Hi,
The time domain delay is inherent in the time it takes (see Red Book) to treat the signal.
That's no problem for as long as the output equals the input and the only difference is time.
Now if the process causes a frequency dependent delay, IOW, not all frequencies are processed simultaneously, then (and this is where the "smearing" term stems from) you have a distortion mechanism that is occurring in the time domain.
The best analogy would be somewhat similar to a multi-way speaker system where the individual units are misaligned causing gross group delays.
None of the speaker units themselves are a source of distortion, their combined outputs are just not coherent in arrival time, it's "smeared".
Sure, it's another one of these audio vocabulary boiler plate words that are used to graphically describe whatever message they try to convey.
They should add a glossary at the bottom of each article but then that would kind of dig their own grave, wouldn't it?
Cheers,
The time domain delay is inherent in the time it takes (see Red Book) to treat the signal.
That's no problem for as long as the output equals the input and the only difference is time.
Now if the process causes a frequency dependent delay, IOW, not all frequencies are processed simultaneously, then (and this is where the "smearing" term stems from) you have a distortion mechanism that is occurring in the time domain.
The best analogy would be somewhat similar to a multi-way speaker system where the individual units are misaligned causing gross group delays.
None of the speaker units themselves are a source of distortion, their combined outputs are just not coherent in arrival time, it's "smeared".
Sure, it's another one of these audio vocabulary boiler plate words that are used to graphically describe whatever message they try to convey.
They should add a glossary at the bottom of each article but then that would kind of dig their own grave, wouldn't it?
Cheers,
Frank, I don't disagree on the theory he mentioned, but disagree on the way he drew a conclusion.
Here's where "irreparable" word has been used:
"As will be explained shortly, all upsampling DACs employ slow roll-off digital reconstruction filters as opposed to conventional DACs, which employ sharp roll-off or brick wall reconstruction filters. Since slow roll-off filters show less time smear than brick wall filters when measured using artificial digital test signals, many have concluded that reduced time smearing is responsible for the subjective improvement in sound quality when playing CDs through upsampling DACs. Although this seems logical and is an appealing explanation, as will be shown, it falls apart upon closer examination. The main reason is that the digital audio data residing on a CD is already irreparably time smeared. No amount post processing of the digital audio data by the playback system can possibility remove or reduce this time smearing."
Here is the reason why sharp filter has been used in producing CD, and its consequences:
"If a sharp anti-aliasing filter is not used during recording at 44.1 kHz or when down-converting a 96 kHz master to 44.1 KHz, folded images of the baseband frequencies will fall right back into the audio baseband instead of being frequency-shifted to ultrasonic frequencies as occurs in playback. These in-band images or aliases are clearly audible even at extremely low levels. To prevent audible in-band images from appearing, the decimating anti-aliasing filter used to down-convert 96 kHz to 44.1 kHz, must be very sharp indeed and must also severely attenuate all frequencies above 22.05 kHz. Thus, anti-aliasing filters as well as decimation filters must introduce time smear at least as severe as that introduced by sharp anti-imaging filters used in playback. What’s more, the time smearing due to the anti-aliasing filters cannot be removed. This is so because the time smearing on CDs is a consequence of the fact that the signal has been strictly bandlimited to 22.05 kHz. Any frequency components above that frequency which may have been recorded on the original the 96 kHz master are gone forever."
Frankly, I'm a bit confused regarding which one is Rife opinion and which one is the one he is trying to disprove.
And a "conclusion" which I believe is his:
"Whatever role time smearing plays in audio quality it is not sufficient to
explain the increase in sound quality due to upsampled digital audio. The main reasons are: a) the time smearing of the kind produced by a sharp anti-imaging filter is already present in the digital data on every CD ever made and; b) the time smearing added by the vast majority of loudspeakers is much larger than the time smearing caused by the combined time smearing of both the anti-aliasing and anti-imaging filters."
Last edited:
That's not distortion, it's phase, easily seen in the frequency domain. It may require a Maxwell Demon to tell the processor what is the "frequency" of the data.
Well, all I can see is that some phase shift may occur, that's how I interpret "time smearing" - and if that phase shift is present in the digital data it can be "unshifted", if need be to perfectly compensate for any further "shifting" that may occur; no audible frequencies will be lost, or harmed , in this exercise. So, I don't see the need to use the word "irreparable".
In my personal experience, I have never heard "bad" sound from phase "misbehaviour" - plenty of other things cause unsatisfactory playback, which can be fixed; but I have never deliberately fiddled with the equipment to try and improve "group delay" aspects ...
, in this exercise. So, I don't see the need to use the word "irreparable".In my personal experience, I have never heard "bad" sound from phase "misbehaviour" - plenty of other things cause unsatisfactory playback, which can be fixed; but I have never deliberately fiddled with the equipment to try and improve "group delay" aspects ...
