In article <u5a2mk$2pmo6$1@dont-email.me>,
NY <me@privacy.invalid> wrote:
> "Bob Latham" <bob@sick-of-spam.invalid> wrote in message
> news:5aadacc905bob@sick-of-spam.invalid...
> > In article <5aad1b258cnoise@audiomisc.co.uk>,
> > Jim Lesurf <noise@audiomisc.co.uk> wrote:

> >> More accurate to say that 'Vinyl' *can* sound very good. Just as CD
> >> can. But not all examples of either will or do.

> > I'll accept that.

> Both *can* sound good, but CD will always win because it is not
> marred by dust/scratch/surface-roughness noise. I have yet to hear
> any vinyl record which even comes *close* to a CD for lack of
> background noise. A vinyl record may have been mastered perfectly,
> but as soon as physical contact between groove and needle is
> involved, very intrusive background noise enters the frame.

I've never argued that Vinyl can get near CD for unwanted noise or
durability or playing time or a host of other things.

However, It's not hard to find CDs that don't sound as good as the
Vinyl. I refer to one in another post. For the majority of recording
that don't have clicks, pops etc. telling which you're listening to
is tricky, remove all the tiny clicks digitally and it's very
difficult certainly on rock/pop with decent gear.

Once more! Vinyl is badly flawed, it has a multitude of sins but on
good gear and a clean record it sounds excellent. I'm not in anyway
advocating vinyl but it is being unfairly rubbished by prejudice and
lack of experience.

Bob.

Bob.

In article <5aadb31d4cbob@sick-of-spam.invalid>,
Bob Latham <bob@sick-of-spam.invalid> wrote:
> In article <u5a2mk$2pmo6$1@dont-email.me>,
> NY <me@privacy.invalid> wrote:
> > "Bob Latham" <bob@sick-of-spam.invalid> wrote in message
> > news:5aadacc905bob@sick-of-spam.invalid...
> > > In article <5aad1b258cnoise@audiomisc.co.uk>,
> > > Jim Lesurf <noise@audiomisc.co.uk> wrote:

> > >> More accurate to say that 'Vinyl' *can* sound very good. Just as CD
> > >> can. But not all examples of either will or do.

> > > I'll accept that.

> > Both *can* sound good, but CD will always win because it is not
> > marred by dust/scratch/surface-roughness noise. I have yet to hear
> > any vinyl record which even comes *close* to a CD for lack of
> > background noise. A vinyl record may have been mastered perfectly,
> > but as soon as physical contact between groove and needle is
> > involved, very intrusive background noise enters the frame.

> I've never argued that Vinyl can get near CD for unwanted noise or
> durability or playing time or a host of other things.

> However, It's not hard to find CDs that don't sound as good as the
> Vinyl. I refer to one in another post. For the majority of recording
> that don't have clicks, pops etc. telling which you're listening to
> is tricky, remove all the tiny clicks digitally and it's very
> difficult certainly on rock/pop with decent gear.

> Once more! Vinyl is badly flawed, it has a multitude of sins but on
> good gear and a clean record it sounds excellent. I'm not in anyway
> advocating vinyl but it is being unfairly rubbished by prejudice and
> lack of experience.

Please remember that until 1979, all we had to listed to were LPs.

--
from KT24 in Surrey, England - sent from my RISC OS 4té
"I'd rather die of exhaustion than die of boredom" Thomas Carlyle

On Thursday, 1 June 2023 at 14:30:06 UTC+1, charles wrote:
> In article <5aadb3...@sick-of-spam.invalid>,

Snip

>
> Please remember that until 1979, all we had to listed to were LPs.

and then in 1985 along came something better...

"R. Mark Clayton" <notyalckram@gmail.com> wrote in message
news:582911af-b735-4784-b6b5-cfd3a62f4064n@googlegroups.com...
> On Thursday, 1 June 2023 at 14:30:06 UTC+1, charles wrote:
>> In article <5aadb3...@sick-of-spam.invalid>,
>
> Snip
>
>>
>> Please remember that until 1979, all we had to listed to were LPs.
>
> and then in 1985 along came something better...

It was before 1985. I remember in my first year at university (so 1982)
going with some friends to hear a demo of a CD player at a hifi shop.

In article <unqg7i989pk4juuplikqlkve2m7rn4tj4l@4ax.com>, Roderick Stewart
<rjfs@escapetime.myzen.co.uk> wrote:

> There's an even bigger advantage in listening to a recording that
> doesn't have any clicks etc in the first place, because the digital
> recording process doesn't create them. Then you don't have to do
> anything to it at all.

> A lot of LPs are made from digital masters, and are thus effectively
> copies made by an analogue process that is full of distortions. How
> can these copies possibly sound better than the originals?

Matter of what someone feels is "better".

BTW Today I've been listening to a 96k transfer I made of a "Queen" double
LP. It sounds better than the CD that was released with it of the same
material. One reason being that the CD has flat top clipping on loud
sections. Result of the pop/rock biz obsession with LOUDNESS SELLS I guess.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics https://www.st-andrews.ac.uk/~www_pa/Scots_Guide/intro/electron.htm
biog http://jcgl.orpheusweb.co.uk/history/ups_and_downs.html
Audio Misc http://www.audiomisc.co.uk/index.html

"Jim Lesurf" <noise@audiomisc.co.uk> wrote in message
news:5aadcb266enoise@audiomisc.co.uk...
> BTW Today I've been listening to a 96k transfer I made of a "Queen" double
> LP. It sounds better than the CD that was released with it of the same
> material. One reason being that the CD has flat top clipping on loud
> sections. Result of the pop/rock biz obsession with LOUDNESS SELLS I
> guess.

Could the down-conversion from 44 kHz uncompressed to 96 kbps
lossy-compressed actually round-off some of the flat-topped peaks in the
original?

On Friday, 2 June 2023 at 10:30:05 UTC+1, Jim Lesurf wrote:
> In article <unqg7i989pk4juupl...@4ax.com>, Roderick Stewart
> <rj...@escapetime.myzen.co.uk> wrote:
>
> > There's an even bigger advantage in listening to a recording that
> > doesn't have any clicks etc in the first place, because the digital
> > recording process doesn't create them. Then you don't have to do
> > anything to it at all.
>
> > A lot of LPs are made from digital masters, and are thus effectively
> > copies made by an analogue process that is full of distortions. How
> > can these copies possibly sound better than the originals?
> Matter of what someone feels is "better".
>

Er less noise and more accurate reproduction aka Hi-Fidelity

OTOH a digital copy is lossless. Even an MP3 compressed copy (with some loss) sounds better than vinyl..

SNIP

On Fri, 02 Jun 23 09:30:02 UTC, Jim Lesurf <noise@audiomisc.co.uk>
wrote:

>> A lot of LPs are made from digital masters, and are thus effectively
>> copies made by an analogue process that is full of distortions. How
>> can these copies possibly sound better than the originals?
>
>Matter of what someone feels is "better".

In the home, perhaps it depends on what sort of music you like. If
your preference is for music that has been created acoustically, then
the old Quad advertising phrase could apply - "The closest approach
to the original sound". This has always been my understanding.

Music that has been created electronically or partially electronically
will never have been an original sound so you could adopt a different
meaning for 'better' - something more like 'sounds more pleasing to
me' which isn't really a definition because it can mean anything.

In the world of broadcasting (where I used to work) the same signal is
sent to many people who will proably have different preferences, so
the general aim is that the equipment should affect the signal as
little as possible (after which it's up to the customers what they do
with it). In this setting, more objective measures can be appilied,
and digital wins hands down.

Maybe I'm biased, but I like real music that still sounds as real as
possible even when it's squeezed out of boxes.

Rod.

On 01/06/2023 10:45, Jim Lesurf wrote:
> More acurate to say that 'Vinyl'*can* sound very good. Just as CD can. But
> not all examples of either will or do

To me the great advantage of CD is that it will sound pretty good, all
the time. I still have an LP which I listened to once, decided I wanted
an easy to use copy, and scratched it while playing a second time to
make a cassette. Cassette isn't great, but it's better than a scratch
_THAT_ big :(

(I've since digitised it, and the scratch isn't audible any more)

Andy

On 02/06/2023 13:35, Roderick Stewart wrote:
<snip>
> In the world of broadcasting (where I used to work) the same signal is
> sent to many people who will proably have different preferences, so
> the general aim is that the equipment should affect the signal as
> little as possible (after which it's up to the customers what they do
> with it). In this setting, more objective measures can be appilied,
> and digital wins hands down.
</snip>
I do listen to broadcast music from time to time. My experience is that
the lossy compression applied to get the sound into a much too low
bandwidth causes audible distortion.

It _could_ be really good, given the technology. But the dictates of
commerce mean more stations make more money than fewer better ones.

Andy

On Friday, 2 June 2023 at 16:58:17 UTC+1, Vir Campestris wrote:
> On 02/06/2023 13:35, Roderick Stewart wrote:
> <snip>
> > In the world of broadcasting (where I used to work) the same signal is
> > sent to many people who will proably have different preferences, so
> > the general aim is that the equipment should affect the signal as
> > little as possible (after which it's up to the customers what they do
> > with it). In this setting, more objective measures can be appilied,
> > and digital wins hands down.
> </snip>
> I do listen to broadcast music from time to time. My experience is that
> the lossy compression applied to get the sound into a much too low
> bandwidth causes audible distortion.
>
> It _could_ be really good, given the technology. But the dictates of
> commerce mean more stations make more money than fewer better ones.
>
> Andy

Stereo FM better than vinyl, almost as good as direct CD (see comments early in thread about listening to records over it), although signal needs to be fairly strong not to get noticeable hiss. Some young people can hear the 19kHz tone used to separate the channels.

