SY said:
I thought it was some theorem by some Norman about
coordinates in the information space or something. Well, well.
You americans, well not so much you, have a tendency to
make a lot of jokes which are so related to american society
and culture that they are incomprehensible for many of us
others. I guess that is one reason why it is often so hard to
understand if Fred actually has something to say or not, one
doesn't quite know it is a joke or some obscure reference
to something actually relevant.
Guess I could slap back with this joke, that you only
understand if you understand swedish.
"Good day, do you have lightning eliminators?"
"No, but we've got zippers."
Have fun
It depends on what the definition of sinewave is
Sine wave for D.I.Y. poets and other non-technical majors.
I am waiting for someone to equate the dynamic range with noise
floor and wring another 10dB out of it. You guys should be in advertising not engineering......
We call this cooking the data in conversations with people with any degree of intellectual honesty. You guys should feel right at home with the dubious claims in high end audio that you find such difficulty with.
Sine wave for D.I.Y. poets and other non-technical majors.
I am waiting for someone to equate the dynamic range with noise
floor and wring another 10dB out of it. You guys should be in advertising not engineering......
We call this cooking the data in conversations with people with any degree of intellectual honesty. You guys should feel right at home with the dubious claims in high end audio that you find such difficulty with.
Attachments
SY said:
Yes, that is basically the assumption, but I suppose the reason
we can recover info in the noise is that we exploit the fact that
music is somewhat periodical (that is a very sloppy term, I know)
and this is sufficient although it doesn't work as well as for
real periodic signals.
My understanding is that, in the digital case, this all relies on
what I discussed earlier about information content in music vs.
noise, that is, we exploit the fact that the music is not
optimally encoded from an information-theoretic point of view,
while for instance noise would be near-optimally encoded and
nothing can be gained from adding noise (to the noise).
WATCH OUT....
Hi,
Just put my Dieckmann cap on...
From what I read here I can only conclude that most people forgot how those 16 bits are achieved.
Maths are one thing, reality is another.
Now go figure it out for yourself and think hard on how that 16 bit claim came about and maybe, just maybe someone might see the light.
Deep sigh...no pun intended.
Hi,
Just put my Dieckmann cap on...
From what I read here I can only conclude that most people forgot how those 16 bits are achieved.
Maths are one thing, reality is another.
Now go figure it out for yourself and think hard on how that 16 bit claim came about and maybe, just maybe someone might see the light.
Deep sigh...no pun intended.
Re: It depends on what the definition of sinewave is
No, I don't think I could convince the potential customers
that a bit has only two values and that this is better than
the three values it actually has. You know, customers usually prefer the higher figures.
BTW, I haven't been in engineering professionally for over
15 years. On the other hand I am in computer science and
do know a bit or two about binary numbers that comes from
credible textbooks and not from EEs aspiring to revolutionize
information theory.
Fred Dieckmann said:
No, I don't think I could convince the potential customers
that a bit has only two values and that this is better than
the three values it actually has. You know, customers usually prefer the higher figures.
BTW, I haven't been in engineering professionally for over
15 years. On the other hand I am in computer science and
do know a bit or two about binary numbers that comes from
credible textbooks and not from EEs aspiring to revolutionize
information theory.
Christer said:BTW, I still don't get it how kuei can get differenc dynamics
for peak and RMS values?
Which post was that?
What we've been discussing (or at least what I've been discussing) has been an undithered digital system. In an undithered system, anything below -96dB gets quantized to 0 and indeed you get nothing.
In a dithered system, a small amount of random noise is added during the analog to digital conversion. The noise adds to the signal which allows signals that would otherwise be below +/-1/2LSB to be encoded. Since the signal is periodic and the noise is random, when you average the noise, your signal appears from up out of the grass.
Our ear/brain system does the averaging which is why we're able to detect signals well below the noise floor.
Dithering allows signals well below -96dB to be recorded. It also decorrelates the quantization error from the signal so there's no distortion per se, just signal and noise. So in the end, you ultimately have what amounts to an analogue system with some noise. And using noise shaping techniques, you can make the noise even less intrusive.
se
Re: Re: Re: It depends on what the definition of sinewave is
I think it is still classiFieD until a certain EE in Texas has FileD his
patent on his newly invented three-value bits, which, by the way,
I suspect he stole the idea for from the Soviet ternary-computer project back in the sixties. I am sure he will get a well-paid
job att TI/BB designing ternary output DACs.
Steve Eddy said:
I think it is still classiFieD until a certain EE in Texas has FileD his
patent on his newly invented three-value bits, which, by the way,
I suspect he stole the idea for from the Soviet ternary-computer project back in the sixties. I am sure he will get a well-paid
job att TI/BB designing ternary output DACs.
Shannon saves
Personally, I prefer "nats", with ~2.7 values each. But, then, I'm a linguist.
