Question about a DVD-A
Greg
capabilities. My first DVD-A was R.E.M.'s "Automatic For The People."
Upon opening it, I noticed that it was mixed at 48kHz. I assumed
DVD-As were all mixed at 192kHz. But this one was only slightly
better than a CD!
Why??????
Do people generally buy DVD-As for the multi-channel sound? Maybe the
bonus features took up a lot of space?
Are DVD-As ever actually mixed (or, insert proper verb here) at
192kHz?
Any help is appreciated.
Thanks,
Greg
Steven Sullivan
> I recently bought a new DVD player because it had DVD-Audio
> capabilities. My first DVD-A was R.E.M.'s "Automatic For The People."
> Upon opening it, I noticed that it was mixed at 48kHz. I assumed
> DVD-As were all mixed at 192kHz. But this one was only slightly
> better than a CD!
> Why??????
Possibly to fit a larger amount of other content.
In any case, do you really think you could tell the difference
between 44, 48, 96 and 192 kHz sampling?
--
-S.
Colin
in blue, was recorded at 96khz 24 bit 6 channel and 2 channel and it has a
dolby digital music video. YES, Magnification, is 96khz 24 bit 6 channel
and 2 channel plus a dolby digital live performance. Fleetwood Mac,
Rumours, is 96khz 24 bit 6 and 2 channel. The other disk is a surround
setup and sampler disk. I don't know what to make of it yet. These all
sound great, so far I'm just using it as a 2 channel player, but soon I'll
hook up the 6 channels.
I guess at the lower sampling frequencies you may loose or have distortions
on some of the audible high frequency components of the music and loose of
ambiance too. try a different DVD-A disk. 48khz at 24 bit will still sound
better than a CD.
Shaun
"Greg" <gregm...@mindspring.com> wrote in message
news:bbs1r...@enews2.newsguy.com...
Schizoid Man
> capabilities. My first DVD-A was R.E.M.'s "Automatic For The People."
> Upon opening it, I noticed that it was mixed at 48kHz. I assumed
> DVD-As were all mixed at 192kHz.
Refer to a thread on the Google archive called 'SACD Does Not Make Technical
Sense'. It deals with the effect of sampling rates on the overall sound
produced by Red Books CDs, SACDs and DVD-As. It also has an interesting
discussion on whether sampling rates about 44.1kHz make any difference at
all, and whether higher sampling rates violate Shannon's theorem.
> But this one was only slightly better than a CD!
How are you making this qualified statement? The reason I am asking is
because the red book mixing of Automatic For The People is superb. If you're
hearing something that is not audible on the CD, then either your DVD-A
player's resolution is astonishing or your hearing is acute. Please
elaborate.
> Do people generally buy DVD-As for the multi-channel sound? Maybe the
> bonus features took up a lot of space?
It is quite possible that the bonus features did take up a lot of space.
> Are DVD-As ever actually mixed (or, insert proper verb here) at
> 192kHz?
Do you know whether the data was encoded in 1-bit word or 16-bit word
lengths? Without this information the sampling rate would be meaningless. I
do know that it is 16-bit for red books and 1-bit for SACDs.
Anyway I seriously doubt whether you will be able to make a difference at
various sampling rates without the use of a superb pair of full range
loudspeakers. What are you using?
Steven Sullivan
> I just bought a DVD-A player too and I have a four disks so far. The Corrs,
> in blue, was recorded at 96khz 24 bit 6 channel and 2 channel and it has a
> dolby digital music video. YES, Magnification, is 96khz 24 bit 6 channel
> and 2 channel plus a dolby digital live performance. Fleetwood Mac,
> Rumours, is 96khz 24 bit 6 and 2 channel. The other disk is a surround
> setup and sampler disk. I don't know what to make of it yet. These all
> sound great, so far I'm just using it as a 2 channel player, but soon I'll
> hook up the 6 channels.
> I guess at the lower sampling frequencies you may loose or have distortions
> on some of the audible high frequency components of the music and loose of
> ambiance too. try a different DVD-A disk. 48khz at 24 bit will still sound
> better than a CD.
Redbook was designed to capture all the audible high frequency components --
which is to say, everything up to 20 kHz -- with no distortion.
The relationship between 'ambience' and sampling rate and bit depth is
unclear to me, and i' d be interested to see the data. Ditto
the evidence that 48 kHz at 24 bit will necessarily
sound better -- or even different from -- 44 kHz at 16 bit,
Stewart Pinkerton
>I just bought a DVD-A player too and I have a four disks so far. The Corrs,
>in blue, was recorded at 96khz 24 bit 6 channel and 2 channel and it has a
>dolby digital music video. YES, Magnification, is 96khz 24 bit 6 channel
>and 2 channel plus a dolby digital live performance. Fleetwood Mac,
>Rumours, is 96khz 24 bit 6 and 2 channel.
The 'Yes' and 'Fleetwood Mac' albums are analogue tape recordings with
significantly less than 14 bits of dynamic range, so please discount
any claims of '24/96' *recording* quality for any album recorded
before the late '80s. In fact, given that almost all modern studio
microphones roll off above 18kHz, and no live performance has more
than 80-85dB dynamic range (many have less than 40dB!), we can safely
discount fanciful claims of '24/96' *content* for *any* commercial
recording. It's just a numbers game, a way of selling that back
catalogue one more time...................
OTOH, multi-channel may indeed be a genuine improvement, if
sensitively handled.
>I guess at the lower sampling frequencies you may loose or have distortions
>on some of the audible high frequency components of the music and loose of
>ambiance too. try a different DVD-A disk. 48khz at 24 bit will still sound
>better than a CD.
A most unreasonable claim, since this is for all practical purposes
identical to CD. Note that '24-bit' recordings never have dynamic
ranges greater than 80dB, i.e. less than 14 bits.
