More audio tomfoolery
 

On Thu, 16 Jul 2015 10:13:29 +0100, Jim Lesurf wrote:

In article , Johnny B Good
wrote:
On Wed, 15 Jul 2015 20:56:38 +0200, John R Leddy wrote:

'Jim Lesurf[_2_ Wrote:
;94195']FWIW I've never felt that going as far as 192k/24 made much
sense for home replay. 96k/24 seems a convenient 'compromise' to me
given the use of decent replay equipment. But YMMV.

It is perhaps worth pointing out to people that if you covert to
flac you will usually find that the resulting 96k/24 file is *not*
twice as big as a 48k/24 flac from the same source.

In general there isn't a lot in the ultrasonic region, and the flac
compression can take advantage of this.

The main difference tends to be that there are more bits devoted to
'noise' in 24bit than 16bit. And flac will faithfully keep those
details.
I can't bring myself to allocate over a gigabyte of storage space to
a single-CD album. 24-bit 96kHz albums seem to average just under a
gigabyte which suits me fine. This aspect, and the fact I was willing
to convert my 24-bit 192kHz files to 24-bit 96kHz, allowed me to
change my first 24-bit 192kHz network audio player for one which has
a maximum 24-bit 96kHz playback. Truth be told, until participating
in this thread, I would've quite happily converted my files to 16-bit
48kHz if I had to and not thought any more about it.

I'd much rather a good quality production and master of a 24-bit
96kHz album, than a 24-bit 192kHz album of poor quality. Shame
someone decided it was easier to sell numbers than improved quality.
I would've preferred the better quality no matter what numbers were
associated with the file. Maybe that's a giveaway when thinking about
the relevant skills within the industry. To fall back on the public's
lack of knowledge seems a bit defeatist and insecure to me. That
said,
I guess we do tend to believe anything we're told and spend our money
accordingly.

Fortunately, I have such appalling taste in music none of this
probably matters a great deal anyway.

I've been following this discussion with a growing dismay as phrases
such as "96k/24 seems a convenient 'compromise' to me" started to rear
their ugly heads.

A guy by the name of Monty Montgomery presented a couple of very
interesting videos that nicely relate to the whole business of digital
audio (and video). The links to those videos can be found on this page:

http://xiph.org/video/

Yes, I've seen them in the past and would recommend them with one
caveat. cf below.



I read about halfway through to the key facts - I'll read the rest
later on- where he states unequivocally that 16 bit 44.1 CD audio far
exceeds the capabilities of even the most superhuman of hearing
abilities. IOW, once you're dealing with a finalised music performance
properly committed to CD, that's it as far as 'perfection' is
concerned.

The problem is that he does omit various factors that make reality
different from 'perfection'. This includes the DAC used as that's the
'final' version produced by the digital chain.

The key point to keep in mind for real engineers is that as a general
rule *every* process or conversion in a chain can be expected to degrade
or alter the information.


The only way that a 24/96 "Hi Definition" version is going to sound
any
better is if the final mixdown processing used to create the CD had
been comprehensively buggered up.

Afraid that isn't an absolute truth. The reality is more complex. Even a
technically perfect downconversion for the CD exposes the listener more
to any imperfections in their DAC. And alas, no practical engineered
system will be perfect.

He did make the point in his "Digital Show & Tell" video that the USB
eMagic unit he was using was already right on the edge of perfection
despite its ten year vintage. You might never be able to create an
absolutely perfect engineering solution but you can certainly get so
close that the resulting distortion products are at ludicrously low
levels of audibility, in this case by at least two or more orders of
magnitude.

You just need to ensure that the inevitable imperfections in the
technology can't produce detectable errors or undesirable behaviours. In
short, absolute perfection is not a requirement. You just need to make
sure that it's good enough by a wide enough margin, in this case, by
several orders of magnitude as it happens.


The point is that each stage will tend to alter the results. A perfect
Audio CD is only a beermat or car scarer if you ignore the stage of
being able to play it. :-) So the aim if you're concerned with quality
is to keep the problems well clear of the audible result *at every stage
along the way*. i.e. inc your DAC, etc.

Sadly, for most popular music and 'digital re-masterings' of analogue
studio recordings and professional multi-track recordings of live
performances, the 'buggering up' is the result of deliberate vandalism,
often in the name of 'winning the loudness wars'.

