 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
70.213.249.103
In Reply to: RE: Upsampling question posted by greg7 on November 10, 2007 at 15:29:43
"Sorry for the recent barrage of questions but here's another: when is upsampling subjectively 'good' and 'bad'?"
Upsampling's or oversampling's intended purpose was to digitally filter out the ultrasonic frequencies above 22 kHz. And unless the device is explicitly stated as non-oversampling (or is the initial Sony design), they all upsample/oversample.
Although there is no official distinction between upsampling and oversampling, the most-common is oversampling being "synchronous," and upsampling being "asynchronous." (Synchronous is often stated as 4x/8x/16x etc. oversampling. Which was predominant during the 1990s.) I personally think asynchronous upsampling is very fatiguing to listen to. It injects a narrow band of HF noise that to me sounds "detailed" initially, but over time, I realize that it's actually noise riding on the music. And I ultimately become fixated on that band of noise. (See link in regard to how this noise occurs.)
Conversion from 16/44 to 24/96 in real time was *not* the original intention of sample rate converters. It was to enable music to be transferred from one medium to another. Like from CD to DVD. These converters only became popular in their current form because of marketing and a couple rave reviews of the dCS Purcell upsampler. (I saw the "bogus" in the product from the very beginning.) Then it was all downhill from that point.
"I've been under the impression that all upsampling is 'bad' (un-musical) in general?"
Not only do I think it's bad, I think it's a technological fraud..... (As a real-time processor.) There is nothing, technically speaking, where such process would suggest improved signal fidelity. The review of the dCS Purcell was what made "24/96" or "24/192" upsampling the hottest new product in audio.
"Two players I'm considering use 192 kHz and 96 kHz, respectively. Both use current generation 24 bit chipsets. Are both at least synchronous? And is 'less' better (i.e., lower upsampling frequency)?"
I've fallen into this trap personally. The chips are often *capable* of playing up to 96 or 192 kHz, but unless a sample-rate converter (SRC) chip is part of the chipset, it's *not* asynchronous. But I do think these newer chips sound worse than the 20-bit chips used during the 1990s.
All else being equal, I also prefer 24/96 over 24/192. But that's not an endorsement of 24/96.
"Are there any general rules of thumb here? I realize upsampling is only one of a multitude of variables lending to sound quality but I'm curious to understand."
My only suggestion is to gather the information, ask questions regarding the validity of marketing claims, and audition as many different players as you can. (You may end up liking asynchronous upsampling.) But I will add that you should extend your listening sessions to at least an hour, to get the big picture. Since a lot of the ills of digital playback don't become apparent in brief listening.
Follow Ups:
It seems that advanced technical expertise can get in the way of just plain listening. Both Gordon Rankin and Todd Krieger know a great deal about a field in which my knowledge does not extend beyond the digital chapter in Robert Harley's hi-fi book, and I mention their names with the greatest respect. But the dCS Purcell with its 192 upsampling made an astonishing difference in my system, and when dCS went all the way with DSD upsampling in the Scarlatti the sound became even better. Haven't seen any reviews of my Scarlatti yet, but the Meitner uses a similar DSD upsampling, and all its reviews have been raves.
What's the big difference between "DSD upsampling" and the old-fashioned single-bit delta sigma processing that most CD players have been using for the last couple decades? Isn't the output a PDM bitstream in both cases? Although in the past they used to run at a much higher bit rate. In the early 90s, Sony was using a 90MHz clock on their top of the line ES machine.
DSD was born out of SACD playback, in the quest for greater compatibility. The conversion from 16/44 to 24/96 has been done using SRC chips, which are continuously variable with input and output frequency. (The output clock runs independent on the input rate, hence it's asynchronous.) The DSD engine can play any of the commercial formats synchronously, since its base rate is a common multiple of all rates.
2.8224MHz isn't a multiple of 96KHz. But that really has nothing to do with my question about whether DSD "upsampling" of CDs (which I believe is what we were talking about here) is really just another name for the PCM to bitstream conversion process that has been used at oftentimes much higher clock rates since the early 90s.
Ted Smith originally pointed me to this, when I had the exact same skepticism that you currently have. This link is an eye-opener.
That link just shows what I said, 96K isn't a multiple of the dsd sample rate. If you want to upsample 96K to dsd, you have to go higher than the dsd base rate to get a common multiple, and then decimate. I think that's well understood by now.Thanks for the link, but again, that didn't really have anything to do with my question about CD upsampling, and whether dsd upsampling of 44.1K pcm was anything new.
There is a "5x oversampling" node in the diagram..... The base DSD rate needs to be 5x oversampled to attain synchronous conversion to the DAT formats (I also notice 24/192 is not in the diagram either, which would require "10x oversampling"). So the base rate of DSD is *not* really a "common multiple" of all format rates, like I erroneously stated earlier..... Sorry about that.....
I presumed it was running without a need for an intermediate "oversample" conversion at an odd integer rate.
I think DSD would be the best "single-unit solution" for those who want a system for playing SACD, DVDA, and CD. But for those solely into CD, there are too many conversions (albeit synchronous) relative to the CD-dedicated oversampling and non-OS which can provide satisfying playback.
| FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
||||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
