Realistic Frequency Response?

[kspv]

I am told that a new born baby can hear frequencies up to 22 KHz, and as one crosses 16 years of age, the upper most audible frequency comes down to about 18 KHz. It was said, that in an experiment with a “golden eared” audiophile, it turned out that the highest frequency he could hear was 18.18 KHz, and no more. This too was thought possible only if the person concerned consciously protects his hearing (=avoids high volumes, wears ear-plugs in noisy environments etc.)

Regarding the spectacular claims of some audiophiles that they could hear frequencies as high as 22, 33 or 44 KHz, it was mentioned that this is a mirage created due to the specific structure of ear-canal in those people, and when these people claim that they have heard 44 KHz, what actually happens is that they hear the “shadow” (the exact term used was “hydanated" which I did not understand!) of the sound at its fundamental frequency of 11 KHz, and therefore this is not true high-frequency hearing. (when the same thing happened in digital domain, we said it was "aliasing" and wanted badly to remove it, right?)

It was said that anything above 10 KHz is generally not pleasing to human ear, and useful only during transients, and human ear gets fatigued if these sounds are over-emphasized.

If this is true, then what is the meaning of priding upon flat frequency responses of speakers upto 50 KHz (KEF), or even something like 42Hz-26 KHz quoted for some of the DIY models (musical instruments may produce frequencies upto 50 KHz, but we can not hear them, right?)

Shouldn’t the speaker enthusiasts aim at a more realistic upper frequency response (=18 KHz?), and concentrate more on other aspects of fidelity instead of committing excesses on frequency response front?

[MBK]

Shouldn’t the speaker enthusiasts aim at a more realistic upper frequency response (=18 KHz?), and concentrate more on other aspects of fidelity instead of committing excesses on frequency response front?

Actually, that's what most loudspeaker designers do anyway. You need a certain overhead for transient response, but other than that, factors such as dispersion, energy storage and of course, low distortion, all usually optimized up to a maximum of 10 kHz,.

The most sensitive frequenzcy of human hearing BTW is 3-4 kHz (see Fletcher Munson curves), the range of human voice is about 200-3000 Hz, and the timbre of a human voice is strongly influenced by frequencies in the 100's of Hz (mid bass). And CD bottoms out at 22 kHz, with a useable transient precison far below that. So, due to all that, only a few exotic brands try to extend frequency respons into ultrasound

[CeramicMan]

Originally posted by MBK

...You need a certain overhead for transient response...

I'm not sure if I agree with that. If a full-range audio signal was subjected to a brick-wall low-pass filter at 22kHz, how would we hear any difference, and how would we be able to tell if there's no "overhead" for those transients?

For example, a "brickwall" filter might have full amplitude at 21999Hz and attenuate by, say, 80dB at 22001Hz. This means that any sound that has a 22kHz frequency component would produce slow-decaying ringing at that frequency, similar to that of a musical instrument. However, because we can't hear that frequency, would the very sharp cutoff (and therefore pathetic transient response) make any difference?

Even assuming that people's eardrums are pliable enough to vibrate at those frequencies, judging by the Fletcher Munson curves and the laws of physics they must form a mechanical low-pass filter at a much lower frequency. Do we actually have any receptors and nerve endings to tell the brain that 22kHz is being heard?

Alternatively, we can say that our ears produce large amounts of distortion, and therefore sub-harmonics and difference tones may be heard by some of the lower frequency receptors when ultrasound is produced.

CM

[MBK]

Ceramicman,

High frequencies and Transients:

I wasn't so much concerned about the exact cutoff value at the upper end, this remark was meant to say that you'd need some more than 10 or 12 kHz response even though the real important spectrum is below 10 kHz.

To reproduce a step response you need high order harmonic content, and of course there is a limit to that, so no step response will reproduce perfectly.

Timing and attack cues are important in music. The timbre of instruments is given by harmonic contents, and I've heard of experiments where cutting off the initial transient of an instrument attack (in time domain) made it impossible for pro musicians to tell instruments apart. So that would mean that a better step response, and therefore some kind of extended HF response, should matter indeed - even though we may not be able to hear a constant non-transient sine wave of corresponding frequency very well.

[hunter audio]

hi

its a trade off

we cant drive above 100 km / h normally but awe cars rated way above it ,

its a engineering goal , not considering the entire freq range to be produced by a single driver

it is quite possible to cover the entire - audio range of 20 to 20 with multiple drive units (i dont want to debate on the factors determining this range)

but it is a way of showing a company s expertise - by being able to produce drivers which will near either extreme of the audio range - with a good spl - phase and linear responce

there are other applications which require highly accurate drivers too - critical sound refence monitors , nva applications - research oriented applications

suranjan

transducer design engineer

[phase_accurate]

I'm not sure if I agree with that. If a full-range audio signal was subjected to a brick-wall low-pass filter at 22kHz, how would we hear any difference, and how would we be able to tell if there's no "overhead" for those transients?

Fist: Why worsen anything that is already bad, by the use of an unnnecessarily restricted frequency response ?

Second: Why design loudspeakers in such a restricted way that they only support an old-fashioned system, that is already a quarter of a century old ????

Regards

Charles

[CeramicMan]

Originally posted by phase_accurate


Fist: Why worsen anything that is already bad, by the use of an unnnecessarily restricted frequency response ?

Second: Why design loudspeakers in such a restricted way that they only support an old-fashioned system, that is already a quarter of a century old ????

Regards

Charles

I'm not sure what the effect is called, but it's used extensively to compress music 10 to 1 with only fractional losses. It's a masking effect (may that's just what they call it?) where a loud tone blocks the ear's ability to hear a relatively quiet tone. This is due to things like: the ear's own distortion, noise-floor, and compromised hearing abilities in order to find a good mix between timing accuracy and frequency accuracy.

Think of the ears as being like large graphic equalisers with over a 1000 bands. To get more accurate frequency information, we would have to add more bands and make each one respond to a narrower range by making it even more resonant than it already is. With an extremely large number of frequency bands, our ability to hear transients would be very sluggish. For example, 2 closely spaced frequencies (eg: 440Hz and 443Hz) would not be audible as a single intermediate frequency with a tremelo (as is ordinarily the case), we would hear 2 distinct and constant frequencies instead. This shows that our ears have "frequency bands" with significant overlap, in order to hear good transient information at the cost of frequency accuracy.

The overlap is the reason why masking of sounds does occur: the information never even gets to the brain because it it drowned out by louder information. The effect is much more pronounced at high frequencies, up to the point where a complete absence of background noise is required to hear some very high frequencies, and I've experienced this myself. This is all happening below, say, 20kHz. Beyond that it gets worse, (ignoring the fact that eventually our ears have to run out of "sensors" anyway) ultrasonic sounds are generally accompanied by much louder sounds below 20kHz, and are therefore effectively masked. Any possible intermodulation distortion (or other possible excuses to make ultrasound audible) is likely to be much quieter than the accompanying real sounds below 20kHz. Eg: crash cymbals, they're so damn loud that if I wanted to copy their sound, I'd start by looking for extremely loud speakers rather than worry about the academics of extended frequency response.

Why limit the frequency response? I'd rather have twice as many songs on a disc that's limited to 22kHz, than only have half the number of songs because of a highly questionable improvement. 2, you'd get improved performance from electronic hardware if it doesn't have to cope with ultrasonics. Parasitic capacitance is always a problem with ultrasonic signals. In amplifiers et al, the mosfets and pcb traces and so on all tend to have a lot of parasitic capacitance of a nasty non-linear kind. While this may not be enough to produce much distortion at low frequencies, even an otherwise superbly designed amp could still have audible intermodulation distortion from ultrasound. A few tiny PP caps could be added to filter out those frequencies and reduce the problems that they cause.

One should not assume that ultrasound necessarily comes from an audio signal. Large quantities can originate from switching DACs and other devices, RFI, SAW devices, current-induced noise etc. Sometimes it's amplitude roughly depends on the amplitude of audio signals at lower frequencies, making it appear to be part of the audio.

The other thing is, the benefits that people claim to hear due to an extended frequency response are highly questionable. For example, a high sampling rate like 192k doesn't necessarily mean that anything beyond 20kHz was actually recorded. It may simply mean that newer ADCs and DACs were used in the process, resulting in better sound for a variety of different reasons.

CM

[phase_accurate]

Some sources claim that it's mainly the lower time-smear being the main advantage of the new hig-resolution formats. So far they, you and me seem to generally agree. They even have advantages when not used up to the Nyqvist frequency, because one can use more relaxed anti-aliasing filters or ones with less pre-ringing.

But IMO a speaker schould not introduce too much errors in the time-and frequency domain within the audio range. And this goal cannot be reached without a frequency response that extends reasonably above the audio band.

Regards

Charles

[Gregm]

Originally posted by phase_accurate
a speaker should not introduce too much errors in the time-and frequency domain within the audio range. And this goal cannot be reached without a frequency response that extends reasonably above the audio band.

I agree. There is a subjective "harshness" or "smearing" otherwise (speakers falling off ~12db at 18kHz).

OTOH, I read somewhere that we "perceive" supersonic frequencies even if we don't hear them. I.e., listeners can tell when they're absent
Of course, in the first place, this all begs the question if there IS musical info from the source (or just noise) over, say, 18kHz.

[phase_accurate]

Of course, in the first place, this all begs the question if there IS musical info from the source (or just noise) over, say, 18kHz.

This is definitely NOT an open question. There are measurements showing significant overtones way above 20 kHZ, depending upon instrument. The question remains how relevant this spectral content might be.

Regards

Charles