Certainly true. And one of the problems with downconversion is that it
tends to generate *higher* peak values in between the sampled instances.
So the simple act of downconversion can lead to a clipped result if the
source material was 'as loud as possible' without itself being clipped.
Again, what you get out here may depend on your DAC.

There is also a more basic problem people don't seem fully aware about.
But which does cause them to engage in activities like 'which
reconstruction filter do I like?'. 8-]

The optimum choice of reconstruction filter (and resampling filters)
depends on the filtering used in the ADC when the digital samples were
made from the incoming audio during recording. The meaning (information
payload)
of the sampled data values is determined by the ADC filtering. This is a
fundamental Information Theory point about which many engineers, etc,
seem totally unaware. To reconstruct an analogue shape you need to know
what input filer was used. Otherwise the result will be altered in ways
you can't predict.

That's more or less 'a given', especially in the early days when times
one sampling with brick wall filtering may have been used (I don't know
whether or not Philips' novel use of 4 x oversampling with 14 bit DACs
was inspired by the not so novel concept of a pre-existing oversampling
technique in the digital capture process).


Given that you usually have no idea what filter was used, and it changes
from one recording to another, this is a poser for making a 'perfect'
DAC. But, again, you can help shove away from audibility such issues by
keeping with high rates until you get to the final DAC.

TBH, I don't think that matters except with perhaps earlier recordings
that may have used brickwall filtering into a non oversampling 16 bit ADC.

I'm pretty certain that even modern prosumer recorders use oversampling
and on the fly reduction to 16/44.1K or 16/48K just to neatly sidestep
the issue of analogue filtering effects.

Once you're dealing with professional kit where the lowest recording
standard might start at a humble 24/48K and run all the way up to
24/192K, I've no doubt the input is done using the highest 24/192K
regardless and down converted to the selected storage format making the
effects of the input filter totally immaterial as far as playback of a
CDDA based music file is concerned.

I'm afraid you'll have to offer an explanation (or a link to an
explanatory article) to convince me as to how a low pass filter with a
turnover frequency in the region of 30 to 50KHz or higher can impact the
replay of a 20Hz to 20KHz band of signals in a CDDA replay system.


--
Johnny B Good

There are no Adobe products on our computers so I'm currently unable to use iPlayer.

Out of idleness I did add the following stations to foobar yesterday afternoon:
Audiophile Jazz (Jazz) 320kbit/s 44.1kHz MP3 - http://yp.shoutcast.com/sbin/tunein-....pls?id=160671
BBS (Swing) 320kbit/s 44.1kHz MP3 - http://yp.shoutcast.com/sbin/tunein-...n.pls?id=44608
Dusty Vinyl Radio (Big Band) 320kbit/s 48kHz MP3 - http://yp.shoutcast.com/sbin/tunein-....pls?id=460269
Groove City FX (Smooth Jazz) 320kbit/s 44.1kHz MP3 - http://yp.shoutcast.com/sbin/tunein-....pls?id=535237
Jazz Piano Magic 320k (Hard Bop) 320kbit/s 44.1kHz MP3 - http://yp.shoutcast.com/sbin/tunein-....pls?id=336469
NuJazz.net (Smooth Jazz) 320kbit/s 44.1kHz MP3 - http://yp.shoutcast.com/sbin/tunein-....pls?id=643517
Quietmoney Radio (Jazz) 320kbit/s 48kHz MP3 - http://yp.shoutcast.com/sbin/tunein-....pls?id=591415
SomehowJazz (Jazz) 320kbit/s 44.1kHz MP3 - http://yp.shoutcast.com/sbin/tunein-...n.pls?id=13786
~bEE (Jazz) 320kbit/s 44.1kHz MP3 - http://yp.shoutcast.com/sbin/tunein-....pls?id=167874

You could keep me straight on a few things:
· If replay files are to be the same spec as working files then 24-bit 96kHz is a reasonable compromise to address the requirements of studio work.
· There was mention of conversion introducing destructive effects to files.
· It was suggested conversion may be unnecessary due to large capacity storage devices being of little cost.
· The proposition was to avoid potential destruction due to conversion.
· The suggestion was further processing via hardware and software should be considered with care and preferably avoided.
· It was accepted practical issues and personal preferences would take precedence over the minutiae.
· The exercise of avoiding conversion was one of conservation and good practice.

More audio tomfoolery
 

In article , Johnny B Good
wrote:


The point he was making was that once the final mix was down mixed and
normalised to fit within the dynamic range of the CD format (with shaped
dithering this is a massive 120dB,

Erm. It is misleading to call that the 'dynamic range'.

