This is just the sort of answer I was hoping for.
Especially, your observation about close mic'd solo piano playback artifacts via the 3rd order filter. I can only surmise that even though the lowest image frequency should be inaudible at 22kHz, it becomes audible via some other mechanism. Which, I can't imagine, is due to inter-modulation within the amplification chain, as the signal should still be sufficiently bandlimited by the 3rd order filter to prevent that.
While the electronics should be free of inter-modulation effects from the image-bands, I wonder whether audible inter-modulation might yet be occurring directly within the ear, even though the images are ultrasonic? I'll have to do a little bit of research on whether or not that's possible. Also, I suppose that Don's suggestion about transient phase-modulation, due to insufficient image suppression, might possibly also be responsible.
What are your thoughts on what mechanism is causing audible artifacts with the 3rd order image-filter, but not with the 7th?
I've now re-read your post from which the above excerpt is taken. I suspect that language issues made reading the entire post something less than clear to me. However, the above paragraph does seem clear, so I'll respond to it.
My question stems from the postulate that if the sound is truly different, than the signal, in some relevant way, MUST also be different. Audio recording/playback comes down to physical laws. This is simply because the necessary devices are physical. Classical physical laws are cause-and-effect in nature. Those physical devices must, of course, be manufactured. To do that they must be specified, then engineered to meet those specifications, then manufactured in to existence. Which requires knowledge of the physics on which they operate.
To declare that one component sounds better than another means nothing from a bringing-it-in-to-existence perspective. The designer should fully understand the WHY something is done, as well as the HOW. Knowing only the how, we are little more than Chimpanzees copying the actions of other chimpanzees without truly understanding why. Without fully understanding why, there is no full control over producing the exact physical devices (DAC's, amplifiers, speakers, etc.) which would completely satisfy the end goals. In the case of audiophiles, those end goals are human perceptual in nature. Otherwise, we are stuck with trial and error, hoping to hit the center of a target while in a partially dark room (an exaggeration, I admit). Lord only knows, we in the DIY community are familiar with the process of trial-and-error. Which means spent time, and spent money. Not to mention, lost opportunities for, perhaps, more satisfying home listening experiences.
NOS and OS DACs tend to present distinct sound characters (in my, admittedly, limited experience). Characters which tend to withstand allowing for sound differences among individual DACs within it's respective group. I'm only hoping to gain a better understanding of the physical causes behind their sound characters. Especially, because, it seems illogical that there should be any, and fairly obvious at that, difference at all. We are presently lacking a full physical (technical) explanation for that, at least, as far as I'm aware. No, I'm far from being a pure objectivist. I do, however, suspect that once all factors are fully understood, which I define as, having a complete understanding and control over the device's physical causes producing a human's perceptual effects. An complete understanding also includes, fully how those causes are dynamically perceived by the human ear-brain system. Then, I suspect that I will become a pure objectivist. But I don't think we are there yet, I know I'm not.
Last edited:
That's a great question.
Many of us, at one time or another, suspected maybe that it was causing perceptual problems. I think this was largely due to analog signal based thinking being applied toward digital signals. Digital signal processing is often counter-intuitive, leading to assumptions which while logical, are also false. For some time now, I've mostly only read anecdotal reports on listening experiments which supposedly showed that the ringing is not audible. I have no experience to back that conclusion, but only because my experimental DACs have featured only NOS, OS with slow roll-off, and OS with fast brick-wall roll-off. Theory shows that ringing (either pre or post) is not audible. Which makes sense, because, ringing, being an inherent part of a theory which perfectly reconstructs the original signal, can therefore leave no 'residue' of itself.
Pre-ringing is what renders the filter as linear phase. The ringing is simply what the signal must do because it's been band-limited. Meaning, all the frequencies above the Nyquist point have been removed. This sharp discontinuity produces ringing, just it would in an sharply band-limited analog filter. For properly bandlimited signals, an absolute constraint on signals in order to satisfy perfect reconstruction, the ringing (which really is only the digital filter's impulse response) is, in fact, necessary. I think that most experts in the field (which, I'm not) have concluded that the subjective problems with consumer digital audio isn't that it too closely adheres to sampling theory, but that it doesn't adhere closely enough. I do know this that is true in regards to most chip based digital interpolation filter implementations.
Many of us, at one time or another, suspected maybe that it was causing perceptual problems. I think this was largely due to analog signal based thinking being applied toward digital signals. Digital signal processing is often counter-intuitive, leading to assumptions which while logical, are also false. For some time now, I've mostly only read anecdotal reports on listening experiments which supposedly showed that the ringing is not audible. I have no experience to back that conclusion, but only because my experimental DACs have featured only NOS, OS with slow roll-off, and OS with fast brick-wall roll-off. Theory shows that ringing (either pre or post) is not audible. Which makes sense, because, ringing, being an inherent part of a theory which perfectly reconstructs the original signal, can therefore leave no 'residue' of itself.
Pre-ringing is what renders the filter as linear phase. The ringing is simply what the signal must do because it's been band-limited. Meaning, all the frequencies above the Nyquist point have been removed. This sharp discontinuity produces ringing, just it would in an sharply band-limited analog filter. For properly bandlimited signals, an absolute constraint on signals in order to satisfy perfect reconstruction, the ringing (which really is only the digital filter's impulse response) is, in fact, necessary. I think that most experts in the field (which, I'm not) have concluded that the subjective problems with consumer digital audio isn't that it too closely adheres to sampling theory, but that it doesn't adhere closely enough. I do know this that is true in regards to most chip based digital interpolation filter implementations.
Last edited:
You re right, it should be an language issue, you didn't get my point or maybe half of it if I understood your answer here . I believe about basic knowledge and global methodology. I'm thinking the same that your last paragraph that I find to be a rephrasing without the filter issue of my basic english knowledge
! Try & error btw I use is just a way to reach or approach a goal with limited toolbox. Though can work very well, although almost impossible to be sure or a knowledge acquisition here that can be always reprated -i.e. a law-My point was, but on a knowledge perspective you hilight, the goal is to make something good -somewhere mastering what is aesthetic or not from ear/
brain substract-. It can sometimes be done with poor ways : try & error, parts, trade off (sacrifice something to pick up another thing much appreciated). And at the end it can sound good. Of course it has few thing to see with a controlled engineering process from A to Z, though some little hifi brands, sometimes very expensive, can use that mix focusing only on the goal, avoiding all explanation about what is done - cause not enterily controlled by themselves, just picked up cause it works often enough to be in their toolbox-
Coming back to your question, we often see w trie to explain something because we are not liking grey area and whatever the knowledge level we want to give an explanation... which of course is a verity only if it's checked by the experiment and repeatable - ok we will not talk about uncertainty here in audio-
. But a human default is wanting to glue something with the knowledge material avaliable, sometimes sort of made smart opinion that can sounds good for the brain though still not the true explanation.Which is fascinating me is very knowledgeable here can come close by a creative process coming from experiments and try & error without theory before and find something by luck as for instance Pasteur did by luck with the vaccine process : pure hazard
From my point of view over sampling or filter introduced make it not always good, however being not knowledgeable I of course can change opinion. again the concept is good, maybe this is a non controled effect of its application that harms when it sounds bad. Again I dunno... just curious it is still discussed as an old knowledge now.
btw Ken, sorry for my english level, I try to do better.
Last edited:
We largely agree about digital reconstruction filters. Many DACs do, or would, sound better without their on-chip filter. I suspect those of producing artifacts stemming from an implementation which prioritizes chip cost over other objectives. One approach is address that problem is to just accept that many digital filters are problematic and dispense with them entirely. I find doing that (NOS) brings certain subjective benefits, but also leaves other subjective benefits behind. So, I'm hoping to marry the subjective benefits of both NOS and OS, while divorcing each from their respective faults. Which requires an understanding of exactly why each produces the sound character it does. If I found eliminating the digital filter to be entirely satisfying, I would simply do that and be done.
Unfortunately, I hear some characteristics from OS sound which are desireable to my ears, and that NOS doesn't seem to provide. Which doesn't preclude me from enjoying my present DAC, which is switchable between NOS or OS mode.
By the way, no need for apologies regarding your english skills. I'm quite certain they are far superior to my skill with your native language.
Unfortunately, I hear some characteristics from OS sound which are desireable to my ears, and that NOS doesn't seem to provide. Which doesn't preclude me from enjoying my present DAC, which is switchable between NOS or OS mode.
By the way, no need for apologies regarding your english skills. I'm quite certain they are far superior to my skill with your native language.
One of the things we don't know so much about in digital audio is what effect the ADC used to produce the digital audio has on the ultimate best obtainable sound. Here in the forum we tend to focus much more on the reproduction aspect, as though the digitization process were done flawlessly. My own feeling would be that we may need to start taking a look at ADC implementation to help unravel some of puzzles observed in dac behavior.
Last edited:
Mark, @Lampie519,
I've no doubt about problems at the front end of the chain. While I don't know which chips are utilized within studio A/D units, such as Apogee's, I have always been bothered by the fact that most(all?) standard product ADC chips offered by the big semiconductor vendors seem to feature the audiophile's nemesis in DACs, the half-band digital Equiripple filter, for anti-alias duty.
A likely insurmountable problem in our attempting to question the industry mastering chain, is that I doubt anyone on that side cares, at all, what a few 'crazy' audiophiles think, even should we identify legit technical problems. The stereo glory days of Wilma Cozart Fine and 'Mercury Living Presence', where excellence was placed first out of respect for the music, are long gone I'm afraid. Replaced by Lawyers, MBAs, and Lawyers with MBAs - the worst combination!
There are forums for recording engineers that helps them as we have our forum here...
Music recording | The Audio Blog
Music recording | The Audio Blog
None are perfect, of course. Here in the 'digital line level' subforum we tend to focus on that aspect.
The most 'real' sounding recording I have heard was a directly to disk lathe 45rpm, 12". It was played back on a very high end system, optical phono, electrostat speakers, treated room, customized preamp and power amps. Easily more than a $50k system but I'm guessing under $100k. On that system no digital has ever compared in terms of sounding like live sound. Of course, really good digital, especially DSD, can make you think its the best there can be, but when compared to good analog it doesn't sound as real.
