I don't know what kind of frequency stability you really need (please enlighten me), but I checked on one of the components retailer websites, and found that even the most stable MEMS oscillators are only a few bucks a piece, any DAC or ADC manufacturer that claims that's worth a $3000 markup is running a scam.
> Software people don’t understand that human hearing, and human sensitivity to timbre and distortion, is spectacularly non-linear. Hearing isn’t a linear counting device with x volume levels.
As a "software person" I take exception to this. Who exactly of my colleagues claimed that it was? Not even the cheapest, most terribly sounding, incompetently designed child's toy's "what sound do cows make" 8-bit 8kHz DAC produces output with x discrete volume levels, it produces a continuous signal. A bad sounding signal perhaps, but a continuous one nonetheless. Plenty of software people know this, but do audiophiles?
No, this is pure steaming BS. $0.2 crystal has 50 ppm stability.
>Some people’s brains and ears are better at all of this than others.
Better than radio receivers operating at GHz ranges I guess.
Well, a lot of this Audiophile stuff is voodoo but if it makes them happy, who cares. But I am skeptical about your statement. The easiest way is to always do a blind or double blind study. A German computer magazine did this once with CD vs. mp3:
https://www.heise.de/ct/artikel/Kreuzverhoertest-287592.html
The guy who was best in spotting the difference and assign it correctly had an ear damage. Since mp3 was developed for "healthy" ears this is not too surprising.
"People who have neither shouldn’t be telling people who do what they can and can’t perceive. "
Let's agree on the scientific method. Anything that shows a significant difference in a double blind study, even for a specific individual I accept.
By the way, I hear my music on my computer via mp3 and send it into my super old tube amplifier. :-)
Even if every properly conducted blind placebo test seems to indicate otherwise?
There’s certainly woo and outright fraud in audio. But there’s also attention to detail.
Professional studio engineers will spend $3000 on a high-precision digital clock for their DACs because it makes a clearly audible difference.
Software people don’t understand that human hearing, and human sensitivity to timbre and distortion, is spectacularly non-linear.
Hearing isn’t a linear counting device with x volume levels. It’s an incredibly sophisticated neural network that performs real time classification and source separation based on spatial and ambient cues, spectral analysis, predictive modelling, and entropy estimation.
Some people’s brains and ears are better at all of this than others. Those people can hear things other people can’t.
In music, someone with a trained ear can name the notes in a chord. Someone with exceptional skill can name the notes in a random cluster of pitches on the lowest octave of a piano.
Audio quality has analogous levels of sensitivity. People who have neither shouldn’t be telling people who do what they can and can’t perceive.