SOUNDSTAGE! ON HIFIFeatures Archives

March 1, 2001

 

Upsampling: Threat or Menace?

There's a problem with writing under a byline. People know how to reach you when they think you're full of it. And one of the things I love about my readers is how personally they take it when they think I'm wrong.

Which I frequently am. I'm not an E.E., so a lot of the time when I report on a new technology or an unexpected consequence of an old one, I have to rely on what I do know and understand. Whenever a manufacturer makes a claim for a product that I do not know for a fact to be true, I try to indicate that the claim is just that, a claim. And I try not speculate about why something happens when I don't understand it -- I simply report what I have heard.

One area where I was beginning to feel uncomfortable was the whole upsampling/oversampling business. So far I've reviewed the Bel Canto and Perpetual Technologies "upsampling" DACs -- and I heard improvements using them and reported on those perceived improvements. And in at least one conversation with me, a Musical Fidelity executive represented the A3CD player as capable of upsampling to 96/24. But something didn't seem quite right. I'm not saying that anyone was deliberately misleading me, but there seemed to be a lot of foot-shuffling and throat-clearing whenever I asked anyone about exactly how it all worked.

I remember being at Madrigal one day and asking if the 30.6 "upsampled." Looks of consternation all around -- what exactly did I mean? I explained upsampling as it had been explained to me and received blank looks of incomprehension from some of the finest digital engineers on the planet. "But every DAC does that!" one of them blurted.

So I wasn't exactly unprepared when I called Ayre's Research Director, Charles Hansen, last week to confirm some facts about his K-1x preamplifier and, instead, received an earful. "I'm concerned about the things that have been written about upsampling," Charlie said. "There's a lot of misinformation being spread around and you've got a great opportunity to clear the air on this issue."

Okay, Charlie. What's your point?

"Upsampling almost always makes a difference, and it can make an improvement -- but it's not at all for the reasons most people are saying it is. Upsamplers come in two varieties -- built-in and external. Any external upsampler is going to affect jitter, for better or for worse.

"Every CD player ever built, except the very first Sony -- the CDP-101 -- had a digital filter in it, and they all oversample. Upsampling and oversampling are the very same thing and anybody who tries to tell you differently is misstating the case. What 'upsamplers' do is basically put two digital filters in a row -- one brings it from 44.1kHz to 88.2kHz or whatever it does, and then the other filter brings it up some more.

How's that, Charlie?

"The composite filter has more computational horsepower and will perform better than a single filter, according to traditional measurements. You can get the same improvement by using a better (more powerful) digital filter in the first place or by rolling your own. But there's no magic in it.

"And there's nothing unusual about putting two digital filters in a row -- virtually every digital filter is a cascade of 2x stages, because it costs less than accomplishing the entire filtering process in one go."

"An additional digital filter will have two effects: First, there's usually a reduction in out-of-band noise components. While this is generally true, its impact will depend on how effective the original digital filter is, and how sensitive the rest of the playback chain is to out-of-band noise.

"And also, the overall response of the digital filter will be altered. The first stage of a digital filter has virtually all of the effect on both the in-band audio frequency response and on the transition-band, which will affect the impulse response. The subsequent stages only affect the stop-band response (out-of-band noise).

"When you add an 'upsampler,' it then becomes the new first stage of the composite digital filter and now determines the overall audible characteristics. The first stage of the existing digital filter now becomes the second stage, and so forth.

"Depending on the 'upsampler,' and depending on the original digital filter, the new first stage will have either more or fewer taps. So, adding an 'upsampler' will almost certainly change the characteristics of the first stage (because it is very unlikely to exactly duplicate the first stage of the existing digital filter)."

Is this good?

"A traditional digital engineer would assert that more taps are a good thing, as they reduce out-of-band noise, as well as allowing for a sharper transition from the passband to the stop-band. However, the experienced audio engineer might well be skeptical about adding additional taps, as the time (impulse) response is degraded. The fact is, we don't really know what a good digital filter is. For an extreme viewpoint on this topic, you can refer to www.sakurasystems.com/reviews/a2.kusunoki.html, where the author advances the viewpoint that any digital filter is a bad idea."

So, there is controversy over what filters do?

"It's perfectly legitimate to say that, if you add this box to your setup, there's an improvement in sound. If you've got a setup with an off-the-shelf digital filter and you add a better digital filter to it, you'll get an improvement in sound.

"The best place to read about this right now, since we haven't written anything about it, is on the Madrigal website (www.madrigal.com/UPCONVERSION.htm). It's technical and it's long, but everything in it is complete and accurate and pretty much covers the full scope of the issue."

Do you feel that companies are misleading the public on the subject?

"Well, not exactly. I think some companies are being pretty coy about it. But the public really wants something like this. It's like trying to sell a seven-year-old on Santa Claus -- it doesn't take much selling. There's an incredible urge to believe in free toys. Well, audiophiles have thousands of CDs and they have thousands of LPs and the last thing they want to be told is they need to buy a new format to get better sound out of their favorite performances. Their urge to believe in a magic converter is incredibly powerful.

"And there really are benefits from improving digital filters. Please don't misunderstand me, I'm not saying you should avoid any product that 'upsamples,' I'm just saying you shouldn't buy into any claims about the benefits of 'upsampling.' The public thinks that 'upsampling' is a new technology that offers new benefits that are not available any other way. They have been (mis)led into thinking that they can turn their Red Book CDs into 96/24-quality discs. And that's simply untrue."

So now you know what I know. Do I believe Charlie and disbelieve the manufacturers of "upsampling converters?" Not entirely. What Charlie says makes a lot of sense to me -- as does Madrigal's posting on the subject. But Charlie and Madrigal also manufacture digital products -- products in competition for the same market -- so it's conceivable they're grinding their own axes.

But Charlie's message of "believe what you hear but doubt the explanations you're being given" is good advice for any audiophile. It's the only way to cut through the claims and counter-claims -- at least the only way to do so without going crazy or becoming a cynic. (But that's the subject of another column.)

I'm glad he took the time to explain his views to me (us!) and I'd be glad to publish any responses from any of the companies producing upsampling converters, especially if they can shed further light on this controversial subject.

...Wes Phillips
wes@onhifi.com


SOUNDSTAGE! ON HIFIAll Contents Copyright 2001
Schneider Publishing Inc., All Rights Reserved
Any reproduction of content on
this site without permission is strictly forbidden.