Just to cut to the chase, what timing differences or time differences will someone hear with an in-ear monitor, transducer spaced equally at 5mm from the tymphanic membrane?
No crossfeed, binaural, impulse convolvers et cetera, just normal stereo or mono music.
The first one will illustrate what micro dynamic range sounds like.
The second is a string of fast impulses, in x / y / z transducer the impulses will converge at different speeds.
Anything unnecessary or already available in the data in such a test?
Typically 0.1 ms or so of delay between channels/ears, and maybe 10 ms or more when uniform between the ears (but that is out of memory, haven't checked the textbooks).
Yes, useful for everyone to do just to hear fro themselves that the only thing affected by the number of bits is signal-to-noise ratio.
Yes, at the speed where you run out of high frequency range. The convergence will be due to the low pass filter that your transducer is.
The tests might be useful for you so you can hear for yourself. It won't reveal anything new not already part of common knowledge to engineers.
Wikipedia doesn't really explain dynamic range or audio bit depth in full detail so we can conceptualize it.
A 4-bit audio file without dither sounds pretty normal, it seems like 16-bit / 96 kHz is the next step above redbook, not 24-bit / 48 kHz, like certain blu-ray discs.
Minimum-phase should be considered in a DAC, you achieve more realistic sound quality,
despite a little phase error which is trivial.
Non-oversampling / filterless is the most useful / accurate in low latency applications requring a digital pathway / software.
Nyquist-Shannon needs infinity on paper, in
virtual simulation or in calculus, not in reality.
The Metrum Acoustics Octave, Phasure NOS1 and Chord Hugo are taking part in fake measurements or fake advertising in a few areas.
The PCM1794/-5/-2 is a 6-bit thermometer DAC hybrid, perhaps this makes it sound better than Sigma-Delta, just a theory. Look up "mother of tone conversion".
Electrostatic / isodynamic have perceivable directivity / radiation patterns, thus sounding different.
The settling time difference in OPA2111KP versus AD797, AD8610 et cetera is very easy to hear.
I try to correlate what I hear to the PDF datasheets and that is my theory.
The positive feedback article on DSD provides a theory why video op-amp slew rate / settling time should be used in the LPF of a Current-out DAC, or elsewhere.
Imagine a pulse train at 1 kHz, spaced at intervals like 50 ms, 10 ms, 1 ms, 500 us, 15 us, 7 us, 750 ns, et cetera.
You're saying the 1 kHz convergence is the equivalent to the high frequency extension?
Like normal SPL at 25 kHz then nearly silent at 27 kHz.
A pulse train (as opposed to a single pulse) at 1 kHz has a period of 1 ms. Of course it "converges" with another pulse train, spaced at less than one cycle apart. What is your point?
Two short pulses converge at an interval that corresponds to the highest frequency the system can reproduce - as they get low pass filtered/integrated.
Those are just random statements. Not sure they are very meaningful.
You really don't know what calculus is, do you?
And some of us have first studied the theory for years at university level, and then spent even more years applying it. The data sheets assume a certain level of fundamental theoretical knowledge. They are not there to teach you the basics.
A bit like how it must be true if it is on TV?
Sorry if I sound a bit annoyed, but you seem to come up with all these rather random claims that basically show you are throwing around words you don't really understand. It would be OK if you presented it as a real willingness to learn, but instead you throw out claims and statements totally unsupported by either theory or evidence - and some of them just don't make any sense.
Alright so when using an in-ear monitor i.e. uniform between the ears, then the fastest ITD is around 10 ms at 1 kHz, plus the fastest pulse train perceivable is around 1ms intervals at 1 kHz, is this right.
What is the fastest settling time humanly perceivable in your view, around 20 usec?
This is how long it takes for the pulse to settle to 0.01, so it's like decay and should fuzz the time up in a pulse train......
What is the fastest settling time humanly perceivable in your view, around 20 usec?
This is how long it takes for the pulse to settle to 0.01, so it's like decay and should fuzz the time up in a pulse train......
If you are talking about interaural time difference (ITD), the signal is, by definition, *not* uniform between the ears.
And your "pulse train" at 1 kHz has 1 ms intervals by definition too, so yes, if you can hear an 1 kHz signal, you can hear something with a period of 1 ms
. I guess I have to ask you "intervals between what and what?".It will be limited by the 20 kHz limit, so 25 usec or so in the most ideal conditions.
Hi Kastor L
Unfortunately no one useful info is provided into this thread.
Regarding WM8741 and PCM1792a:
WM8741 can be configured to work either in hardware mode (no need for microcontroller) or in software mode. Not all functionality is available in hardware mode but OSR setting it is, through the use of a 3 position switch.
PCM1792a is working only in software mode, i.e. there is need of microcontroller to access the configuration registers of all functions provided by this DAC.
Both DACs allow the user to select among 3 OSR levels for the Σ - Δ modulator. Low (32 to 48KHz), Mid (96KHz) and High (192KHz).
WM8741 looks like more flexible as except the hardware mode it also offers an extra set of "PCM digital filters" before the Σ - Δ modulator (in which only the OSR is applied).
When the High OSR is selected in WM8741 it converts flawlesly signals of 192KHz, that is checked on actual hardware. For PCM1792a i do believe the same, but i haven't tried this DAC so far.
In both WM8741 and PCM1792a their internal filters can be bypassed.
Particularly, in WM8741 this function is called 8FS mode and is accessible only in software mode (microcontroller use). When is activated, the PCM signal is applied only up to the volume control module and is returned to the analog section of DAC.
A simillar process named DTFH applies to PCM1792a.
Both DACs offer this functionality to allow the use of an external DSP or digital filter to perform the OSR.
I don't know if and how this functionality could be used for a NOS DAC.
Thanks
"what timing differences or time differences will someone hear with an in-ear monitor, transducer spaced equally at 5mm from the tymphanic membrane?"
"Typically 0.1 ms or so of delay between channels/ears, and maybe 10 ms or more when uniform between the ears"
"Imagine a pulse train at 1 kHz, spaced at intervals like 50 ms, 10 ms, 1 ms, 500 us, 15 us, 7 us, 750 ns, et cetera."
^spaced = intervals of silence
"A pulse train (as opposed to a single pulse) at 1 kHz has a period of 1 ms. Of course it "converges" with another pulse train, spaced at less than one cycle apart. What is your point?"
"Alright so - the fastest pulse train perceivable is around 1ms intervals at 1 kHz, is this right."
^Just repeating your answer
"This is how long it takes for the pulse to settle to 0.01, so it's like decay and should fuzz the time up in a pulse train...... "
^You can not skip this decay part
"your "pulse train" at 1 kHz has 1 ms intervals by definition too, so yes, if you can hear an 1 kHz signal, you can hear something with a period of 1 ms."
^so funny
"I have to ask you "intervals between what and what?""
^to start with, I think it's better to change this to 10 kHz, 125 usec.
The intervals are silence like 10 ms, 1 ms, 500 usec, 125 usec, 75 usec.
125 usec silence = 250 usec peak to peak.
Is it all clear now?
I want to write more but it's a real hassle using my phone, I think I'll leave the forum for a while now.
Anyone else is free to write anything as well you know, it's usually a two-way conversation in here like a phone call.
C u
Does pre-Columbian trans-Atlantic trade have anything to do with your university study or applied field of work?
Honest question by the way.
A single-cycle gated burst is not bandlimited, as it contains frequency components up to infinity. (Oops - I alarmed Kastor again!). Therefore the signal does not meet Nyquist. Therefore anything is possible. In reality what you will get depends on exactly where the sampling points lie on the waveform; for a bandlimited signal this is not the case.Ken Newton said:
This discussion is beginning to resemble one I once had with a friend. He was a dental technician, and probably did no science or maths beyond about 16. At that time I was in the middle of doing postgraduate research on quantum gravity. He asked me to explain quantum mechanics to him and then on the basis of that brief explanation (which he could not understand, as he lacked the required background) he assured me that QM was obviously bunk.
I would suggest that someone who has never studied calculus is not able to properly understand Fourier or Nyquist, so is not able to properly understand sampled data systems. Orthogonal basis sets on an inner product space is part of the stuff to be understood. FFT, which is what most engineers use in practice, is itself a sampled data system - which causes no end of confusion to those who don't realise this. Even with the right mathematical background it took me a few years to understand this stuff, and I am still hazy on some of the finer detail so I listen carefully to what others say here.Kastor L said:
Without the right background you can at most acquire a superficial knowledge based partly on analogy. This is a poor basis from which to form conclusions in a field with counter-intuitive truths.
No, apart from briefly taking up smoking.
I first studied electrical engineering, then computer science, with a brief stint of studying music. Work has been mostly computers and telecommunications, with bursts of pro audio and broadcast.
And if it is band-limited to 20 kHz, it will be a sine wave.
No. Because I think you are confusing two separate things. As DF96 pointed out, a pulse (or pulse train) contains an infinite number of harmonics. It is not band-limited. Any pulse shorter than 22 usec *only* contains frequencies above the band limit, so will/should have been filtered away at the input.
But if you have two separate 1 kHz sine waves, one that starts 1 usec later than the other, a 44.1 kHz system will reproduce the two waves (and their time difference) just fine.
You are introducing factors to limit the test.
You could have just as equally said 192 kHz.
It seems like you are not neutral here.
Julf said:
That's pleasant, but the pulse is 125 usec, it's the silence from pulse to pulse, transducer performance and amplifier settling time I'm interested in here.
You can reduce the silence to 100 usec, 50 usec, 25 usrc, 5 usec, 750 ns.
At 50 usec with a settling time of 150 usec then the settling time connects the pulses, it can't reach zero since the next pulse starts rising.
No, you are the one who said you couldn't handle infinity.
Pick your sample rate, the arguments don't change.
Neutral? No, I am trying to explain how things actually work. Not sure how "neutral" or "not neutral" applies to that.
My mistake. I thought this thread was about "highest resolution without quantisation noise".
No, you can't, not with a band-limited signal.
Alright, I can see the "black and white thinking" now which needs infinity.
No grey zones allowed.
______
"I thought this thread was about "highest resolution without quantisation noise""
It's conversational.
Settling time, transducer speed are "highest resolution".
Either way this thread will not suddenly turn into diamonds anytime soon.
It almost seems like your position is to defend lower sound quality.
I can't respect that.
______
Julf said:
I'll try to remember you when I have the blind test data.
Then again, it looks like you won't believe that Minoan culture reached America and excavated Copper even with dozens of different data.
Does it need to go through "the scientific method" until it becomes "standard-issue fact", is that the viewpoint I'm not connecting with very well here? Seems likely.
That process can be agonizingly slow in some fields.
Pretty sure science will look pretty different in the year 2070.
Alright, summarized that.
Last edited:
Nothing could be further from the truth. I defend sound quality, but I am sceptical about a lot of superstition, folklore and pseudoscience that is spread in audiophile circles.
Not my area of expertise, but one TV program is not proof of anything at all (except maybe the IQ level of the average TV viewer).
Until then it is random speculation - and how do you know what random speculation is true and what isn't?
Indeed. Most of the theoretical background to digital signal processing was developed in the 1800's and early 1900's.
I am sure it will, but even in 2070 it will probably be based on people putting in the effort to learn the intellectual tools needed, and learning about prior work, and then standing on the shoulders of giants applying the scientific method.
Unfortunately it is equally possible that by 2070 the "my opinion is just as important as yours despite the fact that I actually have no idea what I am talking about, but it feels right" truthiness brigade has won, and science has been outlawed for questioning authority...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.