jkenny
Full Member
Posts: 83
About Me: Audio equipment designer forever in pursuit of more realistic & engaging music reproduction purely because of the extra enjoyment of music created by such reproduction.
http://Ciunas.biz
|
Post by jkenny on Mar 12, 2020 6:41:30 GMT 10
I see John Swenson's white paper on the causes of digital audio audible differences without bit changes called "Understanding how perturbations on digital signals can affect sound quality without changing bits," - found here. remember this is not JUST about EtherRegen but "also about any digital audio delivery system". I have some experience in this regard, in the case of USB signal isolation & reclocking so it resonates with me. I call this thread "the bleeding edge" as I believe what it is about is exactly that in so many ways. I believe this is the area where the greatest clash between objectivists & subjectivists will take place & it will eventually reveal the issues about audibility thresholds & about the shortcomings of existing standard measurements. From my first reading of the paper, I agree with almost all statements in it as I have been saying pretty much the same thing for quite some time now. I need to re-read it but the last sentence I'm not sure I agree with "If the receiver reclocks the data with a local clock, you still have the effects of the ground plane-noise from the data causing threshold changes on the reclocking circuit, thus overlaying on top of the local clock." My experience does not correlate with this as I find reclocking does improve matters - I'm not sure what is meant by "still have the effects of the ground plane-noise from the data". I know this is related to the previous thread about measurements Vs perception but this is a specific example rather than a general discussion & so may be easier to make salient points & illustrate with real world examples. So let me give an examples to illustrate what I believe are the shortcomings in the objectivists/measurists approach. I think Amir represents a pretty good example of this category & his response here is worth examining as it is revealing. Amir dismisses the paper with a "jitter" measurement which he says shows no difference with/without EtherRegen & the statements "Translation, we have an extremely, extremely sensitive test here, way beyond anyone's hearing threshold. Yet, the Etherregen fails to show any improvement whatsoever." And conclusion "John's paper points to a truism in system design. What it misses is that the problem he thinks exists, is already solved." Maybe if Amir would zoom in on the jitter plot so we can see the spread at the base of the 12KHz signal (or whatever he uses) - otherwise know as the close-in phase noise - we might see differences between with/without ER in line. I reckon the differences will be there but will be low enough that the argument will be they are inaudible . However, that is a different argument to the one currently offered by the measurists which is that there is no difference - yea, no difference visible on the plots when you use a gross enough x-axis to hide any possibility of seeing 1Hz or <1Hz differences in signal tones. Are these FFT measurements falling into the averaging issue where close-in phase noise will be random enough to appear as "noise" & averaged out by FFTs? My understanding of close in phase noise is that it is a statistical measure of how often the clock signal will be wrong by 1Hz or less ( the 1Hz is just my random choice - it could well be more than this) - it doesn't mean that for every clock tick it is off by 1Hz. If this was the case it would be evident on an FFT as a side spur in close to the main test tone. But when it is random, sometimes off by 1Hz, sometimes 0.1Hz & everything timing error from 1Hz down, I think FFTs will see this as random noise. Amir always demonstrates to me what i consider, experimenters bias in action - no real attempt to examine what is being claimed - when this is rarely pointed out, he offers the excuse that he is not chasing measurements at others whim - in other words, it's very like 5th elements post - "we have our fascist regime & don;t need anybody questioning it's rules" Amir either knowingly or lazily using the standard measurements & offering this as "proof" of no change when the measurement is patently unable to reveal what is being claimed. I think this is the crux at the heart of a lot of objectivists - they have set beliefs about audibility & are not in the least inquisitive enough to examine these beliefs & hence all their "evidence" is curtailed within this belief system - in other words Amir won't a zoomed in FFT or question whether an FFT will miss what is being claimed because he believes even if it did show some differences it would be below audibility & not worth bothering about. I don't want this to be a dump on Amir but he does present the standard arguments from that camp that many listen to & that seem reasonable when they are presented with measurements Correct me if my understanding is wrong but FFTs (as normally run) are a flawed attempt to try to measure what is claimed - close-in phase noise. Wouldn't it require a very long time frame over which the FFT is run to reveal these semi-random signals? I know Jocko uses a very long FFT run to reveal close-in phase noise on clocks. Here's one of his measured & selected low phase noise NDK clocks: oops, how do I upload/link to an image from Dropbox/Google drive or some other hosting site ?   link |
|
sandyk (RIP Alex, 1939 - 2021)
Global Moderator 
Posts: 226
About Me: Retired ex Principal Telecommunications Technical Officer with 43 years at Telstra (Australia)