I once had a long, pointless argument with a colleague about whether it's possible to build a unary computer which uses single-valued units of information. My position was no: yin can't mean anything unless you've also got yang.
-- Rob<space>M (not in New Jersey)
Steve Eddy said:
Personally, I prefer "nats", with ~2.7 values each. But, then, I'm a linguist.
I once had a long, pointless argument with a colleague about whether it's possible to build a unary computer which uses single-valued units of information. My position was no: yin can't mean anything unless you've also got yang.
-- Rob<space>M (not in New Jersey)
Steve Eddy said:
I saw it not long ago, but I can't seem to find it now. I am sure
Kuei will be happy to tell us again how many dB difference there
was between peak- and RMS-value dynamics.
Good, it seems I had about the right idea of it. BTW, what is
the difference between dithering and noise-shaping? Is it just
different noise spectra, or is there something more to it?
Re: Shannon saves
I haven't a clue what a nat is, but the reason the russians
experimented with ternary (ie. three-valued logic) computers
was that 3 is closer to e (which is ~2.7) than 2 is, and that
this would be beneficial. Of course, here it does take some
EE to see why it is not such a good idea after all.
Well, both yes and no, in a sense. you can encode numbers
in unary notation, one usually writes them as a string of '1's
such that the number 7 is written as 1111111 for instance.
That works well in theory (or rather, depends on what we
are using them for) but couldn't be implemented as a computer
either in theory or in practice. You need som kind of storage
space with well-defined information containers (bits in a normal
computer). So we could, in principle. put each of the '1's in their
own unary containers. however, you must be able to somehow represent where one number ends and another begins, which
you cannot do with the unary value '1'. That is, the containers
must be at least binary.
However, the axiomatic foundation for ordinary integer algebra
in mathematics is based on a kind of unary notation, based
on the successor function, but that's perhaps getting too much
off-topic.
Rob M said:
I haven't a clue what a nat is, but the reason the russians
experimented with ternary (ie. three-valued logic) computers
was that 3 is closer to e (which is ~2.7) than 2 is, and that
this would be beneficial. Of course, here it does take some
EE to see why it is not such a good idea after all.
I once had a long, pointless argument with a colleague about whether it's possible to build a unary computer which uses single-valued units of information. My position was no: yin can't mean anything unless you've also got yang.
Well, both yes and no, in a sense. you can encode numbers
in unary notation, one usually writes them as a string of '1's
such that the number 7 is written as 1111111 for instance.
That works well in theory (or rather, depends on what we
are using them for) but couldn't be implemented as a computer
either in theory or in practice. You need som kind of storage
space with well-defined information containers (bits in a normal
computer). So we could, in principle. put each of the '1's in their
own unary containers. however, you must be able to somehow represent where one number ends and another begins, which
you cannot do with the unary value '1'. That is, the containers
must be at least binary.
However, the axiomatic foundation for ordinary integer algebra
in mathematics is based on a kind of unary notation, based
on the successor function, but that's perhaps getting too much
off-topic.
Re: It depends on what the definition of sinewave is
You're the one who should be in advertising.
You're constantly throwing up stuff like this as if on its face it's supposed to be immediately obvious what it is you're trying to say yet you never actually come out and say what it is you're trying to say leaving people to try and second guess.
This allows you to make it appear that you might actually know something without having to commit to making an actual argument which might show that you don't.
Nice and safe.
Are you taking to heart Mark Twain's adivce? That it's better to keep quiet and let people suspect you're an idiot than to open your mouth and remove all doubt?
Look, just put your coloring book away for a while and say what it is you're trying to say. If must use your coloring book, at least acommpany your pictures with an accompanying argument rather than mindless insults.
Like how Kuei cooked the data here by leaving out the fact that there are TWO times 2^15 discrete levels?:
As said, 16 Bit are realy 15Bit plus "sign", that translates into a
system capable of representing 2^15 discrete levels.
Your illustration is actually quite a good illustration here. That half-sine pulse at the bottom illustrates quite well how Kuei wheh he made his original claim was only looking at half the picture.
se
Fred Dieckmann said:
You're the one who should be in advertising.
You're constantly throwing up stuff like this as if on its face it's supposed to be immediately obvious what it is you're trying to say yet you never actually come out and say what it is you're trying to say leaving people to try and second guess.
This allows you to make it appear that you might actually know something without having to commit to making an actual argument which might show that you don't.
Nice and safe.
Are you taking to heart Mark Twain's adivce? That it's better to keep quiet and let people suspect you're an idiot than to open your mouth and remove all doubt?
Look, just put your coloring book away for a while and say what it is you're trying to say. If must use your coloring book, at least acommpany your pictures with an accompanying argument rather than mindless insults.
Like how Kuei cooked the data here by leaving out the fact that there are TWO times 2^15 discrete levels?:
As said, 16 Bit are realy 15Bit plus "sign", that translates into a
system capable of representing 2^15 discrete levels.
Your illustration is actually quite a good illustration here. That half-sine pulse at the bottom illustrates quite well how Kuei wheh he made his original claim was only looking at half the picture.
se
Re: WATCH OUT....
No, not forgotten at all, merely out of the equation since what
is discussed is the actual capability of the medium, or rather
of 16-bit PCM code, not what is theoretically or practically
possible to achieve as input and not how perfect DACs we
can make in practice. Those are all very relevant factors, but
not what this discussion is about, and some people seem to
think those are the only relevant factors. But if the theoretical
capability of the medium is irrelevant, then we could just as
well use 1 bit and 1kHz sampling frequency, and these guys
would be happy anyway, cause their DACs measure so well.
BTW, there is nowadays people who use computers to produce
certains kinds of noise, often advertised as music, that is
entirely synthesized digitally and without any analogue source
at all. This source material could of course be made to arbitrary
precision, though I doubt even you could hear the difference
between 128 and 256 bits PCM.
fdegrove said:From what I read here I can only conclude that most people forgot how those 16 bits are achieved.
Maths are one thing, reality is another.
Now go figure it out for yourself and think hard on how that 16 bit claim came about and maybe, just maybe someone might see the light.
Deep sigh...no pun intended.
No, not forgotten at all, merely out of the equation since what
is discussed is the actual capability of the medium, or rather
of 16-bit PCM code, not what is theoretically or practically
possible to achieve as input and not how perfect DACs we
can make in practice. Those are all very relevant factors, but
not what this discussion is about, and some people seem to
think those are the only relevant factors. But if the theoretical
capability of the medium is irrelevant, then we could just as
well use 1 bit and 1kHz sampling frequency, and these guys
would be happy anyway, cause their DACs measure so well.
BTW, there is nowadays people who use computers to produce
certains kinds of noise, often advertised as music, that is
entirely synthesized digitally and without any analogue source
at all. This source material could of course be made to arbitrary
precision, though I doubt even you could hear the difference
between 128 and 256 bits PCM.
Christer said:
Noiseshaping is somewhat akin to oversampling in that it moves some of the noise specta out of the audio band.
If you've much interest in digital audio, I highly recommend picking up a copy of Pohlmann's book. It costs about $50US but it's more than worth it in my opinion. It covers everything from the basics of sampling theorem to how the various hardware actually works and puts it into one convenient volume.
se
Steve Eddy said:
OK, I'll have to think a bit about that and read up a bit on this
some day. I have no pressing need to learn about this for the
moment, I concentrate on reading up on EE stuff for the moment.
I'll try to remember your advice, though, should I want a book
about digital audio later on.
Re: Re: Shannon saves
The nat is the unit of entropy when you use natural logs rather than base-2 logs. If you're only interested in information from a theoretical point of view, it makes the math easier. Of course, if you've got a unary computer, then you'll want base-1 logs, which makes the math even easier.
Bonus question: what's the unit of entropy when you use base-10 logs?
Christer said:
The nat is the unit of entropy when you use natural logs rather than base-2 logs. If you're only interested in information from a theoretical point of view, it makes the math easier. Of course, if you've got a unary computer, then you'll want base-1 logs, which makes the math even easier.
Bonus question: what's the unit of entropy when you use base-10 logs?
Re: Re: Re: Shannon saves
Well, an educated guess this time of the night would be 10, but
that's too simple.Can't remember I have if I have read about
information theory entropy or not, I only know we read
about entropy in thermodynamics, but that's over 20 years
ago, and thermodynamics has not been a particular interest
of mine since.
Now, if we talk Kolmogorov complexity of strings of information,
then I'd know a bit about what we're discussing. It would be
interesting, actually, to know the Kolmogorov complexity of
the data on a couple of various music CDs.
Edit: I do know, though, that entropy is a measure of either
the amount of order or disorder in thermodynamics (can't
remember which) so it should be related to the amount of
structure in information, just as Kolmogorov complexity.
Rob M said:
The nat is the unit of entropy when you use natural logs rather than base-2 logs. If you're only interested in information from a theoretical point of view, it makes the math easier. Of course, if you've got a unary computer, then you'll want base-1 logs, which makes the math even easier.
Bonus question: what's the unit of entropy when you use base-10 logs?
Well, an educated guess this time of the night would be 10, but
that's too simple.Can't remember I have if I have read about
information theory entropy or not, I only know we read
about entropy in thermodynamics, but that's over 20 years
ago, and thermodynamics has not been a particular interest
of mine since.
Now, if we talk Kolmogorov complexity of strings of information,
then I'd know a bit about what we're discussing. It would be
interesting, actually, to know the Kolmogorov complexity of
the data on a couple of various music CDs.
Edit: I do know, though, that entropy is a measure of either
the amount of order or disorder in thermodynamics (can't
remember which) so it should be related to the amount of
structure in information, just as Kolmogorov complexity.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.