KeithEmo
Member of the Trade: Emotiva
- Joined
- Aug 13, 2014
- Posts
- 1,698
- Likes
- 868
When I read your post, I see many "assertions" - which is something quite different than facts.
For example, you assert that "filter ringing should be clearly inaudible".
Why do you assume that this would be the case?
Ringing produces clearly audible effects in some speakers; and its effects were clearly audible in old-style analog equalizer circuits with L-C filters.
In fact, many EQ plugins for DAWs offer the option of choosing "vintage EQ sound" - which includes filters that introduce ringing and other sorts of "vintage circuit sounds".
Many folks assert that "finally, now, it is inaudible in most modern equipment, unless it's put there intentionally" - but that is simply an assertion.
And, yes, most modern equipment measures frequency response over time.
However, the time constants used vary quite widely.
An oscilloscope trace, especially on a modern digital oscilloscope, will show that firecracker quite clearly - it you look for it.
But the standard spectrum measurements used to measure THD, and the sort of noise sidebands used to quantify jitter effects, are usually averaged over several seconds.
There's a good reason for this: time averaging allows greater accuracy, and helps eliminate random noise and other random fluctuations.
However, it also tends to ignore, or avoid showing, short-term effects like ringing on transients (because, FOR THAT PARTICULAR TEST, those are NOT what we're trying to measure, so they're considered to be "distractions".)
And, of course, any measurement done using steady state sine waves fails to experience or show anything whatsoever about transient performance.
In fact, in order to measure things like transient response, and filter ringing, you have to use special tests and test signals designed for that purpose.
And, yes, many of the microphones and other equipment used to record audio introduce various alterations to the sound; some of which are intentional, and some of which are not.
And, yes, some of them "can... produce greater amounts of artifacts than any well designed filter... and in ranges to which we are sensitive."
However, all of those are part of the music PRODUCTION process; somebody chose that microphone because they like the way it sounds, or because they consider its flaws acceptable, NOT because they imagine it's "perfect".
When REPRODUCING that music, our goal is to play it as it is provided to us, and to avoid introducing more and different alterations.
Van Gogh may have used cheap paint, scraped off old buildings, for one of his paintings; however, when I reproduce it, I still try to do reproduce his painting, original errors and all, just as he painted it, as accurately as possible.
One time one of my customers sent in a piece of equipment for repair of a "problem" he'd discovered.
He also included a copy of the file where he'd noticed it.
His particular piece of equipment would clip this particular test signal - quite obviously - even though we were unable to notice any odd sounds with any other music.
Would you suggest that "it doesn't really matter if that sound is clearly distorted - because it's not something you encounter very often in popular music"?
(That sound was included on that test disc, along with a few other odd sounds, specifically BECAUSE it stresses equipment, and so tends to emphasize flaws that might otherwise not be obvious.)
In fact, when designing tests, we often create specific unusual test signals, designed to maximize the visibility or audibility of certain flaws.
For example, when testing a camera lens for distortion, we use a black and white checkerboard or crosshatch pattern...and, when testing it for color accuracy, we use a checkerboard of accurately known colors.
This is not something that is limited to audio testing....
The very popular JPG image compression format was designed to work well with images with continuous tones - like human faces - for which it works VERY well.
However, it works very poorly with images that include lines or sharp edges - like cartoon images and text.
If you apply a high level of JPG compression to a variety of images, faces will look very good, but text will display obvious artifacts (little ghosts around the edges of the letters).
For this reason, JPG compression is often used for photographs, but would be a bad choice for "compressing an unknown mixture of images and pictures".
(Because, not only do we KNOW that it will perform badly on certain specific types of images, but, if we see artifacts on some of our images, we won't know if they were there originally, or if the JPG processing caused them.)
So, likewise, when attempting to accurately reproduce complex signals, we generally prefer to do so in ways that we KNOW won't alter them audibly, rather than in ways that we hope or assume will not.
(And, yes, one thing that differentiates "audiophiles" from "regular consumers" is that audiophiles are willing to expend more effort and money to avoid issues that occur infrequently or aren't especially annoying.)
You really do need to differentiate between "fixing problems that really don't exist" and "fixing problems that ARE quite real, but aren't bad enough that MOST PEOPLE care about them".
For example, you assert that "filter ringing should be clearly inaudible".
Why do you assume that this would be the case?
Ringing produces clearly audible effects in some speakers; and its effects were clearly audible in old-style analog equalizer circuits with L-C filters.
In fact, many EQ plugins for DAWs offer the option of choosing "vintage EQ sound" - which includes filters that introduce ringing and other sorts of "vintage circuit sounds".
Many folks assert that "finally, now, it is inaudible in most modern equipment, unless it's put there intentionally" - but that is simply an assertion.
And, yes, most modern equipment measures frequency response over time.
However, the time constants used vary quite widely.
An oscilloscope trace, especially on a modern digital oscilloscope, will show that firecracker quite clearly - it you look for it.
But the standard spectrum measurements used to measure THD, and the sort of noise sidebands used to quantify jitter effects, are usually averaged over several seconds.
There's a good reason for this: time averaging allows greater accuracy, and helps eliminate random noise and other random fluctuations.
However, it also tends to ignore, or avoid showing, short-term effects like ringing on transients (because, FOR THAT PARTICULAR TEST, those are NOT what we're trying to measure, so they're considered to be "distractions".)
And, of course, any measurement done using steady state sine waves fails to experience or show anything whatsoever about transient performance.
In fact, in order to measure things like transient response, and filter ringing, you have to use special tests and test signals designed for that purpose.
And, yes, many of the microphones and other equipment used to record audio introduce various alterations to the sound; some of which are intentional, and some of which are not.
And, yes, some of them "can... produce greater amounts of artifacts than any well designed filter... and in ranges to which we are sensitive."
However, all of those are part of the music PRODUCTION process; somebody chose that microphone because they like the way it sounds, or because they consider its flaws acceptable, NOT because they imagine it's "perfect".
When REPRODUCING that music, our goal is to play it as it is provided to us, and to avoid introducing more and different alterations.
Van Gogh may have used cheap paint, scraped off old buildings, for one of his paintings; however, when I reproduce it, I still try to do reproduce his painting, original errors and all, just as he painted it, as accurately as possible.
One time one of my customers sent in a piece of equipment for repair of a "problem" he'd discovered.
He also included a copy of the file where he'd noticed it.
His particular piece of equipment would clip this particular test signal - quite obviously - even though we were unable to notice any odd sounds with any other music.
Would you suggest that "it doesn't really matter if that sound is clearly distorted - because it's not something you encounter very often in popular music"?
(That sound was included on that test disc, along with a few other odd sounds, specifically BECAUSE it stresses equipment, and so tends to emphasize flaws that might otherwise not be obvious.)
In fact, when designing tests, we often create specific unusual test signals, designed to maximize the visibility or audibility of certain flaws.
For example, when testing a camera lens for distortion, we use a black and white checkerboard or crosshatch pattern...and, when testing it for color accuracy, we use a checkerboard of accurately known colors.
This is not something that is limited to audio testing....
The very popular JPG image compression format was designed to work well with images with continuous tones - like human faces - for which it works VERY well.
However, it works very poorly with images that include lines or sharp edges - like cartoon images and text.
If you apply a high level of JPG compression to a variety of images, faces will look very good, but text will display obvious artifacts (little ghosts around the edges of the letters).
For this reason, JPG compression is often used for photographs, but would be a bad choice for "compressing an unknown mixture of images and pictures".
(Because, not only do we KNOW that it will perform badly on certain specific types of images, but, if we see artifacts on some of our images, we won't know if they were there originally, or if the JPG processing caused them.)
So, likewise, when attempting to accurately reproduce complex signals, we generally prefer to do so in ways that we KNOW won't alter them audibly, rather than in ways that we hope or assume will not.
(And, yes, one thing that differentiates "audiophiles" from "regular consumers" is that audiophiles are willing to expend more effort and money to avoid issues that occur infrequently or aren't especially annoying.)
You really do need to differentiate between "fixing problems that really don't exist" and "fixing problems that ARE quite real, but aren't bad enough that MOST PEOPLE care about them".
That is virtually never the case. For example, what "you DO hear" is some external soundwaves hitting your eardrums plus a rather large amount of heartbeat sound, the sound of blood moving though blood vessels and even the sound of your nervous system. What we think we hear is typically NOT closely related to what "you do hear" because, except for some rare circumstances, we never think we hear these constant, relatively loud body function sounds because our brain eliminates them from our perception, and this is just one of many similar examples. The reality of the situation is that almost never is what we think we hear "closely related" to what we do hear, the best we can truthfully say is that occasionally, what we think we hear aligns quite well with what other people think they hear.
1. True but in this particular case we're talking about filter ringing, something which should be clearly inaudible, unless it's been specifically designed to be audible.
1a. OK, let's run with your analogy. By orders of magnitude, the most common type of equipment used to examine sound is the graphical representation of sample data over time (in say a DAW/audio editor) and graphical representations of frequency content over time, both of which would blatantly obviously "reflect how audible the firecracker really is". In fact, you'd have to doctor the firecracker recording in the case of the SPL measurement or use a measurement type specifically designed not to report amplitude variations (only an average), all while avoiding the almost unavoidable other types of measurements. In other words, one would need to be deliberately trying to avoid a measurement that would "accurately reflect how audible that firecracker really is"!
1. The controlled tests I'm aware of used standard type filters, with a transition band of around 2kHz or so. What about the inverse of the question, how steep were the filters used in controlled tests where a difference could be detected? As far as I'm aware the answer is either: Steeper than is ever found in consumer equipment or extremely shallow and well into the audible range (for some "warmth" or other lower audible fidelity and supposedly subjectively better result).
2. A very good point and one often omitted in the arguments made by audiophiles (or marketed to them). How many pieces of music do you know that were NOT recorded by mics and had no EQ (or other ringing/phase inducing effect) applied during mixing or mastering? All of which can not only produce far greater amounts of ringing/phase related artefacts than any well designed/standard reconstruction filter but also in a freq range to which we're actually sensitive!
We're trying to determine scientifically/factually whether there is any audible difference between DACs designed for high-fidelity conversion when reconstructing commercial digital audio.
Why would anyone be trying to "determine, scientifically, if there is ANY audible difference between different DACs, even if it can only be detected with certain test signals"? Of course there is, that's already been determined scientifically, decades ago! It's easy to deliberately make a DAC sound different with certain test signals and there are actual, commercially released, deliberately obvious examples of this. For example there were some which did not oversample and had no reconstruction filter (the NOS/Filterless DACs) and it's simple to design a test signal which would result in the DAC producing alias images which are clearly audible or at least, easily differentiated from a DAC with a competent filter.
The thread is: "Testing audiophile claims and myths". What audiophiles are we talking about? How many audiophiles are there who listen exclusively to test signals and are not interested in fidelity? Even if there are some, the audiophile claims and myths clearly relate to reproducing commercial audio.
The above points are all misrepresentations or obfuscations, although I can't be sure if they're deliberately so or inadvertent but they are obfuscations, as per my last post (#9650), and I fail to see how any of this makes your "interest more scientific". If anything, it demonstrates to me "less scientific" or at least, less factual.
G