--
Stewart Pinkerton | Music is Art - Audio is Engineering
Steven Sullivan
> On 9 Jun 2003 04:05:30 GMT, "Colin" <sc...@mts.net> wrote:
>>I just bought a DVD-A player too and I have a four disks so far. The Corrs,
>>in blue, was recorded at 96khz 24 bit 6 channel and 2 channel and it has a
>>dolby digital music video. YES, Magnification, is 96khz 24 bit 6 channel
>>and 2 channel plus a dolby digital live performance. Fleetwood Mac,
>>Rumours, is 96khz 24 bit 6 and 2 channel.
> The 'Yes' and 'Fleetwood Mac' albums are analogue tape recordings with
> significantly less than 14 bits of dynamic range, so please discount
> any claims of '24/96' *recording* quality for any album recorded
> before the late '80s.
Magnification is their most recent studio album, from
a year or so ago, and was likely recorded
digitally. Perhaps you're thinking of 'Fragile', the 1971 album that's also
been released on DVD-A.
--
-S.
Stewart Pinkerton
Oooops... :-(
What, they're still alive?!? :-)
Steven Sullivan
> On 9 Jun 2003 17:50:20 GMT, Steven Sullivan <ssu...@panix.com> wrote:
>>Stewart Pinkerton <pat...@dircon.co.uk> wrote:
>>> On 9 Jun 2003 04:05:30 GMT, "Colin" <sc...@mts.net> wrote:
>>
>>>>I just bought a DVD-A player too and I have a four disks so far. The Corrs,
>>>>in blue, was recorded at 96khz 24 bit 6 channel and 2 channel and it has a
>>>>dolby digital music video. YES, Magnification, is 96khz 24 bit 6 channel
>>>>and 2 channel plus a dolby digital live performance. Fleetwood Mac,
>>>>Rumours, is 96khz 24 bit 6 and 2 channel.
>>
>>> The 'Yes' and 'Fleetwood Mac' albums are analogue tape recordings with
>>> significantly less than 14 bits of dynamic range, so please discount
>>> any claims of '24/96' *recording* quality for any album recorded
>>> before the late '80s.
>>
>>Magnification is their most recent studio album, from
>>a year or so ago, and was likely recorded
>>digitally. Perhaps you're thinking of 'Fragile', the 1971 album that's also
>>been released on DVD-A.
> Oooops... :-(
> What, they're still alive?!? :-)
After a fashion. ;>
(They're touring near you right now.)
--
-S.
Colin
news:bc281...@enews2.newsguy.com...
Well 44.1khz 16 bit music divides your sound into 65,536 individual level
steps, or quantizations, and 24 bit is a whooping 16,777,216 steps
(individual voltage levels that can be reproduced). There are claims
especially from the turn table fanatics that they can hear more from a
record than a CD because they can hear more than 16 bit digitization on
CD's. This "more sound is the ambiance that record lovers are most likely
talking about. If 16 bit was adequate, then why is the music industry
producing music and players that are higher bit rate. (or maybe it's just
to sell more product?).
As for the 44.1khz being adequate, I firmly believe that this is a mis
understanding of sampling theory. Nyquist states that to sample a
waveform,
you need to sample at a rate of at least 2 times the highest frequency of
the waveform you want to reproduce. But this only captures the fact that a
high frequency exists but you cannot accurately reproduce its amplitude or
phase relationship to the fundamental frequency. I'm sure I'm going to get
flak for this statement.
If you sample at 4 times 20khz or more ( 88.2, 96, 192khz), the phase and
amplitude of the music's higher frequencies are accurately reproduced and
will sound more realistic.
As for the bit depth, a record or a tape is analog and doesn't have voltage
steps, it's continuous. So if you digitize a master recording for DVD-A or
SACD or enhanced CD's from master tape, you can get a better recording than
regular CD.
Try drawing a sinewave on a piece of paper and draw it on zero line and a
scale. Then make two equally spaced points on 1 cycle of the wave so that
your simulating about 2 samples (not on the peaks). Now draw a second
scale
below the first one with the wave and place the points at the same
location.
Now from the zero point draw a line to the first point, then a line from
that to second point. You should have a saw tooth wave form. The filters
inside of the CD player will convert this saw tooth into a sine wave again
but notice that the peaks are at the the sampling points and not the peaks
of the original waveform. If you have 4 sample per cycle or even 10. Then
the true amplitude and phase relationship of the original signal can be
accurately reproduced. Try in again with more samples, you'll see that you
can see the original waveform in the simulated sample. I've done this
experiment on mathcad in college 9 years ago.
Colin
Stewart Pinkerton
>"Steven Sullivan" <ssu...@panix.com> wrote in message
>news:bc281...@enews2.newsguy.com...
>> Redbook was designed to capture all the audible high frequency components --
>> which is to say, everything up to 20 kHz -- with no distortion.
>> The relationship between 'ambience' and sampling rate and bit depth is
>> unclear to me, and i' d be interested to see the data. Ditto
>> the evidence that 48 kHz at 24 bit will necessarily
>> sound better -- or even different from -- 44 kHz at 16 bit,
>
> Well 44.1khz 16 bit music divides your sound into 65,536 individual level
> steps, or quantizations, and 24 bit is a whooping 16,777,216 steps
> (individual voltage levels that can be reproduced).
A common misconception, for in fact properly made 16-bit divides a
93dB dynamic range into a *continuous* series of levels whose
*uncertainty* is 1/65,536 of the full-scale voltage. 24-bit
quantisation does *exactly* the same, but with a 144dB theoretical
dynamic range (about 110-120dB in practice).
Unfortunately, no one has yet been able to identify a master tape with
more than 75-80dB dynamic range...............
> There are claims
> especially from the turn table fanatics that they can hear more from a
> record than a CD because they can hear more than 16 bit digitization on
> CD's. This "more sound is the ambiance that record lovers are most likely
> talking about.
Actually, the 'ambience' is midrange phasiness and the 'extra' detail
is just artificial compression used to lift detail above LP surface
noise. No LP has an intrinsic dynamic range of more than 80dB or so,
even discounting tape and microphone noise.
> If 16 bit was adequate, then why is the music industry
> producing music and players that are higher bit rate. (or maybe it's just
> to sell more product?).
Well spotted! :-)
> As for the 44.1khz being adequate, I firmly believe that this is a mis
> understanding of sampling theory. Nyquist states that to sample a
>waveform,
> you need to sample at a rate of at least 2 times the highest frequency of
> the waveform you want to reproduce. But this only captures the fact that a
> high frequency exists but you cannot accurately reproduce its amplitude or
> phase relationship to the fundamental frequency. I'm sure I'm going to get
> flak for this statement.
You can definitely reprduce its amplitude and phase accurately, so you
were right to expect flak for this flat-out wrong statement.
> If you sample at 4 times 20khz or more ( 88.2, 96, 192khz), the phase and
> amplitude of the music's higher frequencies are accurately reproduced and
> will sound more realistic.
No, they won't. Examine a 15kHz signal on a CD or on a 24/192kHz
DVD-A, and you'll find that they are identical - and essentially
perfect.
> As for the bit depth, a record or a tape is analog and doesn't have voltage
> steps, it's continuous.
Utter rubbish! Consider how tape works - magnetic domains.
> So if you digitize a master recording for DVD-A or
> SACD or enhanced CD's from master tape, you can get a better recording than
> regular CD.
Not within a 90dB dynamic range and a 20kHz frequency range, you
can't.
> Try drawing a sinewave on a piece of paper and draw it on zero line and a
> scale. Then make two equally spaced points on 1 cycle of the wave so that
> your simulating about 2 samples (not on the peaks). Now draw a second
>scale
> below the first one with the wave and place the points at the same
>location.
> Now from the zero point draw a line to the first point, then a line from
> that to second point. You should have a saw tooth wave form. The filters
> inside of the CD player will convert this saw tooth into a sine wave again
> but notice that the peaks are at the the sampling points and not the peaks
> of the original waveform. If you have 4 sample per cycle or even 10. Then
> the true amplitude and phase relationship of the original signal can be
> accurately reproduced. Try in again with more samples, you'll see that you
> can see the original waveform in the simulated sample. I've done this
> experiment on mathcad in college 9 years ago.
You misused the program, because the peaks should certainly *not* have
been in the positions you claim. This is partly because the internal
signal is *not* a saw tooth, but a series of steps with flat tops set
at the amplitude of the sampled values. Must try harder.......
Richard D Pierce
> As for the 44.1khz being adequate, I firmly believe that this is a mis
> understanding of sampling theory. Nyquist states that to sample a
>waveform,
> you need to sample at a rate of at least 2 times the highest frequency of
> the waveform you want to reproduce. But this only captures the fact that a
> high frequency exists but you cannot accurately reproduce its amplitude or
> phase relationship to the fundamental frequency. I'm sure I'm going to get
> flak for this statement.
You should expect flak, because your "interpretation" of the
sampling theorem is completely wrong, a point that is amplified
by the fact that the MEASURED phase relationships on 44.1 kHz
sampled systems is preserved quite accurately up to its Nyquist
limit.
>> If you sample at 4 times 20khz or more ( 88.2, 96, 192khz), the phase and
>> amplitude of the music's higher frequencies are accurately reproduced and
>> will sound more realistic.
False. Completely false. If your assertion were true, then it
would be trivial to measure increasing phase and amplitude
errors in such systems. When, in fact, we DO the measurements
(as opposed to simply specualting about them from a position of
misonformation), we find, in fact, that at ALL levels within the
dynamic range of the system, BOTH phase AND amplitude
relationships are preserved to within limits set only by the
noise of the system.
> As for the bit depth, a record or a tape is analog and doesn't have voltage
> steps, it's continuous.
First, you're wrong: master tapes are NOT a continuous
representation. Second, continuous DOES NOT MEAN infinitesimal
increments in meaningful levels: all media, INCLUDING CD, are
limited in their resolution by their fundamental noise levels
over the detection bandwidth. NO useful information exists below
the noise level within that detection badnwidth. (Before you fly
off with the "I can hear below the noise in analog, consider the
fact that you CAN ALSO hera below the quantization level in
dithered CD, within precisely the same limits set by the
detection bandwidth).
The notion that simly by virtue of the fact that a medium MAY be
continuous DOES NOT AUTOMATICALLY MEAN that the medium thus has
infinite resolution. Indeed, the total resolution of the system
is proprtional to the product of its bandwidth and its dynamic
range.
>> Try drawing a sinewave on a piece of paper and draw it on zero line and a
>> scale. Then make two equally spaced points on 1 cycle of the wave so that
>> your simulating about 2 samples (not on the peaks). Now draw a second
>>scale
>> below the first one with the wave and place the points at the same
>>location.
>> Now from the zero point draw a line to the first point, then a line from
>> that to second point. You should have a saw tooth wave form. The filters
>> inside of the CD player will convert this saw tooth into a sine wave again
>> but notice that the peaks are at the the sampling points and not the peaks
>> of the original waveform. If you have 4 sample per cycle or even 10. Then
>> the true amplitude and phase relationship of the original signal can be
>> accurately reproduced. Try in again with more samples, you'll see that you
>> can see the original waveform in the simulated sample. I've done this
>> experiment on mathcad in college 9 years ago.
Yes, and by your description you clearly did it wrong. I assume
you were graded accordingly.
Now, let's do the experiment with a 16-bit A/D-D/A system and,
guess what, it does NOT behave as you describe.
Instead of GUESSING or ASSUMING how things behave, is there an
issue with avtually TRYING it?
--
| Dick Pierce |
| Professional Audio Development |
| 1-781/826-4953 Voice and FAX |
| DPi...@world.std.com |
Colin
news:bc51p...@enews3.newsguy.com...
If you are right and claim to be, then why are people claiming better sound
quality from DVD-A and SACD 2 channel and 6 channel than with standard
CD's. My DVD-A player also upsamples CD's and the sound better after
upsampling?
