Questions about equivalents of audio/video and digital/analog.

[Scott Seidman]

Really? Then I shouldn't be able to find any standards
organizations which use a conflicting definition, right?

The nicest thing about standards is that there are so many from which to
choose.

 

[Floyd L. Davidson]

Nonsense. You don't believe it is possible to sample
an "analog" signal and have it remain analog?

I didn't say that an analog signal is always present. I
said the value representation is by definition
continuous.

If you cannot understand these simple things without it
being spelled out in detail what they mean...

Well, THAT certainly makes it clearer....

And of course you had no idea until now that that is
what we've always been talking about, right?

The fact that this sort of trivial detail has to be
explained to you certainly indicates just how uninformed
you are on this topic. You should not be arguing with
anyone about it, but asking questions and learning.

Since you seem to be so hung up on definitions, Floyd,
try this one on for size:

Continuous: unchanged or uninterrupted: continuing
without changing, stopping, or being interrupted in space
or time.

Now apply that to the value of an analog signal, and you
too can have a good understanding of the definitions
used to distinguish between analog and digital.

Please note that a non-existent signal cannot be either
analog or digital. Hence when you try to weasel out of
valid definitions for a signal by claiming that pulsed
samples are not continuous, it might seem cute to you,
but it is trivially childish.

The definition does not say that the signal's
*existence* is continuous, it says that the *value* of
the signal is continuous for analog while digital has a
finite set of discrete values.

Note that this does not say anything at all about the range
of possible values being "continuous" (which is what you
seem to be trying to say in the above).

True, it doesn't say anything at all about the
definitions of digital and analog, so one wonders why
you would bring up such a childish and trivial attempt
at obfuscation. Other than the obvious fact that it has
finally dawned on you that the definitions provided are
in fact rigorously correct...

 

[Floyd L. Davidson]

I aced "Signals And Systems Analysis" and placed high in "Linear Stochastic
Optimal Control" - probably the heaviest signals trips I ever took.

Then you *should* know that you cannot look at signals
on a scope and tell if they are analog or digital.

How do you explain the conflict between what you claim
your education is and the practical level your previous
post demonstrated?

 

[Floyd L. Davidson]

Really?

Do you need the definitions repeated? If the values are
discrete and from a finite set, that is defined as digital.

Crawl into a corner and deny standard definition of terms
if you like, but it means that nothing you say about the
topic is credible.

Suppose I show you an oscilloscope screen which
is displaying a single line of video, which happens to be
carrying an 8-level gray-scale pattern. It clearly shows
a set of discrete levels.

It shows no such thing. You cannot tell from looking at
it how many levels it could possibly have. You can't
actually tell (just from looking at it) how many levels
it has at any given instant.

The scope simply does not tell you that sort of
information. It only shows you what currently exists,
and over a period of time you can (perhaps) get an idea
of at least some of the possible range of variations.
But you *cannot* tell if the variations are or are not
discrete or continuous.

Only if you already know absolutely what the format is,
can you determine what part of the format you are
seeing.

Further, since this video happened
to be created by a D/A converter with only three bits at
the input (our video generator was built on the cheap!), those
are the ONLY levels this signal may exhibit. Is this a
"digital" signal?

You have said nothing that makes it necessary one or the
other.

Ah, Floyd - argument from authority again, huh?

Ah, Bob... that is very clearly a *valid* appeal to
authority. The quotes are indeed from an expert,
virtually *all* experts agree with what that
particular one said, and the quotes were in context and
meant to be what I claim they are. Look it up in any
reference work on logic, and you'll find that is
precisely what makes such an argument valid.

You, on the other hand, have a totally worthless
opinion, with nothing at all to support it. And that's
why you continue to make such gross errors!

 

[Floyd L. Davidson]

Well it shows what was once a set of discrete levels. Since it is now in the
analog domain, there will be rise time, overshoot, tilt, simple inaccuracy,
and etc. .

That has no relevance to whether it represents a discrete set
of values.

It's an analog signal that represents something that was once quantized.

It could also be a digital signal. You simply cannot
tell from looking at a scope.

The definitions are fine, it is the misapplication of them that sticks.

Oh, are we back to the idea that the NTIA had never heard of
PCM when they came up with those definitions?

In fact, *you* are not able to apply them, as is obvious from
what you said above.

 

[Piergiorgio Sartor]

Sheesh! That *is*, by definition a digital signal.

Maybe you should go back to look at the root of the word "analogue".
It is the same as "analogy".
And, usually, it means that a physical entity is represented (by
analogy) by an other physical entity.
For example, a mercury thermometer uses the "length", a physical
entity, to represent the "temperature", an other physical entity.
The analogy is "inches" ("millimeters") to "degrees".
A pressure gauge uses "angle degrees" as analogy for "pascal".
So it does an analog volt meter.

On the other side we have digital, where a number is used to
represent a physical entity. And this is a just plain number, so
the signal represented has no physical energy, only statistical.
A digital thermometer shows directly the temperature as a number.

All this has nothing to do with sampled, discreet, quantized,
continuous an so on.

