Hi Bill,
The available literature mentions ITD (phase dependent) up to 1Khz, a mix of ITD and IID from 1Khz to 3.5Khz and pure IID (volume dependent) above 3.5Khz.
Forget the harmonics, just the fundamental.
Hans
Last edited:
The signal with all its mirrors is multiplied with a sin(x)/x with its first zero at Fs.
Hans
Guys, about dsd...
The base sample rate is 2.8MHz, or 44100kHz x 64.
For a 22.05 kHz signal, that means there are 128 samples available for a complete period. This limits the maximum available number of samples to 64 from peak to peak.
Does that mean that DSD cannot reproduce a 22.05 kHz (or 16kHz or even lower) signal at 0dB?
The base sample rate is 2.8MHz, or 44100kHz x 64.
For a 22.05 kHz signal, that means there are 128 samples available for a complete period. This limits the maximum available number of samples to 64 from peak to peak.
Does that mean that DSD cannot reproduce a 22.05 kHz (or 16kHz or even lower) signal at 0dB?
Yes and how much actual difference on a DAC processing something like Redbook vs higher such as 24/192k.
Nordmark's 1972 testing showed ITD response out to 12khz. However, that was a jittered test signal. Unjittered, he showed (IIRC) a dropping response out to 5K.
He did not test real music however, so his data does not truly represent (IMHO) how well we can discriminate. His intent was to find the absolute limits regardless of the test signal, real or synthetic.
Also, I agree with forget the harmonics, as long as the fundamental is over 2.5Khz give or take.
jn
My first thought on analysis would be to do a harmonic subtraction analysis, jpg attached.
Start analog with multiple frequencies of interest added together, convert via redbook, play them back in a system, take the resultant analog waveforms and do a subtraction.
The example I post (I cropped all the scales and titles for IP reasons) shows the result of one analysis. the dark blue line is the signal of interest. I sequentially subtracted appropriate amplitude/phase waveforms , minimizing the residual error (the bottom orange trace). The grey was first removed it is the fundamental of the waveform, then the yellow third harmonic, then grey fourth harmonic.
What is really interesting is that the residue is actually a product of first and fourth harmonic. That information would be lost to a mechanic if I simply used an FFT. What it shows is that the fourth harmonic is being gated, or modulated, by the first harmonic, or vise versa. Mechanically, that is consistent with a torque ripple at first harmonic being put through a mechanism that has a fourth harmonic structure and issue...for example, an involute straight cut gearset, or an poorly chosen coupler that is being overstressed as a result of misalignment (the actual cause of this waveform btw..)
While the FFT is a very strong tool, in this particular case the resultant FFT output would not be very easy for a technician to understand. Showing that a specific mechanism which causes first and a specific mechanism which causes fourth are somehow interacting is far more powerful a diagnostic.
With a CD output, I would first do an analog subtraction at various frequencies for examination of modulations.
jn
Start analog with multiple frequencies of interest added together, convert via redbook, play them back in a system, take the resultant analog waveforms and do a subtraction.
The example I post (I cropped all the scales and titles for IP reasons) shows the result of one analysis. the dark blue line is the signal of interest. I sequentially subtracted appropriate amplitude/phase waveforms , minimizing the residual error (the bottom orange trace). The grey was first removed it is the fundamental of the waveform, then the yellow third harmonic, then grey fourth harmonic.
What is really interesting is that the residue is actually a product of first and fourth harmonic. That information would be lost to a mechanic if I simply used an FFT. What it shows is that the fourth harmonic is being gated, or modulated, by the first harmonic, or vise versa. Mechanically, that is consistent with a torque ripple at first harmonic being put through a mechanism that has a fourth harmonic structure and issue...for example, an involute straight cut gearset, or an poorly chosen coupler that is being overstressed as a result of misalignment (the actual cause of this waveform btw..)
While the FFT is a very strong tool, in this particular case the resultant FFT output would not be very easy for a technician to understand. Showing that a specific mechanism which causes first and a specific mechanism which causes fourth are somehow interacting is far more powerful a diagnostic.
With a CD output, I would first do an analog subtraction at various frequencies for examination of modulations.
jn
Attachments
Last edited:
That is a good read. I used cymbal on purpose as to just how bad CD is in reproducing the highs accurately - realistically. .
They go out far above hearing. Some are heavy and thick and some are light and thin and go higher.
There is an enormous amount of spectral content above the fundamental, all the way to ultrasonic frequencies.
. Cymbals can go up to about 120 kHz if you want all the sound it produces.
you'll hear the "clang" of a cymbal even on a cheap clock radio that can only play midrange frequencies, but all the shimmer and tone of the cymbal will be lost. Most of the tone of a cymbal is in the high frequencies.
On a high res system and close miced recording you can easy hear the wood hitting the metal... (or nylon tip). The tapping of the sick on the metal before the shimmer.
Music isnt the only thing recorded, of course. Everything from gun shots to steam whistles and locomotive engines get recorded.
JN, what the best way to keep in tact the original phase information all the way thru? Digitally.
Audibility of phase shifts and time delays
THx-RNMarsh
Last edited:
My wheelhouse is more towards understanding what needs to be understood and then tested.
In my opinion, the only way to fix anything or make it better is to be able to measure it. If I need accuracies at the 100 nanometer level, I need measurement capability in the 10 nanometer range. If I need 5 uSec stability, I need to measure at 500 nSec.
I work on the metrics. Until something has been measured and verified as amiss, your question cannot be answered.
Should the metrics find something amiss, there are many here more skilled in the digital domain to realize solutions.
jn
Last edited:
He guys, do-you realize we all are arguing deaf, without reading what the others wrote ?
One party (A) pretends the 44.1 sampling rate is not high enough for some aspects of musical signals. Non symmetrical, fast, ultra short duration: transients, micro dynamic, attacks, call-it.
It is the 20 000Hz as highest frequencies that humans can hear that is questioned. RNM has described-it with the cymbals and the stick example.
The other party (B) looks at Nyquist law, and answers in loop that Nyquist is right while nobody pretended the contrary.
I vote for the A party. For several reasons.
First i can feel the difference. Why some does not ? It depends on your system and the details of the reproduction you are sensible to. On most of the speakers (non time aligned, cones), it is not obvious. It is, on some instruments, with horns, carefully time aligned when the record is good enough.
But this is highly subjective. We just have to look at the election results.
Am-I able (at my age), to hear even a sinusoidal frequency at 15kHz ? No.
So why ?
Because, strangely, I can feel the difference if I introduce a brick wall filter in a system at 20KHz (even if no phase turns difference). An illusion ? It is not your business, the earth keeps spinning.
I believe our auditory system works a little like the separating power of our vision. We will not notice a very little detail in a landscape if it don't move. We will if it moves. But, here, it is subjective too and we can argue at nauseum. During evolution, our hearing system has been refined to be able to perceive the slightest crack, essential for our survival.
My point is that, as long as mics and tweeters are able to go to 40 000Hz, It is the responsibility of the audio industry not to reduce the quality of a sound recording by inappropriate material when it can be avoided. If only for reasons of conservation of a work of art for future generations.
24-96 is more accurate to mimic a complex short duration wave form than 44.1-16 bits.
No one can say otherwise. So what ?
It seems to me that it is time to find another topic to dive into, by throwing our phantasms in the face, some having classified the others in the camp of the audiophile superstitious, amateurs of snake oil, influenced by the literature of some dishonest gurus, the others having classified the first ones in that of the above ground deafs with small glasses in tortoiseshell, believers blindly (an other kind of superstition) in a religion that they were taught at school, that they defend like the members of a sect.
The bell at the end of recess has not rang yet in this kindergarten courtyard.
But we could continue our fights in a more sporty way.
One party (A) pretends the 44.1 sampling rate is not high enough for some aspects of musical signals. Non symmetrical, fast, ultra short duration: transients, micro dynamic, attacks, call-it.
It is the 20 000Hz as highest frequencies that humans can hear that is questioned. RNM has described-it with the cymbals and the stick example.
The other party (B) looks at Nyquist law, and answers in loop that Nyquist is right while nobody pretended the contrary.
I vote for the A party. For several reasons.
First i can feel the difference. Why some does not ? It depends on your system and the details of the reproduction you are sensible to. On most of the speakers (non time aligned, cones), it is not obvious. It is, on some instruments, with horns, carefully time aligned when the record is good enough.
But this is highly subjective. We just have to look at the election results.
Am-I able (at my age), to hear even a sinusoidal frequency at 15kHz ? No.
So why ?
Because, strangely, I can feel the difference if I introduce a brick wall filter in a system at 20KHz (even if no phase turns difference). An illusion ? It is not your business, the earth keeps spinning.
I believe our auditory system works a little like the separating power of our vision. We will not notice a very little detail in a landscape if it don't move. We will if it moves. But, here, it is subjective too and we can argue at nauseum. During evolution, our hearing system has been refined to be able to perceive the slightest crack, essential for our survival.
My point is that, as long as mics and tweeters are able to go to 40 000Hz, It is the responsibility of the audio industry not to reduce the quality of a sound recording by inappropriate material when it can be avoided. If only for reasons of conservation of a work of art for future generations.
24-96 is more accurate to mimic a complex short duration wave form than 44.1-16 bits.
No one can say otherwise. So what ?
It seems to me that it is time to find another topic to dive into, by throwing our phantasms in the face, some having classified the others in the camp of the audiophile superstitious, amateurs of snake oil, influenced by the literature of some dishonest gurus, the others having classified the first ones in that of the above ground deafs with small glasses in tortoiseshell, believers blindly (an other kind of superstition) in a religion that they were taught at school, that they defend like the members of a sect.
The bell at the end of recess has not rang yet in this kindergarten courtyard.
But we could continue our fights in a more sporty way.
Last edited:
We've been waiting decades for a real example of this ultra short duration even that needs capturing. Still nothing but arm waving. If it exists someone must have measured it surely?
I am trying to imagine a musical instrument which produces 3/4 of a cycle at 10kHz, but I can't come up with one. What is your example above supposed to tell us about audioo reproduction?
May I ask what DAC you used for that?
Also, you did all the subtraction in the analog domain then digitized the result?
There are much more recent tests with adapted and newer insights.
Scan the internet, starting with Wiki, and you will find quite a few.
Pure ITD seems to stop at 1 Khz, with 1.5Khz being the frequency where it takes exactly 180 degrees to travel from one ear to the other when coming from the side of your head.
Hans
Someone else stated that all that was needed to reconstruct any sine being sampled at just over 2x was three samples.
So my question was, can this be shown using redbook.
I did not state that any source can produce 3/4 of a cycle at 10Khz.
Again, my statement is pulled out of context, where the context is so important.
jn
The part I underlined does not encourage me to answer you, do-you understand why ?
However, I will try to do so, one step in your direction..
First, please, accept the assumption that, no, we cannot measure everything. We just try our best. In audio, most measurements are made from long-lasting sinusoidal signals. Music is not of this nature.
We try to measure what we observe or even guess. We don't know how to measure everything. not yet and not the opposite. We don't necessarily are able to observe what we measure.
Then I'm asking myself why do you rely on the literature ? If you are interested in a subject, why don't you verify yourself as i do as much as possible, because I believe it is the correct scientific approach.
Surely you have a good electret microphone ? Take any pencil or stick. Hit a mat surface. Record in 44.1, and in 48X96. Compare the waveforms. (As I did thousand times in my job).
But, simple is beautiful, listen. Because it is the target of the game.
Don't you hear any difference? The 48x96 does nothing for you.
Why don't have respect for those that feel otherwise. The kind of people with enough culture or good equipment to make the difference between a cimbal and an other (all drummers do), a nylon tip of a stick and the wooden one (all drummers do), and the weight of the stick itself (all drummers do), that RNM enlighted.
This is not an aggression, we all are different, there is no scale of value here, neither some kins of judgment.
On my big system, I can feel an obvious difference. On my PS sound system, not.
I spend X10 time listening music on my PC, see what i mean ?
Last edited:
That is not the result of a DAC.
The waveform is the positional output of a linear encoder capable of 1 nanometer resolution subtracted from the calculated precise location it should be at during motion. The calcs of the positional error are done using 48 bit precision. For the actual subtraction technique I simply used an excel spreadsheet, where the three waveforms I subtracted had phase and amplitude handles, the residual was minimized by eye.
Since I know the base frequencies, I could have done a simple multiply using sines and cosines to grab phase and amplitude (NIST uses that), but excel is quick and dirty. Converting to a useable GUI I would rely on some young coders here. My intent was to prove the usefulness.
I used that crop merely to show the advantage of this totally old school technique, where a new school one (FFT) would muddy the causal effect.
jn
- Status
- Not open for further replies.