4 ms·

Yes, mystery still remains in some parts of the audio world - in particular when we're talking about transducer theory, there are some aspects of loudspeaker an

Yes, mystery still remains in some parts of the audio world - in particular when we're talking about transducer theory, there are some aspects of loudspeaker and microphone performance that we can reliably characterize with derived equations, but where we still don't totally understand how those models translate back to the underlying physics fundamentals... but these are relatively specific edge cases in speaker design. Amplifiers and DACs have much less mystery about them - the math is solid, extremely well understood, and tested extensively in related fields like radar and sonar where the tolerances tighter and incentives for getting it right are stronger. Marketing absolutely drives specs (and especially specs without tolerances) when pushing gear, but that doesn't mean that the measurements are somehow invalid. Just expect the specs to tell you how close to the goal of a "wire with gain" you're getting, not to tell you specifically how the measured performance deviates from that goal. It also gets a lot more complicated when you're dealing with devices like headphones that have widely variable impedences - an amplifier that measures relatively flat into a medium-impedance load might fall apart when given dramatically higher or dramatically lower input impedance, and impedance matching is still a somewhat overlooked issue in the industry. The elephant in the room is always that some people just don't want to admit they think colored sound is better, so they convince themselves that there must be some aspect to audio performance that the science can't quantify - in reality, they're just avoiding quantifying a warm EQ curve, or nonlinear distortion, etc. I've studied audio electronics and DSP for audio at length, write plugins for audio software, have worked as a recording engineer, etc... and have no problem saying that mild nonlinear distortion and a bass boost sounds nice. The industry knows what "sounds good" and uses it to sell gear, but consumers think they want accuracy, so they get marketed to with terms like "transparency" or "resolution", and explain away unpleasant results with terms like "synergy" (or lack thereof).

beat 7y ago
Ok, since I brought up THD as a bad measurement, let's take a look at what, exactly, is being measured. A standard THD test measures harmonics generated from a sine wave at 1khz, into a resistive output load. So, how much does a 1khz sine wave represent an actual musical signal? And how much does a resistive load represent speaker impedance? The answers are "not at all", and "not at all". Do you see the problem yet? Of course you do. And that's the problem with amplifier specs. They're measuring what is easy to measure, not what reflects sonic behavior. And it gets worse, because it leads to designing for specs rather than designing for sonics. The easiest way to get "low distortion" spec-wise in an amp is large amounts of global negative feedback. But the negative feedback loop isn't just the outgoing signal. It also picks up back EMI from the speaker. A speaker is an electric motor, but it's also an electric generator hooked to a spring. Energy put into it is stored by and distorted by the suspension, then fed back into the output of the amplifier in a time and phase delayed manner. Yuck. And unless the output of the amplifier has infinite impedance (which it does not), then that back EMI is getting injected into the global negative feedback loop, doing things an 8 ohm resistor in the THD test doesn't do to the signal. The negative feedback loop is reducing input distortion, but increasing speaker EMI. At which point the "distorted" impedance-matched output transformer of an old school tube amp starts looking like a much better proposition...
holy_city 7y ago
I'd disagree with you on the point of transducer theory. It's well understood and poorly published. In particular, proprietary models are significantly more reliable than what is publicly available (both for old school lumped-element and newer finite-element analysis), and the major manufacturers keep those close to their chest. Even the published models by those engineers are glorified approximations of internal research. The limiting factors are in manufacturing, where DIYers and small shops do not have access to the materials and testing equipment of the big players. I think the real elephant in the room is that audiophiles spend a lot of money and don't like being told they're wrong. There's a weird psychology in high end audio (both professional and consumer) - sales are made based on the users' preconceived notions about the product/brand, and you make money by playing into those notions instead of trying to subvert them. PS: based on your username, I think I apprenticed under you a few times.
- sethhochberg 7y ago
  If our mutual boss was named Paul, quite likely :)

beat 7y ago

Ok, since I brought up THD as a bad measurement, let's take a look at what, exactly, is being measured. A standard THD test measures harmonics generated from a sine wave at 1khz, into a resistive output load. So, how much does a 1khz sine wave represent an actual musical signal? And how much does a resistive load represent speaker impedance? The answers are "not at all", and "not at all". Do you see the problem yet? Of course you do. And that's the problem with amplifier specs. They're measuring what is easy to measure, not what reflects sonic behavior. And it gets worse, because it leads to designing for specs rather than designing for sonics. The easiest way to get "low distortion" spec-wise in an amp is large amounts of global negative feedback. But the negative feedback loop isn't just the outgoing signal. It also picks up back EMI from the speaker. A speaker is an electric motor, but it's also an electric generator hooked to a spring. Energy put into it is stored by and distorted by the suspension, then fed back into the output of the amplifier in a time and phase delayed manner. Yuck. And unless the output of the amplifier has infinite impedance (which it does not), then that back EMI is getting injected into the global negative feedback loop, doing things an 8 ohm resistor in the THD test doesn't do to the signal. The negative feedback loop is reducing input distortion, but increasing speaker EMI. At which point the "distorted" impedance-matched output transformer of an old school tube amp starts looking like a much better proposition...

holy_city 7y ago

I'd disagree with you on the point of transducer theory. It's well understood and poorly published. In particular, proprietary models are significantly more reliable than what is publicly available (both for old school lumped-element and newer finite-element analysis), and the major manufacturers keep those close to their chest. Even the published models by those engineers are glorified approximations of internal research. The limiting factors are in manufacturing, where DIYers and small shops do not have access to the materials and testing equipment of the big players. I think the real elephant in the room is that audiophiles spend a lot of money and don't like being told they're wrong. There's a weird psychology in high end audio (both professional and consumer) - sales are made based on the users' preconceived notions about the product/brand, and you make money by playing into those notions instead of trying to subvert them. PS: based on your username, I think I apprenticed under you a few times.

sethhochberg 7y ago
If our mutual boss was named Paul, quite likely :)