[Barry B]

I stumbled across the following article while, as usual, researching something else. The writer, Monty from Redcat, asserts that 16/44.1 is actually the best format sonically. I'm not quite schooled enough to be able to see holes in his theory, yet it goes counter to what my ears tell me. If any of you, more learned and versed in digital theory than I, have the time to peruse the article and comment I'd be very interested.

Here it is..

https://xiph.org/~xiphmont/demo/neil-young.html


As an aside.. Mike, thanks for the recommendation on the Tascam US-366. I picked one up and it does the job nicely. Much appreciated.

Best,

Barry

[Barry B]

On Sunday, January 24, 2016 at 12:40:48 PM UTC-5, Barry B wrote:
I stumbled across the following article while, as usual, researching something else. The writer, Monty from Redcat, asserts that 16/44.1 is actually the best format sonically. I'm not quite schooled enough to be able to see holes in his theory, yet it goes counter to what my ears tell me. If any of you, more learned and versed in digital theory than I, have the time to peruse the article and comment I'd be very interested.

Here it is..

https://xiph.org/~xiphmont/demo/neil-young.html


As an aside.. Mike, thanks for the recommendation on the Tascam US-366. I picked one up and it does the job nicely. Much appreciated.

Best,

Barry

My apologies.. I just saw that this very article was also mentioned in the "Sample Rates" thread back in December.. I hadn't read that thread yet. Still... I'm interested in any further discussion.

Thanks..

Barry

[Peter Larsen[_3_]]

On 24-01-2016 18:40, Barry B wrote:

I stumbled across the following article while, as usual, researching something else.
The writer, Monty from Redcat, asserts that 16/44.1 is actually the
best format sonically.

He writes about just about anything and the kitchen sink but not about
what is important: the AD conversion.

https://xiph.org/~xiphmont/demo/neil-young.html

Didn't read with any great care, stopped bothering when I saw the
Fletcher Munson threshold diagram. Its relevance is in the context of
perceptual encoding.

However the issue is that the bad stuff that some converters do is less
audible if the sample rate is higher. Because of this it has merit to
record with the sample rate that gives the best sounding recording.
Usually that will be the hardware maximum. It also has merit to do the
post at a high sample rate in case transient type noise is to be
algorithmically removed, because that facilitates detection.

I had preferred to use 64 kHz as house sample rate, but it appears that
it is not a widely held viewpoint. So I stick to 48 kHz because of the
storage space requirements of the higher rates. It is well worth
determining with the actual equipment whether recording at a higher rate
and then converting to a lower rate sounds better than recording at the
target sample rate for distribution.

Barry

Kind regards

Peter Larsen

[Peter Larsen[_3_]]

On 24-01-2016 20:06, Peter Larsen wrote:

On 24-01-2016 18:40, Barry B wrote:

I stumbled across the following article while, as usual, researching
something else.

The writer, Monty from Redcat, asserts that 16/44.1 is actually the
best format sonically.

He writes about just about anything and the kitchen sink but not about
what is important: the AD conversion.

https://xiph.org/~xiphmont/demo/neil-young.html

Didn't read with any great care, stopped bothering when I saw the
Fletcher Munson threshold diagram. Its relevance is in the context of
perceptual encoding.

However the issue is that the bad stuff that some converters do is less
audible if the sample rate is higher. Because of this it has merit to
record with the sample rate that gives the best sounding recording.
Usually that will be the hardware maximum. It also has merit to do the
post at a high sample rate in case transient type noise is to be
algorithmically removed, because that facilitates detection.

I had preferred to use 64 kHz as house sample rate, but it appears that
it is not a widely held viewpoint. So I stick to 48 kHz because of the
storage space requirements of the higher rates. It is well worth
determining with the actual equipment whether recording at a higher rate
and then converting to a lower rate sounds better than recording at the
target sample rate for distribution.

Barry

Oh, and yes, I agree that distributing 192 kHz samplerate files with 24
bit wordlength is a waste of CO2.

Kind regards

Peter Larsen

[Scott Dorsey]

Barry B wrote:
I stumbled across the following article while, as usual, researching someth=
ing else. The writer, Monty from Redcat, asserts that 16/44.1 is actually =
the best format sonically. I'm not quite schooled enough to be able to see=
holes in his theory, yet it goes counter to what my ears tell me. If any o=
f you, more learned and versed in digital theory than I, have the time to p=
eruse the article and comment I'd be very interested.

1. Having more bandwidth than you actually need means having to deal with
distortion products from any signal that might be part of that extended
bandwidth. So, if say you have tones at 25 Khz and 26 KHz from a fiddle
being recorded on your wideband system, you need to be damned sure that
you aren't creating a mixing product at 1 KHz. If you're cutting everything
off at 22 KHz you don't have to worry about that.

2. Higher sampling rates mean having converters that need to be much better
behaved. The settling time acceptable at 44.1 is not acceptable at 96.
The degree of jitter which is inaudible at 44.1 may be audible at 96.
If you actually do an A/B test you may find many commercial converters
sound better at 44.1 than they do at 96, because of the added stress put
on the technology at the higher rate.

Now.... none of these are problems with the format, they are problems with
the implementation. The first one can be fixed by low-passing everything
although of course that eliminates whatever benefit one might get from
the wider bandwidth. The second one can be fixed by putting more money into
converter design.

But, this being the real world where we have to deal with real implementations
there are good reasons to stay at 44.1. Still, I record at 96 when I have
customers who demand it.
--scott


--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Edi Zubovic]

On Sun, 24 Jan 2016 09:40:45 -0800 (PST), Barry B
wrote:

I stumbled across the following article while, as usual, researching something else. The writer, Monty from Redcat, asserts that 16/44.1 is actually the best format sonically. I'm not quite schooled enough to be able to see holes in his theory, yet it goes counter to what my ears tell me. If any of you, more learned and versed in digital theory than I, have the time to peruse the article and comment I'd be very interested.

Here it is..

https://xiph.org/~xiphmont/demo/neil-young.html


As an aside.. Mike, thanks for the recommendation on the Tascam US-366. I picked one up and it does the job nicely. Much appreciated.

Best,

Barry


--While I think that 192 kHz sampling rate is a bit an overkill, I
always work with 96 kHz and 64-bit depth. at 64 bit depth, you get a
complete mirroring at Nyquist frequency so if you know how to
linearize all in between, you have a flat, few-dB deviation, frequency
spectrum up to abt. 48 kHz. And no, these HF harmonics aren't
artifacts any more, there are more or less all there, no missing parts
and they are all leveled. Later, resampled to 44,1/16, things sound
different.

Work files ie. temporary files are huge per a minute of recording.
Therefore I have a standard hard disk for interim storage and a 13 GB
ramdrive. For interim storage, any standard hard disk drive is
sufficient and safe enough.

A big amount of data, repetive i/o of some 600 - 700 MB for average 3
min. of work, would kill a SSD sooner or later so this is why I
wouldn't recommend a standard SSD for this purpose.

With a ramdrive, I have some 7 to 8 GB per sec. I/O. You need a
stable, UPS powered PC and I still can't recommend a ramdisk for any
critical work. Obviously, after a restart after a sudden crash, the
"drive" is alwayas empty and this is as if you never did your steps
after those last saved to your hard disk.

Edi Zubovic, Crikvenica, Croatia

[Les Cargill[_4_]]

Edi Zubovic wrote:
On Sun, 24 Jan 2016 09:40:45 -0800 (PST), Barry B
wrote:

I stumbled across the following article while, as usual, researching something else. The writer, Monty from Redcat, asserts that 16/44.1 is actually the best format sonically. I'm not quite schooled enough to be able to see holes in his theory, yet it goes counter to what my ears tell me. If any of you, more learned and versed in digital theory than I, have the time to peruse the article and comment I'd be very interested.

Here it is..

https://xiph.org/~xiphmont/demo/neil-young.html


As an aside.. Mike, thanks for the recommendation on the Tascam US-366. I picked one up and it does the job nicely. Much appreciated.

Best,

Barry


--While I think that 192 kHz sampling rate is a bit an overkill, I
always work with 96 kHz and 64-bit depth. at 64 bit depth, you get a
complete mirroring at Nyquist frequency

How is it that there is an *in*complete mirroring @ Fs/2 for 64 bit?
Is that 64 bit linear? Floating point?

so if you know how to
linearize all in between, you have a flat, few-dB deviation, frequency
spectrum up to abt. 48 kHz. And no, these HF harmonics aren't
artifacts any more, there are more or less all there, no missing parts
and they are all leveled. Later, resampled to 44,1/16, things sound
different.


Yeah, resampling is frequently pretty nasty; especially between the 48k
family and the 44.1k family of sampling rates.

Work files ie. temporary files are huge per a minute of recording.
Therefore I have a standard hard disk for interim storage and a 13 GB
ramdrive. For interim storage, any standard hard disk drive is
sufficient and safe enough.

A big amount of data, repetive i/o of some 600 - 700 MB for average 3
min. of work, would kill a SSD sooner or later so this is why I
wouldn't recommend a standard SSD for this purpose.

With a ramdrive, I have some 7 to 8 GB per sec. I/O.

Yarg.

You need a
stable, UPS powered PC and I still can't recommend a ramdisk for any
critical work. Obviously, after a restart after a sudden crash, the
"drive" is alwayas empty and this is as if you never did your steps
after those last saved to your hard disk.

Edi Zubovic, Crikvenica, Croatia


--
Les Cargill

[Edi Zubovic]

On Mon, 25 Jan 2016 08:15:09 -0600, Les Cargill
wrote:

Edi Zubovic wrote:
On Sun, 24 Jan 2016 09:40:45 -0800 (PST), Barry B
wrote:

I stumbled across the following article while, as usual, researching something else. The writer, Monty from Redcat, asserts that 16/44.1 is actually the best format sonically. I'm not quite schooled enough to be able to see holes in his theory, yet it goes counter to what my ears tell me. If any of you, more learned and versed in digital theory than I, have the time to peruse the article and comment I'd be very interested.

Here it is..

https://xiph.org/~xiphmont/demo/neil-young.html


As an aside.. Mike, thanks for the recommendation on the Tascam US-366. I picked one up and it does the job nicely. Much appreciated.

Best,

Barry


--While I think that 192 kHz sampling rate is a bit an overkill, I
always work with 96 kHz and 64-bit depth. at 64 bit depth, you get a
complete mirroring at Nyquist frequency

How is it that there is an *in*complete mirroring @ Fs/2 for 64 bit?
Is that 64 bit linear? Floating point?
Yes it's floating point...


-----------------8-----------------------

Edi Zubovic, Crikvenica, Croatia

[[email protected]]

Edi Zubivic:

64BIT! That's practically analog in
terms of accuracy.

[Edi Zubovic]

On Mon, 25 Jan 2016 06:48:53 -0800 (PST),
wrote:

Edi Zubivic:

64BIT! That's practically analog in
terms of accuracy.

--And you have a _realy big_ safety margin! To clip that thing you
have to cross a lot over 0 dB dBFS. 24 bit is OK, but this...

Gosh if I had 128 bits, I'd use them... but that thing is AFAIK
non-existent.

Edi Zubovic, Crikvenica, Croatia

[John Williamson]

On 25/01/2016 14:59, Edi Zubovic wrote:
On Mon, 25 Jan 2016 06:48:53 -0800 (PST),
wrote:

Edi Zubivic:

64BIT! That's practically analog in
terms of accuracy.

--And you have a _realy big_ safety margin! To clip that thing you
have to cross a lot over 0 dB dBFS. 24 bit is OK, but this...

Gosh if I had 128 bits, I'd use them... but that thing is AFAIK
non-existent.


Given that no real world recording environment lets you make use of more
than 20 bits or so, and if you're in the lower 30 bits of a 64 bit
resolution recording, it will be pure Brownian motion noise anyway, even
in the complete absence of background sounds such as air conditioning
and distant traffic noise, the current 24 bits is more than good enough
to capture any sound perfectly.

You'd also need to cool the analogue stages and ADC with liquid nitrogen
to get the various sources of electronic noise down to that level.

--
Tciao for Now!

John.

[Scott Dorsey]

Edi Zubovic edi.zubovic[rem wrote:

--And you have a _realy big_ safety margin! To clip that thing you
have to cross a lot over 0 dB dBFS. 24 bit is OK, but this...

BUT, due to the nature of floating point, the louder the signal, the less
resolution you have.

With fixed point, it's very easy to know where you are and what your
precision is. With floating point you have an enormously increased range
but it can be difficult to know exactly how much range you have and what
your resolution at a given level is.

Floating point is a good thing, but too many people just handwave away all
the precision issues rather than actually do the numeric analysis, and
that's going to get people into trouble.

Of course, going to floating point intermediates has been the best thing that
ever happened to DAW systems and it's made it possible to do extensive
processing and be reasonably sure you aren't inadvertently truncating
anything. But it is not without its own set of risks.

Gosh if I had 128 bits, I'd use them... but that thing is AFAIK
non-existent.

You don't need any more exponent, but it wouldn't hurt to have more mantissa.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Scott Dorsey]

John Williamson wrote:
Given that no real world recording environment lets you make use of more
than 20 bits or so, and if you're in the lower 30 bits of a 64 bit
resolution recording, it will be pure Brownian motion noise anyway, even
in the complete absence of background sounds such as air conditioning
and distant traffic noise, the current 24 bits is more than good enough
to capture any sound perfectly.

The high resolution isn't needed for recording, it's needed for intermediate
processing. 16 bits is more than enough to capture the full dynamics of a
concert hall, but once you start doing equalization and (horrors) reverb
in software, you need much larger intermediate values in order to maintain
the full 16 bits of resolution.

That is why the push for huge floating point intermediate values inside the
DAW.
--scott


--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Barry B]

Scott and others,

Thanks for responding. I've inserted a few more questions below, not by way of debate (I don't know enough to debate anything re/ this yet), but just to help me understand the physics behind this better..

On Sunday, January 24, 2016 at 5:04:06 PM UTC-5, Scott Dorsey wrote:


1. Having more bandwidth than you actually need means having to deal with
distortion products from any signal that might be part of that extended
bandwidth. So, if say you have tones at 25 Khz and 26 KHz from a fiddle
being recorded on your wideband system, you need to be damned sure that
you aren't creating a mixing product at 1 KHz. If you're cutting everything
off at 22 KHz you don't have to worry about that.

Although the energy present at those upper frequencies is of course minimal compared to the fundamental and first few partials, frequencies in that range are always present. If you sample at 44.1khz, w/ the resulting upper frequency limit of 22khz, wouldn't you still have the same dilemma? There would be still be resultant frequencies caused by interaction between, say, 18khz and 18.5khz, at 500hz. The variances in amplitude of the upper harmonics are what, in large part, give instruments their own unique timbre. Wouldn't reproducing those extreme upper partials, as high as theoretically possible, result in a sound that is truer to the original tone, even though it may be only a very miniscule improvement? I know it's beyond the range of human hearing, so maybe this point is completely academic rather than practical. I'm just curious. I also realize that pretty much no audio equipment has a smooth frequency response (or any, in some cases, I guess) out to the 96khz limit of 192kz sampling.


2. Higher sampling rates mean having converters that need to be much better
behaved. The settling time acceptable at 44.1 is not acceptable at 96..
The degree of jitter which is inaudible at 44.1 may be audible at 96.
If you actually do an A/B test you may find many commercial converters
sound better at 44.1 than they do at 96, because of the added stress put
on the technology at the higher rate.

Sort of the digital analog (see what I did there? ) to the age-old axiom... put as little in between the mic and the listener as possible to achieve your goals, yes? I've done a little bit of comparison between recordings I've done at 16/44.1 and 24/96 (I don't have anything that'll record at a 192 rate, and probably wouldn't bother if I did), and played them both back at their original rates. The 24/96 recordings sound noticeably better to me, especially in terms of front-to-back depth and timbre. The difference is not as consistent among the few commercial 24/96 and 24/192 recordings I have. Some sound wonderful, some do not. There's also a marked difference in quality depending upon what DAC I use. The Tascam US-366 that I bought just so I'd have a digital in to my laptop works pretty good as a quick'n'easy usb recording interface, but the little Cambridge Audio DacMagic sounds less harsh to me, fuller, more natural, and still "airy." Sorry... pretty subjective terms. And yes, I know I'm hardly discussing "pro" audio gear; both of those pieces of equipment are sub-$200. Once I get the kinks out of this setup, I'll pull up my old Apogee PSX-100 and run a couple tests with that.

Now.... none of these are problems with the format, they are problems with
the implementation. The first one can be fixed by low-passing everything
although of course that eliminates whatever benefit one might get from
the wider bandwidth. The second one can be fixed by putting more money into
converter design.

But, this being the real world where we have to deal with real implementations
there are good reasons to stay at 44.1. Still, I record at 96 when I have
customers who demand it.
--scott

Makes perfect sense. I'd love to sit and pick your brain for at least a few days. I've been using Samplitude since I first found this newsgroup back in the 90s, and still use it. My old path for location symphonic recordings used to be two mics (Earthworks qtc-1 w/ Jecklin disc, ORTF w/ Neumann KM184, or pseudo-Blumlein using two AT-1050 in figure 8 mode) to a Great River preamp, Apogee PSX-100, to Sony PCM-R700 DAT. From there, I'd load the dat digitally to my desktop via a DELTA1010 interface, edit and "master" it w/ Samplitude,
then burn. For a duffer, I got some really good results, including one REALLY nice recording of the Bartok Roumanian Dances for string orchestra. One of the nicest string sounds I've heard recorded. Got lucky.. it sure wasn't any skill on my part (although I did play viola in the orchestra..grin ) .

Now I'm trying to keep the same mic/pre/apogee front end, but I'd like to record to hard disk or other media that allows for much quicker transfer and, assuming it's worth it, a higher sampling rate than DAT allows. I did a big band recording using the two Neumanns direct into the Tascam US-366 and my Win 7 laptop running Samplitude at 24/96 last week, and the results were encouraging, although I still don't trust the laptop for any "gotta get it" scenarios yet.. or maybe ever. In any case, when I burned a quick CD copy of the session, I just let Samplitude take care of the rate conversion with it's default settings. In hindsight, I should have recorded 1/2 the session at 16/44.1, then compared the final result once the 24/96 was converted to 16/44.1. Next time.

ANYTHING you'd like to offer would be wonderful.

Thanks in advance,

Barry

[John Williamson]

On 25/01/2016 15:48, Scott Dorsey wrote:
John Williamson wrote:
Given that no real world recording environment lets you make use of more
than 20 bits or so, and if you're in the lower 30 bits of a 64 bit
resolution recording, it will be pure Brownian motion noise anyway, even
in the complete absence of background sounds such as air conditioning
and distant traffic noise, the current 24 bits is more than good enough
to capture any sound perfectly.

The high resolution isn't needed for recording, it's needed for intermediate
processing. 16 bits is more than enough to capture the full dynamics of a
concert hall, but once you start doing equalization and (horrors) reverb
in software, you need much larger intermediate values in order to maintain
the full 16 bits of resolution.

That is why the push for huge floating point intermediate values inside the
DAW.

I don't have a problem with that, but I may have misread the thread,
which gave me the impression someone claimed to be recording at 64 bit
depth, and 96kHz, rather than upscaling and using that as an
intermediate step.


--
Tciao for Now!

John.

[Mike Rivers[_2_]]

On 1/25/2016 4:40 AM, Edi Zubovic wrote:
While I think that 192 kHz sampling rate is a bit an overkill, I
always work with 96 kHz and 64-bit depth.

Where are you getting 64-bit A/D converters? Or do you mean you're
allowing numbers to lengthen up to 64 bits during processing?

--
"Today's production equipment is IT based and cannot be operated without
a passing knowledge of computing, although it seems that it can be
operated without a passing knowledge of audio" - John Watkinson

Drop by http://mikeriversaudio.wordpress.com now and then

[Edi Zubovic]

On 25 Jan 2016 10:46:11 -0500, (Scott Dorsey) wrote:

Edi Zubovic edi.zubovic[rem wrote:

--And you have a _realy big_ safety margin! To clip that thing you
have to cross a lot over 0 dB dBFS. 24 bit is OK, but this...

BUT, due to the nature of floating point, the louder the signal, the less
resolution you have.

With fixed point, it's very easy to know where you are and what your
precision is. With floating point you have an enormously increased range
but it can be difficult to know exactly how much range you have and what
your resolution at a given level is.

Floating point is a good thing, but too many people just handwave away all
the precision issues rather than actually do the numeric analysis, and
that's going to get people into trouble.

Of course, going to floating point intermediates has been the best thing that
ever happened to DAW systems and it's made it possible to do extensive
processing and be reasonably sure you aren't inadvertently truncating
anything. But it is not without its own set of risks.

Gosh if I had 128 bits, I'd use them... but that thing is AFAIK
non-existent.

You don't need any more exponent, but it wouldn't hurt to have more mantissa.
--scott
No, it wouldn't be bad... but hey what a payload per a minute this
would be. Nowadays, we have far more processor and I/O power than that
of ten or fifteen years ago -- I smile while writing this -- but there
are limits. Of course, I use 64 bit-float for processing. In practice,
I sometimes get a silly looking over-the-top envelope. At 16 and even
24 bit, this would be a square wave generator output or so. At 64-bit,
I just normalize to zero-dB level and go on. Nice!

Edi Zubovic, Crikvenica, Croatia

[Edi Zubovic]

On Mon, 25 Jan 2016 08:36:52 -0800, Mike Rivers
wrote:

On 1/25/2016 4:40 AM, Edi Zubovic wrote:
While I think that 192 kHz sampling rate is a bit an overkill, I
always work with 96 kHz and 64-bit depth.

Where are you getting 64-bit A/D converters?

----- Or do you mean you're
allowing numbers to lengthen up to 64 bits during processing? ------

--Exactly that!

Edi Zubovic, Crikvcenica, Croatia

[Les Cargill[_4_]]

Scott Dorsey wrote:
John Williamson wrote:
Given that no real world recording environment lets you make use of more
than 20 bits or so, and if you're in the lower 30 bits of a 64 bit
resolution recording, it will be pure Brownian motion noise anyway, even
in the complete absence of background sounds such as air conditioning
and distant traffic noise, the current 24 bits is more than good enough
to capture any sound perfectly.

The high resolution isn't needed for recording, it's needed for intermediate
processing. 16 bits is more than enough to capture the full dynamics of a
concert hall, but once you start doing equalization and (horrors) reverb
in software, you need much larger intermediate values in order to maintain
the full 16 bits of resolution.


But 32 bit float should be 24 bits of mantissa, or so they say.


That is why the push for huge floating point intermediate values inside the
DAW.
--scott



--
Les Cargill

[Scott Dorsey]

Barry B wrote:

Although the energy present at those upper frequencies is of course minimal=
compared to the fundamental and first few partials, frequencies in that ra=
nge are always present. If you sample at 44.1khz, w/ the resulting upper f=
requency limit of 22khz, wouldn't you still have the same dilemma? There w=
ould be still be resultant frequencies caused by interaction between, say, =
18khz and 18.5khz, at 500hz.

Right, so distortion is still a problem. But the wider your bandwidth, the
less distortion you can get away with to get the same amount of distortion
products.

The variances in amplitude of the upper harmo=
nics are what, in large part, give instruments their own unique timbre. Wo=
uldn't reproducing those extreme upper partials, as high as theoretically p=
ossible, result in a sound that is truer to the original tone, even though =
it may be only a very miniscule improvement?

It would, if you could hear it. Same goes for stringed instruments too;
fiddles produce a whole lot of ultrasonic stuff.

The thing is, if it is low-passed accurately and well in the digital domain,
I can't hear the effect of cutting those off. Someone out there might be
able to hear them, I don't know. But I couldn't, and the people that were
tested in the Meyer and Moran study couldn't.

So... I'm not saying there is no possibility of any benefit, but I am saying
any possibility is very slim.

I know it's beyond the range =
of human hearing, so maybe this point is completely academic rather than pr=
actical. I'm just curious. I also realize that pretty much no audio equip=
ment has a smooth frequency response (or any, in some cases, I guess) out t=
o the 96khz limit of 192kz sampling.

Oh, we can build stuff that is smooth out to there if necessary, the problem
is that we have to sacrifice something else to do it. My worry is that people
are sacrificing things they can hear for things they cannot.

Makes perfect sense. I'd love to sit and pick your brain for at least a fe=
w days. I've been using Samplitude since I first found this newsgroup back=
in the 90s, and still use it. My old path for location symphonic recordin=
gs used to be two mics (Earthworks qtc-1 w/ Jecklin disc, ORTF w/ Neumann K=
M184, or pseudo-Blumlein using two AT-1050 in figure 8 mode) to a Great Riv=
er preamp, Apogee PSX-100, to Sony PCM-R700 DAT. From there, I'd load the =
dat digitally to my desktop via a DELTA1010 interface, edit and "master" it=
w/ Samplitude,=20
then burn. For a duffer, I got some really good results, including one REAL=
LY nice recording of the Bartok Roumanian Dances for string orchestra. One=
of the nicest string sounds I've heard recorded. Got lucky.. it sure wasn=
't any skill on my part (although I did play viola in the orchestra..grin=
) .

Now I'm trying to keep the same mic/pre/apogee front end, but I'd like to r=
ecord to hard disk or other media that allows for much quicker transfer and=
, assuming it's worth it, a higher sampling rate than DAT allows. I did a =
big band recording using the two Neumanns direct into the Tascam US-366 and=
my Win 7 laptop running Samplitude at 24/96 last week, and the results wer=
e encouraging, although I still don't trust the laptop for any "gotta get i=
t" scenarios yet.. or maybe ever. In any case, when I burned a quick CD co=
py of the session, I just let Samplitude take care of the rate conversion w=
ith it's default settings. In hindsight, I should have recorded 1/2 the se=
ssion at 16/44.1, then compared the final result once the 24/96 was convert=
ed to 16/44.1. Next time.

90% of your worries revolve around the converters, and maybe 10% of them
revolve around making sure everything getting data from the converters to
the final CD is bit-accurate. If you keep using your existing Apogee
converters than that 90% is taken care of.

ANYTHING you'd like to offer would be wonderful. =20

You need better microphones. No matter HOW good your microphones are, you
still need better microphones. If you have the best microphones in the world,
you still need better microphones. And a better room, too, of course. The
transducers and the rooms have been the limiting factor for the past fifty
years and I don't see that changing any time soon.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[JackA]

Man already attempted to impress using 35mm film for audio. It did't last long. Bottom line, few people care about superior audio quality.

I like the site that stated most would have to struggle at 160kbps and next to impossible at 256kbps, compared to CD, hearing the difference.

Jack

[Trevor]

On 26/01/2016 3:36 AM, Mike Rivers wrote:
On 1/25/2016 4:40 AM, Edi Zubovic wrote:
While I think that 192 kHz sampling rate is a bit an overkill, I
always work with 96 kHz and 64-bit depth.

Where are you getting 64-bit A/D converters?

Anywhere where BS sells more than fact.
IF you mean genuine 64 bit conversion, not in this universe!

Trevor.

[Scott Dorsey]

Les Cargill wrote:

But 32 bit float should be 24 bits of mantissa, or so they say.

The friendly people in the IEEE 754 committee say that is the best
format for the widest variety of uses.

There are plenty of other floating point formats out there, although
they have mostly fallen by the wayside since IEEE 754 is very good and
very standardized and there is a lot of commercial hardware available
for it.

It used to be common for people to build specialized hardware with
particular floating point formats designed for a single application, but
that's not going to happen anymore. So we get IEEE 754, which is never
the best but is always pretty good.
--scott


--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Wolfgang Schwanke]

Barry B wrote in
:

There would be still be resultant
frequencies caused by interaction between, say, 18khz and 18.5khz, at
500hz. The variances in amplitude of the upper harmonics are what, in
large part, give instruments their own unique timbre. Wouldn't
reproducing those extreme upper partials, as high as theoretically
possible, result in a sound that is truer to the original tone, even
though it may be only a very miniscule improvement?

Assuming this effect exists at all, then the beat frequency would
already be present in the air of the recording room and the membrane of
the microphone, and would be recorded as such without having to record
ultrasonics.

Always remember: The human ear has the same lowpass filter as digital
(and analogue) audio. Therefore whatever shortcomings you attribute to
the lowpass in audio equipment, the human hearing process has it too
and thus won't suffer any more if the recording process does the same.

--
John Peel is not enough

http://www.wschwanke.de/ http://www.fotos-aus-der-luft.de/
usenet_20031215 (AT) wschwanke (DOT) de

[Mike Rivers[_2_]]

On 2/2/2016 2:53 AM, Wolfgang Schwanke wrote:
Assuming this effect exists at all, then the beat frequency would
already be present in the air of the recording room and the membrane of
the microphone, and would be recorded as such without having to record
ultrasonics.

But they aren't. They're created by distortion in the signal path.

--
For a good time, visit http://mikeriversaudio.wordpress.com

[Scott Dorsey]

Wolfgang Schwanke wrote:
Barry B wrote in
:

There would be still be resultant
frequencies caused by interaction between, say, 18khz and 18.5khz, at
500hz. The variances in amplitude of the upper harmonics are what, in
large part, give instruments their own unique timbre. Wouldn't
reproducing those extreme upper partials, as high as theoretically
possible, result in a sound that is truer to the original tone, even
though it may be only a very miniscule improvement?

Assuming this effect exists at all, then the beat frequency would
already be present in the air of the recording room and the membrane of
the microphone, and would be recorded as such without having to record
ultrasonics.

No, we're talking about beat products that are produced by nonlinearity
in the electronics.

Now, there ALSO are apt to be beat products produced by nonlinearity in
the instrument mechanically, and maybe some produced by nonlinearity in
the ear. This is part of why it's hard to hear low order harmonic distortion
in the electronics, since they are masked by these nonlinearities.

Always remember: The human ear has the same lowpass filter as digital
(and analogue) audio. Therefore whatever shortcomings you attribute to
the lowpass in audio equipment, the human hearing process has it too
and thus won't suffer any more if the recording process does the same.

This is absolutely true. My point here is that if the audio system has
wider bandwidth, it needs to maintain more linearity than if it had more
narrow bandwidth.

The degenerate case of course is a system that has very narrow bandwidth,
such as the final in a broadcast transmitter. It can be very nonlinear,
since the harmonics will all be out of band and can be easily filtered out.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."