Sometimes the analogy is one to one (length for length, for example),
but the concept is the same.

bye,

 

[Bob Myers]

Apparenlty Claude Shannon didn't agree with you on that.

And apparently you don't understand what was meant in
the above.

It is theoretically impossible for any real-world communications
channel to be noise-free or possessed of infinite bandwidth.
Do you disagree with this statement? If so, please show the
error.

This does not prevent a noise-free channel from being IMAGINED,
and used as the basis for a mathematical analysis, which is
what Shannon did. But Shannon most definitely did NOT
expect any such thing to be realized, and fully understood
why it could not be.

Have you even read Shannon's paper? In section V (27),
Shannon makes virtually the same statement I gave earlier
re the notion of "infinite" levels:

"This means that to transmit the output of a continuous
source with *exact recovery* [emphasis Shannon's]
at the receiving point requires, in general, a channel of
infinite capacity (in bits per second). Since ordinarily
channels have a certain amount of noise, and therefore
a finite capacity, exact transmission is impossible."

Gee, here's another puzzler for you - throughout his
paper, Shannon discusses channel capacity in terms of
"bits per second." Does this mean that his work is
applicable only to digital systems? If not, why not?

Bob M.

 

[Bob Myers]

That doesn't make it quantised.

To borrow a notion from our mutual friend Floyd,
who appears to be infallible in terms of definitions
and such, that is the very definition of quantized!

Whether or not a signal is "digital" depends upon what you intend to do
with it. A PCM signal conveying the digits 01010101... is digital; a
square wave generated within an audio synthesiser isn't, even if the
signals are identical.

Precisely! It is purely a matter of intent at the transmitting
end, of how the information conveyed by the signal in
question is SUPPOSED to be interpreted. The two signals
described above - actually, this could be considered as one
signal, since the two ARE "identical" - carry vastly different
meanings, depending solely on the intent of the transmitter.

Bob M.

 

[Bob Myers]

That's what I was talking about. Reducing the temporal frequency of
the video w/out low-pass filtering or increasing the length of the
movie.

And THAT is simply using a lower frame rate in the first
place. No "filtering" involved, per se, but it can have some
undesired results in terms of the portrayal of motion, etc..

Bob M.

 

[Floyd L. Davidson]

Nope, the phase shift is appreciable down to Fc/10. Ever hear of Bode plots?

"Appreciable" is not specific... The point is that in
practical applications there are few effects at Fc/2
which are unreasonable. Measurable? Sure, but so what.

For example... A typical voice channel provided by the
telecommunications industry is often stated as a 4Khz
wide channel. In fact of course it is not. The lower
limit is determined by high pass filters designed to
reduce 60Hz power line interference. Cutoff filters are
usually at about 80 Hz for that. The high end is
limited in digital systems by the need to avoid aliasing
of frequencies above 4000Hz, so the LP filters generally
have a cutoff between 3750 and 3850. Essentially that
provides a 80-3750Hz channel, maximu.

The industry only guarantees that they will provide
400-2800 Hz of bandwith over the Public Switched
Telephone Network. Much of the reason for that is
that phase shifts and envelop delay at frequencies
from 3000 to 3750 are sometimes (but not always) high
enough to be a problem (which would require special
conditioning to correct for).

You are saying the effects are "appreciable" down to
375 Hz, which means the _entire_ 400-2800Hz bandwidth
is suffers "appreciable" effects.

It can be measured, but it does *not* cause appreciable
effects.

Most digital audio is brick walled at 0.95 * Fs/2 or about 21 KHz. Most
power amps are pretty flat up to about 50 KHz.

So, as I've stated... your example shows that the
filters are placed at 0.95 times the maximum possible
frequency that can traverse the channel. And those
filters are on the *input* to the channel.

The fact that the power amp, the last part active of the
system, actually has a greater bandwidth (even then,
only 2x), is insignificant. What about the bandwidth of
the output stages of the D to A converter... That is
where it makes the most difference in regard to
conversion of digital to quasi-analog to analog without
appreciable artifacts. But secondarily, the speakers
are also an important part of the overall channel... and
ultimately the point at which a digital signal is
absolutely converted to analog (perhaps artifacts and
all).

For people who don't already know?

You said it was simple. Apparently you should read up
on the topic.

Well, by the brick wall filter which is usually set as stated above.

The brick wall filtering is on the channel *input*, to
prevent aliasing. It does not limit what could show up
on the output, as if it were not there the frequencies
that it removes would be folded at the output anyway.
The absolute upper limit (at which point it generates
100 percent distortion) is Fs/2, for input bandwidth.

DACs can't alias. Only ADCs can alias. Improperly filtered DACs may produce
images.

Okay... if you want to limit this to separating the ADC
from the DAC, the DAC doesn't have the brick wall
filter, the ADC does.

How do you get an image out of a DAC, if it wasn't
produced at the ADC?

Improperly filtered channel output might have artifacts
from the sampling though, but no HP output filtering is
necessary to remove any signal resulting from the input
signal.