The most 'real' sounding recording I have heard was a directly to disk lathe 45rpm, 12". It was played back on a very high end system, optical phono, electrostat speakers, treated room, customized preamp and power amps. Easily more than a $50k system but I'm guessing under $100k. On that system no digital has ever compared in terms of sounding like live sound. Of course, really good digital, especially DSD, can make you think its the best there can be, but when compared to good analog it doesn't sound as real.
Last edited:
Sure analog can sound incredible and better than digital (a real photographer does not need a Leica to shoot great images).
Unfortunately i have never heard ANY DSD recording that could outperform a standard 16bit recording (at my place). Anyone that was convinced this would be the case was very disappointed when listening to my dual mono DAC's (NOS 16 bit 176.4Khz).
Only 1 note was required (every time, because i like to demonstrate this every time this discussion arrises) to hear the difference...
Unfortunately i have never heard ANY DSD recording that could outperform a standard 16bit recording (at my place). Anyone that was convinced this would be the case was very disappointed when listening to my dual mono DAC's (NOS 16 bit 176.4Khz).
Only 1 note was required (every time, because i like to demonstrate this every time this discussion arrises) to hear the difference...
I haven't heard the best of NOS dacs, but I know someone who has heard some of the best out there regardless of technology. He was a consultant to Wadia in their heyday. Then chief design engineer at Cary Audio. Later he was designing for Pass Labs. Although we argue about audio reproduction at times, in the end he always turns out to be right. Very humbling. Anyway, he tells similar types of stories about astonishing digital audio engineers who fully expected their digital to blow away any analog reproduction. He says they were shocked, then angry when they heard how much better good phono still sounds. He also thinks good DSD is better than good PCM, even if the audio is converted to DSD256 from 16-bit CDs. Its all a matter of how well its done. OTOH, he said there is a trade off between the best of oversampling sigma-delta and the best discrete resistor dacs (upsampled to 172.6 or not). He would like to have the best of both worlds, but we don't have that yet.
Last edited:
My opinion (for what it's worth) is that we do not need any better then 14 bit audio files for play back. So 16bit is on the safe side (as we know all 16bit dacs in the past could not deliver anyhow). Now (since 1995 or so) we can create better ladder dacs so this is no issue anymore. Delta sigma is only better on paper (still my opinion).
A Nos DAC by nature does not over- or up- sample so it needs to be able to handle the signal offered as pure as possible to be able to outperform any other. This includes the power supplies and ground handling. Here is a lot to gain against all other.
My DAC is dual mono and balanced (the latter is not a requirement), no ground connection between source and dac chips (totally floating, and i mean totally floating) The DAC can be seen as a phono cartidge with 2 individual coils, only when connecting it to a pre amp the 2 grounds will meet.
No reclocking is required but i2s needs to be time correct (between the 3 signals the clock as reference). This is most important to pevent any glitches especially if 176.4khz (or higher) signals are expected.
My transport can deliver at this moment 2x 176.4Khz (dual mono SPDIF) so here is my limit. I hope to test the new Hugo M Scaler soon.
The power supply is something else alltogether as i use a generated HV RF sinewave (with very low noise) and transform it to the desired voltages without the use of any regulators (no need as the sinewave is super stable and can deliver enough power)
As the noise of the HV sine wave will be transformed to a low voltage so will the already low noise on the wave, leaving almost no noise at all (this is the advantage of all transformers when transforming high voltages to low voltage but not the other way around, maybe nothing new here). The other advantage is that by using RF the transformer cores are small and unable to pass any 50 or 60Hz noise other then modulated on the RF sinewave.
A Nos DAC by nature does not over- or up- sample so it needs to be able to handle the signal offered as pure as possible to be able to outperform any other. This includes the power supplies and ground handling. Here is a lot to gain against all other.
My DAC is dual mono and balanced (the latter is not a requirement), no ground connection between source and dac chips (totally floating, and i mean totally floating) The DAC can be seen as a phono cartidge with 2 individual coils, only when connecting it to a pre amp the 2 grounds will meet.
No reclocking is required but i2s needs to be time correct (between the 3 signals the clock as reference). This is most important to pevent any glitches especially if 176.4khz (or higher) signals are expected.
My transport can deliver at this moment 2x 176.4Khz (dual mono SPDIF) so here is my limit. I hope to test the new Hugo M Scaler soon.
The power supply is something else alltogether as i use a generated HV RF sinewave (with very low noise) and transform it to the desired voltages without the use of any regulators (no need as the sinewave is super stable and can deliver enough power)
As the noise of the HV sine wave will be transformed to a low voltage so will the already low noise on the wave, leaving almost no noise at all (this is the advantage of all transformers when transforming high voltages to low voltage but not the other way around, maybe nothing new here). The other advantage is that by using RF the transformer cores are small and unable to pass any 50 or 60Hz noise other then modulated on the RF sinewave.
Attachments
Last edited: