All depends on what the objective is.
For telephony, a signal to distortion ratio of 35dB is seen as sufficient for signal levels of -10dB to –30dB below clipping [2]. Sounds horrible, but its enough to render intelligible speech.
NICAM (Near Instantaneous Companding Audio Multiplex) is a digital sound transmission system that uses only 10 bits for representing the signal waveform. But these 10 bit are combined with a gain scaling factor of 4 bit that is updated at 1ms intervals. This increases the total dynamic range via simple companding and expanding (actually the same idea as is used by tape noise reduction systems). So NICAM can offer the total dynamic range of a 14 bit system (86dB), but depending on the signal amplitude it has the quantization error that is no better than that of a 9 to 10 bit system and therefore the maximum signal to distortion ratio varies between 56dB and 62dB. This can hardly be called high fidelity, but appears to be sufficient for the needs of stereo TV sound.
What if we want audio heaven? What is needed for absolute perfect high fidelity reproduction?
Well, a reasonable requirement would be that all distortions are small enough to fall below the human hearing threshold . There can be no doubt that this will make any kind of distortion inaudible. But that is a quite tough requirement. The human ear is extremely sensitive: at the hearing threshold, the magnitude of ear drum vibration is less then the diameter of a hydrogen atom (!). And it has an enormous dynamic range: the pain threshold is 140dB [1] above the hearing threshold, which is a power ratio of ten million times ten million. No music reproduction chain can achieve that dynamic range. However, at sound levels of 120dB above threshold or higher, even very short exposure can cause permanent hearing damage [3], so it is a bit ludicrous to aim for such high playback levels. A "safe" dynamic range requirement might therefore be set at for example 120dB. This is still a power ratio of a million times a million (*).
No amplifier on the market can guaranty that level of performance under all conditions all the time for any input signal and any load. Except perhaps the amplifiers that I designed and demonstrated in the '90s (which were never on the market). These could probably (**) do just that, and not only in special lab setups but robustly guarantied by design, whatever happens. These amplifiers achieved up to 120dB loopgain at 10kHz, or a gain-bandwidth product of 10GHz (!). And for achieving this speed they needed only very humble power transistors with an ft of 3MHz. Assuming for the sake of argument that the open loop distortion is 1%, then the closed loop distortion of such an amplifier would be -160dB. As listening test showed, the sound quality of these amplifiers reached absolute perfection, far beyond what any existing high end amplifier technology can ever deliver.
Refs
[1] Hearing and Deafness, Third edition, Davis & Silverman, figure 2-6.
[2] Speech Processing, Chris Rowing, figure 3.2.
[3] Audio Engineer’s Reference Book, M Tabot-Smith.
Notes
(*) A practical example. Assume a 100W amplifier is driving a loudspeaker with 90dB/W sensitivity, a fairly typical situation. At 100W amplifier output power, this combination is capable of producing a sound pressure of 110dB. To produce a sound at hearing threshold level at 3kHz, the amplifier output is only 3nW.
(*) This was never proven by measurements due to lack of sensitive enough measurement equipment. But it was shown by simulations.
For telephony, a signal to distortion ratio of 35dB is seen as sufficient for signal levels of -10dB to –30dB below clipping [2]. Sounds horrible, but its enough to render intelligible speech.
NICAM (Near Instantaneous Companding Audio Multiplex) is a digital sound transmission system that uses only 10 bits for representing the signal waveform. But these 10 bit are combined with a gain scaling factor of 4 bit that is updated at 1ms intervals. This increases the total dynamic range via simple companding and expanding (actually the same idea as is used by tape noise reduction systems). So NICAM can offer the total dynamic range of a 14 bit system (86dB), but depending on the signal amplitude it has the quantization error that is no better than that of a 9 to 10 bit system and therefore the maximum signal to distortion ratio varies between 56dB and 62dB. This can hardly be called high fidelity, but appears to be sufficient for the needs of stereo TV sound.
What if we want audio heaven? What is needed for absolute perfect high fidelity reproduction?
Well, a reasonable requirement would be that all distortions are small enough to fall below the human hearing threshold . There can be no doubt that this will make any kind of distortion inaudible. But that is a quite tough requirement. The human ear is extremely sensitive: at the hearing threshold, the magnitude of ear drum vibration is less then the diameter of a hydrogen atom (!). And it has an enormous dynamic range: the pain threshold is 140dB [1] above the hearing threshold, which is a power ratio of ten million times ten million. No music reproduction chain can achieve that dynamic range. However, at sound levels of 120dB above threshold or higher, even very short exposure can cause permanent hearing damage [3], so it is a bit ludicrous to aim for such high playback levels. A "safe" dynamic range requirement might therefore be set at for example 120dB. This is still a power ratio of a million times a million (*).
No amplifier on the market can guaranty that level of performance under all conditions all the time for any input signal and any load. Except perhaps the amplifiers that I designed and demonstrated in the '90s (which were never on the market). These could probably (**) do just that, and not only in special lab setups but robustly guarantied by design, whatever happens. These amplifiers achieved up to 120dB loopgain at 10kHz, or a gain-bandwidth product of 10GHz (!). And for achieving this speed they needed only very humble power transistors with an ft of 3MHz. Assuming for the sake of argument that the open loop distortion is 1%, then the closed loop distortion of such an amplifier would be -160dB. As listening test showed, the sound quality of these amplifiers reached absolute perfection, far beyond what any existing high end amplifier technology can ever deliver.
Refs
[1] Hearing and Deafness, Third edition, Davis & Silverman, figure 2-6.
[2] Speech Processing, Chris Rowing, figure 3.2.
[3] Audio Engineer’s Reference Book, M Tabot-Smith.
Notes
(*) A practical example. Assume a 100W amplifier is driving a loudspeaker with 90dB/W sensitivity, a fairly typical situation. At 100W amplifier output power, this combination is capable of producing a sound pressure of 110dB. To produce a sound at hearing threshold level at 3kHz, the amplifier output is only 3nW.
(*) This was never proven by measurements due to lack of sensitive enough measurement equipment. But it was shown by simulations.