Here is what Kusunoki has said regarding time-domain accuracy versus frequency-domain accuracy:
"The difference between the non-oversampling DAC and the conventional DAC with the digital filter lies whether you attach importance on the accuracy in the time domain or in the frequency domain. In other words, whether you choose the musical performance or the quality of a sound. This trade-off line defines the boundary of the current digital audio format .
A natural, stress-free sound that communicates the musicians' intention directly to you. That is the sound of non-oversampling DAC. The feel of this sound is closer to that of analog reproduction."
Later on Doug Rife mentioned about the "superiority" of upsampling filter, or shallow slope filter. Also, Dejan's NAD C565 has the option to choose the slope. I'm curious to know which setting that he prefers.
Hi,
Well, it's still a distortion somehow, no?
I don't think the processor cares about what it is being fed. It's not the processor but the process (or so it seems). Where and how, we'd like to know.....
The good news is that with the "right" accessories (theirs, not yours) it can all be dramatically reduced.
Cheers,
Well, it's still a distortion somehow, no?
I don't think the processor cares about what it is being fed. It's not the processor but the process (or so it seems). Where and how, we'd like to know.....
The good news is that with the "right" accessories (theirs, not yours) it can all be dramatically reduced.
Cheers,
IME, similar to Kusunoki's "experience", phase misbehaviour is the reason of all non-musical system.
Now think about this Frank:
If your speaker is a computer speaker, plenty of things cause unsatisfactory playback, which can be fixed as you always said. Now imagine if you have gone all the troubles to ensure a perfect system... your problem now becomes more difficult to define... And this is where I have found the importance of phase behavior.
As also mentioned by Rife, group delay in speaker is so big and common that phase delay in digital should be considered immaterial. And that is true. Also why many TL speakers don't sound good to my ears with such a big group delay (not to mention Doppler effect).
Not that I can see. No new frequencies are added or subtracted. There's no irreversible information loss.
re Kusunoki (whoever he is): What some guy says in an ad is not terribly meaningful, especially when either he doesn't understand basic Fourier math or (worse) he does and assumes that his audience won't.
Fair enough, Jay, if you are very sensitive to phase anomalies then you will need to consider the impact of all things related to that. So far it hasn't been an issue for me, so maybe I'm just lucky,
Or you don't know yet. You still rely on powerful amplifier damping quality and very probably have NO big issue with phase. Try to have a wide band speaker (3-way) using magnesium cone drivers for example. Or what you have seen on Audio Shows.
Or if you believe that class B amplifiers have that fatiguing crossover distortion, try to build a class-A amplifier with its commonly found limited power and damping (and your speaker is that typical low sensitivity drivers).
Converters have linear phase FIR filters, there's no trade-off between time and frequency domain at all. Frequency is ruler flat and phase is ruler flat.
With a nos converter you must make a very steep analogue anti imaging filter, this is almost impossible but with great effort it can be done. See my previous post.
Apparently some people like the sound of IMD.
For completeness, I will put up the alternative to 2 stage design, and is sometimes used in hi end designs, including SY's, Walt Jungs, and even me (on occasion).
The problem with this design is the input stage gain needed to keep the high frequencies from falling into the noise region, ultimately, because we can use only 30dB or so of input gain.
SY's approach is optimum for tubes. For solid state, we have to work with lower impedances for the EQ, and the following stage very quiet, in order to keep the high frequency signal out of the noise floor.
It can be done with solid state, but it is not as easy as the passive-active EQ approach used in the Vendetta, JC-3 or the Constellation. Entirely open loop operation, however, is probably best. That is what Charles Hansen and I strive for, and SY seems to do well with it as well.
The problem with this design is the input stage gain needed to keep the high frequencies from falling into the noise region, ultimately, because we can use only 30dB or so of input gain.
SY's approach is optimum for tubes. For solid state, we have to work with lower impedances for the EQ, and the following stage very quiet, in order to keep the high frequency signal out of the noise floor.
It can be done with solid state, but it is not as easy as the passive-active EQ approach used in the Vendetta, JC-3 or the Constellation. Entirely open loop operation, however, is probably best. That is what Charles Hansen and I strive for, and SY seems to do well with it as well.
Attachments
Some of us prefer to live in a world where causality rules.Jay said:
No, silver has slightly higher conductivity than copper is experimental fact - but I'm sure someone has a good theory of metallic conduction to explain it too.
Only for people who know its silver, and whose ears somehow circumvent the theory of potential dividers on which audio interconnects rely (for those of us who prefer causality).
It is possible to conceive of circuits which modify the sound without adding harmonic distortion, but these would need to be carefully engineered (e.g. AGC). It is hard to conceive that these could accidentally arise, so looking for a mismatch between Fourier components in the input and output will always be a useful way of finding problems.nigel pearson said:
Filterless NOS DAC cannot reproduce the signal at the output of the recording anti-aliasing filter. It does not attempt to. The best it can do is provide bursts of image HF noise at the same time as bursts of signal HF noise. Fortunately this may mimic the sound of percussive notes, which is about all there is at those frequencies.Jay said:
Last edited:
- Status
- Not open for further replies.