[stolio]

I almost get a kick out of the cries of "science" debunking the Pono player.

Yes, a 192k playback sampling rate is a joke and maybe even superstitious. There's not much of a reason for anything to play back at over 96k and many argue 41k is sufficient for mixed/mastered material. Point granted.

However good engineering is most definitely not a joke. Using higher quality DAC's (digital-to-analog converters) and better designed circuitry, spending the extra few pennies and dollars here and there to get the parts with lower tolerances, designing the shape of the product around the circuits instead of cramming circuits into the shape of the product, these are the exact things I would expect to see in a better music player.

[badsock]

The thing is, we've long past the point where it's a difficult task to exceed the the perceptual limits of the human ear. Even the middle-of-the-road DACs are indistinguishable from perfection now, and the rest of the circuit is just ensuring that any noise in the power supply and the output traces are below a certain threshold - not trivial, but the industry has built techniques and for doing this in much, much more demanding applications than audio reproduction, and so it's not like you have to use exotic components or circuit design techniques.

Yes, there's lots of crappy audio kit out there, but it's not hard to get superfluously good stuff either.

I think people just liked the fact that in the 60's and 70's you could make a hobby out of actively pursuing a better sound in amplifiers, and are disappointed that some time in the 80's it became possible to buy gear that was indistinguishable from perfection, and these days it's not even expensive.

Neil Young should have focused on headphones, or speakers - that's an area where there's still detectable amounts of distortion. But that would require something more difficult that attaching a big branding effort to a solved problem.

[stolio]

Isn't this the middlebrow dismissal we're supposed to avoid here? I'm not sure where the ubiquitous, portable, high-quality music players are that make Pono redundant. Are you referring to smartphones?

If you look at actual studies of the performance of smartphone audio you see it's not a trivial task to get right[0], if we're seeing problems on a large company's flagship model like Samsung's Galaxy S5 then this isn't a solved problem.

Playback at home while plugged into a 120V power grid with equipment that only needs to fit into a shoebox is a bit different from playback from a device that's simultaneously a computer and a phone which also happens to have severe space and power constraints (and a giant color touch-screen to boot.)

[0] - http://www.anandtech.com/show/8078/smartphone-audio-testing-...

[tonyarkles]

> Yes, a 192k playback sampling rate is a joke and maybe even superstitious. There's not much of a reason for anything to play back at over 96k and many argue 41k is sufficient for mixed/mastered material. Point granted.

Please don't interpret what I'm about to say as audiophoolery. I'm an EE. I haven't worked in audio a whole lot, but I've looked at things enough that I feel like I could do a reasonable job at designing a DAC.

The one thing that a higher sampling rate does is that it makes good engineering easier. Designing an antialiasing filter that both prevents aliasing and still sounds good at a 41k sample rate is hard. There's just not a whole lot of room between your passband and stopband, so you need a really sharp filter. Add in component tolerances, and your sharp filter is probably going to be a set of mediocre tradeoffs.

I suspect when the author talks about "smoother treble", they're probably referring to the fact that the anti-aliasing filter doesn't need to have a weird phase response to be able work properly.

[badsock]

But that's why we've got oversampling. The higher sample rate can be in the conversion stage, it doesn't have to be in the transmission format.

[tonyarkles]

I'd be super curious to see what the frequency spectrum of the input files looks like too. If your source material is being directly sampled at 44k1, then you're going to need a brick wall filter on the input of that too (or you can oversample with a high rate ADC and do the brick wall digitally). If the source material is cut off hard due to the sample rate, no amount of oversampling can fix that.

For robustness sake, I really really love keeping as many filters as possible in the "gentle" range of design specs. Sometimes that's not possible, but if it's possible to do it start to finish through the signal chain, that makes me very happy.

[badsock]

Agreed, which is why I don't think anyone argues against, say, 192kHz/24bit on the production side. Unfortunately, I think that people look at what the studios use and interpret that as a quality signifier when it really has no use once the signal is mixed down for transmission.

[tonyarkles]

What I'm getting at though, is that if the source material goes from 192/24 in the studio and then downsampled to 44.1/16 into the final file that gets distributed, that material is going to have a brick wall filter applied to it in the downsampling process. If new systems like this are actually distributing real 192/24 streams that haven't ever been downsampled to 44.1/16, I'd expect to get better results from it than from 44.1/16 oversampled back up to 176.4/16.

I'm not saying that we can hear stuff on the recording at 30kHz, but the stuff right at the edge of our hearing around 20kHz will get less messed up.

My gut feeling is that it's way easier to make a well-engineered audio system when there aren't any sample frequencies that cause any of the stopbands to be close to audible frequencies.

[badsock]

Ah, I get what you're saying now. On the downsampling side, correct me if I'm wrong, but my understanding is that resampling in general is a well-solved problem (brick wall included), given reasonable target rates.

In terms of it being easier to filter an orginal signal than to filter a downsampled->upsampled one... my gut goes the other way, but it's something I'd definitely have to look into before I'd comment either way. Interesting question though.

[tigeba]

I'm personally torn about 192K. Some folks that know a ton more about the recording process than I swear by it. On occasion I will source things at 96/24 if the track count is going to be low. Generally I just stick with 44.1/24. The higher sample rates also use a lot more space and processing power.

[stolio]

For reference, here's Dan Lavry's whitepaper on the subject: http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf

It's possible there are counterpoints to this that I don't know about.