As a followup to his detailed explanation of why 24/192 downloads are complete and utter snake oil, Chris “Monty” Montgomery of Xiph.org has produced a video showing you, live, using spectroscopes and oscilloscopes, why digital wave forms aren’t stairsteps and why 16/44 really is enough for the ears of any human ever measured. It’s an incredibly lucid 23 minutes that I recommend heartily, and if you don’t believe anything he says there, the source code is available for you to try it all yourself. (Hat tip to Unter.)
(Greg Wadley asks me to note that this is for the end listener — you should be recording at 24/96 or 32/96 on the assumption it’ll subsequently go through several hundred DSP filters, and headroom for mistakes is cheap these days.)
Monty is full of the proverbial.
The issue is not hearing over 20kHz, the issue is building a brickwall filter to get 16 bits down in 2kHz. It can’t be done.
Normal “rule of thumb” in engineering is that five times oversampling is “about right”. Which brings us up to 100kHz – and that’s for instrumentation, not HiFi. 192 is starting to get in the right area.
And yes, I am an electronics /communications engineer. And old enough to remember when a 750MB SCSI drive cost as much as a new Ducati.
16/44 was NOT picked for reasons of “perfect sound forever” – it was picked because it suited the equipment available at the time (PCM on VHS). It was and is a god-awful compromise.
Oh, I’m quite aware of the compromises involved in picking 16/44!
But is there a human measured in the last century who can tell the difference? There’s been mountains of A/B/X testing of this stuff in the last twenty years. Remember we’re not talking here about the recording stage – we’re talking about an end-user listener, which is a quite specific application where we know quite a lot about what they can hear and what they really just can’t hear.
16 bits 20kHz would be fine – but you can’t get that from 16/44. There is no way in hell you can get a filter that drops a full 16 bits between 22khz and 20khz and keeps phase/group delay anything like in control. And stays 16 bits down. Go and do the math. (Whatever you try ends up in compromise. Pick your compromise.)
As for humans who can, there’s a bunch of people who’ve trained their ears to pick up subtle distortions. And like other people who train (and have talent to start with) they are waaay ahead of the rest of us.
And that assumes you’re able to not clip your 16 bits when recording. Tape is much more forgiving than an ADC when you overdrive*.
192/24 gives the recording and design engineers a bucket load more space to breath.
As for A/B/X, when most pax don’t listen to actual live, unamplified music regularly, I’ll take that will a LARGE pile of salt. Esp when same pax pick Bose. Or cask over Grange.
On which point I’ll also note that different folks care about different faults. Once competently designed, you are into nuances.
Let me also restate – there’s a HUGE pile of snake oil out there and I have no problem undressing that emperor. And there I agree with Monty – any extra fidelity is utterly wasted in an i-thingy with e-buds. Especially if you’re listening to anything intended to be played on “popular” radio.
* I have a natty little digital effects unit for my guitar. But when the input clips, it clips with a CLACK. Fine for blues but fugly for acoustic music – I then lose several bits avoiding clipping.
>16 bits 20kHz would be fine – but you can’t get that from 16/44. There is no way in hell you
>can get a filter that drops a full 16 bits between 22khz and 20khz and keeps phase/group
>delay anything like in control. And stays 16 bits down. Go and do the math.
Not in analog, true! In digital it’s easy, and I even do it in my video if you missed that part. The lowpass that I circle in the last section rolls off 100dB between 20kHz and 21kHz, stays 160dB down, and has linear phase / zero group delay (though I did not mention these last two bits). This is not that unusual a filter, it’s a mundane part of any software resampler.
The great advance of the 1980s was the realization that didn’t have to be done with a physical circuit!