[Geoff]

Eiron wrote:
David Looser wrote:
worked quite well, but no match for a
software solution.

Scratch filters such as that in Goldwave are excellent; there is no
reason not to use them on a whole album. I tried subtracting the
'cleaned' version from the 'raw' one and was left with just the
clicks and scratches on a background of perfect silence, thus showing
that the filter didn't remove any music.

Waves X-Click/Noise/Crackle allow you to listen to the output audio, or the
'removed' audio realtime.

geoff

[David Looser]

"geoff" wrote in message
...
David Looser wrote:
"Peter Larsen" wrote in message
...


I have found it most useful to sample them at 96 kHz 16 bit so as to
save disks space, I don't see any logical reason in wasting it for
writting 16 binary ones for each sample, but I want a good sharp and
undistorted clicks in case automated click removal is relevant.
Mostly I just take the big ones out with fix single click
functionality,

Does anyone remember the Garrad "Music Recovery Module"? It was
designed to remove the big clicks in real time by briefly shunting
the audio with a light-dependent resistor when a click was detected.
Click detection was based on the idea that clicks were of large
amplitude, had a fast rise-time and had a significant out-of-phase
component. It actually worked quite well, but no match for a software
solution.

That would freak out on modern hip-hop stuff that has surface noise/clicks
as part of the 'music' !

It could be switched to "bypass", in which case it became merely a
high-quality RIAA pre-amp.

David.

[David Looser]

"geoff" wrote in message
...
Peter Larsen wrote:
Steven Sullivan wrote:


The peak signal when recording vinyl comes frm the clicks, usually 6
dB higher than the signal for the large ones. Grammophone records and
quality playback of them can provide a very high quality sound IF and
only IF all links are good.

I have found it most useful to sample them at 96 kHz 16 bit so as to
save disks space, I don't see any logical reason in wasting it for
writting 16 binary ones for each sample, but I want a good sharp and
undistorted clicks in case automated click removal is relevant.
Mostly I just take the big ones out with fix single click

Record at 24 bits, then once you've got rid of your clicks, then you can
raise the overall level with less degradation.

Why 96/16 rather than 44k1/24 ? I don't follow that logic. The highest
freq recorded on most LPs was around 15KHz, apart from clicks of course...


I'm not sure I follow yours. Whilst the highest *recorded" frequency was
around 15kHz, the click spectrum would go much higher than that so
preserving the fast risetime of the clicks would be of value to automatic
click detection software. On the other hand the S/N ratio of no better than
70dB requires only a 13-bit ADC, leaving a margin of 3 bits (18dB) for click
headroom/ post digitising amplification even when using a 16-bit converter.
And it doesn't matter if high-amplitude clicks are clipped, as long as the
rise-time is preserved.

I would be astonished if anyone could tell the difference between an
original 24-bit digitisation and a 16-bit one when digitising vinyl.

In the old days of 405-line TV the (AM) sound channel would have a simple
impulse interference reduction limiter fitted. This worked on the fact that
the impulse would have a far faster rise-time than any audio content. For
this reason the band-width of the sound IF channel was kept far wider than
need for the audio, around 100kHz.


David.

[Peter Larsen[_2_]]

David Looser wrote:

What millenium are you living in guys, look at the waveform, does it
hit 0 dB FS?

How do you look at an analogue waveform?,

Surely the screen image of the audio editor package is good enough.

we are talking about
setting the analogue level into the ADC.

And?

You can, of course, do a transfer, look at the resulting digital
waveform, and then re-do it if the levels are way off, but generally
it's easier to get it more or less correct the first time.

Correct transfer means that the loudest click is not clipped, you only need
to make that adjustment once.

David.


Kind regards

Peter Larsen

[Arny Krueger]

"David Looser" wrote in
message
"Peter Larsen" wrote in message
...
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0 dBFS,
unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can be
just shy of 0dBFS. I agree in practice a 3dB or so
margin between *apparent* peak and 0dBFS is advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?


How do you look at an analogue waveform?, we are talking
about setting the analogue level into the ADC.

You can, of course, do a transfer, look at the resulting
digital waveform, and then re-do it if the levels are way
off, but generally it's easier to get it more or less
correct the first time.

Most audio capture software has a real time display.

I set levels using a trackability test track. If the cartridge is
mistracking, tain't no need for that much more headroom in the preamp!

