The Myth of 'High Resolution' audio

[Clive197]

I think there is a degree of truth and sanity in Macca's post.

A number of things come to mind. When I rip a CD to my NAS I am convinced the streamed music sounds better than the original CD, it shoudnt but it does!

Many so called Hi-Rez downloads are taken from a 16/44.1 files and upscaled. That is NOT Hi-Rez. This has been proved by a number of Hi Fi magazines. HiFi News now test downloads (somehow) and give the results.

SACD has a smoother sound IMO but I also have a number of SACD's which sound no different to the original CD (Blood,Sweat & Tears 3 and Abraxas come to mind) which could easily be down to mastering.

I think the jury is probably still out on this one and opinions will go back and forth for an eternity. Please don't start me on BluRay 24/96 files of which I have a number.

Clive

[Marco]

Hi Clive,

A number of things come to mind. When I rip a CD to my NAS I am convinced the streamed music sounds better than the original CD, it shoudnt but it does!

That's also been my experience, since using the Pi. Don't underestimate the (what can only be negative) influence the transport mechanism in the CD player is having on proceedings...

Many so called Hi-Rez downloads are taken from a 16/44.1 files and upscaled. That is NOT Hi-Rez.

That I buy, which I'd term loosely as 'psuedo hi-res'. So what *is* 'bonafide hi-res', then?

Marco.

[Marco]

You need to get some cream for that

Yeah... The bonus is, I found a bit of pizza I'd been saving!

Marco.

[AlexM]

Think -10db is the human threshold of normal hearing so I would doubt it

Db level of the threshold of hearing is not the same as relative amplitude (e.g. -xDb).

Distortion is also a funny thing - the human response to it is complex and some types of distortion can be heard at low levels whereas we are tolerant of high levels of even order harmonic distortion (hello valve lovers LOL!). I am prepared to believe that low quantisation resolution and low temporal resolution for low level signals is one of the reasons why 16/44 audio sounds 'odd' to many. Yes, I know about information theory, Nyquist-Shannon's sampling theorem etc but in practice decoding isn't the same as having sufficient bits to accurately encode a band-limited signal, otherwise you would see a nice clean sine at -90db, not that noisy pile of shite. That issue can be dealt with simply by going to 24bit and it should be where possible.

So, from my point of view, 'Hi-Res' PCM starts at 24bit 48Khz, and heads rapidly into diminishing returns at higher sample rates. there are reasons to use high sample rates, but they aren't to do with extending HF response, and these can be realised with upsampling. My CD player upsamples to 24bit/352Khz to allow gentle filters to be used that don't have pre-ringing without risking aliasing in the audio band. Even bats have no use for 192Hkz bandwidth!.

Cheers,
Alex

[Clive197]

........ So what *is* 'bonafide hi-res', then?

Marco.

Now that's a bloody good question!

Clive

[Rothchild]

Interesting discussion, sorry to pop in with a couple of points of pedantry.

The measure is dB (deciBel as in 'One tenth of a Bel')

If one is offering for instance -10dB it has to be referenced to something:

0dBFS = Full Scale digital, the largest number that can be represented digitally - Works 'down' because over 0dBFS is clipped/distorted so measurements will generally be -xdBFS
0dB SPL = 20 micropascals (Sound Pressure Level) - Works 'up' 0dB SPL is the quietest registerable sound in the real world, and as you can see relates to a real measurable phenomena ie air pressure
0dBu = .775Volts RMS
0dBV = 1Volts RMS, so 0dBV means X=1V RMS, +X dBV means greater than 1V RMS, and -X dBV means less than 1V RMS. Are both measures of voltage gain based on established engineering standards.

I would suggest that if you calibrated your listening environment such that 0dBFS (in your 16 bit system) = 96dBSPL then at -90dBFS things are going to start getting confused with the quiescent noise of your analogue stages, before you'll really hear any of the type of artifacts shown in the linked graphs.

Finally, for clarity on the first post most DACs run at either 16 or 24bit (not 48 96 which is I think just a typo/confusion with sample rate) however most modern studio software will put those 'fixed point' (ie there are a fixed 24bits) in to a 'floating point' container (normally at 32 or 64bit) this enables files to be manipulated well above 0dBFS without detriment (a 32bit float file has ~1500dB of range) however because converters operate with 'fixed point' numbers all signals must be reduced to 0dBFS or less before conversion to avoid clipping and distortion.

It's easy to tell if your 'hi rez' file is the real deal, simply put it through any number of spectrum analyser tools available (I like voxengo SPAN) and see if there's any signal above 22kHz, if it rolls of steeply at that point odd on it's been 'upsampled' from a 44.1kHz source, if there is signal above that point then chances are it's the real mcoy.

[Oldpinkman]

In your opinion, not mine

Marco.

It's a simple test really. Take the high-res sample and have it down-sampled to red book using good software, and then compare the hi-res original with the down-sampled copy. Both will be the same master, and so Macca's objection will not be valid. I am still playing with this (well was before I got distracted by France) and the jury is out

I think it is absolutely true to say that there is a great deal of potential variability of quality due to mastering techniques and so it is quite likely that a lot of high-res sounds better due to better mastering than due to the medium itself. I'm not yet sure that means high-res is the same as red book when they're both done properly. And I expect to see later on in this thread others challenging Macca's sums

[Marco]

Sure, I'd be up for carrying out any tests, as long as they involve using your ears, and not graphs or measurement equipment...

Like I said, at the moment, I'm pretty confident I can differentiate between 'bonafide high-res' recordings and those that have simply been mastered better, although I couldn't say for sure what the mechanism responsible for causing that difference is.

Marco.

[awkwardbydesign]

This is what you said:


Quite simply, the proof of the pudding is in the listening, as I've got identical albums, both on CD, and high-res versions of such on my hard-drive. There are numerous examples where the latter significantly outperform the former, sonically, with no other variables in the equation. I'll demonstrate that to you when you next visit

Different masters. That's all there is to it.

I just watched all 1hr 19min of this. https://www.youtube.com/watch?v=SXbH-yzGNfg From an engineer who cares, and there is a LOT more going on than just the mastering.
Recommended.

[pitadavespa]

I'm scratching my bits now.

Marco.

I hope you use a lower frequencie than 44.1, 96 and 192 KHz...

--

Martin, that's a good point, I guess you are probably right.