sam9 said:
Yes, I didn't mean certain groups of non-western people would
start off higher up than 20kHz in general, just that their hearing
doesn't deteriorate as quickly. My only source for that though
was based on testing aboriginese (spelling?) people in Australia
not living in cities,
where 60-70 year olds has as good hearing as a 20 year old
in Europe or the US or something like that.
Those cases I referred to were done by audiograms, ie. listening
to one single frequency at a time, so I don't quite see that IM
would be an error source. On the other hand, I cannot in any
way know for sure that any of these studies were correctly
conducted. Equipment or handling of it could have been flawed,
of course.
It is also worth pointing out that we can actually hear frequencies
up to some 200kHz if the sound source is pressed against the
skull, which I understand is well-known in medicine. That is
somewhat different, though, since it is not about sound
transmitted through the air and via the eardrum.
As for traderbams comment about CDs, yes, of course that media
is limited, so this discussion is mainly relevant, if at all, for live
music and media with potentially better frequency response like
LP and SACD.
thoriated said:
For the second time in a few days I suspect this old thread might
be of some interest to some people:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=6309&highlight=iverson
Hmm. Suppose we can hear intermodulation artifacts of multi-frequencies over 20kHz. If our CD player cuts out above 20kHz then is it not even more important to suppress any amplification of noise/distortion that the amp produces above 20kHz incase the intermodulation artifacts are audible? This would add new signals to the original recording. Which isn't what we want at all.
After reading on the website suggested by another diyaudio member earlier in this thread, I'm not sure about much at all:
http://www.nutshellhifi.com
1 MHz bandwidth may be excessive and just create more noise. I should be careful throwing around numbers that I've not verified. I've been reading too much marketing hype, probably. No need to have excess bandwidth for the application. Knowing what bandwidth is appropriate and designing for it, may by good.
Thanks!
http://www.nutshellhifi.com
1 MHz bandwidth may be excessive and just create more noise. I should be careful throwing around numbers that I've not verified. I've been reading too much marketing hype, probably. No need to have excess bandwidth for the application. Knowing what bandwidth is appropriate and designing for it, may by good.
Thanks!
Traderbam wrote:
I take it then that you regard an old-fashioned standard, that is already a quarter of a century old, as a reference ?!?
Keep in mind that there are better standards out there now.
But besides this: The ear can measure inter-aural time delays of transient signals with a resolution in the the microseconds range. It does achieve this resolution by correlating the different spectral parts of such a transient against each other. If your reproduction chain is too much restricted at the upper end you introduce time-smearing that will also smear spatial reproduction.
Keep in mind that we didn't get our ears to listen to music. Neither were they developed to be able to talk to each other. They were developed as a means of survival. The direction of a sound is the first thing our hearing system recognizes.
So everything should be done to avoid such restrictions. It is impossible to improve on that within a CD player. It is difficult (i.e. i don't say impossible !!) to do it within our speakers. And it is most easily avoided within amplifiers !!!!
So don't build amps with an upper cutoff frequency of 20 kHz, but don't build those with 2 MHz cutoff (used together with 20 kHz speakers !!!) either, because you will most probably have to pay for this extreme feature somewhere else.
Regards
Charles
I take it then that you regard an old-fashioned standard, that is already a quarter of a century old, as a reference ?!?
Keep in mind that there are better standards out there now.
But besides this: The ear can measure inter-aural time delays of transient signals with a resolution in the the microseconds range. It does achieve this resolution by correlating the different spectral parts of such a transient against each other. If your reproduction chain is too much restricted at the upper end you introduce time-smearing that will also smear spatial reproduction.
Keep in mind that we didn't get our ears to listen to music. Neither were they developed to be able to talk to each other. They were developed as a means of survival. The direction of a sound is the first thing our hearing system recognizes.
So everything should be done to avoid such restrictions. It is impossible to improve on that within a CD player. It is difficult (i.e. i don't say impossible !!) to do it within our speakers. And it is most easily avoided within amplifiers !!!!
So don't build amps with an upper cutoff frequency of 20 kHz, but don't build those with 2 MHz cutoff (used together with 20 kHz speakers !!!) either, because you will most probably have to pay for this extreme feature somewhere else.
Regards
Charles
traderbam said:
I seem to recall that Baxandall did some work on this many years ago with records. The conclusion was that you only needed a few volts per microsecond (about 5?) at powers of 100 watts. He did it by measuring the slew rate of music coming off the record.
It would be interesting to do the same with CDs.
cheers, Keith
PS I can't remember the article but twas in Wireless World.
Pjotr said:
Note that the phase starts changing (roughly) a decade lower than the gain starts to drop.
I have to ask the question here. How so? In a signals and Systems course I did some years ago (just dug out the notes) it was indicated that frequency and phase responses are transforms of one another and that one can be entirely derived from the other using basically simple tools (we even did experiments on it). Then, if the frequency response is not changing, how can the phase response change?
cheers, Keith
Question:
Answer:
However music in not a repeating time domain waveform in the sense of Fourier. So my previous post reflects to your statement:
I wonder how you manage to faithfully reproduce a real played piece of Bartok by means of a collection of sine wave generators
Please enlighten me
Keith,
You are right in the relation between phase and amplitude. What I meant is what you see on a Bode plot. There phase starts changing earlier than amplitude. If it makes sense to how good an amp sounds … ? Who knows? But this is also an already very long ongoing debate without agreement.
Cheers
Answer:
However music in not a repeating time domain waveform in the sense of Fourier. So my previous post reflects to your statement:
I wonder how you manage to faithfully reproduce a real played piece of Bartok by means of a collection of sine wave generators
Please enlighten me Keith,
You are right in the relation between phase and amplitude. What I meant is what you see on a Bode plot. There phase starts changing earlier than amplitude. If it makes sense to how good an amp sounds … ? Who knows? But this is also an already very long ongoing debate without agreement.
Cheers
I'd have to search quite long for this, maybe there's another way to find it ?
Concerning Fourier: One has to distinguish between Fourrier analysis and Fourier transform. The former applies to periodic signals and gives a discrete spectrum whil the latter applies to one-time events and gives a density spectrum. Simple as that !
Concerning phase response and amplitude response: Amplitude response starts to change quite far away from the cutoff-frequency of a filter and therefore phase-response as well.
Regards
Charles
Concerning Fourier: One has to distinguish between Fourrier analysis and Fourier transform. The former applies to periodic signals and gives a discrete spectrum whil the latter applies to one-time events and gives a density spectrum. Simple as that !
Concerning phase response and amplitude response: Amplitude response starts to change quite far away from the cutoff-frequency of a filter and therefore phase-response as well.
Regards
Charles
thoriated said:
Now, this is where things get interesting. The 1 kHz tone (perhaps talking about kilocycles would be cooler
) he heard is the beat frequency of two different tones. Let's assume a recording (be it CD or any other media) contains tones of 19 kHz and 20 kHz (I think most of us agree they can be heard) and the beat frequency of 1 kHz is audible in the recording. Now, if the gain of the amplifier reaches unity at 20 kHz, it must be already dropping at 19 kHz. The recording engineering has expected the end of the reproduction chain to have a constant gain at these frequencies, but if it isn't, the 1 kHz tone will not sound the same if it was. I bet nobody will argue whether we are able to hear 1 kHz Regarding the Fourier things, any signal can be made periodical. Just use your imagination and add periodical extension to it. I refuse to comment the mechanisms of the human hearing but I think the popular opinions are too easily accepted without criticism.
Why not? Take all the samples at 44.1KHz you have, transform them with FFT into spectrum (you'll have helluva lot spectral lines - as many as there are samples). If you then care to put up as many sine wave genetrators with precise amplidudes and phases (and weird periods), you'll reproduce a real played piece of Bartok okay, by simply running them for the duration of it. Bizarre as it sounds, its the very essence of Fourier - its always possible to decompose any signal into frequency domain, and back.Pjotr said:I wonder how you manage to faithfully reproduce a real played piece of Bartok by means of a collection of sine wave generators
loop it and it is.
It relates to signal delay through the amp, that depends on frequency. 12ns delay compares to 90 degrees phase shift wrt 20KHz sine. Given that phase is important component in spectral restoration of say Bartok, it seems natural to assume its audible, somehow. Predictable phase errors should be compensatable by DSPs quite well though.
In regards to harmonics, power needed to produce SPL is dependant on frequency, sonic energy is dependant on frequency. 100W speakers won't survive submission of 100W 20KHz into tweaters, and you'd never ever want to get that to your ears. Our hearing is more sensitive to higher harmonics. Sounds reasonable that amps with miserable 2nd harmonics may actually sound better than halfdecent amps with quite some spread of N upper harmonics.
As to audible 21KHz signals, imo, its quite irrelevant if we can conciously recognize presence of 21KHz sine signal, what we perceive is more like sonic energy shocks, that lacks or has convincing levels depending on spectral layout of the complex signal. Don't believe in MHz though. Would rather assume that 100KHz would be about dead stop for our dna.
Amps of course need MHz for precision. If one amp can't follow at its rated power the slew rate of a complex signal, then about what kind of precision can we even talk about? Overprovisioning amp bandwidth is like a means to avoid tradeoff to precision. What was the audible jitter of CD's? That amp bandwidth is of sorta same kind.
wimms said:
Audio codecs exist that make use of this principle (sinusoidal coding):
http://www.s3.kth.se/sip/publications/publication_files/Vafin_AES99.pdf
Steven
phase_accurate said:
Now it is really getting interesting. I suppose it is too much to
ask for a reference, but it would be very interesting to read more
about this.
BTW, I had a hearing test done today by a physician who, although
not a specialist in audiology, has worked with therapy using
processed sound for several years and also recently attended
a seminar on the recent advances in audiology. I briefly discussed
some of these super-20kHz issues etc. with her. I asked her for
instance if it is known whether the sensory cells in the cochlea
react to frequencies or rather to some something else, like
transients, and the frequencies are "computed" by the brain.
As far as she could tell, this is probably not known. I also asked
if she thought it could be possible that we can hear above 20kHz
but the brain filters out non-fundamentals as irrelevant, she
agreed that yes, that might be a possibility.
Please don't get me wrong. I am not trying to suggest a new
theory of hearing. These hypotheses I posted earlier and discussed
witht the physician today were meant just as hypotheses with
no assignment of probablility to them. One could most certainly
come up with other possible alternative hypotheses too. The
point is just that maybe we make to many assumptions about
how human hearing works, and Charles' piece of info above is
very interesting in this respect.
Nope, the spectrum you get then is the average of each frequency along the FFT length window. It tells you nothing of the momentarily amplitude of each frequency.
So if you take an FFT over 1 sec samples, the magnitude of i.e. a 1 KHz component is the average over 1 sec. It can be done but you also need to store the magnitude of each sample at each frequency. And is an arbitrary modulated sine wave still a sine wave?
Long time ago there was a device that measured the amplitude of a large number of frequency bands. These amplitudes were used to modulate sine generators of same frequencies. That device was called “Vocoder” and was used many times to generate robot voices in film. But the thing had nothing to do with FFT’s
Cheers
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.