But most amps don't measure 120 db, right? Its between 80 and 100 mostly. And crosstalk is also very variable.
All this has been said before, many times:
Amir says he can detect distortions down to nearly -120 dB, and I believe him. He received special training. But I seem to recall his description of how he did it: He gain-rides the system during the quiet bits.
I've heard this myself. I was listening to the quiet part of a recording of Yes's
Close to the Edge. In the recording, the music starts with a natural-sound sample, which fades in from nothing. Right just as it is fading in when the signal is probably at -90 dB, cranking it WAY up at that point to bring it to full scale will reveal all sorts of distortions and artifacts. But one must crank it back down
immediately, because when the signal grows, it will quickly blow up the speakers at amplification levels that high. I think Amir talked of listening in the tails of reverberation as the sound was dying away. With the volume set for the music, these noises and distortions will be well below the ambient noise level in the space.
And as for
harmonic distortion, I can't detect it at all, even in accurate headphones at elevated levels, when it is worse than about 1%, -40 dB. For me, an amp with harmonic distortion at -70 or -80 dB will be crystal clear, and I would bet that every amp made today for use in a proper sound system achieves that threshold. Non-harmonic distortion is easier to hear, but again I don't know of any studies at normal listening levels (that is, without gain-riding and other special techniques) where
any distortion lower than -80 dB can be detected.
So, an amp that measures at -120 dB reflects engineering prowess but it has no practical value beyond that. I would challenge anyone to show me they can hear the difference between, say, my Buckeye Hypex NC502MP amp (SINAD around -100 dB) and a similarly powered amp of
any design, when played at normal (LOUD!) listening levels using real speakers. And especially using real speakers, which probably have distortion levels orders of magnitude higher than the amp.
But I bet there are a lot of amps out there that when played into typically inefficient speakers are overdriven on the sharpest peaks of dynamic music. Let's say my loudspeakers are able to produce 85 dB SPL at one meter when driven by 1 watt of power (at whatever impedance--let's say 8 ohms). If I play dynamic music that averages 90 dB SPL, which will sound reasonably loud, it may have peaks of 105 dB very easily. Those peaks will only need a little over 30 watts driving one speaker at a meter away. That crappy SMSL amp might be acceptable for that application--speakers very close to the listener and listening levels moderated. But we need to add 6 dB if we are at 2 meters, though we can subtract 3 if we are listening in stereo. That net 3 dB gain doubles the power requirement to about 65 watts.
But if we want to play the same music at 100 dB average, we'll need amps that can deliver the power needed to drive peaks at 118 dB SPL. 100 dB on average is very loud, but I daresay not beyond what many audio enthusiasts expect from their systems. Those peaks will require (118 - 85 =) 33 dB of gain to play those peaks without clipping. That will demand 2000 watts. Very few people have that much power available. So, we move to more efficient speakers. Mine produce 90 dB with that 1 watt of input, and that helps. But those peaks will still require 630 watts. Into these 6-ohm speakers, my amp can produce maybe 400 watts at 1% distortion. At 8 ohms, it's 330, and at 4 ohms, it goes into protection at less than 1% distortion, but is supplying over 500 watts when doing so. These are per-channel outputs and I'm driving two speakers, we can effectively double these outputs. So, this amp is barely large enough to play this scenario's loud, dynamic music into these 90-dB, 6-ohm speakers without clipping. I know this is true, because if I play dynamically recorded percussion instruments, I can make the clipping indicators flash without the sound being intolerably loud by any means. This is an amp rated at 350 wpc into 8 ohms--impossibly large by the standards of a few decades ago.
Again, we grossly underestimate required power if we intend to play dynamic music (or soundtracks) at high levels.
When the amp can't produce enough power to fill out the waveforms in those peaks, it suffers, and this is where amps differ quite a lot. Some will just cut off the peak with a hard clip and flat-spot the wave form with sharp corners. That produces a whole range of spurious frequencies that will measure well over 1% distortion. If the clipping noise (which may sound like a loud click) is as loud as the music being clipped, the distortion is 100%. I've heard it. But if it's percussion being clipped like that, we may not particularly notice, because the click might sound a bit like the instrument.
The amp may have gentler ways of responding to the clipping. It may be much smoother than a hard clip, which may produce noise/distortion byproducts that are much less objectionable. It may even simply compress the peaks so that they don't change the sound much but do compress the dynamics. Or, it may have enough reserve power to fill out the waveform for a few milliseconds, which might be enough for that percussion instrument, but maybe not for the sound of the imperial starcruiser passing overhead. When playing real music, those different behaviors might well make audible differences.
I think this happens a lot more often during amp testing than people realize, and may well account for those observations of differences that are real. (Recognizing that most aren't.) Almost anybody asked to audition an amp is going to really crank it up at least part of the time, because the whole purpose of an amp is to make the music loud.
But audiophiles are often into their tiny tube amps that may put out 15 or 20 or 50 watts. For driving desktop speakers that are a meter away to play moderately loud music they may be fine. But for a larger system in a larger room playing music or sound effects at realistic levels, no way. So, instead of getting a more powerful amp, the audiophile get more efficient speakers, like horn-loaded compression drivers, which add orders of magnitude more distortion on their own.
The usual claim that all amps sound the same include this unrealistic qualifier: "when driven within their limits." It's my view that these limits may be the proximate cause of amps sounding different (when they do sound different) in typical audition scenarios, which include cranking them up to full power.
The good news is that inexpensive, clean power is readily available these days.
On the crosstalk issue, do we doubt that crosstalk less than about -25 or -30 dB is inaudible?
Rick "clipping performance is rarely measured, but it is easily measured" Denney