Transient is the changing of amplitude of a specific frequency component. Changing the amplitude of a frequency component can violate Nyquist even if the frequency itself does not. That violation is dependent on the frequency and rate of change. While you were typing, I posted, detailing what it would take to have lower frequencies not violate nyquist by changing envelope.
Increasing the rate of primary sampling (as you mentioned is becoming standards) will set a frequency limit below which ITD issues are moot. If the rate is high enough, it will take the issue clear of our discernment bandwidth.
Jn
Ps.. What I really enjoy about this discussion is..it is exactly what I have encountered and solved at work. It feels good to pass it out there for all. Granted, I've been doing it at the 10 millihertz to 1 kHz range, but as Scott inferred, theory is infinite...so is math.
Oh, also having the privilege of discussion with people who have a high level of knowledge outside my personal wheelhouse, thanks to all.
Last edited:
Again, outside the wheelhouse of my discussion. I discuss the fundamental issue with the initial rate of 44.1.
Jn
Hum ...
The most used is Newman u67 or , similar, U 87
For cymbals, and some acoustic guitars, I used MK 41 Shoeps or other similar capsules.
My favorites with moving coils were AKG D202 for electric guitars (Indeed -20dB @20K
and Sennheiser MD441
My favorite Ribbon, Beyer M260
Last edited:
The link to JASA paper on:
"Binaural beats at high frequencies: Listeners’ use of envelope‐based interaural temporal and intensitive disparities" didn't woik?
My initial problem from the beginning, if this so called transient content is above Nyquist and you want to capture it then what you need to do is obvious, what's the discussion for? If the data from any experiment processed by the ideal low pass anti-aliasing filter does not give you the information you want then you need to change something.
Remember the same maths apply to both channels and I assume the localization stuff is more related to the difference, I would think this needs factored in.
Simple. When I initially questioned the ability of 44.1 to capture ITD level fidelity, darn near everybody either provided FFT plots out to 20khz, claimed it was non existent, or just badgered me as stoopid, idiotic, or just crazy.(I make no claims as to the accuracy of those btw..)
Overall, everybody fought the premise without a clear understanding. It required a lengthy discussion to get to the understanding, and along the way I learned much more about what the industry is currently doing.
The beautiful take away is, it is now trivial to calculate the sampling rate required to guarantee that human ITD discernment levels are not violated. Don't even have to take the shoes off, and no guessing.
While your point on interchannel is indeed accurate, to guarantee an interchannel ITD of 5 uSec at a specific frequency for example, just make sure each channel is capable of 2.5 usec. Simple, and doesn't require trying to figure out any fancy waveform pair that might test the limits.
Scott W...no,it's not the filter problem, it's the sampling rate problem removing transient information.
Jn
Scottjoplin, no it didn't. I hate IPads, will try monday
Last edited:
Did you read the guy that made the cymbal recordings? He wants 60kHz and above, these mics have serious phase/amplitude issues beyond 15/17k. I would expect the matching and exact orientation of a stereo pair would have a high variance considering the wavelength at 20k is much smaller than the transducer. They are so far from a point source 2us timing differences would be nearly impossible to maintain.
I did neither, I presented an ideal transient (we are talking theory not what any snapshot of technological capability can do) and ideal anti-aliasing filter, in fact no FFT's at all. A simple appeal to superposition is all that is needed, if superposition does not apply to simple multiply and adds then all bets are off.
This discussion on transients is hopeless and endless and useless. It only discovers basic misunderstanding, ignorance and missing knowledge of some actors. Yes, transient signals, i.e. the signals starting and ending in zero level, have infinite spectrum. So what? They must be frequency limited before they are sampled. So what? A week of discussion?? I am not getting it.
Your example of a waveform as "transient" simply because it is comprised of multiple frequencies is not sufficient. It is trivial to imagine infinite trains comprised of many frequencies.
Do not paint the picture as black and white.
Jn
If it's beyond Nyquist it has to. You need to do some examples with real "transient" audio waveforms not time windowed sine waves.
Scott, you are the reason I said darn near, you were the pleasant exception.
In fact, you were the first to point out that sufficiently fast changes in the envelope can violate nyquist (using other words of course).
That said, I did not ever make any kind of claim that simple multiply or add could violate superposition, so have no idea why you have created that as a strawman argument.
Jn
Why?
You just said "if it's beyond nyquist it has to be" so you have completely supported my statement..😕
I just put some math to your statement. Look, I know everybody prefers the frequency domain, I get it.
Jn
There are measurement mics out there which were capable to do the job above 20 kHz at least since the 1960's, and some of these were indeed often used for recording purposes, sometimes for special purposes sometimes even as a stereo pair without any further mics.
Surely the majority of records is done with other mics, but again does it matter?
In a bandwidth restricted system, rate of change (rise time) and bandwidth are linked, so if a system is filtered properly (according to the Nyquist criterion) the rate of change of any inband signal is limited.
As said before, I've still to reread Nyquist publications, but I somehow doubt that arbitrarily probing of a signal very near the limit at the most unfavourable instances was, what he had on his mind. 🙂
Surely the majority of records is done with other mics, but again does it matter?
In a bandwidth restricted system, rate of change (rise time) and bandwidth are linked, so if a system is filtered properly (according to the Nyquist criterion) the rate of change of any inband signal is limited.
As said before, I've still to reread Nyquist publications, but I somehow doubt that arbitrarily probing of a signal very near the limit at the most unfavourable instances was, what he had on his mind. 🙂
Wonderful.
The sum of two sines is not transient because there are two sines.
Nor 3,4,5...if it were, all music would violate nyquist.
Your stance of attacking me rather than actually discussing, (should be) below you. Forums are full of people who switch to attacking the messenger when they are unable to support their premise with knowledge, this thread has a few. I just assumed you were different.
Jn
Last edited:
- Status
- Not open for further replies.