Stereo in DVB/T or DVB/T - second best, but not far behind.

High bit rate DAB - OK, probably comparable to vinyl.

Standard DAB - poor - significantly worse than vinyl.

AM - very poor and only mono, but when I was a kid listening to Radio Luxembourg under the bed covers I thought it was great...

In article <65e60678-3733-40d4-a126-138b1af64d40n@googlegroups.com>,
R. Mark Clayton <notyalckram@gmail.com> wrote:

> OTOH a digital copy is lossless. Even an MP3 compressed copy (with
> some loss) sounds better than vinyl..

More tosh from the man who's never heard a good vinyl system. Keep
digging.

Bob.

On Fri, 2 Jun 2023 10:37:52 +0100, NY <me@privacy.invalid> wrote:

>> BTW Today I've been listening to a 96k transfer I made of a "Queen" double
>> LP. It sounds better than the CD that was released with it of the same
>> material. One reason being that the CD has flat top clipping on loud
>> sections. Result of the pop/rock biz obsession with LOUDNESS SELLS I
>> guess.
>
> Could the down-conversion from 44 kHz uncompressed to 96 kbps
> lossy-compressed actually round-off some of the flat-topped peaks in the
> original?

He didn't say 96 kbps. He said 96k. This is presumably sloppy for 96 kHz
sampling.

In article <slrnu7nnms.3q8k.abuse@news.pr.network>,
Paul Ratcliffe <abuse@orac12.clara34.co56.uk78> wrote:
> On Fri, 2 Jun 2023 10:37:52 +0100, NY <me@privacy.invalid> wrote:

> >> BTW Today I've been listening to a 96k transfer I made of a
> >> "Queen" double LP. It sounds better than the CD that was
> >> released with it of the same material. One reason being that the
> >> CD has flat top clipping on loud sections. Result of the
> >> pop/rock biz obsession with LOUDNESS SELLS I guess.
> >
> > Could the down-conversion from 44 kHz uncompressed to 96 kbps
> > lossy-compressed actually round-off some of the flat-topped peaks
> > in the original?

> He didn't say 96 kbps. He said 96k. This is presumably sloppy for
> 96 kHz sampling.

Yes, I'm sure. Quite rightly, Jim wouldn't ever use lossy compression.

Bob.

"Paul Ratcliffe" <abuse@orac12.clara34.co56.uk78> wrote in message
news:slrnu7nnms.3q8k.abuse@news.pr.network...
> On Fri, 2 Jun 2023 10:37:52 +0100, NY <me@privacy.invalid> wrote:
>
>>> BTW Today I've been listening to a 96k transfer I made of a "Queen"
>>> double
>>> LP. It sounds better than the CD that was released with it of the same
>>> material. One reason being that the CD has flat top clipping on loud
>>> sections. Result of the pop/rock biz obsession with LOUDNESS SELLS I
>>> guess.
>>
>> Could the down-conversion from 44 kHz uncompressed to 96 kbps
>> lossy-compressed actually round-off some of the flat-topped peaks in the
>> original?
>
> He didn't say 96 kbps. He said 96k. This is presumably sloppy for 96 kHz
> sampling.

Ah, my mistake. I hadn't realised that some sound was even sampled at 96 kHz
(ie 96,000 samples per second per channel).

My comment about rounding-off flat-topped peaks wasn't as daft as it sounds.
CoolEdit, a sound package that pre-dates Audacity, had the ability to do
"clip restoration" which somehow, magically, managed to produce waveforms
which had some semblence of varying levels even when there were sections
where the samples had maxed-out. Pure f-ing magic ;-)

In article <u5cd9l$34vl9$1@dont-email.me>, NY <me@privacy.invalid> wrote:
> "Jim Lesurf" <noise@audiomisc.co.uk> wrote in message
> news:5aadcb266enoise@audiomisc.co.uk...
> > BTW Today I've been listening to a 96k transfer I made of a "Queen"
> > double LP. It sounds better than the CD that was released with it of
> > the same material. One reason being that the CD has flat top clipping
> > on loud sections. Result of the pop/rock biz obsession with LOUDNESS
> > SELLS I guess.

> Could the down-conversion from 44 kHz uncompressed to 96 kbps
> lossy-compressed actually round-off some of the flat-topped peaks in the
> original?

Not sure what you are asking/suggesting.

However I made the 96k version from the LP. Used a Benchmark (studio/lab
quality) ADC and took care not to clip.

