|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
70.190.203.46
In Reply to: RE: What do you think of upsampling ?................... posted by Cut-Throat on July 15, 2014 at 14:16:28
"I got this new DAC (Teac UD-301) this week and it has the capability to upsample. I have been using it mostly from 44.1K to 176.4K and I think the upsample sounds better."This is a "synchronous" upsample. Which in essence is the same thing as "4x oversampling", in your case. Depending on the filter algorithm, the end result can indeed be very satisfying.
"I have tried to find the explanation for this, but I after reading various articles on up-sampling, I usually end up more confused."
You're confused because there is no technically sound explanation for "asynchronous" upsampling, which is the most-common form of upsampling, as it is called.
Most asynchronous conversion is from 44.1k to 96k or 192k. (Benchmark DACs asynchronously upsample 44.1k to 110k.) It was basically the use of (asynchronous) sample rate converter (SRC or ASRC) chips in the digital-to-analog conversion chain.
The original intended purpose of these chips was to convert files to playback formats that would have otherwise been incompatible with DACs unable to play at the native sample rate.
When dCS originally developed DACs with ASRC, there was a rave review from one of the foremost magazines. And ASRC was then presumed to be the newest thing for CD playback. But those at dCS admitted they couldn't even explain the "improvement". (Which I personally didn't think was actually an improvement.) Unlike some later adopters of ASRC, dCS never conjured up any questionable theories.... They just used the reviews to market their products.
Over time, claims surfaced trying to "explain" why ASRC sounded better. Initially, the claim was "resolution enhancement", people at the time didn't realize that resolution couldn't be improved upon its native resolution. (Otherwise it would be possible to create a high-rez recording from a 64kbps MP3 master, if you get the gist.) Once that claim was debunked, there were claims of "breakthrough interpolation" and "jitter rejection". The "interpolation" claim tried to give readers the impression that existing digital filters didn't "interpolate", but all digital filters for CD playback interpolate. (Interpolation is the calculation of values between samples of the original signal. The filter is in essence depicted in software/firmware.) The "jitter rejection" claim was also bogus because the jitter error at the input gets embedded into the signal as noise artifacts during conversion. (Once this noise is embedded, it cannot be undone.) It also presumes that the the clock inside the SRC chip is reliably more precise than the clock of the device sending the digital signal. (An explanation of how ASRC converts jitter to noise in link.)
And finally, there are companies like Wadia which use the term "upsampling" to describe its synchronous oversampling. Basically a more modern term than "oversampling". But the same conversion/filter methods as classic oversampling. (Which seems like what you're doing.) Synchronous oversampling/upsampling, unlike asynchronous upsampling, does not introduce noise artifacts during conversion. Because synchronous oversampling acts purely on the data, independent of timing/jitter.
"Can someone give me a fairly simple explanation?"
See above.... ;-]
The fortunate part is that the industry has ultimately come to the realization that asynchronous conversion (the most common form of "upsampling") didn't really improve anything, from a technical standpoint. So "upsampling" products aren't nearly as widespread as they were 5 to 10 years ago. (This is why I no longer bring it up, unless one explicitly inquires about it.)
Now if you like how it sounds (even if it's "asynchronous"), that's the most important part. Regardless of the technical workings behind it. Or how I or someone else thinks about it.
Edits: 07/15/14Follow Ups:
"The "jitter rejection" claim was also bogus because the jitter error at the input gets embedded into the signal as noise artifacts during conversion. (Once this noise is embedded, it cannot be undone.)"
You are confusing two different potential sources of jitter: (1) timing errors in the conversion of the original analog waveform to a series of digital samples, (2) timing errors in the arrival of digital samples at the input of the ASRC block in a DAC.
The former occurs during record. Timing errors at this point are embedded into the digital file and will be there regardless of how the file is played.
The latter occurs during playback. Errors in timing at this point have at best a minute effect on the samples output by the ASRC converter. They have an effect only in as much as jitter affects the determination of the "average" input sample frequency. This average sample frequency is calculated using a very narrow bandwidth digital phase lock loop. To the extent that the DPLL bandwidth approaches zero the timing of input samples will have no effect on the values output by the sample rate converter, hence the presence of jitter rejection. To the extent that the bandwidth is not zero the jitter rejection will be less than perfect. All audio is less than perfect, so this is not an argument that the jitter rejection is "bogus". Unfortunately, doing the necessary complex interpolation is computationally complex. This offers lots of opportunity for ASRC to sound poor for reasons other than jitter rejection. Many of these problems exist with any sample rate converter.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Thank you, Tony. I was just about to provide that same correction, before I saw your response. I've unsuccessfully attempted to enlighten Todd on how ASRC can reduce playback jitter over the past few years now. Perhaps, you will have success, but somehow, I doubt it.
_
Ken Newton
Edits: 07/16/14
"I've unsuccessfully attempted to enlighten Todd on how ASRC can reduce playback jitter over the past few years now. Perhaps, you will have success, but somehow, I doubt it."
No, I've posted on this subject every few years after Todd makes his claim. Perhaps it will be necessary to work out a simple example of why ASRCs work in theory.
How well ASRCs work in practice is a different question, and there I tend to agree with Todd's subjective comments. For that matter, I've yet to see a commercial SRC (SRC, not ASRC) that will upsample and downsample back to the original format and demonstrate that the Sampling Theorem works to a high level of accuracy, as measured by a deep null. Residual errors at -75 dB or worse tend to be typical, even when tested on sine waves that don't come close to the Nyquist frequency. (My metric is the L-infinity norm, i.e. largest absolute value of the error signal.)
Some audiophiles have complained that changing the low order bit of samples of a 24 bit sound file "completely trashed" the sound. How this can possibly be is a related subject.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
ASRC has become an "obsolete" technology.... To where I'd only respond to explicit queries about it (or "upsampling"). Only few companies still use it in their current digital audio processing products. (It's an option in the Sabre DAC, but I think it isn't utilized in most products with the chip.)So if you want to tout the virtues, have at it. ;-]
My only concern now is on the recording side (CDs mastered from 24/192 using ASRC instead of 147/640 or DSD conversion), not so much the playback side.
Edits: 07/16/14
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: