1
\$\begingroup\$

I am curious, whether extending audio amplifier's BW (bandwidth) beyond 20 kHz has any benefits compared to 2Hz-20kHz audio BW amplifier?

Some say that high-frequency -3dB point should should be 30kHz, others say 50 kHz, some say that it shouldn't go further than 100kHz.

I have read somewhere that low high-frequency roll-off point attenuates higher audio frequencies that cannot be heard by human and that there is "void" in output signal, due to high audio frequencies attenuation (higher than 20kHz).

Is very high BW in audio amplifier a good thing (the higher the better) or is it only asking for trouble? Could the amplifier start oscillating on some high frequency with very high BW?

enter image description here

\$\endgroup\$
5
  • 1
    \$\begingroup\$ Why would you want to provide an extended bandwidth to frequencies beyond the hearing of humans? \$\endgroup\$
    – Andy aka
    Nov 10, 2018 at 18:39
  • 2
    \$\begingroup\$ In general, the tighter the bandwidth the lower the noise floor (other things equal.) Also, pushing the high end frequency out also tends to put pressure on the possibility of oscillating and the need to take extra measures (meaning cost), and local and global feedback choices need to be more carefully examined. It's not a cakewalk. (Hi, Keno!) \$\endgroup\$
    – jonk
    Nov 10, 2018 at 18:54
  • \$\begingroup\$ Even the most expensive concert amplifiers rated for 2,000 to 5,000 watts of power have a -3dB limit of 50 KHZ to 70 KHZ, due to custom designs for ultra-low THD. For home use 30 KHZ to 50 KHZ is the upper limit. At 100 KHZ it would be tough to keep it stable under various loads. \$\endgroup\$
    – user105652
    Nov 10, 2018 at 20:15
  • \$\begingroup\$ @Andyaka Maybe you want to provide music to dogs? \$\endgroup\$
    – Hearth
    Nov 10, 2018 at 20:33
  • \$\begingroup\$ @sparky I agree but, at those powers, the amplifier’s slew rate requirements mean that inevitably, the frequency response is wide. It’s impossible to design an amp with high slew rate that has got a poor frequency response. \$\endgroup\$
    – Andy aka
    Nov 10, 2018 at 21:37

4 Answers 4

3
\$\begingroup\$

In parallel with high frequency attenuation audio amplifiers have another "lack of speed" error. Power amplifier circuits use feedback to make Uout follow Uin, but amplified. When the intermediate or output stage of the power amp attenuate high frequencies, the input stage tries to drive more strongly to force Uout to follow rapid changes of Uin. That easily causes overdrive clipping in intermediate stages. The phenomena was known as Transient Intermodulation Distortion and it was really a problem in the past.

It popped out at first when mass produced transistor audio amplifiers started to appear. The industry used used low cost transistors which caused serious attenuation (compared to DC) even at few kHz. With feedback everything still seemed good when the sound had high power peaks only at low frequencies. But percussive hits sounded dull. Second time the same happened when 741 opamps started to be used in audio applications.

Some propellerheads (for ex. Tapio Köykkä) knew the problem as soon as it appeared and even explained it ok, but big consumer electronics companies stamped them of course paranoidic maniacs. Fortunately also academic works appeared (for ex. Matti Otala) and today amps where every stage can handle 20kHz or more at full power are common.

In theory with good microphones one can catch percussive sounds which have substantial energy beyond 20kHz. That can cause transient intermodulation distortion, if the amp can handle only up to 20kHz. But common digital sound recording and distribution systems have steep lowpass filters which wipe off everything beyond 20kHz, so only direct concert sound amplifiers and and some special high bandwidth recording playing systems can benefit having amps with higher full power bandwidth than 20kHz.

\$\endgroup\$
1
  • \$\begingroup\$ Solid answer that covers most all of the important stuff. \$\endgroup\$
    – user105652
    Nov 11, 2018 at 1:22
1
\$\begingroup\$

Besides the increases in possible complexity and cost, there is very little to be gained.

Humans that can hear anything above 20kHz are very rare, and above 22kHz nearly non-existing. So, unless your intention is to design a sound system for your dog, it is just wasted bandwidth.

The -3dB point, which is the standard used for bandwidth, does imply that there is appreciable attenuation of frequency components that we can actually hear. So having some extra bandwidth so that at 20kHz you have less than 1dB of attenuation can somehow be justified. But the ear response is not constant either. As you approach the high end, your hearing is less sensitive and might even have nulls in the response. And if you are not a teenager you are very likely not to have that range at all.

However you must keep in mind that a high-C (soprano C) is barely above 1kHz. The highest note in a piano is ~4.2kHz. That means that 20kHz is nearly 5 harmonics of the highest note. The only sounds that might have significant content at that high range would be percussive (noisy/impulsive) ones.

But what about the source material? Standard audio CDs and most musical material are sampled at 44kS/s (so as to allow for filter roll-off) that means that physically there are no frequency components above 22kHz, other than distortion and noise. And the needed reconstruction filtering would reduce this bandwidth by itself.

So, in conclusion:

  • unless you are into high resolution or high definition audio, the source material has no content on that bandwidth. All you would do is amplify (inaudible) noise and distortion. Thus wasting power.
  • higher cutoff frequencies can provide for a flatter response in the high end, but so would a higher-order system response.
\$\endgroup\$
2
  • 1
    \$\begingroup\$ Edgar, just a side note that CD's are a outdated standard. Most people who want better quality can download 24 bit to 32 bit audio with as much as 130 dB of dynamic range. Sampling rate (in the studio) is 176 KHZ or 352 KHZ per channel. Playback in a good smart phone is typically 88KHZ to 176 KHZ, so the LPF can be tiny. My cell phone is rated for 130 dB dynamic range, way beyond what a CD can do. \$\endgroup\$
    – user105652
    Nov 10, 2018 at 22:13
  • \$\begingroup\$ @Sparky256 there are practical reasons for using wider bandwidths in both recording and reproduction equipment (easier more consistent and cheaper filtering mostly), and wider dynamic ranges can always be desirable. But there is really no need for frequencies above 20kHz. Even our ear pinna causes spectral nulls at those frequencies.. \$\endgroup\$ Nov 10, 2018 at 23:15
1
\$\begingroup\$

Nobody mentions phase shifts. In binaural hearing, timing is everything. If you limit bandwidth to only 20kHz, the phase shifts are already quite significant at 20 kHz and even at 15 kHz. For this reason, keep you phase response flat up to 20 kHz, which means a bandwidth of 10 times 20kHz or more to secure proper transparency and imaging.

\$\endgroup\$
1
\$\begingroup\$

Also... the artikel in link below shows compelling proof that our brain does actually respond to sound above 20 kHz, even if we can't consciously hear it. It's a great read.

https://journals.physiology.org/doi/full/10.1152/jn.2000.83.6.3548

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.