The CD was the original release made alongside the LP. i.e. The music
company released that to have CD/LP version on offer.

FWIW the DACs I use will cope with 'intersample overs' and - for individual
examples, give peaks above 0dBFS as relevant. But the problem with many
'popular' CDs has been sections where a *series* of successive samples are
nailed to the max value.

You can find examples documented on my webpages. IIRC the 'Queen' example
is just one of them. I also have pages showing various other gross problems
that the music company releases on CD display. e.g. examples where their
ADC wasn't linear at a fine-grain level.

The reality is that the music biz often makes crappy CDs just has it often
made/makes crappy LPs. GIGO then rules, disrupting any attempt to just
either format 'better'.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics https://www.st-andrews.ac.uk/~www_pa/Scots_Guide/intro/electron.htm
biog http://jcgl.orpheusweb.co.uk/history/ups_and_downs.html
Audio Misc http://www.audiomisc.co.uk/index.html

In article <u5d3in$37oim$2@dont-email.me>, Vir Campestris
<vir.campestris@invalid.invalid> wrote:
> On 02/06/2023 13:35, Roderick Stewart wrote: <snip>
> > In the world of broadcasting (where I used to work) the same signal is
> > sent to many people who will proably have different preferences, so
> > the general aim is that the equipment should affect the signal as
> > little as possible (after which it's up to the customers what they do
> > with it). In this setting, more objective measures can be appilied,
> > and digital wins hands down.
> </snip> I do listen to broadcast music from time to time. My experience
> is that the lossy compression applied to get the sound into a much too
> low bandwidth causes audible distortion.

Depends on what pathway you employ how much the levels get fiddled with.

iPlayer from R3 tends to mean less level compression for BBC material. Of
course, with CDs they have to start with what th music company sold. But in
general, if you want optimum quality go for iPlayer and skip FM or DAB or
'TV' versions. The 320k aac is pretty close to flac.

For R4 level control has become a bit of a shambles in recent years, alas.
No-one seems now to check and deal with programme material levels. And
radio items that have added tops/tails promoing other things can have
wildly different levels to the main content they sandwitch, etc. I suspect
the last engineer to leave BBC radio switched off the lights leaving
producers in the dark! 8-[

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics https://www.st-andrews.ac.uk/~www_pa/Scots_Guide/intro/electron.htm
biog http://jcgl.orpheusweb.co.uk/history/ups_and_downs.html
Audio Misc http://www.audiomisc.co.uk/index.html

In article <5aaf2a16abbob@sick-of-spam.invalid>, Bob Latham
<bob@sick-of-spam.invalid> wrote:
> In article <slrnu7nnms.3q8k.abuse@news.pr.network>, Paul Ratcliffe
> <abuse@orac12.clara34.co56.uk78> wrote:
> > On Fri, 2 Jun 2023 10:37:52 +0100, NY <me@privacy.invalid> wrote:

> > >> BTW Today I've been listening to a 96k transfer I made of a "Queen"
> > >> double LP. It sounds better than the CD that was released with it
> > >> of the same material. One reason being that the CD has flat top
> > >> clipping on loud sections. Result of the pop/rock biz obsession
> > >> with LOUDNESS SELLS I guess.
> > >
> > > Could the down-conversion from 44 kHz uncompressed to 96 kbps
> > > lossy-compressed actually round-off some of the flat-topped peaks in
> > > the original?

> > He didn't say 96 kbps. He said 96k. This is presumably sloppy for 96
> > kHz sampling.

> Yes, I'm sure. Quite rightly, Jim wouldn't ever use lossy compression.

Basically, yes. For LP transfers I use 96k/24 so I can give the recording
lots of 'space'. I then clean up any noticable clicks and process down to
48k for undemanding material.

Given flac I'd use that. However I'm happy with BBC generated 320k aac. And
*have* used *very* low rate lossy for some old files to put poor recordings
of speech on the web. If anyone's interest in the life of an 'east ender'
born well before WW1 then they are at

https://jcgl.orpheusweb.co.uk/history/people/Tapes/Tapes.html

The recordings were done using old kit, not very well, decades ago. So of
'historic' interest rather than musical content! The earliest example are
particularly poor quality, sadly. Later ones were a bit better as I built a
mic preamp and use a better recorder and mics. But bodged home recordings
not pro studio work.

I kept the digital 'masters' as flac, though. Just in case I can ever do a
better clean-up. But probably never will. Remaining life too short, alas!

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics https://www.st-andrews.ac.uk/~www_pa/Scots_Guide/intro/electron.htm
biog http://jcgl.orpheusweb.co.uk/history/ups_and_downs.html
Audio Misc http://www.audiomisc.co.uk/index.html

In article <u5hkkr$3tpau$1@dont-email.me>, NY <me@privacy.invalid> wrote:

> Ah, my mistake. I hadn't realised that some sound was even sampled at 96
> kHz (ie 96,000 samples per second per channel).

'High end' tends to be even higher. 192k 24bit is a minimum for many
studios, etc. Look for the '2L' label's website and you can find some free
samples.

FWIW even cheap-as-chips USB ADC/DACs for individual/home use will work
nicely up to 192k/24bit stereo in and out. e.g. the Scarlett 2i2 3rd gen
Quite impressive for the price.

Did some tests on one and it works well. So even musos who are working
out of their bedroom can have good kit in terms of sample rate, etc.
Same maker does similar items with more channels if the user needs them.

> My comment about rounding-off flat-topped peaks wasn't as daft as it
> sounds. CoolEdit, a sound package that pre-dates Audacity, had the
> ability to do "clip restoration" which somehow, magically, managed to
> produce waveforms which had some semblence of varying levels even when
> there were sections where the samples had maxed-out. Pure f-ing magic
> ;-)

Yes, you can do that by 'projection' from the waveforms before and after
the gap. It then makes the 'least harm' estimate of the lost part. May
improve the sound quality but is guesswork. Just as the click 'repair'
does. Gets rid of the click assuming the smoothest transition estimated,

Alas, some CDs have such long 'saturated overs' that this can be expected
to be a poor cousin of the real waveform.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics https://www.st-andrews.ac.uk/~www_pa/Scots_Guide/intro/electron.htm
biog http://jcgl.orpheusweb.co.uk/history/ups_and_downs.html
Audio Misc http://www.audiomisc.co.uk/index.html

"Jim Lesurf" <noise@audiomisc.co.uk> wrote in message
news:5aaf482ce6noise@audiomisc.co.uk...
>> Ah, my mistake. I hadn't realised that some sound was even sampled at 96
>> kHz (ie 96,000 samples per second per channel).
>
> 'High end' tends to be even higher. 192k 24bit is a minimum for many
> studios, etc. Look for the '2L' label's website and you can find some free
> samples.

What is the advantage of sampling at greater than about 45,000 samples/sec?
(44.1 for CD, 48 for most MPEG)

Nyquist says that if you sample at n samples/sec, you can reproduce signal
frequencies up to n/2 Hz. The highest frequency that humans can hear is
around 20 kHz, and this gradually decreases with age. It is *essential* that
the signal does not contain frequencies higher than n/2, otherwise you get
aliasing (the audible equivalent of wagon wheels appearing to run
backwards), so a low-pass filter is needed which passes everything up to 20
kHz and then rapidly rolls off above this to it attenuates to negligible
level by the time the sampling frequency is reached. (*) That's why CDs use
44.1 kHz and MPEG uses 48 kHz rather than 40 kHz sampling for reproducing
sounds up to 20 kHz.

What is the benefit of sampling at roughly four times the maximum signal
frequency - ie twice what Nyquist requires?

I can see the benefit in the size of those samples being larger than the
default 16 bits at the mastering stage in a studio, as it allows lower-level
signals to be boosted if required by multiplying the digital sample rather
than requiring greater analogue gain before sampling, while still staying
within the 16 bit range of the final release of the track (ie it doesn't
introduce quantisation noise).

(*) A brick wall transfer characteristic (gain=100 % for f < cutoff, and 0%
for f> cutoff) would be the ideal but is impossible to to achieve with
analogue components.

>> My comment about rounding-off flat-topped peaks wasn't as daft as it
>> sounds. CoolEdit, a sound package that pre-dates Audacity, had the
>> ability to do "clip restoration" which somehow, magically, managed to
>> produce waveforms which had some semblence of varying levels even when
>> there were sections where the samples had maxed-out. Pure f-ing magic
>> ;-)
>
> Yes, you can do that by 'projection' from the waveforms before and after
> the gap. It then makes the 'least harm' estimate of the lost part. May
> improve the sound quality but is guesswork. Just as the click 'repair'
> does. Gets rid of the click assuming the smoothest transition estimated,

Yes, I've only tried it with recordings where occasional peaks try to go
"over" 100% and you get brief periods of maxed-out samples. The result
sounds better and looks to be an improvement if you look at a waveform of
sample value versus time.

I have noticed something weird. A lot of recordings from BBC radio stations,
especially BBC Radio 4 which is the main thing I record, appear to have
asymmetric gain: the positive samples appear to have a different gain to the
negative ones, about the zero point. It doesn't look as if it's just DC
bias.

https://i.postimg.cc/BZ1fDvnx/Cool-Edit-Radio-4.png is an example. Recorded
from BBC R 4 on Freeview, audio extracted from the TS file and saved as WAV
file. 5 minute chunk displayed in CoolEdit. Both channels appear to have
consistently larger positive excursions than negative excursions about the 0
point - particular around times 10:10-10:20 and 12:20-12:40.

