opinions of InfoComm International.

¹ Both definitions courtesy of the RANE Pro Audio Reference – Copyright 2005 by the Rane Corporation. This is available online and was updated February 2009. I would not use the 2005 reference.
² Definition courtesy of the RANE Pro Audio Reference – Copyright 2005 by the Rane Corporation. This is available online and was updated February 2009. I would not use the 2005 reference.