[Peter Larsen[_2_]]

Eiron wrote:

Scratch filters such as that in Goldwave are excellent; there is no
reason not to use them on a whole album. I tried subtracting the
'cleaned' version from the 'raw' one and was left with just the
clicks and scratches on a background of perfect silence, thus showing
that the filter didn't remove any music.

Take a known good recording with high quality transients, say a chamber
music recording made with DPA 4006 mikes. Pass it through a declicker,
notice the number of reported "fixed clicks". End of story, except that
there is incredibly poor audio around that will become less offensive via
good automated declick.


Kind regards

Peter Larsen

[David Looser]

"Peter Larsen" wrote in message
...
David Looser wrote:

What millenium are you living in guys, look at the waveform, does it
hit 0 dB FS?

How do you look at an analogue waveform?,

Surely the screen image of the audio editor package is good enough.

But you don't get to see that until *after* you've made the recording!!!

we are talking about
setting the analogue level into the ADC.

And?

BEFORE you make the recording (so you haven't got a screen image from the
audio editing package to look at yet!)

You can, of course, do a transfer, look at the resulting digital
waveform, and then re-do it if the levels are way off, but generally
it's easier to get it more or less correct the first time.

Correct transfer means that the loudest click is not clipped, you only
need to make that adjustment once.

Yes of course, did I say anything different?

David.

[David Looser]

"Arny Krueger" wrote in message
. ..
"David Looser" wrote in
message
"Peter Larsen" wrote in message
...
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0 dBFS,
unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can be
just shy of 0dBFS. I agree in practice a 3dB or so
margin between *apparent* peak and 0dBFS is advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?


How do you look at an analogue waveform?, we are talking
about setting the analogue level into the ADC.

You can, of course, do a transfer, look at the resulting
digital waveform, and then re-do it if the levels are way
off, but generally it's easier to get it more or less
correct the first time.

Most audio capture software has a real time display.

But it's not predictive. If it indicates clipping you can only go back and
start again.

I set levels using a trackability test track. If the cartridge is
mistracking, tain't no need for that much more headroom in the preamp!

Fine, but bearing in mind that there are large variations in maximim level
from LP to LP you will probably need to apply amplification to the digital
file on some (most?) of the recordings later.

David.

[Arny Krueger]

"David Looser" wrote in
message
"Arny Krueger" wrote in message
. ..
"David Looser" wrote in
message
"Peter Larsen" wrote in message
...
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0 dBFS,
unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can be
just shy of 0dBFS. I agree in practice a 3dB or so
margin between *apparent* peak and 0dBFS is advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?


How do you look at an analogue waveform?, we are talking
about setting the analogue level into the ADC.

You can, of course, do a transfer, look at the resulting
digital waveform, and then re-do it if the levels are
way off, but generally it's easier to get it more or
less correct the first time.

Most audio capture software has a real time display.

But it's not predictive. If it indicates clipping you can
only go back and start again.

Strictly speaking, nothing is predictive. But here is something that is
indicative:

The loudest trackability track on a test LP.

I set levels using a trackability test track. If the
cartridge is mistracking, tain't no need for that much
more headroom in the preamp!

Fine, but bearing in mind that there are large variations
in maximim level from LP to LP you will probably need to
apply amplification to the digital file on some (most?)
of the recordings later.

Something I happily do with 20:20 hindsight.

[Eiron]

Peter Larsen wrote:
Eiron wrote:

Scratch filters such as that in Goldwave are excellent; there is no
reason not to use them on a whole album. I tried subtracting the
'cleaned' version from the 'raw' one and was left with just the
clicks and scratches on a background of perfect silence, thus showing
that the filter didn't remove any music.

Take a known good recording with high quality transients, say a chamber
music recording made with DPA 4006 mikes. Pass it through a declicker,
notice the number of reported "fixed clicks". End of story,

Is there any particular LP you had in mind for this test?

--
Eiron.

[Peter Larsen[_2_]]

geoff wrote:

Why 96/16 rather than 44k1/24 ? I don't follow that logic.

Because the treble sounds cleaner with better inter-transient silence, and
that really matters with decayed audio, it gets less splatty.

One more reason for this practice was that the cpu load on the Celeron was
20 to 25 percent when recording at 96-16 with CE2k, I didn't want to push
the issue, life is too short for worrying about clicks. That machine is now
en route to become volkswagens and mobile phones, the only thing that was
broken was the CPU fan .... really really bad noise occasionally, but it had
a 12 year active life.

The
highest freq recorded on most LPs was around 15KHz, apart from clicks
of course...

Exactly. It is only an asumption - ie. I haven't asked on the adobe forums -
but said asumption is that the better the clicks are recorded the easier
they are to identify correctly. I don't care much about this, simply because
I don't see much relevance of doing anything but fix single click's on most
vinyl. I have however encountered one disk that was so decayed that it
needed treatment as if 78 rpm ... ie. center channel extract with suitably
modified settings. It was a mono record and the result was amazingly fine.

geoff


Kind regards

Peter Larsen

[Peter Larsen[_2_]]

David Looser wrote:

Surely the screen image of the audio editor package is good enough.

But you don't get to see that until *after* you've made the
recording!!!

Bad choice of software then. But it doesn't matter much. Take a record you
don't like and lower the cartridge on it with the lift while recording. When
that records cleanly the level adjustment is done.

Correct transfer means that the loudest click is not clipped, you
only need to make that adjustment once.

Yes of course, did I say anything different?

You do seem to consider it to be "an issue".

David


Kind regards

Peter Larsen

[Peter Larsen[_2_]]

David Looser wrote:

Fine, but bearing in mind that there are large variations in maximim
level from LP to LP you will probably need to apply amplification to
the digital file on some (most?) of the recordings later.

Of course you will, you have to record clicks that are 6 to 8 dB above the
strongest cut audio signal cleanly so that the are easy to identify for
click removal software. That too is the good practice reason for using
longer than 16 bit wordlength when recording.

David


Kind regards

Peter Larsen

[Peter Larsen[_2_]]

Eiron wrote:

Take a known good recording with high quality transients, say a
chamber music recording made with DPA 4006 mikes. Pass it through a
declicker, notice the number of reported "fixed clicks". End of
story,

Is there any particular LP you had in mind for this test?

I did it with a chamber music recording made with known minimalist
equipment. I don't think you got the point btw. ... "known good" was used in
the sense of a recording that is known to be good in terms of freedom from
clicks. That would not be something that was played back mechanically.


Kind regards

Peter Larsen

[Steven Sullivan]

In rec.audio.tech Peter Larsen wrote:
David Looser wrote:

Not necessarily ideal, due to the possibility of intersample peaks. It's
advisable to
record with peak samples a dB or three shy of 0 dBFS, unless you have
accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually know exactly
what the "peak of peaks" is, that can be just shy of 0dBFS. I agree
in practice a 3dB or so margin between *apparent* peak and 0dBFS is
advisable.

What millenium are you living in guys, look at the waveform, does it hit 0
dB FS?

Look at the waveform when? After it's recorded? By then it's too late.

Are you familiar with the concept of intersample peaks? It's mainly a
monitoring problem. Read more he

http://www.cadenzarecording.com/pape...distortion.pdf


___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

In rec.audio.tech David Looser wrote:
"geoff" wrote in message
around 15kHz, the click spectrum would go much higher than that so
preserving the fast risetime of the clicks would be of value to automatic
click detection software. On the other hand the S/N ratio of no better than
70dB requires only a 13-bit ADC, leaving a margin of 3 bits (18dB) for click
headroom/ post digitising amplification even when using a 16-bit converter.
And it doesn't matter if high-amplitude clicks are clipped, as long as the
rise-time is preserved.

I would be astonished if anyone could tell the difference between an
original 24-bit digitisation and a 16-bit one when digitising vinyl.

You must not visit 'audiophile' forums much. Such claims are routine
-- as is the claim that neither digitization will sound as good as the
vinyl. They';re never backed up with anything like hard evidence, of course
but they're not at all uncommon. So if you ever feel like being thus astonished,
or perhaps depressed, visit audioasylum.com or stevehoffman.tv



___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

In rec.audio.tech Arny Krueger wrote:
"David Looser" wrote in
message
"Peter Larsen" wrote in message
...
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0 dBFS,
unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can be
just shy of 0dBFS. I agree in practice a 3dB or so
margin between *apparent* peak and 0dBFS is advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?


How do you look at an analogue waveform?, we are talking
about setting the analogue level into the ADC.

You can, of course, do a transfer, look at the resulting
digital waveform, and then re-do it if the levels are way
off, but generally it's easier to get it more or less
correct the first time.

Most audio capture software has a real time display.

Yes, but is it accurate real time display, modeling
a reconstruction filter?

I don't know that those are so common. HEre's one:

http://www.secaudio.ch/side122.html

scroll down to the 'TL Mastermeter'



___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

In rec.audio.tech Peter Larsen wrote:
geoff wrote:

Why 96/16 rather than 44k1/24 ? I don't follow that logic.

Because the treble sounds cleaner with better inter-transient silence, and
that really matters with decayed audio, it gets less splatty.

Do tell. Your proof of this is....?




___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Peter Larsen[_2_]]

Steven Sullivan wrote:

In rec.audio.tech Peter Larsen wrote:
geoff wrote:

Why 96/16 rather than 44k1/24 ? I don't follow that logic.

Because the treble sounds cleaner with better inter-transient
silence, and that really matters with decayed audio, it gets less
splatty.

Do tell. Your proof of this is....?

I stated an opinion. I do not waste time proving recording choices, I make
them based on what sounds best.


Kind regards

Peter Larsen

[Peter Larsen[_2_]]

Steven Sullivan wrote:

Look at the waveform when? After it's recorded? By then it's too
late.

Are you familiar with the concept of intersample peaks? It's mainly a
monitoring problem. Read more he

http://www.cadenzarecording.com/pape...distortion.pdf

IF there is an inter-sample peak issue, then there will be two consecutive
samples at 0 dB FS. Quality audio software considers such samples to be
clipped and can provide a count of probably clipped samples.


Kind regards

Peter Larsen

[Peter Larsen[_2_]]

Steven Sullivan wrote:


[someone typed]

I would be astonished if anyone could tell the difference between an
original 24-bit digitisation and a 16-bit one when digitising vinyl.

You must not visit 'audiophile' forums much. Such claims are routine

For the wordlenght difference to matter in RECORDING there has to be signal
that is either truncated or hit by converter unlinearities. The latter
hardly applies in case of the same converter and the former does not apply
for vinyl. The only advantage in sampling at 24 or 32 bits is in the
workflow because a file format conversion can be skipped.

-- as is the claim that neither digitization will sound as good as the
vinyl. They';re never backed up with anything like hard evidence, of
course but they're not at all uncommon.

First of all it will not sound like grammophone playback that is influenced
via the grammophone hearing the loudspeakers. This is not a new issue. For
quality playback the grammophone has to be in another room than the one you
listen in. Next there is the issue of number of analog components the signal
passes through prior to being digitized.

The issue that matters is that the digitized recording does not deteriorate
further and that the life of the backup version, the vinyl disk, is greatly
extended by it not being played back.

-S


Kind regards

Peter Larsen

[Arny Krueger]

"Steven Sullivan" wrote in message

In rec.audio.tech Arny Krueger wrote:
"David Looser" wrote in
message
"Peter Larsen" wrote in message
...
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0 dBFS,
unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can be
just shy of 0dBFS. I agree in practice a 3dB or so
margin between *apparent* peak and 0dBFS is advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?


How do you look at an analogue waveform?, we are talking
about setting the analogue level into the ADC.

You can, of course, do a transfer, look at the resulting
digital waveform, and then re-do it if the levels are
way off, but generally it's easier to get it more or
less correct the first time.

Most audio capture software has a real time display.

Yes, but is it accurate real time display, modeling
a reconstruction filter?

Yes.

I don't know that those are so common. HEre's one:

http://www.secaudio.ch/side122.html

Scroll down to the 'TL Mastermeter'

Adobe Audition, AKA CoolEdit is widely used for transcribing LPs, and
provides a scrolling real-time display.

[Arny Krueger]

"Steven Sullivan" wrote in message

In rec.audio.tech Peter Larsen
wrote:
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0 dBFS,
unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can be
just shy of 0dBFS. I agree in practice a 3dB or so
margin between *apparent* peak and 0dBFS is advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?

Look at the waveform when? After it's recorded? By then
it's too late.

Are you familiar with the concept of intersample peaks?
It's mainly a monitoring problem. Read more he

http://www.cadenzarecording.com/pape...distortion.pdf

In the real world they happen, but they are relatively rare. If you record
with levels near FS, there might be one or two such clipped peaks every 3-5
minutes. Their duration is by definition very short. They are very unlikely
to be heard. They are more of an intellectual nuisance than anything else.

[Arny Krueger]

"Steven Sullivan" wrote in message

In rec.audio.tech David Looser
wrote:
"geoff" wrote in message
around 15kHz, the click spectrum would go much higher
than that so preserving the fast risetime of the clicks
would be of value to automatic click detection software.
On the other hand the S/N ratio of no better than 70dB
requires only a 13-bit ADC, leaving a margin of 3 bits
(18dB) for click headroom/ post digitising amplification
even when using a 16-bit converter. And it doesn't
matter if high-amplitude clicks are clipped, as long as
the rise-time is preserved.

I would be astonished if anyone could tell the
difference between an original 24-bit digitisation and a
16-bit one when digitising vinyl.

Agreed.

You must not visit 'audiophile' forums much. Such claims
are routine -- as is the claim that neither digitization
will sound as good as the
vinyl. They';re never backed up with anything like hard
evidence, of course
but they're not at all uncommon. So if you ever feel like
being thus astonished, or perhaps depressed, visit
audioasylum.com or stevehoffman.tv

I was discussing the results of some of my recent tests of MP3 coders with a
friend who had a long, sucessful career transferring analog tape to movie
film optical sound tracks. When I described how modern MP3 coders tend to
reduce information content by bringing up the noise level between musical
tones, he said: "You mean like vinyl or analog tape"?

LOL!

[Arny Krueger]

"Peter Larsen" wrote in message

Steven Sullivan wrote:

In rec.audio.tech Peter Larsen
wrote:
geoff wrote:

Why 96/16 rather than 44k1/24 ? I don't follow that
logic.

Because the treble sounds cleaner with better
inter-transient silence, and that really matters with
decayed audio, it gets less splatty.

Do tell. Your proof of this is....?

I stated an opinion. I do not waste time proving
recording choices, I make them based on what sounds best.

Almost right. When we record we make choices based on what sounds best to
us, or what we think is the best reasonble practice.

Admittedly, we don't subject every choice we make while recording to DBTs on
the spot.

However, a great amount of wisdom has forced itself into my life when I did
subject recording certain choices to DBTs.

[[email protected]]

On Dec 14, 3:17 am, "Peter Larsen" wrote:
Steven Sullivan wrote:
Look at the waveform when? After it's recorded? By then it's too
late.
Are you familiar with the concept of intersample peaks? It's mainly a
monitoring problem. Read more he
http://www.cadenzarecording.com/pape...distortion.pdf

IF there is an inter-sample peak issue, then there will be two consecutive
samples at 0 dB FS. Quality audio software considers such samples to be
clipped and can provide a count of probably clipped samples.

Why, two adjacent full-scale sample DO NOT a priori
mean that clipping occurred. Indeed, in the realm of
PCM stream, two such samples are perfectly legal
and represent v a completely valid waveform. Now,
yyour DAC may not be able to handle them, but that's
the DAC's problem, not a sample representation
problem.

MORE than two consecutive full-scale samples may be
another issue altogether.

Secondly, it does NOT require, as you claim, that the
two samples be at 0 dB FS. You could have two of them
a mere 1 LSB lower than 0 dB FS and STILL have the
same issue, only the peak will be proportionally lower,
i.e., not much. That's the difference between 0 dB
and -0.00027 dB FS.

Thirdly, if your "quality software" automatically assumes
that two adjacent o dB FS samples unambiguous clipping,
I'd posit it is NOT "quality software."

[Steven Sullivan]

In rec.audio.tech Peter Larsen wrote:
Steven Sullivan wrote:

In rec.audio.tech Peter Larsen wrote:
geoff wrote:

Why 96/16 rather than 44k1/24 ? I don't follow that logic.

Because the treble sounds cleaner with better inter-transient
silence, and that really matters with decayed audio, it gets less
splatty.

Do tell. Your proof of this is....?

I stated an opinion. I do not waste time proving recording choices, I make
them based on what sounds best.

Figured as much.


___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

In rec.audio.tech Peter Larsen wrote:
Steven Sullivan wrote:

Look at the waveform when? After it's recorded? By then it's too
late.

Are you familiar with the concept of intersample peaks? It's mainly a
monitoring problem. Read more he

http://www.cadenzarecording.com/pape...distortion.pdf

IF there is an inter-sample peak issue, then there will be two consecutive
samples at 0 dB FS. Quality audio software considers such samples to be
clipped and can provide a count of probably clipped samples.

Did you read any of the link I posted? Here's another

0dbFS+ Signals in Digital Mastering
http://www.tcelectronic.com/media/Le...per_AES109.pdf

The two consec samples needn't be AT 0 dBFS, for there to be intersample overs. With
contrived signals the ISOs can be as high as +6dBFS. Anecdotally, I see reports of +1 dBFS.
So consec samples at -0.5dB can still flank an ISO.


___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

In rec.audio.tech Peter Larsen wrote:
Steven Sullivan wrote:


[someone typed]

I would be astonished if anyone could tell the difference between an
original 24-bit digitisation and a 16-bit one when digitising vinyl.

You must not visit 'audiophile' forums much. Such claims are routine

For the wordlenght difference to matter in RECORDING there has to be signal
that is either truncated or hit by converter unlinearities. The latter
hardly applies in case of the same converter and the former does not apply
for vinyl. The only advantage in sampling at 24 or 32 bits is in the
workflow because a file format conversion can be skipped.

Look, I'm just reporting what *they* claim. Not what I think.

-- as is the claim that neither digitization will sound as good as the
vinyl. They';re never backed up with anything like hard evidence, of
course but they're not at all uncommon.

First of all it will not sound like grammophone playback that is influenced
via the grammophone hearing the loudspeakers. This is not a new issue. For
quality playback the grammophone has to be in another room than the one you
listen in. Next there is the issue of number of analog components the signal
passes through prior to being digitized.

If you record the output of the grammophone/cart/pre, you are capturing whatever the
grammaphone is 'hearing' from the loudspeakers.

The issue that matters is that the digitized recording does not deteriorate
further and that the life of the backup version, the vinyl disk, is greatly
extended by it not being played back.


You seem to be confusing me with someone who actually *believes* the audiophile
nosnense.



___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

In rec.audio.tech Arny Krueger wrote:
"Steven Sullivan" wrote in message

In rec.audio.tech Arny Krueger wrote:
"David Looser" wrote in
message
"Peter Larsen" wrote in message
...
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0 dBFS,
unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can be
just shy of 0dBFS. I agree in practice a 3dB or so
margin between *apparent* peak and 0dBFS is advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?


How do you look at an analogue waveform?, we are talking
about setting the analogue level into the ADC.

You can, of course, do a transfer, look at the resulting
digital waveform, and then re-do it if the levels are
way off, but generally it's easier to get it more or
less correct the first time.

Most audio capture software has a real time display.

Yes, but is it accurate real time display, modeling
a reconstruction filter?

Yes.

I don't know that those are so common. HEre's one:

http://www.secaudio.ch/side122.html

Scroll down to the 'TL Mastermeter'

Adobe Audition, AKA CoolEdit is widely used for transcribing LPs, and
provides a scrolling real-time display.

Are you sure Audition shows you when you're generating intersample overs?

It's not enough just to offer scrolling real-time display. The display has to 'know' what
consumer DAC/output stages do to a signal with high consecutive sample peaks -- most of the
ones tested by Nielsen and Lund had no headroom for properly reconstructing ISOs between such
sample peaks. Good peak monitoring should warn you when an 'illegal' reconstruction peak will
be generated between two 'legal' sample peaks.

Not saying Audition doesn't do it...just wondering how it's been determined
that it does.



___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[David Looser]

"Peter Larsen" wrote in message
...
David Looser wrote:

Surely the screen image of the audio editor package is good enough.

But you don't get to see that until *after* you've made the
recording!!!

Bad choice of software then.

You have *predictive* software? Just like the lifts in "Hitch-Hiker" it
knows what's going to happen *before* it happens? I'm amazed, does this
software also predict the numbers for next weeks lottery?

David.

[Arny Krueger]

"Steven Sullivan" wrote in message

In rec.audio.tech Arny Krueger wrote:
"Steven Sullivan" wrote in message

In rec.audio.tech Arny Krueger wrote:
"David Looser" wrote in
message
"Peter Larsen" wrote in message
...
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0
dBFS, unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can
be just shy of 0dBFS. I agree in practice a 3dB or
so margin between *apparent* peak and 0dBFS is
advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?


How do you look at an analogue waveform?, we are
talking about setting the analogue level into the ADC.

You can, of course, do a transfer, look at the
resulting digital waveform, and then re-do it if the
levels are way off, but generally it's easier to get
it more or less correct the first time.

Most audio capture software has a real time display.

Yes, but is it accurate real time display, modeling
a reconstruction filter?

Yes.

I don't know that those are so common. HEre's one:

http://www.secaudio.ch/side122.html

Scroll down to the 'TL Mastermeter'

Adobe Audition, AKA CoolEdit is widely used for
transcribing LPs, and provides a scrolling real-time
display.

Are you sure Audition shows you when you're generating
intersample overs?

Yes.

In fact I just pulled CE 2.1 up on this computer I'm typing one and made a
few instesample overs by hand, just to be zillion-times sure.

It's not enough just to offer scrolling real-time
display. The display has to 'know' what consumer
DAC/output stages do to a signal with high consecutive
sample peaks -- most of the ones tested by Nielsen and
Lund had no headroom for properly reconstructing ISOs
between such sample peaks. Good peak monitoring should
warn you when an 'illegal' reconstruction peak will be
generated between two 'legal' sample peaks.

I've been aware of intersample overs for years - based on experience with my
own work.

Not saying Audition doesn't do it...just wondering how
it's been determined that it does.

Real world experience. Just happened to have my eyes open while working on
some recordings that I made.

[Arny Krueger]

"Steven Sullivan" wrote in message

In rec.audio.tech Peter Larsen
wrote:
Steven Sullivan wrote:

Look at the waveform when? After it's recorded? By
then it's too late.

Are you familiar with the concept of intersample peaks?
It's mainly a monitoring problem. Read more he

http://www.cadenzarecording.com/pape...distortion.pdf

IF there is an inter-sample peak issue, then there will
be two consecutive samples at 0 dB FS. Quality audio
software considers such samples to be clipped and can
provide a count of probably clipped samples.

Did you read any of the link I posted? Here's another

0dbFS+ Signals in Digital Mastering
http://www.tcelectronic.com/media/Le...per_AES109.pdf

The two consec samples needn't be AT 0 dBFS, for there to
be intersample overs. With contrived signals the ISOs
can be as high as +6dBFS. Anecdotally, I see reports of
+1 dBFS. So consec samples at -0.5dB can still flank an
ISO.

With 16 bit samples 1 and 4 at zero, samples 2 and 3 at about 24,000 will
just give an over between them. With samples 1 and 4 at -32767, samples at
2 and 4 need to only be about 14,400 for there to be an over between them.
There will also be unders to the left and right of samples 1 and 4.

Again, in the real world this is pretty much moot. The ear can tolerate a
few dozen microseconds of light clipping as long as it is an isolated event.

[Peter Larsen[_2_]]

Arny Krueger wrote:

However, a great amount of wisdom has forced itself into my life when
I did subject recording certain choices to DBTs.

DBT's are good at large differences. It is also an excellent point to make
that a difference by definition is not a major difference if it doesn't show
up in a DBT. It is also extremely worthwhile to remember the differences in
tonality and imaging caused by moving a main pair 2 inches ....


Kind regards

Peter Larsen

[Peter Larsen[_2_]]

Steven Sullivan wrote:

Did you read any of the link I posted? Here's another

No.

0dbFS+ Signals in Digital Mastering
http://www.tcelectronic.com/media/Le...per_AES109.pdf

The two consec samples needn't be AT 0 dBFS, for there to be
intersample overs. With contrived signals the ISOs can be as high as
+6dBFS. Anecdotally, I see reports of +1 dBFS. So consec samples at
-0.5dB can still flank an ISO.

This is no doubt important to remember in an equipment design context. It is
considerably less important in the real world context of digitizing vinyl at
96 kHz sample rate with the clicks defining the record level setting such
that there are two unused bits above the audio. Also btw. the file is not
clipped if the samples are correct, strictly speaking this is a DA
conversion issue.

-S


Kind regards

Peter Larsen

[Peter Larsen[_2_]]

Steven Sullivan wrote:

If you record the output of the grammophone/cart/pre, you are
capturing whatever the grammaphone is 'hearing' from the loudspeakers.

That would be an incompetent thing to do, it is indeed one of the many
errors I too have made, but it is is not new knowledge.

You seem to be confusing me with someone who actually *believes* the
audiophile nosnense.

I did get that impression yes, my apology.

-S


Kind regards

Peter Larsen

[Peter Larsen[_2_]]

David Looser wrote:

"Peter Larsen" wrote in message
...

David Looser wrote:

Surely the screen image of the audio editor package is good enough.

But you don't get to see that until *after* you've made the
recording!!!

Bad choice of software then.

You have *predictive* software? Just like the lifts in "Hitch-Hiker"
it knows what's going to happen *before* it happens? I'm amazed, does
this software also predict the numbers for next weeks lottery?

You DO realize that you are quabling about the possibility of inter-sample
overs in DA conversion of a file that is recorded at 96 kHz sample rate with
2 full bits of headroom above the audio signal to make room for the clicks.
Those large clicks are later removed. The file is eventually as previously
suggested by me normalized to -2.5 dB ref. full scale.

The issue you worry about no doubt is real, but there is no data loss as
long as the sample values are correct. What this is about is that the analog
stage in front of the AD converter needs to have headroom above 0 dB FS and
that the analog stage after the AD converter needs to have headroom above 0
dB FS. The digital file can not be considered clipped if the sample values
are correct no matter how high the intersample peak happens to be.

If you have commercial pop cd's with a large number of consecutive samples
at 0 dB FS, then you have something to worry about and someting to rant away
about.

If you have such productions mp3's without a preceding gain reduction of
some 2 dB, then you will have even more overs ....

David


Kind regards

Peter Larsen

[Steven Sullivan]

In rec.audio.tech Arny Krueger wrote:
"Steven Sullivan" wrote in message

In rec.audio.tech Arny Krueger wrote:
"Steven Sullivan" wrote in message

In rec.audio.tech Arny Krueger wrote:
"David Looser" wrote in
message
"Peter Larsen" wrote in message
...
David Looser wrote:

Not necessarily ideal, due to the possibility of
intersample peaks. It's advisable to
record with peak samples a dB or three shy of 0
dBFS, unless you have accurate peak monitors
that show you what the *output* level is.

That's really a measurement problem. If you actually
know exactly what the "peak of peaks" is, that can
be just shy of 0dBFS. I agree in practice a 3dB or
so margin between *apparent* peak and 0dBFS is
advisable.

What millenium are you living in guys, look at the
waveform, does it hit 0 dB FS?


How do you look at an analogue waveform?, we are
talking about setting the analogue level into the ADC.

You can, of course, do a transfer, look at the
resulting digital waveform, and then re-do it if the
levels are way off, but generally it's easier to get
it more or less correct the first time.

Most audio capture software has a real time display.

Yes, but is it accurate real time display, modeling
a reconstruction filter?

Yes.

I don't know that those are so common. HEre's one:

http://www.secaudio.ch/side122.html

Scroll down to the 'TL Mastermeter'

Adobe Audition, AKA CoolEdit is widely used for
transcribing LPs, and provides a scrolling real-time
display.

Are you sure Audition shows you when you're generating
intersample overs?

Yes.

In fact I just pulled CE 2.1 up on this computer I'm typing one and made a
few instesample overs by hand, just to be zillion-times sure.

How? And what did it show?

Audition Help includes this warning

"If you're planning to put normalized audio on CD, you might want to normalize the waveforms
to no more than 96% as some audio compact disc players have problems accurately reproducing bits
that have been processed to 100% (maximum) amplitude."

And when I've normalize a music track to 0dBFS, I've never seen the Audition peak meter go
into the +0 zone. So I assumed that its peak meter does not model reconstructed output.




___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

In rec.audio.tech Peter Larsen wrote:
Arny Krueger wrote:

However, a great amount of wisdom has forced itself into my life when
I did subject recording certain choices to DBTs.

DBT's are good at large differences.

Actually, they're really quite useful for verifying audibility of *small* differences.
Hence their use in codec development and psychoacoustics research.

___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

In rec.audio.tech Peter Larsen wrote:
Steven Sullivan wrote:

If you record the output of the grammophone/cart/pre, you are
capturing whatever the grammaphone is 'hearing' from the loudspeakers.

That would be an incompetent thing to do, it is indeed one of the many
errors I too have made, but it is is not new knowledge.


? How do *you* digitize an LP, if not from the analog output?

Or are you just saying that when you do, you make sure there is not acoustic feedback from
nearby loudspeakers to the turntable?




___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason