Home
AudioAsylum Trader
Digital Drive: Re: Erroneous Stuff... by csown

Upsamplers, DACs, jitter, shakes and analogue withdrawals, this is it.

For Sale Ads

FAQ / News / Events

 

Re: Erroneous Stuff...

207.89.144.49


[ Follow Ups ] Thread:  [ Display   All   Email ] [ Digital Drive ]
[ Alert Moderator ]

> I don't think the problem he was addressing was phase distortion. He basically thought upsamplers *themselves* applied a gradual filter (as opposed to a "brickwall" filter) to reduce "time smear," which is an artifact of "brickwall" filters. All I was saying is the claim is unequivocally false.

I'm only "mad" because those marketing upsamplers are touting them as "new technology," but in reality is merely another method of oversampling. And I really think people are blowing thousands on upsamplers based on deceptive premises. And all this article does is *advance* this deception. (Aside from the Ah Tjoeb, many of the people here ultimately find out the hard way that upsampling is nothing more than a marketing gimmick.)

I need to read the article again, because we should take a second and make sure we're talking about the same things here (I'm not on a stable connection right this minute, so this'll have to wait). I equate phase distortion with time smear. Or rather, phase distortion is what gives rise to time smear.

If you oversample at a large integer multiple, the amplitude response is basically unity across the audible band. However, with upsampling (non-integer oversampling), you will get a damping envelope that begins to manifest itself in the audible band. The damping envelope of the upsampler is steeper than that of typical zero-order-hold, thus it gives an effect of being a gradual filter, but one that allows frequency-dependent high frequency images. And with this particular type of digital filter, the phase error is not terribly bad. In this sense, it is a "gradual filter that reduces time smear". I might be missing some details about the behavior of upsampling LPF filter behavior. I'll have to think more about it and get back to you on this one a little later...

You are right though, this is not new technology - in fact, at face value it is a flawed implementation of old technology. And definitely, people are being misled about this as a solve-it-all scheme for digital audio. But do note this - it is not well-understood why various sampling schemes sound the way they do, and the field is much more confusing because implementation (capacitors, PCB trace routing) can have a greater effect than simply the filtering methods. I like to substantiate my claims with fundamental theory and sound engineering, but not all manufacturers do - buyer beware! Upsampling can sound awesome if done correctly, but of course, this is not always the case.


> What is "differential linearity?"

There is a section in the article in which Doug talks about this. He describes two types of linearity error. One is an overall deviation from linearity (i.e., a linear DAC follows a smooth curve that is not exactly a straight line). The other - differential linearity - is a fine-scale linearity error, meaning that if you take a random value at some point in time, increasing this signal by one bit does not give the right amount of change in voltage (or current, if it's a current DAC). So if you look at the output curve of a presumably linear DAC real close up, you can see that the line is really quite granular, and this jagged curve will change from day to day (thermal noise, vibrations, diffusion over time, etc..).


> I realize better linearity is a big deal, but how is this achieved with an upsampler??

This is probably the biggest point Doug makes in his article. If you read carefully the section on differential linearity, he elucidates that the upsampling (non-integer oversampling) allows frequency-dependent high-frequency image information at the output. The differential error in the DAC can be partially corrected by this because you are essentially adding dither to a hold operation (or ramp, or whatever, depending on the type of DAC). In essence, the high frequency bands give an averaging effect to the "fundamental" frequency. This is what "linearizing" means here.


> By the way, phase accuracy in Redbook playback is a misnomer. While the **average** phase error is zero degrees, the **instantaneous** phase error can be as great as + 90 degrees.

Which type of Redbook playback are you talking about? Are you referring to redbook playback as a whole?

I recall (correct me if I'm wrong on this) that one of the strengths of FIR filters is pure phase response. This type of filter is quite common in oversampling DACs. But the analog post-filter mucks up this phase performance really bad, and depending on how it's implemented you can get quite crazy phase errors all over the band. For non-oversampling DACs such as my dAck!, I cannot claim zero phase error, but because I don't use a steep analog post-filter, I still have better phase response than most every D/A out there. The ZOH operation in non-oversampling DACs results in total phase error of 90 degrees across the band; in practice it is not quite this bad. No steep analog post-filter means that this is the worst the phase error gets. That's part of why the imaging with the dAck! is so good; most digital designers are so concerned about flat frequency response that by ensuring it, they compromise imaging performance.

-Chris


This post is made possible by the generous support of people like you and our sponsors:
  Kimber Kable  



Topic - upsampling article - solidgore 08:08:38 08/22/03 ( 47)