The basic dynamic range is more like 90 to 96 dB depending on the choice of
dither, etc. But because of the nature of human hearing the perceived range
can be larger, depending on a lot of details.

What is correct is that - correctly dithered/noise shaped - the system
allows the recording and detection on replay of signal patterns well below
-96dBFS. However that comes with a lot of "provided that..." qualifiers.
Simply stating the above value without explaining that can be very
misleading and confuse people.

The "120dB" figure tends to stem from people quoting against
power-frequency spectra that divide the entire range into enough bins that
each individual frequency bin only has an amount of noise/dither that is at
-120dBFS. But the human isn't only hearing a single bin. What the human
*is* doing is employing complex signal processing in their brain/ear to
'pattern recognise' so the 'recognised' waveforms sound clearer above
the noise.


comfortably matching the widest limits between the sensitivity and pain
thresholds, marked in red, of human hearing demonstrated by the
Fletcher-Munson equal loudness curve plots) you had every aspect of the
performance you could usefully present (and then some!) nicely
encapsulated by the CD standard.

Again, the FM curves only show the perceived loudness *for isolated single
sustained tones*. Human hearing reacts differently when that isn't the
*total* input. The curves are also averages, so your hearing or mine might
be markedly different.



The only remaining issue, pointed out by Jim, is the question of the
quality of the DACs used to effect a flawless replay of the audio so
carefully encoded into a Music CD. On a technical level at least, this
is a problem that was solved over two decades ago using oversampling
techniques to neatly sidestep both the demands for a 'brickwall'
analogue filter and to push the aliasing/digital artefact noise an
octave or more beyond the 22.05KHz region.

In principle, yes. In practice people still use flawed ADCs, DACs, and
signal processing. This sometimes shows up when the results are analysed.
e..g the recent Verdi Requiem cover CD from BBC Music Magazine. You might
assume these problems ceased decades ago, but alas, real life isn't so
simple. :-/

[snip other examples of how our understanding has developed, but alas not
everyone may have kept up.]


One would hope by now that the more reputable manufacturers of CD
players who subscribe to the best practices of "High Fidelity" have long
since 'put this one to bed', removing any final (misplaced) criticisms
of the, now venerable, CDDA format.

In practice some makers of replay equipment and CDs have dealt with these
issues. But not every one has, in every case.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

More audio tomfoolery
 

In article , Johnny B Good
wrote:

You just need to ensure that the inevitable imperfections in the
technology can't produce detectable errors or undesirable behaviours. In
short, absolute perfection is not a requirement. You just need to make
sure that it's good enough by a wide enough margin, in this case, by
several orders of magnitude as it happens.

Alas "you just need" whilst certainly true in principle may not lead to the
makers invarably doing as required. Afraid its a bungle out there! :-/



Given that you usually have no idea what filter was used, and it
changes from one recording to another, this is a poser for making a
'perfect' DAC. But, again, you can help shove away from audibility
such issues by keeping with high rates until you get to the final DAC.

TBH, I don't think that matters except with perhaps earlier recordings
that may have used brickwall filtering into a non oversampling 16 bit
ADC.

You're still assuming that the DAC (and any prior downcoversion) *is*
essentially perfect. I'm afraid that in reality this isn't invariably true.


I'm pretty certain that even modern prosumer recorders use oversampling
and on the fly reduction to 16/44.1K or 16/48K just to neatly sidestep
the issue of analogue filtering effects.

Many ADC/DAC designs use oversampling (sometimes low-bit high
oversample). These help avoid some classes of problem. But at the expense
of exposing use to other more complicated kinds of flaws.

A fundamental problem here is that such systems tend to end up being
nonlinear 3rd (or higher) order feedback/folding systems. The earliest
practical consequences was that people started hearing 'tones' and 'buzzes'
in the background, or some ADCs/DACs 'locked up'. (Early SACD modulators
and demodulators did this, so Philips/Sony had to keep changing the designs
trying to find ones that didn't - or at least were less likely to do so.)

This is because above 2nd order, such systems can become finite state 'semi
chaotic' processors. In effect, they can become almost impossible to check
for such problems without a brute force check on all possible state/input
situations.



I'm afraid you'll have to offer an explanation (or a link to an
explanatory article) to convince me as to how a low pass filter with a
turnover frequency in the region of 30 to 50KHz or higher can impact the
replay of a 20Hz to 20KHz band of signals in a CDDA replay system.