If you have the hardware to do that with then you can. A computer simultion
program setup based on sampling theory can tell you things too. I wish I
could ge back to what I had done and check.
Colin
news:bc4sg...@enews4.newsguy.com...
> On 10 Jun 2003 04:41:12 GMT, "Colin" <sc...@mts.net> wrote:
>
> >"Steven Sullivan" <ssu...@panix.com> wrote in message
> >news:bc281...@enews2.newsguy.com...
>
> >> Redbook was designed to capture all the audible high frequency
components --
> >> which is to say, everything up to 20 kHz -- with no distortion.
> >> The relationship between 'ambience' and sampling rate and bit depth is
> >> unclear to me, and i' d be interested to see the data. Ditto
> >> the evidence that 48 kHz at 24 bit will necessarily
> >> sound better -- or even different from -- 44 kHz at 16 bit,
> >
> > Well 44.1khz 16 bit music divides your sound into 65,536 individual
level
> > steps, or quantizations, and 24 bit is a whooping 16,777,216 steps
> > (individual voltage levels that can be reproduced).
>
> A common misconception, for in fact properly made 16-bit divides a
> 93dB dynamic range into a *continuous* series of levels whose
> *uncertainty* is 1/65,536 of the full-scale voltage. 24-bit
> quantisation does *exactly* the same, but with a 144dB theoretical
> dynamic range (about 110-120dB in practice).
>
They are still voltage steps, or digitizations and from what audiophiles
seem to be hinting is that they can hear more that what's possible by 16 bit
sound. An analog source is continuous, not steps. I guess one could say
that the induvidual magnitic domains on the tape material could be viewed as
steps...... I wonder what resolution it is equivelent too? Or how about
stiction of the cutter head that makes a record... is it perfectly linear or
does it require a certain current change to move??
But if you look at that signal that the sampling circuit has produced of a
high frequency waveform or component. You can draw lower or higher
amplitude sinewave that FIT those sample.. This is my point. Two samples
(looking at things visuall) doesn't seem to have enough info to say what the
original waveform was.
Nousaine
One of the interesting things about these excercises is that they make some
fundamentally flawed assumptions; not the least of which is that a drawing of a
sine wave, or a scope trace, is "the signal" when it's just an analog
representaion of it.
Like wise the stair-step digital drawing is also not the actual signal or sound
but simply a representation of it.
Aldo Pignotti
>
> Anyway I seriously doubt whether you will be able to make a difference at
> various sampling rates without the use of a superb pair of full range
> loudspeakers. What are you using?
I have a home theater set up with Boston Acoustic VR-M60s up front.
These are a decent $900/pair speakers, perphaps a little bright for
2 channel listening. I have John Coltrane's "Ultimate Blue Train" on
cd which I used to listen to on a very good Sony cd player. Lately,
I've been listening to it on a Panasonic DVD-RP91K DVD player. I
also have a 2 channel, 24 bit, 96KHz DVD-Audio version of the
original recordings just called "Blue Train" and the difference is
quite audible. Of course, there may be other differences with the mix,
etc. but the DVD-Audio sound reminds me a lot of a very good LP,
without the background crackle and pops you get on vinyl.
Stewart Pinkerton
>"Stewart Pinkerton" <pat...@dircon.co.uk> wrote in message
>news:bc4sg...@enews4.newsguy.com...
>> On 10 Jun 2003 04:41:12 GMT, "Colin" <sc...@mts.net> wrote:
>> > Well 44.1khz 16 bit music divides your sound into 65,536 individual level
>> > steps, or quantizations, and 24 bit is a whooping 16,777,216 steps
>> > (individual voltage levels that can be reproduced).
>>
>> A common misconception, for in fact properly made 16-bit divides a
>> 93dB dynamic range into a *continuous* series of levels whose
>> *uncertainty* is 1/65,536 of the full-scale voltage. 24-bit
>> quantisation does *exactly* the same, but with a 144dB theoretical
>> dynamic range (about 110-120dB in practice).
>>
>They are still voltage steps, or digitizations and from what audiophiles
>seem to be hinting is that they can hear more that what's possible by 16 bit
>sound.
Two points - first, they are *not* steps, that's what dither is for,
and secondly, what audiophiles may 'hint' at does not necessarily have
anything to do with the real physical world.
> An analog source is continuous, not steps. I guess one could say
>that the induvidual magnitic domains on the tape material could be viewed as
>steps...... I wonder what resolution it is equivelent too?
Resolution? About 14-15 bits for a top-class analogue deck running two
tracks on 1/2 inch tape at 15-30 ips.
> Or how about
>stiction of the cutter head that makes a record... is it perfectly linear or
>does it require a certain current change to move??
Hardly the point, since there are infintely larger problems with
cutter heads! However, the surface noise limits resolution to the
equivalent of about 13-14 bits absolute maximum.
>> > If you sample at 4 times 20khz or more ( 88.2, 96, 192khz), the phase and
>> > amplitude of the music's higher frequencies are accurately reproduced and
>> > will sound more realistic.
>>
>> No, they won't. Examine a 15kHz signal on a CD or on a 24/192kHz
>> DVD-A, and you'll find that they are identical - and essentially
>> perfect.
>>
>But if you look at that signal that the sampling circuit has produced of a
>high frequency waveform or component. You can draw lower or higher
>amplitude sinewave that FIT those sample.. This is my point. Two samples
>(looking at things visuall) doesn't seem to have enough info to say what the
>original waveform was.
Yes it does, and you are flat out wrong. You can *not* draw a lower
frequency wave that fits those samples, either in frequency, amplitude
or phase, and while you certainly can fit a higher one, this is known
as aliasing, and is the reason why there is a bandwidth limit *before*
the A/D process. Look up Nyquist and Shannon for how sampling systems
work.
