Sorry for the delay, a bit busy.
First lets make sure we are using the same words to mean the same thing.
An analog system is continuous time and level. A digital system is discrete time and level. A sampled (or discrete time) system is discrete time but continuous level.
A linear system follows x=c*y. No system is ever perfectly linear.
A non linear system has some added component x=c*y + d*y*y. ( c and d are any constants)
When we design a filter we can use a mathematically derived filter. There are many of these and most have a name. So if you want a minimum phase, sharp cutoff, or flat pass-band you can in theory select a filter design that will mostly give it to you.
Of course what is unanswered is which filter sounds best in a given application.
In designing or evaluating filters the two classic methods are Bode and Root-Locus. Bode is usually used for filters and Root-locus for control systems.
So when I mention that the settling of any RC filter is based on an inverse exponent I consider that to be classic engineering math. (Classic does not mean simple!)
You can look at the output of a D/A converter either as the step response or as the frequency response. Different view of the same thing in theory!
When you look at the step response of of a signal it will settle say 44 db in one sample time period. If the step is single bit then the error is small and is down by 44 db from the step size. If the step is full scale output the error is the same 44 db down but in terms of absolute output voltage -44 is a big error.
If the filter did not have the response of an inverse exponent but had a constant voltage error then it would be say 40 db down from a single bit and with a 16 bit PCM (Pulse Code Modulation) converter it would be down 136 db from a full scale step.
So the issue is what kind of filter has a fixed error rather that an inverse exponent or even constant percentage error.
A second big issue is adding some signal to the input of an A/D converter to improve the apparent sound quality.
As a given A/D converter is in theory accurate to 1/2 bit then any input should be digitized to with 1/2 bit. But at some levels it may, due to random variation, be + 1/2 bit at other levels it may be -1/2 bit or even somewhere in between. An added signal (dither or similar) bounces this error around in a method that makes this sound better to us.
However in an oversampled A/D converter with a skillfully applied signal that deliberately adds or subtracts a 1/2 bit over the full oversampling cycle, you may be able to improve the resolution of the conversion. (If you are oversampling 16 times with the proper added signal you probably can get 3 more bits of resolution, other caveats apply!)
Now if we want to produce a D/A converter that makes use of this we can use an oversampled PCM converter and some math on the digital signal to produce an output that when passed through a known filter type gives us our desired response.
That next raises the issue of real filters versus the mathematical models of them. All of the analog filters we look at use capacitors. These have "soakage" inherent in the way they are made. This adds another non-linear factor to the filter's unit impulse response.
The best solution is to use as few capacitors as possible and of the best types. These are expensive and more so the larger the values. So for now a good approach is to over-sample, use a very high rate and a small filter capacitor.
This is similar to the single bit ratio-metric approach. The limit to ratio-metric is noise distortion. So by carefull selection of the number of bits in a PCM converter and the oversampling you can design a better D/A converter.
As the technology improves the number of bits and the sampling rate also gets better.
ES
Last edited:
In some senses, yes, since I absolutely trust your ability to set up and verify the test without obvious errors in an electrical sense and to deliberately not choose a nonsymmetrical . No, in that neither you nor he have escaped having human brains and need to verify your impressions with controlled subjective tests (i.e., blind) in order for them to have any credibility. But JJ wouldn't need to be told that, and I'm certain that the first thing he'd do is try to verify that the effect is actually auditory.
My assumption here is that you're talking about an interconnect that's built symmetrically, with a result that's extraordinary. If it's asymmetric (e.g., shield grounded at one end), the result would be totally unremarkable.
j.j. (no Caps) writes in a pro-audio forum, about how difficult it is to setup an ABX test. So many unintended variables to be accounted for. As "X" can be easily identified even before getting to the heart of the test.
Absolutely. And as we are humans, the same mechanisms could help to grasp the ´real thing´ or could led to a totally false perception.
And quite the same mechanisms are still at work under blind test conditions.
In a blind test we block the "peeking" part but add a "destabilization" part as we are (normally) not used to do decisions only by _one_ sense.
Our conscious perception is the result of the reaction of our sensory system to a stimulus and the following quite complex filtering of this reaction.
Reg. se´s post; you simply can´t have it both ways
first- you have to do controlled blind tests (despite of all of your listening experience)
and
second- no we don´t need training and controls in our tests (because of all of your listening experience)
Yes. Another interesting thing I recently learned:
Did you know that of all the hearing hair cells that are groupd in a certain frequency area in your ears, only about 1/4 functions as inputs to your brain, and that the other 3/4 are actually DRIVEN from your brain to enhance the movements of the 1/4 input hair cells?
It's a positive feedback process that allows the brain to suppress freq bands it just isn't interested in at this particular moment, or to enhance the sensitivity in a particular freq band it IS interested in. So hearing a tiger roar can let the brain adjust your hearing sense to focus on the roar while you become totally unaware (literally momentary deaf) to the other sounds around you that are not so important right now.
It's mostly automatic, but you can use it to conciously 'listen' to a certain sound, for instance. I wonder how this mechanism impacts sighted and DB listening?
Anyway, knowing this, the phrase 'I hear what I hear' takes on a whole new meaning to me.
jd
Did you know that of all the hearing hair cells that are groupd in a certain frequency area in your ears, only about 1/4 functions as inputs to your brain, and that the other 3/4 are actually DRIVEN from your brain to enhance the movements of the 1/4 input hair cells?
It's a positive feedback process that allows the brain to suppress freq bands it just isn't interested in at this particular moment, or to enhance the sensitivity in a particular freq band it IS interested in. So hearing a tiger roar can let the brain adjust your hearing sense to focus on the roar while you become totally unaware (literally momentary deaf) to the other sounds around you that are not so important right now.
It's mostly automatic, but you can use it to conciously 'listen' to a certain sound, for instance. I wonder how this mechanism impacts sighted and DB listening?
Anyway, knowing this, the phrase 'I hear what I hear' takes on a whole new meaning to me.
jd
Janneman's post is exceptionally interesting as it explains what many of us have experienced in certain circumstances. Perhaps it explains why it is that becoming used to one's wife's voice can result in temporary deafness - especially when she wants money for something considered such as yet another pair of shoes!!
This whole matter of DB testing was as most of us know, covered widely if not deeply in the infamous marathon thread on cables - which had to be shut down. Can anyone with wide experience in fact describe the benefits of true DB testing?
This whole matter of DB testing was as most of us know, covered widely if not deeply in the infamous marathon thread on cables - which had to be shut down. Can anyone with wide experience in fact describe the benefits of true DB testing?
That was also done several times in the cable thread, unfortunately to little effect on the pre-Enlightenment thought process rampant in corners of the tiny, niche high end audio community.
It all sums up to this: it will distinguish what you can actually hear from what you imagine you can hear. That frees you to concentrate efforts on variables that actually make a sonic difference.
My experience- a few dozen (amateur) double blind audio tests, a few thousand (professional) double blind haptics tests, and a few thousand (professional) double blind organoleptic tests. Outside of sensory research, a few hundred (professional) double blind biochemical tests. Reasonably "wide experience."
@ janneman,
isn´t that the quite new research result about the possible working mechanism for outer hair cells helping the inner hair cells ?
@ brianco,
a proper test (which means a lot more than only being double blind) can give with a quantifiable probability the result that a certain effect is really audible.
And of course the phrase "is audible" may have a lot of different meanings. It could be "audible to at least one listener (on earth)" or "audible to the average person (means of all persons living on earth)" or "audible to experienced listeners (depending on the definition of experienced)" and so on.
In a _proper_ test it is somewhere descripted what the aim of the investigation was.
isn´t that the quite new research result about the possible working mechanism for outer hair cells helping the inner hair cells ?
@ brianco,
a proper test (which means a lot more than only being double blind) can give with a quantifiable probability the result that a certain effect is really audible.
And of course the phrase "is audible" may have a lot of different meanings. It could be "audible to at least one listener (on earth)" or "audible to the average person (means of all persons living on earth)" or "audible to experienced listeners (depending on the definition of experienced)" and so on.
In a _proper_ test it is somewhere descripted what the aim of the investigation was.
Last edited:
I have talked about that many times and i will do it here again. A hypothetical totally transparent system may not sound totally satisfying because we are listening to a recording and not reality. So it could be that a technically non perfect amplifier sounds better most of the time then a totally transparent ( or call it blameless ) one. When i develop preamps for example i use a straight wire bypass and compare input to output.
Still in the final stage of development i do some tuning so that the preamp gives a subjectively satisfying result in a real world condition.
Still in the final stage of development i do some tuning so that the preamp gives a subjectively satisfying result in a real world condition.
I'm terribly sorry, but until you can show me that a bright and accomplished fellow like jj actually said such a ridiculous thing, I do not believe you.
Now that is something I agree with.
- Status
- Not open for further replies.