Erm. You seem to have your telescope the wrong way around here. My point is
that using higher rates helps avoid the risks. I can't recall claiming that
all low rate DACs *will* sound poor. Indeed, many seem fine to me. What I
have said is that using high rates will help shove any problems further
away from being audible.

If you want to know some of the ways some DACs can foul up the results when
fed with low rates (and/or 16bit) you only have to read some of your own
comments about how people have discovered the problems. Then note that in
reality the relevant solutions haven't *always* been implimented in every
DAC made.

The reality is that any real DAC (or ADC or process) has to be a
compromise. I've lost count of how many measurements I've seen that show
things like anharmonic distortion, aliasing, etc for real world designs.
What is less clear is when this may or may not matter in terms of causing a
significant audible degrading. But I'm just saying that you can help dodge
any such uncertainty by playing what was recorded as 96k *as* 96k. Thus
avoiding any possible damage due to downconversion or the DAC.

Given that modern DACs play 96k/24 (or higher) quite happily I can't really
see much reason *not* to do this. And it would seem annoying to me for
someone to downsample many of their recordings, then later realise it did
have an effect, but that they no longer have the higher rate sources.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

More audio tomfoolery
 

In article , John R Leddy
wrote:

'Jim Lesurf[_2_ Wrote:
;94206']Well, there is jazz on R3. And the BBC do plan to upgrade
their other UK radio stations to 320k aac once the iplayer 'Audio
Factory' changes are completed.

OK, 320k aac isn't 96k/24 flac from an injuneering POV. But as other
discussions (and listening) show, good 320k aac can sound a hell of a
lot better than sloppy 'high rez'.

What use the other BBC radio stations make of it is another issue.
Their rock/pop producers may have as little clue as in the commercial
sector.
There are no Adobe products on our computers so I'm currently unable to
use iPlayer.

get_iplayer doesn't require flash. It can capture live or 'on demand'.

And IIUC ffmpeg can now play the live streams without it.


Out of idleness I did add the following stations to foobar yesterday
afternoon: Audiophile Jazz (Jazz) 320kbit/s 44.1kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-....pls?id=160671 BBS (Swing)
320kbit/s 44.1kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-...n.pls?id=44608 Dusty Vinyl
Radio (Big Band) 320kbit/s 48kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-....pls?id=460269 Groove City FX
(Smooth Jazz) 320kbit/s 44.1kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-....pls?id=535237 Jazz Piano
Magic 320k (Hard Bop) 320kbit/s 44.1kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-....pls?id=336469 NuJazz.net
(Smooth Jazz) 320kbit/s 44.1kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-....pls?id=643517 Quietmoney
Radio (Jazz) 320kbit/s 48kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-....pls?id=591415 SomehowJazz
(Jazz) 320kbit/s 44.1kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-...n.pls?id=13786 ~bEE (Jazz)
320kbit/s 44.1kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-....pls?id=167874

Interesting list. But note that mp3 isn't necessarily as good as aac at the
same rate. That said, 320k mp3 should be capable of decent results, but in
practice this will depend on the care with which the streams are made.


You could keep me straight on a few things: · If replay files are to be
the same spec as working files then 24-bit 96kHz is a reasonable
compromise to address the requirements of studio work.

That's my view, yes.

· There was mention of conversion introducing destructive effects to
files.

From the POV of real engineering *all* conversions should be expected to
lose some detail or introduce some flaws. What is open to question is how
small these degradations may be or if they matter in any given case. Done
*well* the effects shouldn't matter. But my view is that now it is so easy
to keep with 96k/24 that we can simply bypass any of the risk entailed by
downcoversions. The best cables are the shortest ones that reach. :-)

· It was suggested conversion may be unnecessary due to large
capacity storage devices being of little cost. · The proposition was to
avoid potential destruction due to conversion. · The suggestion was
further processing via hardware and software should be considered with
care and preferably avoided. · It was accepted practical issues and
personal preferences would take precedence over the minutiae. · The
exercise of avoiding conversion was one of conservation and good
practice.

Yes, that's my view.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

More audio tomfoolery
 

People might find this page of interest. I did it as a demo of the effects
of resampling in a different context. In particular to show how the
'recipy' used for resampling and the choice of reconstruction filter
altered the results.

http://www.audiomisc.co.uk/software/...psampling.html

Here I used noise as the test input so the main effect shows up as the
creation of the 'out of band' components that weren't in the source
material. But if the test input is something like a pair of tones you can
also see anharmonics folded down into the audible range.

You can reduce these effects by having better filtering. But you can't
actually make them always absolute zero in practice.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

More audio tomfoolery
 

Johnny B Good wrote:


Philips realised that they could use a 4 times oversampling technique
using cheaper yet higher precision 14 bit DACs to achieve exactly the
same dynamic range performance of a conventional 16 bit DAC using a times
one sampling rate.

** Philips/Marantz used a pair of 14bit, TDA1540 DACs in a few early CD players - see:

http://www.lampizator.eu/lampizator/...tda%201540.pdf

To make one of the above convert all the bits on a CD, each 16bit sample was replaced with 4 sequential 14 bit samples with values modulated by a digital filter algorithm. Averaging the result recreated any of the otherwise 4 missing values that a 16bit DCA would normally output.

The same algorithm also included a steep LP filter which had the interesting effect of producing pre and post ringing on each step transition.

It worked well enough and some of the noise was pushed out of the audio band - but the specified THD was rather higher than with good 16 bit DACs as used in the Sony machines of the same era.


Not only where they able to solve the 'monotonicity' issue at a stroke,
the oversampling technique also introduced two additional benefits. The
first being that the inevitable digital hash and aliasing products were
all pushed into the 88KHz part of the spectrum, well clear of the
problematic 22KHz region that had mandated the use of 'brickwall'
analogue filtering required of the primitive methods using expensive 16
bit DACs. The second benefit being that cheaper, less ripply analogue
filtering could be employed to protect the following analogue stages (and
the listener) from both digital artefacts and unwanted aliasing products.

** In fact, CD players using TDA1540 DACs usually had high levels of supersonic hash at the output - viewed on a scope, this amounted to 20 or 30mV rms whenever a CD was playing. This hash defeated most reviewers and others attempts to verify the s/n ratio and THD figures claimed by Philips.

Many early players like the Sony CDP101 included ceramic filter assemblies with 100dB/oct roll off slopes that removed all signs of such hash.



The remaining issue with DACs was the analogue output stage clipping
that afflicted some of the earlier products due to inadequate voltage
rail provisioning derived from the "Join the dots" peak amplitude
calculations by some rather naive designers who didn't fully understand
the process of handling a bandwidth limited analogue signal encoded into
the digital domain.

** Not real sure what you are on about here, but the maximum signal level possible from a CD player is 2Vrms or 2.83V peak. Given that most players have dual 12 or 15 volt supplies for the op-amps, there is no such issue.



.... Phil

More audio tomfoolery
 

In article , Phil
Allison wrote:
** In fact, CD players using TDA1540 DACs usually had high levels of
supersonic hash at the output - viewed on a scope, this amounted to 20
or 30mV rms whenever a CD was playing. This hash defeated most reviewers
and others attempts to verify the s/n ratio and THD figures claimed by
Philips.

I'd echo with that. I used a '1st generation' Marantz player with the same
Philips 14bit x 4 chipset. I added a Toko analogue low pass filter to its
output. This limited the bandwidth to about 19kHz, but reduced the hash. I
felt it sounded better as a result. Used it happily for about a decade.

The filter was a design that used to be made and sold in quantity for use
in the Yamaha-style FM tuners that had active 19kHz pilot tone nulling. So
could use a wider filter to keep a more uniform response up to about 15kHz
than the filters which had to remove the pilot.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

Thanks Jim, that's me up and running with get_iplayer.
There's been no need to look at FFmpeg so far.

More audio tomfoolery
 

On Fri, 17 Jul 2015 08:48:43 +0200, John R Leddy
wrote:

Audiophile Jazz (Jazz) 320kbit/s 44.1kHz MP3 -
http://yp.shoutcast.com/sbin/tunein-....pls?id=160671

This stream is from "Audiophile Stream Network"

http://stream.psychomed.gr/

and they offer five MP3 streams all at 320 kbps, 44100 Hz:

Baroque

http://50.7.173.162:2199/tunein/baroque.pls

Classical

http://50.7.173.162:2199/tunein/classical.pls

Jazz

http://50.7.173.162:2199/tunein/jazz.pls

Live

http://50.7.173.162:2199/tunein/live.pls

Icecast

http://50.7.173.162:2199/tunein/enieopyy.pls

If .pls doesn't work with a particular player try changing it to
one of the following: .asx .ram .qtl

The Live and Icecast streams play a mix of different styles.
--
Roger