Richard D Pierce
>> False. Completely false. If your assertion were true, then it
>> would be trivial to measure increasing phase and amplitude
>> errors in such systems. When, in fact, we DO the measurements
>> (as opposed to simply specualting about them from a position of
>> misonformation), we find, in fact, that at ALL levels within the
>> dynamic range of the system, BOTH phase AND amplitude
>> relationships are preserved to within limits set only by the
>> noise of the system.
>
>If you are right and claim to be, then why are people claiming better sound
>quality from DVD-A and SACD 2 channel and 6 channel than with standard
>CD's.
Why ask me? It's THEIR claim, not mine. And, indeed, it's YOUR
claim.
Can you show, for example, that when comeone is making such a
comparison, they are listening to otherwise identical copies,
differing only in their format, sample rate, wod depth and so
on? Are there NO differences in mastering the two versions, no
differences in levels, no differences OTHER THAN THE FORMAT?
Unless you can show that, all claims to differences in format
are discountable.
>My DVD-A player also upsamples CD's and the sound better after
>upsampling?
And this is the most telling on two fronts.
First, upsampling CANNOT intoroduce any new information, indeed,
it canm't make things any smoother. By your description, ytou
obviously have never observed the proprties you claim exist.
Second, "upsampling" has been the norm in CD players fro well
over a decade, it is IDENTICAL to "oversampling," a valid
technique applied for the implementation of antialiasing and
reconstruction filters.
>> >> Try drawing a sinewave on a piece of paper and draw it on zero line and
>a
>> >> scale.
Here, once again, Colin applies what seems to be a perfectly
valid INTUITIVE process, but ends up witha result that is
completely wrong because there is no assurance that the
intuitive approach is correct. In this case, it's wrong, and the
result he reaches is wrong.
>> Yes, and by your description you clearly did it wrong. I assume
>> you were graded accordingly.
>>
>> Now, let's do the experiment with a 16-bit A/D-D/A system and,
>> guess what, it does NOT behave as you describe.
>>
>> Instead of GUESSING or ASSUMING how things behave, is there an
>> issue with avtually TRYING it?
>
>If you have the hardware to do that with then you can.
Indeed I do, many different versions. NONE of them behave as you
describe. Not one.
>A computer simultion
>program setup based on sampling theory can tell you things too.
A CORRECT computer simualtion will, indeed. A WRONG computer
simulation will give you wrong results, and you have presented
just such an example of doing it wrong. It's a lot easier to do
it wrong.
>I wish I
>could ge back to what I had done and check.
It would be far more productive to move forward and get it
right.
Again, I would suggest Pohlmann's "Pricniples of Digital Audio"
as a start.
Arny Krueger
news:WoyFa.905193$OV.840553@rwcrnsc54
> "Richard D Pierce" <DPi...@TheWorld.com> wrote in message
> news:bc51p...@enews3.newsguy.com...
>> On 10 Jun 2003 04:41:12 GMT, "Colin" <sc...@mts.net> wrote:
>>>> If you sample at 4 times 20khz or more ( 88.2, 96, 192khz), the phase
and
>>>> amplitude of the music's higher frequencies are accurately reproduced
and
>>>> will sound more realistic.
>> False. Completely false. If your assertion were true, then it
>> would be trivial to measure increasing phase and amplitude
>> errors in such systems. When, in fact, we DO the measurements
>> (as opposed to simply specualting about them from a position of
>> misonformation), we find, in fact, that at ALL levels within the
>> dynamic range of the system, BOTH phase AND amplitude
>> relationships are preserved to within limits set only by the
>> noise of the system.
> If you are right and claim to be, then why are people claiming better
> sound quality from DVD-A and SACD 2 channel and 6 channel than
> with standard CD's.
(1) People can claim what they will. Do they claim what they claim because
of hype or reliable perception?
(2) There is no such thing as a 6 channel standard red book CD
(3) DVDs that are neither DVD-A nor SACD but do have > 2 channels typically
use some kind of perceptual coding, which can cause audible degradation of
sound quality.
>My DVD-A player also upsamples CD's and they
> sound better after upsampling?
Tell me about your time-synched, level-controlled, bias-controlled listening
tests.
Other than that, this is just another anecdote.
>> Instead of GUESSING or ASSUMING how things behave, is there an
>> issue with actually TRYING it?
> If you have the hardware to do that with then you can.
Any good DAC performs as Pierce says it does. No breaks, no jumps.
> A computer
> simulation program setup based on sampling theory can tell you things
> too. I wish I could go back to what I had done and check.
Try it again, now that you've heard that it can't be the way that you
thought it was.
Richard D Pierce
>"Stewart Pinkerton" <pat...@dircon.co.uk> wrote in message
>news:bc4sg...@enews4.newsguy.com...
>> > steps, or quantizations, and 24 bit is a whooping 16,777,216 steps
>> > (individual voltage levels that can be reproduced).
>>
>> A common misconception, for in fact properly made 16-bit divides a
>> 93dB dynamic range into a *continuous* series of levels whose
>> *uncertainty* is 1/65,536 of the full-scale voltage. 24-bit
>> quantisation does *exactly* the same, but with a 144dB theoretical
>> dynamic range (about 110-120dB in practice).
>
>They are still voltage steps, or digitizations
No, there are not. You are fundamentally misundewrstanding the
behavior of dithered, quantized systems. From that
misunderstanding, no good comes.
>and from what audiophiles
>seem to be hinting is that they can hear more that what's possible by 16 bit
>sound. An analog source is continuous, not steps.
Neither is a properly dithered digital system: it's just as
continuous with the same resolution as an analog system with the
same noise floor.
And as to what "audiophiles are hinting at," please.
Many audiophiles more than hint at the benefit of green pens
applied to the edges of CDs.
Many audiophiles more than hint at the clear and obvious
benefits of Mpingo disks.
Many audiophiles hint at many things which are shown to be utter
hogwash.
>I guess one could say
>that the induvidual magnitic domains on the tape material could be viewed as
>steps...... I wonder what resolution it is equivelent too? Or how about
>stiction of the cutter head that makes a record... is it perfectly linear or
>does it require a certain current change to move??
It has noise, thus it has limited resolution. Unitl you
undertsnad this fundamental principle, you'll keep making wrong
conclusions.
>But if you look at that signal that the sampling circuit has produced of a
>high frequency waveform or component. You can draw lower or higher
>amplitude sinewave that FIT those sample..
You can "draw" anything you like: your "algorithm" for
reconstruction is just plain wrong. As long as you insist at
maintaining this wrong approach, you will never get the right
answer.
>This is my point.
And your point is wrong.
>Two samples
>(looking at things visuall)
THE DAC DOESN'T LOOK AT THINGS 'VISUALLY.' You insist on this
visual analog and drawing lines and all that is wrong and you've
been told it's wrong and still you insist on using it.
Why? WHy do you like wrong answers?
>doesn't seem to have enough info to say what the
>original waveform was.
Because the sitaution you just describe violates the NByquist
criteria. You need MORE THAN TWO points per cycle. That's
mathematically identical to saying that the sample rate must
exceed twice the sammpled bandwidth.
By insisting on two samples per cycle, YOU have violated the
Nyquist criteria. And, once you've violated it, you complain
things are broken.
Duh!
John Stimson
> If you are right and claim to be, then why are people claiming better sound
> quality from DVD-A and SACD 2 channel and 6 channel than with standard
> CD's. My DVD-A player also upsamples CD's and the sound better after
> upsampling?
The short answer to all this is "implementation."
An improper implementation of a 16-bit 44.1ksps DAC may introduce
audible effects that make you sit up and take notice, while improper
implementation in a 24-bit 96ksps DAC might cause only inaudible
artifacts. Or maybe the CD players in question just aren't of as high
quality as the DVD-A players. After all, the odds are that the DVD-A
players are newer and probably designed for the audiophile market (the
idea of a mass-market DVD-A player is absurd).
Real-time (FIR) digital filtering isn't perfectly accurate. The more
processing power you have available, the more accurate the results.
Processing power has gotten cheaper over time, so you would expect a
more expensive or newer player to have more accurate oversampling and
reconstruction filters.
Dick is right that sampled data contain phase and amplitude
information for all frequencies below the Nyquist frequency. Your
example is precisely at the Nyquist frequency, which is why you can't
uniquely determine the gain and phase.
Colin
You seem to be assuming a quanitization level, to be able to say that a deck
has 14 to 15 bits resolution. If the quantization steps were smaller then
the bits required to cover the full dynamic range of the deck would be
larger, would it not, or is this another matter of noise??
Ok, misunderstanding of sampling theory.
Colin
news:ykLFa.119102$DV.1...@rwcrnsc52.ops.asp.att.net...
> In article <bc6ct...@enews1.newsguy.com>, Colin <sc...@mts.net> wrote:
> >"Stewart Pinkerton" <pat...@dircon.co.uk> wrote in message
> >news:bc4sg...@enews4.newsguy.com...
> >level
> >> > steps, or quantizations, and 24 bit is a whooping 16,777,216 steps
> >> > (individual voltage levels that can be reproduced).
> >>
> >> A common misconception, for in fact properly made 16-bit divides a
> >> 93dB dynamic range into a *continuous* series of levels whose
> >> *uncertainty* is 1/65,536 of the full-scale voltage. 24-bit
> >> quantisation does *exactly* the same, but with a 144dB theoretical
> >> dynamic range (about 110-120dB in practice).
> >
> >They are still voltage steps, or digitizations
>
> No, there are not. You are fundamentally misundewrstanding the
> behavior of dithered, quantized systems. From that
> misunderstanding, no good comes.
So, I never took dithering in college related to audio reproduction. I have
worked on dithering circuits for motor controls. A digital signal is
switched on and off (1 or 0) at appropriate times to recreated the desired
signal, it's modulated with the analog signal. For motor controls, its
usually much more than 10 times the modulation rate, so that it's
transparent to the motor. If this is done to audio on a CD like you say,
How much extra resolution does it give. One level, two levels ??
>
> >and from what audiophiles
> >seem to be hinting is that they can hear more that what's possible by 16
bit
> >sound. An analog source is continuous, not steps.
>
> Neither is a properly dithered digital system: it's just as
> continuous with the same resolution as an analog system with the
> same noise floor.
>
> And as to what "audiophiles are hinting at," please.
>
> Many audiophiles more than hint at the benefit of green pens
> applied to the edges of CDs.
>
> Many audiophiles more than hint at the clear and obvious
> benefits of Mpingo disks.
>
> Many audiophiles hint at many things which are shown to be utter
> hogwash.
>
> >I guess one could say
> >that the induvidual magnitic domains on the tape material could be viewed
as
> >steps...... I wonder what resolution it is equivelent too? Or how about
> >stiction of the cutter head that makes a record... is it perfectly linear
or
> >does it require a certain current change to move??
>
> It has noise, thus it has limited resolution. Unitl you
> undertsnad this fundamental principle, you'll keep making wrong
> conclusions.
>
Are you saying that this applies over the whole dynamic range or just at the
noise floor. If it is just at the noise floor, then there are voltage steps
above that in a digital reproduction that are not continuous.
> >But if you look at that signal that the sampling circuit has produced of
a
> >high frequency waveform or component. You can draw lower or higher
> >amplitude sinewave that FIT those sample..
>
> You can "draw" anything you like: your "algorithm" for
> reconstruction is just plain wrong. As long as you insist at
> maintaining this wrong approach, you will never get the right
> answer.
Ok, my interpretation of the sampling is wrong, You've convinced me.
>
> >This is my point.
>
> And your point is wrong.
>
> >Two samples
> >(looking at things visuall)
>
> THE DAC DOESN'T LOOK AT THINGS 'VISUALLY.' You insist on this
> visual analog and drawing lines and all that is wrong and you've
> been told it's wrong and still you insist on using it.
Visualization is how we learn new concepts, it is a tool that we can use
for our minds.
>
> Why? WHy do you like wrong answers?
I don't like wrong answer, just an explanation!
> >doesn't seem to have enough info to say what the
> >original waveform was.
>
> Because the sitaution you just describe violates the NByquist
> criteria. You need MORE THAN TWO points per cycle. That's
> mathematically identical to saying that the sample rate must
> exceed twice the sammpled bandwidth.
>
> By insisting on two samples per cycle, YOU have violated the
> Nyquist criteria. And, once you've violated it, you complain
> things are broken.
>
> Duh!
No need to get sarcastic!
Richard D Pierce
>"Richard D Pierce" <DPi...@theworld.com> wrote in message
>news:ykLFa.119102$DV.1...@rwcrnsc52.ops.asp.att.net...
>>
>> No, there are not. You are fundamentally misundewrstanding the
>> behavior of dithered, quantized systems. From that
>> misunderstanding, no good comes.
>
>So, I never took dithering in college related to audio reproduction. I have
>worked on dithering circuits for motor controls. A digital signal is
>switched on and off (1 or 0) at appropriate times to recreated the desired
>signal, it's modulated with the analog signal. For motor controls, its
>usually much more than 10 times the modulation rate, so that it's
>transparent to the motor. If this is done to audio on a CD like you say,
>How much extra resolution does it give. One level, two levels ??
Tun understand this completely, you have to understand how the
principle detector we are concenred with functions. That
detector would be the ear. The ear averages signals over a
period of time. It DOES NOT look at a signal at an instant of
time. The mechanism by whgich it does this is often referred to
as cochlear filters. These are, essentially, bandpass filters in
the ear. In the midrange they are about 1/3 octave wide (a good
enough approximation for this discussion).
The ear is capable of detection regular signals below the level
of the noise because of the spectral differences between
"regular" signals and noise. The energy of a regular signal, as
in a musical note, is confined to a fairly narrow badwidth,
while the energy of noise is spread out over a very wide
bandwidth.
Let's assume, for the moment, that our noise has "white"
distribution, that is it has equal energy per cycle of
bandwidth, regardless of frequency. And let's assume that our
"signal" is a simple 1 kHz tone. If I put the 1 kHz tone through
a 1/3 octave wide bandpass filter centered on 1 kHz, since ALL
of the energy of that tone is at 1 kHz, it ALL gets through the
filter unattenuated.
On the other hand, if I put the noise through the same filter,
since the energy is spread out all over the place, some of it
gets through, and sone does not. How much? Well a 1/3 octave
wide filter at 1 kHz is about 230 Hz wide. If we assume the
noise is spread out over a 20 kHz bandwidth evenly, then the
filter allows 231/20000 of the total noise through. That means
the noise portion is attenuated by nearly 20 dB.
So, we can see that the ear, or any other detector of similar
properties, could hear a 1 kHz tone 20 dB below the broadband
noise floor.
Now, what DITHER does, in the audio quantization sense, is the
it linearizes the behavior of the converter over a time-averaged
basis. The way it works is that at every sample, BEFORE we
quantize it, we add a random signal to it of just the right
amplitude and just the right distribution. THEN we quantize it.
What happens is that the suignal level that is less than one
quantization step AND the added random value biases the
resulting truncated value so that the resulting AVERAGE value is
the same as the original signal.
Think of it using this simple gedanken: Imagine our converter
has a fundamental quantization step of 1 uV. WIthout dither, if
we put in 0.25uV, we'll get 0 out of the converter all the time.
You're right: we've losty anythin below 1 uV steps.
However, add a random signal whose level is +-0.5 mV to EVERY
value BEFORE truncating. Now watch what happens: some of thew
resulting values are still too small to triger the converter,
and out comes 0. But some are high enough to trigger, and out
comes 1. In fact, in this example, we'll find that about 3/4 of
the time, the value will be 0, and about 1/4 of the time, it
will be 1. SO we might get a serries of samples out of the
converter like:
0 0 1 0 1 0 0 0 0 1 0 0 1 0 0 0 1 0 0 0 0 0 1 1 ...
Once this is converted back into the analog domain, and run
through the reconstruction filter, guess what shows up on the
analog output? What will show up is a noisy 0.25 uV
Dither doesn't work on individual samples: we don't care,
because we don't (indeed we CAN'T) listen to individual
samples. We listen to the processed stream of air proessure
changes that cause movements of the ear drum: we don't hear
individual positions of the eardrum. Thus, just like we did in
our gedanken above, we AVERAGE the output.
As a result, we can make some reasonable stabs at your question:
how much is the increase in resolution? Well, the answer is as
dependent upon the detector (our ears) as anything. But, IF we
make the assumption that we're looking at mid-band signals, and
we assume the noise distribution of the dither is white, and the
amplaitude and its distribution is sufficient, then we CAN say
and, indeed, we can show, that in such a properly dithered
digital system, the ear CAN detect signals that are approaching
20 dB below the broadband raw quantization floor of the system.
The ultimate resoltuion is limited by the ear, and it's the same
limit for continuous analog systems.
Now, let's set our noise floor at 1 uV. In our system above,
that represented a quantization level of 1 bit. But let's
instead, add more "resolution" to the system. Let's say that 1
uV represents 8 bits of resolution, and that the ultimate
resolution of the system is now not 1 uV but instead its 1/256
(8 bits) lower, or 0.4 PICOvolts (that's billionths of a volt).
BUT LET'S KEEP THE NOISE FLOOR AT THE SAME LEVEL, around 1 uV.
How much more resoltuion would this new system have?
Well, you'd be inclined to say it's resolution is 256 times
smaller, but you'd be wrong, BECAUSE THE RESOLTUION OF THE
SYSTEM, AS FAR AS OUR EARS ARE CONCERNED< IS SET BY THE NOISE
FLOOR. Assume both systems are properly dithered, you're not
going to hear any deeper into one system or the other simply
because THE EAR ITSELF IS THE LIMIT. The inability of the ear
to average any better is what sets the limit.
>Are you saying that this applies over the whole dynamic range or just at the
>noise floor. If it is just at the noise floor, then there are voltage steps
>above that in a digital reproduction that are not continuous.
Nope, since the dithered is applied AL THE TIME, it adds noise
to EVERY step, no matter WHAT its level. Instead of doing our
gedanken at 0.25 uV, do it at 10,000.25 uV, 80 dB higher. You
get precisely the same behavior: that last nit will toggle due
to the dither, and the average will be a noisy 10,000.25 uV (but
the noise is still on the order of 1 uV.
>> >But if you look at that signal that the sampling circuit has produced of
>a
>> >high frequency waveform or component. You can draw lower or higher
>> >amplitude sinewave that FIT those sample..
>>
>> You can "draw" anything you like: your "algorithm" for
>> reconstruction is just plain wrong. As long as you insist at
>> maintaining this wrong approach, you will never get the right
>> answer.
>
>Ok, my interpretation of the sampling is wrong, You've convinced me.
Look at it in the following fashion. The amobguities you are
struggling with, i.e., the ability to draw many curves through
the same points have counterpart in reality: a digital signal
stream contains the baseband signal as well as ALL ITS IMAGES,
and there are an infinite number of images. What's an image?
It's the same as the baseband signal but reflected at multiples
of the baseband. A single sine wave at 20 kHz digitized at 44.1
kkHz has images at 24.1 kHz, 64.1 kHz, and so on. If you let all
that stuff through, then, indeed, you WILL get the wierd things
you draw on paper.
Let's make a leap here, do you now understand why the filter
applied to the output of a DAC is somethimes called an "ANTI-
IMAGING" filter? Get rid of ALL those out-of-band images, and
you will find that there is only one curve that can fit through
the sample points: ONLY ONE, and it is the original signal.
A simple MatLab simulation, UNLESS IT PROPERLY DEALS WITH IMAGE
REJECTION, will result in incorrect answers.
>> THE DAC DOESN'T LOOK AT THINGS 'VISUALLY.' You insist on this
>> visual analog and drawing lines and all that is wrong and you've
>> been told it's wrong and still you insist on using it.
>
>Visualization is how we learn new concepts, it is a tool that we can use
>for our minds.
But there is absolutely NO assurances that any given
visualization, even it is intuitively confortable, is right.
Indeed, the intuitive approach to sampling almost always is
wrong. Consider, also, how "intuitive" quantum mechanics is: it
is totally counterintuitive. But all indications are to date
that despite it's inherent non-visualizability, it is one of the
most correct theories yet developed.
>> Why? WHy do you like wrong answers?
>
>I don't like wrong answer, just an explanation!
Does this help?
>> >doesn't seem to have enough info to say what the
>> >original waveform was.
>>
>> Because the sitaution you just describe violates the NByquist
>> criteria. You need MORE THAN TWO points per cycle. That's
>> mathematically identical to saying that the sample rate must
>> exceed twice the sammpled bandwidth.
>>
>> By insisting on two samples per cycle, YOU have violated the
>> Nyquist criteria. And, once you've violated it, you complain
>> things are broken.
>>
>> Duh!
>
>No need to get sarcastic!
Not sarcastic, merely emphatic.
Bob Olhsson
<cygn...@idsfa.net> wrote:
>The short answer to all this is "implementation."
Implementation is usually the long answer too. Virtually every
generalization you'll read about audio technology is wrong.
--
Bob Olhsson Audio Mastery Recording Project Design and Consulting
Box 90412, Nashville TN 37209 Tracking, Mixing, Mastering, Audio for Picture
615.385.8051 FAX: 615.385.8196 Mix Evaluation and Quality Control
40 years of making people sound better than they ever imagined!
Steven Sullivan
> In article <heOFa.108966$d51.177324@sccrnsc01>, John Stimson
> <cygn...@idsfa.net> wrote:
>>The short answer to all this is "implementation."
> Implementation is usually the long answer too. Virtually every
> generalization you'll read about audio technology is wrong.
Including this one? (uh oh...in the old sci-fi shows,
this is where the computer exploded.)
--
-S.
Colin
Thanks, I think I'm starting to get the idea. More reading involved..
Uhhh!
Yes, I can see why now.
> A simple MatLab simulation, UNLESS IT PROPERLY DEALS WITH IMAGE
> REJECTION, will result in incorrect answers.
>
> >> THE DAC DOESN'T LOOK AT THINGS 'VISUALLY.' You insist on this
> >> visual analog and drawing lines and all that is wrong and you've
> >> been told it's wrong and still you insist on using it.
> >
> >Visualization is how we learn new concepts, it is a tool that we can use
> >for our minds.
>
> But there is absolutely NO assurances that any given
> visualization, even it is intuitively confortable, is right.
> Indeed, the intuitive approach to sampling almost always is
> wrong. Consider, also, how "intuitive" quantum mechanics is: it
> is totally counterintuitive. But all indications are to date
> that despite it's inherent non-visualizability, it is one of the
> most correct theories yet developed.
>
> >> Why? WHy do you like wrong answers?
> >
> >I don't like wrong answer, just an explanation!
>
> Does this help?
Yes, thanks.
Stewart Pinkerton
>In article <heOFa.108966$d51.177324@sccrnsc01>, John Stimson
><cygn...@idsfa.net> wrote:
>
>>The short answer to all this is "implementation."
>
>Implementation is usually the long answer too. Virtually every
>generalization you'll read about audio technology is wrong.
Well no, almost every generalisation you'll read is generally correct,
but there are almost always exceptions to those rules...
Stewart Pinkerton
wrote:
>Bob Olhsson <o...@hyperback.com> wrote:
I am Nomad.
What'sthe difference between theory and practice? In theory, there
*is* no difference, but in practice...... :-)