I am also a Moderator in Hi Fi Critic Forum
Electronics hobbyist for >65 years with DIY projects including Loudspeakers, Stereo FM tuner, S/W Regen Receiver, Superhet AM ,
Synchrodyne PLL AM tuner (Phase Lock Loop),Stereo Tape Deck, Amplifiers including I.C. types, Class A, Class AB 100W/Ch. (ETI5000) 240W/Ch. Mosfet (AEM6000) ,several DACs , numerous PSUs including VERY low noise (<4uV) types etc.for myself and friends
Audio Industry Affiliation: NIL
|
Post by sandyk (RIP Alex, 1939 - 2021) on Mar 12, 2020 7:35:09 GMT 10
I see John Swenson's white paper on the causes of digital audio audible differences without bit changes called "Understanding how perturbations on digital signals can affect sound quality without changing bits," - found here. remember this is not JUST about EtherRegen but "also about any digital audio delivery system". I have some experience in this regard, in the case of USB signal isolation & reclocking so it resonates with me. I call this thread "the bleeding edge" as I believe what it is about is exactly that in so many ways. I believe this is the area where the greatest clash between objectivists & subjectivists will take place & it will eventually reveal the issues about audibility thresholds & about the shortcomings of existing standard measurements. From my first reading of the paper, I agree with almost all statements in it as I have been saying pretty much the same thing for quite some time now. I need to re-read it but the last sentence I'm not sure I agree with "If the receiver reclocks the data with a local clock, you still have the effects of the ground plane-noise from the data causing threshold changes on the reclocking circuit, thus overlaying on top of the local clock." My experience does not correlate with this as I find reclocking does improve matters - I'm not sure what is meant by "still have the effects of the ground plane-noise from the data". I know this is related to the previous thread about measurements Vs perception but this is a specific example rather than a general discussion & so may be easier to make salient points & illustrate with real world examples. So let me give an examples to illustrate what I believe are the shortcomings in the objectivists/measurists approach. I think Amir represents a pretty good example of this category & his response here is worth examining as it is revealing. Amir dismisses the paper with a "jitter" measurement which he says shows no difference with/without EtherRegen & the statements "Translation, we have an extremely, extremely sensitive test here, way beyond anyone's hearing threshold. Yet, the Etherregen fails to show any improvement whatsoever." And conclusion "John's paper points to a truism in system design. What it misses is that the problem he thinks exists, is already solved." Maybe if Amir would zoom in on the jitter plot so we can see the spread at the base of the 12KHz signal (or whatever he uses) - otherwise know as the close-in phase noise - we might see differences between with/without ER in line. I reckon the differences will be there but will be low enough that the argument will be they are inaudible . However, that is a different argument to the one currently offered by the measurists which is that there is no difference - yea, no difference visible on the plots when you use a gross enough x-axis to hide any possibility of seeing 1Hz or <1Hz differences in signal tones. Are these FFT measurements falling into the averaging issue where close-in phase noise will be random enough to appear as "noise" & averaged out by FFTs? My understanding of close in phase noise is that it is a statistical measure of how often the clock signal will be wrong by 1Hz or less ( the 1Hz is just my random choice - it could well be more than this) - it doesn't mean that for every clock tick it is off by 1Hz. If this was the case it would be evident on an FFT as a side spur in close to the main test tone. But when it is random, sometimes off by 1Hz, sometimes 0.1Hz & everything timing error from 1Hz down, I think FFTs will see this as random noise. Amir always demonstrates to me what i consider, experimenters bias in action - no real attempt to examine what is being claimed - when this is rarely pointed out, he offers the excuse that he is not chasing measurements at others whim - in other words, it's very like 5th elements post - "we have our fascist regime & don;t need anybody questioning it's rules" Amir either knowingly or lazily using the standard measurements & offering this as "proof" of no change when the measurement is patently unable to reveal what is being claimed. I think this is the crux at the heart of a lot of objectivists - they have set beliefs about audibility & are not in the least inquisitive enough to examine these beliefs & hence all their "evidence" is curtailed within this belief system - in other words Amir won't a zoomed in FFT or question whether an FFT will miss what is being claimed because he believes even if it did show some differences it would be below audibility & not worth bothering about. Correct me if my understanding is wrong but FFTs (as normally run) are a flawed attempt to try to measure what is claimed - close-in phase noise. Wouldn't it require a very long time frame over which the FFT is run to reveal these semi-random signals? I know Jocko uses a very long FFT run to reveal close-in phase noise on clocks. Here's one of his measured & selected low phase noise NDK clocks: oops, how do I upload an image from Dropbox?link Hi John Very interesting and thanks for posting your perspective on this as well .
Kind Regards Alex
Put ![]() ![]() in front, and raw=1 [/IMG] after the question mark[/p] 
|
|
sandyk (RIP Alex, 1939 - 2021)
Global Moderator
Posts: 226
About Me: Retired ex Principal Telecommunications Technical Officer with 43 years at Telstra (Australia)
I am also a Moderator in Hi Fi Critic Forum
Electronics hobbyist for >65 years with DIY projects including Loudspeakers, Stereo FM tuner, S/W Regen Receiver, Superhet AM ,
Synchrodyne PLL AM tuner (Phase Lock Loop),Stereo Tape Deck, Amplifiers including I.C. types, Class A, Class AB 100W/Ch. (ETI5000) 240W/Ch. Mosfet (AEM6000) ,several DACs , numerous PSUs including VERY low noise (<4uV) types etc.for myself and friends
Audio Industry Affiliation: NIL
|
Post by sandyk (RIP Alex, 1939 - 2021) on Mar 12, 2020 7:51:31 GMT 10
John
I sent you an email as the info I put about inserting images kept getting changed and is incorrect in my reply
Regards
Alex
|
|
jkenny
Full Member
Posts: 83
About Me: Audio equipment designer forever in pursuit of more realistic & engaging music reproduction purely because of the extra enjoyment of music created by such reproduction.
http://Ciunas.biz
|
Post by jkenny on Mar 12, 2020 9:21:15 GMT 10
Thanks Alex - good tip! OK, what I wanted to illustrate with Jocko's FFT of clock phase noise are a number of points about FFTs which I may be wrong about & am open to be corrected. FFT & random signals (noise) or semi-random signals (phase noise) are a bit difficult to get ones head around. Someone gave me a good analogy once which helped me get a handle - FFTs are like long exposure photographs - the parts of the image that are unchanging will be clearly rendered, the parts that move not so clear & even invisible if they are random events happening during the exposure time. Let's take Amir's FFT used as "proof" that the EtherRegen does nothing, as a case study. He runs a 12KHz test signal through his audio system with/without EtherRegen. he calls this "Jitter noise & spectrum, 1 million point FFT, 8 averages" An FFT is a plot of the series of frequency bins into which the signal bandwidth is split - 1 million bins in this case. So 20KHz bandwidth (I presume) splits into 0.02Hz per FFT bin - all good, so far. The 12KHz test signal is clearly shown as a main spur @ 0dB amplitude & at 12KHz frequency i.e it is showing a high energy in all the bins which span this 12KHz freq range - just like the long exposure photo will show the high energy of the accumulated photons hitting the same place on the film or same pixels on the camera sensor. If this was a signal purely rendered, the spur would be a fine, sharp pencil line. At the bottom of this spur we see a spread which is the amount of deviation from this pure 12KHz. Why is it just a widening of the base & not a widening of the whole spur? Because the deviations are more random (i.e noise-like) so the energy in the frequency bins near the 12KHz freq is lower than 0dB. So what if the FFT spur was zoomed into with the x-axis showing the actual bin freqs i.e .02Hz? Would that show a pencil thin spur line or some thickening all through the spur? Would it show a difference between with/without EtherRegen? I'll show a plot below which illustrates this better view into FFTs. An FFT is also run on a signal acquired over a short period of time. To be able to gather a relevant amount of energy in semi-random frequency events, a much longer acquisition time is required. jocko's plots of close-in phase noise needs something like 20 hours runtime. So when someone claims that close-in phase noise is at the heart of the action of a device it requires a different approach then the usual FFT. Now, even if these frequency deviations are shown, what would be the perceptual effect of these deviations? To my mind & I think it's state din Swenson's paper, all frequencies are effected with some level of frequency smear. This resonates with me as in all my experiments with my digital audio devices I & many others experience more solidity to each individual instrument/voice. As a result the soundstage is perceived as much more solid & real with most people also reporting the improved bass now with inner texture & sounding louder. This is the example of the FFT spur being zoomed in - from Joe Rasmussen here (Alex probably knows him or of him?) The first plot is a standard FFT which is actually two FFTs overlaid - one black (different clock) & the other red (stock clock)  And now the zoomed in FFT - the difference can be clearly seen & the clock giving the black FFT was considered to sound better  So what we see is that the red side shoulders of the plot show when the frequency of the test signal is not exact - in error between 0 & 40Hz either side of the main tone. But it looks like it's -90dB from the main tone? But here's where the head twist sets in - remember the FFT is showing the energy collected in bins. So looking at the red section at 40 Hz away from the main tone, the number of times this happens is seldom so the energy in these bins is low. But the tone is replayed/produced at the same level as the main tone, but very seldom, hence the low energy in these bins. It doesn't mean that the tone+40Hz (& tone-40Hz) is produced at -130dB down. I'm pretty sure we perceive the energy in a signal so a couple of random clock cycles that is 40Hz either side of the main tone will not be perceived? As we move closer to the main tone, the frequency of error increases so close-in we have freq errors nearly as regular as the main tone & perceptually more noticeable. I've seen people dismiss these frequency deviations being masked by the main tone but I don't buy this. I would need more in-depth knowledge about the exact characteristics of this jitter - how it arises across time. Depending on the origin of the jitter, my understanding is that it's not purely random, one clock tick being exactly 1KHz & the next 1001Hz & the next being 998Hz, etc. (these random changes do happen because of the physics of crystal structures & how they oscillate). If it comes from internally created ground noise of receiver ICs at the front of DAC devices, then it will have a pattern as the ground noise fluctuates in tandem with the digital signal being processed. If it originates from power supply fluctuations then again it will have a pattern in tandem with the PS fluctuations. In both of these cases the frequency will gradually change from 1KHz to 1010 or whatever the maxima is - same with minima freqs. I doubt this sort of frequency smearing goes unperceived or at least its presence isn't perceived until it is removed & then it is not directly perceived as a lack of freq smear but rather as a better solidity & definition to instruments/vocals & soundstage - just as we see reported. |
|
|
|
Post by ROWUK on Mar 13, 2020 5:48:08 GMT 10
I think that it is a stretch to, at this stage attribute sonic characteristics to the "non-random" jitter.
In music, there is no "smearing" of frequencies for normal instruments rather sum and difference tones. The "original" music is there. if we have simple music like solo flute, we often can hear modulation from other sources - wow and flutter from an analog source for instance. In the case of "non random" jitter, a concert "A" at 440 Hz would be modulated by for instance a 12kHz "tone" producing 11560Hz (12k - 440) and 12440Hz (12k+440). These two resultant tones are NOT harmonically related to the 440Hz from the flute and could perhaps "colour" the tone with "fake" harmonics depending on amplitude. As these are in the upper harmonics area where our sensitivity to pitch is not as great, it would only change timbre slightly - not sense of pitch. Instruments like oboe or trumpet that are very rich in uneven harmonics would surely mask any effect - or be perceived as more brilliant with the additional non harmonic information (if it exists in this form). If the playback system is not phase aligned - all bets are off about any audibility.
To modulate the part of music that defines our sense of pitch, we would need appreciable resultant tones between 100 and 1000 Hz where the fundamentals in real music are. We also would need very low uneven harmonic distortion (3rd/5th/7th)to not mask the effect. This would make it very difficult to hear if we for instance were using PP amplifiers (masking), but probably easier to hear with a single ended amplifier. If we do have modulation, it would be measurable.
Definition of the soundstage in my experience happens below 500 Hz where the spatial cues are. above that, phase is a very big part of our perception. I have trouble imagining where jitter side bands could influence this.
In the 1970s I participated in listening experiments with a device called the "Aphex" aural expander. Here they asymmetrically clipped music signals to add uneven harmonics at low levels. Most listeners liked the effect.
With the various threads here about the quality of our sources, I really wonder about what the threshold of audibility is for the music that we listen to. Any type of digital processing would seem to "colour" our sound to a far greater degree than non random jitter - or the problem lies elsewhere.
I would think that it is possible to construct a test to at least quantify what is different - whether the perceived "differences" are due frequency (resultant tones), distortion (harmonically related or not), noise or phase. Defining the threshold of audibility would be a far more daunting task.
|
|
jkenny
Full Member
Posts: 83
About Me: Audio equipment designer forever in pursuit of more realistic & engaging music reproduction purely because of the extra enjoyment of music created by such reproduction.
http://Ciunas.biz
|
Post by jkenny on Mar 14, 2020 21:48:49 GMT 10
Thanks, it's good to be challenged with alternative reasoned views - it makes one examine one's hypothesis in more detail, from other angles.
I guess my primary evidence for some effect taking place is based on exactly the low phase noise oscillator I showed a graph of above. I had always wanted to do this experiment but could never guarantee that an oscillator qualified as low close-in phase noise until this one (I wasn't willing to spend huge money on OCXOs). Then jocko started to measure NDK oscillators, categorising them into close-in phase noise bands
I simply changed the audio oscillators in my DAC to these 22.5792 & 24.576MHz low phase noise versions (nothing else changed) & the sound was audibly improved in the ways I mentioned.
Now, I agree, I might have settled on the wrong conclusion from this experiment (& others which are no so clear cut) but it became my working premise.
In normal music (being played by humans) is there not fluctuations/modulations in the sound envelopes produced & capturing these micro-variations would be important for creating the illusion during playback of realism? Just like voices in choirs which go in & out of harmonic synch (resonance?) is important to reproduce in replay to more realistically represent a real choir?
So I might be incorrect in stating that it's only smearing of the fundamental freq that is at play, it's undoubtedly more complicated & may involve non-random jitter also slightly modifying sound envelope structure (i.e the timbre).
Richard Dudley - Abraxalito (on DIYA), Opus101(on AS/CA) - has a theory that it is LF noise which is at the heart of the phenomena - higher jitter through IMD folding down into the LF spectrum. I don't discount this either as improving power supply, common mode noise, IMD could all be operating (at least partially) through this mechanism.
|
|
sandyk (RIP Alex, 1939 - 2021)
Global Moderator
Posts: 226
About Me: Retired ex Principal Telecommunications Technical Officer with 43 years at Telstra (Australia)
I am also a Moderator in Hi Fi Critic Forum
Electronics hobbyist for >65 years with DIY projects including Loudspeakers, Stereo FM tuner, S/W Regen Receiver, Superhet AM ,
Synchrodyne PLL AM tuner (Phase Lock Loop),Stereo Tape Deck, Amplifiers including I.C. types, Class A, Class AB 100W/Ch. (ETI5000) 240W/Ch. Mosfet (AEM6000) ,several DACs , numerous PSUs including VERY low noise (<4uV) types etc.for myself and friends
Audio Industry Affiliation: NIL
|
Post by sandyk (RIP Alex, 1939 - 2021) on Mar 15, 2020 7:15:04 GMT 10
Thanks, it's good to be challenged with alternative reasoned views - it makes one examine one's hypothesis in more detail, from other angles. I guess my primary evidence for some effect taking place is based on exactly the low phase noise oscillator I showed a graph of above. I had always wanted to do this experiment but could never guarantee that an oscillator qualified as low close-in phase noise until this one (I wasn't willing to spend huge money on OCXOs). Then jocko started to measure NDK oscillators, categorising them into close-in phase noise bands I simply changed the audio oscillators in my DAC to these 22.5792 & 24.576MHz low phase noise versions (nothing else changed) & the sound was audibly improved in the ways I mentioned. Now, I agree, I might have settled on the wrong conclusion from this experiment (& others which are no so clear cut) but it became my working premise. In normal music (being played by humans) is there not fluctuations/modulations in the sound envelopes produced & capturing these micro-variations would be important for creating the illusion during playback of realism? Just like voices in choirs which go in & out of harmonic synch (resonance?) is important to reproduce in replay to more realistically represent a real choir? So I might be incorrect in stating that it's only smearing of the fundamental freq that is at play, it's undoubtedly more complicated & may involve non-random jitter also slightly modifying sound envelope structure (i.e the timbre). Richard Dudley - Abraxalito (on DIYA), Opus101(on AS/CA) - has a theory that it is LF noise which is at the heart of the phenomena - higher jitter through IMD folding down into the LF spectrum. I don't discount this either as improving power supply, common mode noise, IMD could all be operating (at least partially) through this mechanism. John I changed the 24.576MHZ 50ppm type in my old X-DAC V3 to a .1PPM TCXO with improved power, and noted a marked improvement after a couple of minutes when it suddenly appeared to jump up a notchy in SQ. The result was more obvious with 24/96 and 24/192, but still a worthwhile improvement at 16/44.1. I can't state for sure though that it was a lower phase noise version, as I was unable to find the specifications for the unit used , other than it being marked .1PPM
Thanks for the heads up on Richard. I didn't realise that, or I would have given more respect in some of his A.S. posts,. and not regard him as just like another closed minded Mansr .
Kind Regards Alex
|
|
|
|
Post by ROWUK on Mar 15, 2020 21:58:01 GMT 10
I do not see my post as a "challenge". The "evidence" simply does not line up with my view of the supposed problem. The first issue that I have is the term "Sound Quality". As I have often mentioned, "differences" or "preferences" do not always translate to better or worse in my world. Months go by before I can make a "difference" a "preference". I do not AB/ABX or DBT during this time. I listen to a wide range of music with a large range of personal moods. Then after a couple months, I switch back and see what happens. I believe that acclimation is one of the most necessary processes in great playback. Stereo is missing so much information and relies on our perception to become a plausible 3D image. Getting back to the issue: Here 2 different effects seem to be most prevalent in online reporting. Perception of distortion or harshness (probably a HF phenomenon) and perception of space and geometry (probably a LF phenomenon). Funny enough, I mostly do not hear these - at least not in the posted abundance of others. What I hear is mostly how well defined the "beginnings" or articulation as the start of each note that is played. A trumpet needs several cycles before the pitch has full amplitude. This "colours" the attack in a special way. The same applies to any other acoustic instrument. The hammer of a piano responds differently than the quill of a harpsichord. This is NOT generally a HF or transient issue. The "problem" here is that we have to consider the directional characteristics of the instrument or voice and the position of the microphone relative to that voice. A microphone at 12" will give us a different representation of the same voice compared to a microphone at 30 feet. In most commercial recordings, we get a mix of microphones that dramatically distorts the geometry and harmonic content relative to one another. We have a voice located in a probable position geometrically, but have the harmonic content of an unnatural listening position (close to the source). What would improve this "quality"? Faster or slower transient response? What would be the target of our efforts in this case? Let us take the example of a massed choir or string recording. We talked about "pitch smearing". In the musicians world, this is an expressive technique called vibrato and is unique to each individual musician. Vibrato is varied by loudness, intensity, frequency, position in the musical line. From an engineers standpoint it is random as the human state reacts relative to the moment, not predictably. IN a good choir or string recording, each voice has its own physical position in the geometry, intermodulating with the voices in various degrees around it. In real life, we do not have pinpoint imaging for the most part, still we do not perceive varied vibrato throughout the ensemble as a loss of clarity or quality as the effects are random and masked by the musical intent. If we would take a recording and modulate the entire thing, our brain would not randomise the effect. This would be defined in the analog world as "wow and flutter". This is NOT what I hear in any of the music samples offered. Alex brought up something else recently which was an ear opener for me. People with hearing deficiencies can have a higher sensitivity for certain types of events than others. This makes perfect sense to me. With a handicap, our bodies use the tools left over to complete the expected picture as much as possible. Subconscious strategies are developed to help us protect ourselves (fight or flight). How we could factor these into the difference of perception is worth investigating. So, I am convinced that we are hearing the same differences BUT expressing them in different frameworks. Jitter is a convenient measure, but I cannot see the relationship between its effects and what I hear. I do not hear distortion. What I hear could actually be happening in the frequency response domain. The harshness and space/geometry differences equally so. Thanks, it's good to be challenged with alternative reasoned views - it makes one examine one's hypothesis in more detail, from other angles. I guess my primary evidence for some effect taking place is based on exactly the low phase noise oscillator I showed a graph of above. I had always wanted to do this experiment but could never guarantee that an oscillator qualified as low close-in phase noise until this one (I wasn't willing to spend huge money on OCXOs). Then jocko started to measure NDK oscillators, categorising them into close-in phase noise bands I simply changed the audio oscillators in my DAC to these 22.5792 & 24.576MHz low phase noise versions (nothing else changed) & the sound was audibly improved in the ways I mentioned. Now, I agree, I might have settled on the wrong conclusion from this experiment (& others which are no so clear cut) but it became my working premise. In normal music (being played by humans) is there not fluctuations/modulations in the sound envelopes produced & capturing these micro-variations would be important for creating the illusion during playback of realism? Just like voices in choirs which go in & out of harmonic synch (resonance?) is important to reproduce in replay to more realistically represent a real choir? So I might be incorrect in stating that it's only smearing of the fundamental freq that is at play, it's undoubtedly more complicated & may involve non-random jitter also slightly modifying sound envelope structure (i.e the timbre). Richard Dudley - Abraxalito (on DIYA), Opus101(on AS/CA) - has a theory that it is LF noise which is at the heart of the phenomena - higher jitter through IMD folding down into the LF spectrum. I don't discount this either as improving power supply, common mode noise, IMD could all be operating (at least partially) through this mechanism. |
|
jkenny
Full Member
Posts: 83
About Me: Audio equipment designer forever in pursuit of more realistic & engaging music reproduction purely because of the extra enjoyment of music created by such reproduction.
http://Ciunas.biz
|
Post by jkenny on Mar 16, 2020 6:27:56 GMT 10
John I changed the 24.576MHZ 50ppm type in my old X-DAC V3 to a .1PPM TCXO with improved power, and noted a marked improvement after a couple of minutes when it suddenly appeared to jump up a notchy in SQ. The result was more obvious with 24/96 and 24/192, but still a worthwhile improvement at 16/44.1. I can't state for sure though that it was a lower phase noise version, as I was unable to find the specifications for the unit used , other than it being marked .1PPM
Thanks for the heads up on Richard. I didn't realise that, or I would have given more respect in some of his A.S. posts,. and not regard him as just like another closed minded Mansr .
Kind Regards Alex
That's always been the problem with clock changes - they usually accompany power changes also (3rd party clocks, for instance) & it's difficult to know what is responsible for any SQ improvement. Yes, Richard is one of the good EEs - he has an open mind on many things in audio electronics |
|
jkenny
Full Member
Posts: 83
About Me: Audio equipment designer forever in pursuit of more realistic & engaging music reproduction purely because of the extra enjoyment of music created by such reproduction.
http://Ciunas.biz
|
Post by jkenny on Mar 16, 2020 7:16:23 GMT 10
I do not see my post as a "challenge". The "evidence" simply does not line up with my view of the supposed problem. Sure, challenge was the wrong word & I didn't mean it in any aggressive way. I agree but I do not take a couple of months, depending on playtime usually a week is enough I was just comparing two DACs & the difference that I could spot was that one portrayed the slight vibrato in a solo voice more realistically than the other (seems to curtail or muffle the sweetness) - not outlandishly so but if you listen for it, it's there. But I think that some of the difference is that I could only use Direct Sound with one & could use Kernel Streaming with another. If I switched the KS DAC to DS the difference was lessened & might even be gone? But that small difference in vibrato rendering brings a sweetness & realism to the sound that makes it far more interesting to listen to - so what is a small A/B difference translates into a far different listening experience I don't analyse the recording to the depth that you do, I simply judge the playback - see my example of solo voice slight vibrato. It's a song that you guys might like to check out - a Scottish band - Admiral Fallow - first album "Boots Met My Face" - track "Bomb through the Town" - the intro has a female solo voice with very slight vibrato - you will get a hint of it here I guess this is an example of what I'm talking about - if you can download a WAV version of this song & play it on a good system that delivers the sweetness/realism in the voice & maybe compare to a system that doesn't quiet get there, you will get what I mean. It's this realism that I find differentiates the good systems from the ones that "play all the notes in the right places" but miss this extra "je ne sais quoi". It's the je ne sais quoi in playback that I'm trying to get to the bottom of But I'm not saying the effect of close-in phase noise is wow & flutter Yes, I don't hear distortion either - see above |
|
|
|
Post by ROWUK on Mar 16, 2020 8:11:33 GMT 10
I will let you know about the vibrato as soon as the CD gets here.
|
|
jkenny
Full Member
Posts: 83
About Me: Audio equipment designer forever in pursuit of more realistic & engaging music reproduction purely because of the extra enjoyment of music created by such reproduction.
http://Ciunas.biz
|
Post by jkenny on Mar 16, 2020 9:01:51 GMT 10
Oh, I didn't mean you to buy a CD just for this but if the music is to your liking then it's an investment
I like this album - find it interesting both lyrically & musically. It seems autobiographical about growing up in Scotland. Musically interesting with flute, clarinet & I like the harmonising between the female & male lead singers.
I use it as my test record for the sweetness of the female (Sarah Hayes) voice, the Scottish accent of the main singer/songwriter (Louis Abbot) can sound very sibilant on lesser systems & just right on good systems - as well, a lot of songs build to a dense crescendo of multi instruments which can sound garbled noise-like on lesser systems but good systems reveal the different instrument strands within this crescendo mix
Hope you enjoy it - it's one of my favourites
|
|
|
|
Post by ROWUK on Mar 17, 2020 3:29:19 GMT 10
John,
the beauty of knowing people like you is the ability to share such things. Most forums just share how much we should tighten the screws on the woofer or phono cartridge...
If a recording is special enough for you to mention it in the context of sound quality, I am sure that it will be special for me too - perhaps not daily listening, but we will see (hear). If Corona in Europe keeps going like it is, I could end up with a 4 week stretch of home office. Then there will be an even greater need for variety!
|
|
jkenny
Full Member
Posts: 83
About Me: Audio equipment designer forever in pursuit of more realistic & engaging music reproduction purely because of the extra enjoyment of music created by such reproduction.
http://Ciunas.biz
|
Post by jkenny on Mar 17, 2020 4:59:51 GMT 10
Thanks
Yes, self-isolation will require some variety to keep the brain active & some good music is a start
|
|
jkenny
Full Member
Posts: 83
About Me: Audio equipment designer forever in pursuit of more realistic & engaging music reproduction purely because of the extra enjoyment of music created by such reproduction.
http://Ciunas.biz
|
Post by jkenny on Mar 28, 2020 5:17:42 GMT 10
|
|