Is that to be expected? Do microphones tend to produce larger voltage
excursions on one side of the zero point than on the other side, about the
rest position of the diaphragm which I presume is deemed to be 0 volts. Is
it a myth that a signal will have equal positive and negative excursions for
any given part of the waveform? Is it more common with speech than music to
get unequal excursions?

https://i.postimg.cc/rFHBp0W6/Cool-Edit-CD.png is a sample of music from a
CD, grabbed by iTunes to mp3 and then converted to WAV. Similar horizontal
scale (3m 40s rather than about 5m).

So maybe it's a "funny" with speech - the Radio 4 example was from Last Word
which is mainly speech.

In article <u5krb9$c09b$1@dont-email.me>, NY <me@privacy.invalid> wrote:
> "Jim Lesurf" <noise@audiomisc.co.uk> wrote in message
> news:5aaf482ce6noise@audiomisc.co.uk...
> >> Ah, my mistake. I hadn't realised that some sound was even sampled at
> >> 96 kHz (ie 96,000 samples per second per channel).
> >
> > 'High end' tends to be even higher. 192k 24bit is a minimum for many
> > studios, etc. Look for the '2L' label's website and you can find some
> > free samples.

> What is the advantage of sampling at greater than about 45,000
> samples/sec? (44.1 for CD, 48 for most MPEG)

> Nyquist says that if you sample at n samples/sec, you can reproduce
> signal frequencies up to n/2 Hz.

Yes. However tha assumes some things which people often fail to realise.

e.g. For every instant in the output your DAC is expected to use *all* the
samples in the full-length 'record. i.e. if you have a recording 1000
seconds long then each individual instant you want out for the 'analogue
has to employ *all* 1000 x 44100 samples for that channel!

And all the computation has to use enough precision to avoid *any* data
contributions from going AWOL in the process. So far more precision than,
say 24bit.

Yes, some DAC designers try to get as wide a range employed. But only get
approximations that are "good enough" in someone's judgement.

Also it isn't n/2, but up to very *slightly* below that.

So in mere reality, all DACs and ADCs can only 'approximate' to Nyquist.

> The highest frequency that humans can
> hear is around 20 kHz, and this gradually decreases with age.

Depends on age, exposure to loud sounds, some medical conditions, and
biological sex.

> What is the benefit of sampling at roughly four times the maximum signal
> frequency - ie twice what Nyquist requires?

Makes it easier to get an accurate result for the output as the 'error
level' in any following process can be reduced if done decently.

> I can see the benefit in the size of those samples being larger than the
> default 16 bits at the mastering stage in a studio, as it allows
> lower-level signals to be boosted if required by multiplying the
> digital sample rather than requiring greater analogue gain before
> sampling, while still staying within the 16 bit range of the final
> release of the track (ie it doesn't introduce quantisation noise).

Given methods like 'noise shaping' (cf my webpages) you can 'swap' sample
rate for bit-depth to a significant extend. This also adds to exploiting
'dither' to allow signals *below* LSB amplitude to be preserved and
carried. And helps suppress quantisation distortions when combined with
shaping.

> (*) A brick wall transfer characteristic (gain=100 % for f < cutoff, and
> 0% for f> cutoff) would be the ideal but is impossible to to achieve
> with analogue components.

Also impossible for real-world digital. cf above.

> I have noticed something weird. A lot of recordings from BBC radio
> stations, especially BBC Radio 4 which is the main thing I record,
> appear to have asymmetric gain: the positive samples appear to have a
> different gain to the negative ones, about the zero point. It doesn't
> look as if it's just DC bias.

It is normal for various real world sounds to have asymmetric patterns.
again you can find examples from CDs on my webpages. e.g. Playing a violin
can make the string+bow interaction act like a 'stick - slip' oscillator.
Some notes on brass also do this, etc. There isn't any 'dc bias' as the
'smaller' side lasts longer, so the +v and -v excursions over some cycles
average away.

> https://i.postimg.cc/BZ1fDvnx/Cool-Edit-Radio-4.png is an example.
> Recorded from BBC R 4 on Freeview, audio extracted from the TS file and
> saved as WAV file. 5 minute chunk displayed in CoolEdit. Both channels
> appear to have consistently larger positive excursions than negative
> excursions about the 0 point - particular around times 10:10-10:20 and
> 12:20-12:40.

That does look like one-sided soft clipping. The time scale is too gross to
tell. But speech may well be asymmetric, as may singing. Depends on the
specific instance, etc. Asymmetry in peaks isn't rare.

> Is that to be expected? Do microphones tend to produce larger voltage
> excursions on one side of the zero point than on the other side, about
> the rest position of the diaphragm which I presume is deemed to be 0
> volts. Is it a myth that a signal will have equal positive and negative
> excursions for any given part of the waveform? Is it more common with
> speech than music to get unequal excursions?

Many mics have many defects, and may indeed be nonlinear for loud sounds.
That, and their varying frequency responses, etc, are why some singers have
'favourite' mic types. Various sounds may be asymmetric in terms of their
peaks +/-.

> https://i.postimg.cc/rFHBp0W6/Cool-Edit-CD.png is a sample of music from
> a CD, grabbed by iTunes to mp3 and then converted to WAV. Similar
> horizontal scale (3m 40s rather than about 5m).

> So maybe it's a "funny" with speech - the Radio 4 example was from Last
> Word which is mainly speech.

TBH I've come to the conclusion that the BBC have let most of their sound
engineers leave! If nothing else, the level varies wildly from one R4
programme to another. Often seem recorded by an amateur or the presenter,
now! Then may have a 'top or tail' added at a wildly *different* level to
promo some other 'podcast'. These days it seems routine to use such a
'podcast' version for the file made available from the radio page.

Coming back to the start: I tend to record 96k/24 because it is now easy to
do so and gives confidence in allowing a recorder peak well clear of
clipping and knowing that nothing audible will be lost due to the process
filtering in the sampling. I then may process down to, say 48k/16 the final
version at a sensible conversion gain level. i.e. no samples bigger than c
-3dFS to avoid risk of intersample overs. (In theory you need more
clearance but I've never seen the "Waveform From Hell" on any input,
despite seeing some pretty awful clicks from some LPs. (Again, on my
webpages.) I suspect LP would be incapable of generating the WFH. 8-]

One advantage of recording 96k/24 and then processing down is that it
allows noise shaping to be employed and get a lower audible noise floor
than if you did textbook 48k/16. Not likely to matter with an LP capture,
though. 8->

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics https://www.st-andrews.ac.uk/~www_pa/Scots_Guide/intro/electron.htm
biog http://jcgl.orpheusweb.co.uk/history/ups_and_downs.html
Audio Misc http://www.audiomisc.co.uk/index.html

In article <5ab032a08bnoise@audiomisc.co.uk>,
Jim Lesurf <noise@audiomisc.co.uk> wrote:
> In article <u5krb9$c09b$1@dont-email.me>, NY <me@privacy.invalid> wrote:

> > Nyquist says that if you sample at n samples/sec, you can
> > reproduce signal frequencies up to n/2 Hz.

> Yes. However tha assumes some things which people often fail to
> realise.

I'm certainly failing right now. :-)

> e.g. For every instant in the output your DAC is expected to use
> *all* the samples in the full-length 'record. i.e. if you have a
> recording 1000 seconds long then each individual instant you want
> out for the 'analogue has to employ *all* 1000 x 44100 samples for
> that channel!

I've read that and read it. I cannot understand it.

> And all the computation has to use enough precision to avoid *any*
> data contributions from going AWOL in the process. So far more
> precision than, say 24bit.

I also fail to understand any of that.

> Yes, some DAC designers try to get as wide a range employed. But
> only get approximations that are "good enough" in someone's
> judgement.

Jim, I don't want to attack, I want to understand could you expand
please?

My understanding is obviously very poor. Traditionally, I thought a
single sample when fed to a dac produced a proportional current which
is then converted to a spot voltage which is then fed to the
re-construction filter.

The only calculations I'm aware of are to do with error correction
prior to the dac getting the sample or up-sampling for systems that
do that.

Why more precision that 24 bit?

Myself, I think there is very little difference in hi-res audio ie.
above CD quality but there are advantages to be found in the softer
reconstruction filter with higher sample rates.

Bob.

"Bob Latham" <bob@sick-of-spam.invalid> wrote in message
news:5ab03a3072bob@sick-of-spam.invalid...
> In article <5ab032a08bnoise@audiomisc.co.uk>,
> Jim Lesurf <noise@audiomisc.co.uk> wrote:
>> In article <u5krb9$c09b$1@dont-email.me>, NY <me@privacy.invalid> wrote:
>
>> > Nyquist says that if you sample at n samples/sec, you can
>> > reproduce signal frequencies up to n/2 Hz.
>
>> Yes. However tha assumes some things which people often fail to
>> realise.
>
> I'm certainly failing right now. :-)
>
>> e.g. For every instant in the output your DAC is expected to use
>> *all* the samples in the full-length 'record. i.e. if you have a
>> recording 1000 seconds long then each individual instant you want
>> out for the 'analogue has to employ *all* 1000 x 44100 samples for
>> that channel!
>
> I've read that and read it. I cannot understand it.
>
>> And all the computation has to use enough precision to avoid *any*
>> data contributions from going AWOL in the process. So far more
>> precision than, say 24bit.
>
> I also fail to understand any of that.

Yes, I'm wondering whether I have forgotten or misunderstood what I learned
at university about signal reproduction - it was 40-odd year ago.

>> Yes, some DAC designers try to get as wide a range employed. But
>> only get approximations that are "good enough" in someone's
>> judgement.
>
> Jim, I don't want to attack, I want to understand could you expand
> please?
>
> My understanding is obviously very poor. Traditionally, I thought a
> single sample when fed to a dac produced a proportional current which
> is then converted to a spot voltage which is then fed to the
> re-construction filter.
>
> The only calculations I'm aware of are to do with error correction
> prior to the dac getting the sample or up-sampling for systems that
> do that.
>
> Why more precision that 24 bit?

I wonder I understand this correctly. If you take multiple (eg 2) samples at
24 bits precision and then average them to a single sample at the final
sampling speed for the released CD, then if the consecutive samples were 48
and 49 (for example) you can interpolate that to a single sample of 48.5 -
ie you've "gained" an extra bit of precision that wasn't there in any of the
original samples?

In article <u5na0j$o2gm$1@dont-email.me>,
NY <me@privacy.invalid> wrote:
> "Bob Latham" <bob@sick-of-spam.invalid> wrote in message
> news:5ab03a3072bob@sick-of-spam.invalid...
> > In article <5ab032a08bnoise@audiomisc.co.uk>,
> > Jim Lesurf <noise@audiomisc.co.uk> wrote:
> >> In article <u5krb9$c09b$1@dont-email.me>, NY <me@privacy.invalid> wrote:
> >
> >> > Nyquist says that if you sample at n samples/sec, you can
> >> > reproduce signal frequencies up to n/2 Hz.
> >
> >> Yes. However tha assumes some things which people often fail to
> >> realise.
> >
> > I'm certainly failing right now. :-)
> >
> >> e.g. For every instant in the output your DAC is expected to use
> >> *all* the samples in the full-length 'record. i.e. if you have a
> >> recording 1000 seconds long then each individual instant you
> >> want out for the 'analogue has to employ *all* 1000 x 44100
> >> samples for that channel!
> >
> > I've read that and read it. I cannot understand it.
> >
> >> And all the computation has to use enough precision to avoid
> >> *any* data contributions from going AWOL in the process. So far
> >> more precision than, say 24bit.
> >
> > I also fail to understand any of that.

> Yes, I'm wondering whether I have forgotten or misunderstood what I
> learned at university about signal reproduction - it was 40-odd
> year ago.

> >> Yes, some DAC designers try to get as wide a range employed. But
> >> only get approximations that are "good enough" in someone's
> >> judgement.
> >
> > Jim, I don't want to attack, I want to understand could you
> > expand please?
> >
> > My understanding is obviously very poor. Traditionally, I thought
> > a single sample when fed to a dac produced a proportional current
> > which is then converted to a spot voltage which is then fed to
> > the re-construction filter.
> >
> > The only calculations I'm aware of are to do with error
> > correction prior to the dac getting the sample or up-sampling for
> > systems that do that.
> >
> > Why more precision that 24 bit?

> I wonder I understand this correctly. If you take multiple (eg 2)
> samples at 24 bits precision and then average them to a single
> sample at the final sampling speed for the released CD, then if the
> consecutive samples were 48 and 49 (for example) you can
> interpolate that to a single sample of 48.5 - ie you've "gained" an
> extra bit of precision that wasn't there in any of the original
> samples?

That reminds me of what Philips did with their 14 bit dacs and over
sampling to get the 16 bit precision they needed.

Bob.

On 30/05/2023 09:45 am, Jim Lesurf wrote:

> JNugent <jenningsandco@mail.com> wrote:
>
>> I managed to extract the opinion that "good vinyl playback" was really
>> only available on a system that cost about the same as the price of a
>> nearly-new Mini.
>
> Well, in fairness if someone is willing to pay for a decent stereo based
> on, say CD, the cost of a decent deck, etc, shouldn't be a bar to getting
> decent results *from a good LP*. I doubt the cost of a small car is
> needed.

According to the staff at that shop (and this was 1990), you had to be
looking at £2,000 just for the turntable and cartridge.

> After that - as with cars - if you want something a bit better, the
> cost racks up. But you don't need much for many LPs as they tended to be
> cut to suit what people use, not to the best possible.
>
> FWIW If you want examples of the 'best' LPs, I'd say try the "Dragon's
> Dream" ones. They *are* expensive as they are direct cuts. But done with
> great care, and very impressive on a good system.

I buy according to the content, not the label!
>
> That said, I tend to make a digital copy for 'safety and convenience' and
> then play then instead! 8-] Keep the LPs.

I used to do that pre-1990 (LP to cassette), but with CD, of course, it
isn't necessary.