The input side of a digital system is called an ADC, not a DAC as you just
said.

They are both part of a "digital channel", and *that* is
what I am talking about. Nothing else makes sense in
this context.

 

[Don Pearce]

Claude Shannon proved it mathematically in the late
1940's.

You are free to dispute Shannon...

I don't dispute Shannon and he didn't write the above. Show me how
sampling at a rate equal to twice the highest frequency works. It is
there quite explicitly. Could you not even spot that? The entire
clause "must be equal to" is incorrect.

But I forget, this is authoritative so it must be right. Shame Shannon
got it wrong all those years ago, isn't it?

d

 

[Bob Myers]

Actually I don't want other parts to be unchanged. What I would like
is the temporal frequencies [of all parts of the video] to be
decreased but without decreasing the speed of the video signal.

And here's where you need to be clearer, and very likely
do some more thinking about what you're after - exactly
what do you mean by "speed" of the video signal?

Bob M.

 

[Arny Krueger]

Floating point is analog, integer is digital.

?????????????????????

 

[Don Pearce]

Phase may not affect your ears much, but it does affect
the data. It has nothing to do with flat tops and
vertical edges, which seem to be something you can't get
around.


My point is that there is a lot more to a practical
amplifier than just simple gain. Sorry that you missed
the point, again.

No. You said that a signal that goes into an amplifier digital will
come out analogue. Now justify that piece of shit or go away.

d

 

[Arny Krueger]

True. Theory predicts that you cannot look at a scope
and tell what kind of information, digital or analog, is
carried by a signal. So one wonders why you want to talk
about scopes.

Not true. Theory predicts the scope pattern that various kinds of signals
will produce.

Nobody made a "decent" scope for several years after I
was born.

Yeah, sure.

Bringing up the distraction of looking at such signals
with a scope clearly indicates that you do not
understand it. You *cannot* distinguish between digital
and analog signals with a scope.

As a rule, I can. Been doing it for about half a century, more or less.

(See below, for a very good example of why that is true.)

It *commonly* does. Every time anyone fires up a v.90
modem and downloads *anything*, the waveform on the
signal from the telco to the subscriber's v.90 modem
looks like sine waves.

You've picked a situation where *everything* including no data at all looks
like sine waves. What you see on a modem line is not the actual data, you
see that data modulating sine waves.

That is true whether it is
imaging data, text, voice, or whatever else you can
think of.

So what?

But in fact those "sine waves" are as much as 8000 bytes
per second digital signaling.

So what?

And if you put your scope on the line and look at them,
and then hang up the phone and make a voice call, you
will not be able to see *any* difference between the
analog voice signal and the v.90 protocol digital
signal!

So what?

The reason is because there absolutely is no difference
at all. Both signals are generated in exactly the same
way by the exact same CODEC in the line card at the
telco. They necessarily will look identical on a scope.

So what?

 

[Arny Krueger]

At this point I need to say "who cares how it was generated?" It sure
won't
need to be passed through any analog channel to make it analog. It is
analog by the time it gets off the board that generating it.

At some level every signal is analog. The digital nature of a signal comes
from how it is coded and decoded.

 

[Bob Myers]

And of course you had no idea until now that that is
what we've always been talking about, right?

This is just priceless. Floyd, you really need to have
your Sarcasm Detector looked at because, son, it
ain't a-workin' at all...

Bob M.

 

[Bob Myers]

The nicest thing about standards is that there are so many from which to
choose.

Amen! Finally someone who gives some evidence of
actually having "been there, done that" as opposed to
just looking the answer up in the Holy Technical
Scriptures...

Bob M.

 

[Arny Krueger]

"Appreciable" is not specific...

Thanks for admitting that you have no idea about Bode plots.

The point is that in
practical applications there are few effects at Fc/2
which are unreasonable.

Depends what Fc is. If you are sampling audio at 5 KHz, there are any
number of audible effects at Fc/2 that are pretty unreasonable. One is the
total absence of any signal for much of the audio band.

For example... A typical voice channel provided by the
telecommunications industry is often stated as a 4Khz
wide channel. In fact of course it is not. The lower
limit is determined by high pass filters designed to
reduce 60Hz power line interference. Cutoff filters are
usually at about 80 Hz for that. The high end is
limited in digital systems by the need to avoid aliasing
of frequencies above 4000Hz, so the LP filters generally
have a cutoff between 3750 and 3850. Essentially that
provides a 80-3750Hz channel, maximu.

The industry only guarantees that they will provide
400-2800 Hz of bandwith over the Public Switched
Telephone Network. Much of the reason for that is
that phase shifts and envelop delay at frequencies
from 3000 to 3750 are sometimes (but not always) high
enough to be a problem (which would require special
conditioning to correct for).

You are saying the effects are "appreciable" down to
375 Hz, which means the _entire_ 400-2800Hz bandwidth
is suffers "appreciable" effects.

Voice channels like the one you have described. sound very much like they
are suffering from appreciable effects. They are not hi fi. Their fidelity
is so low that we start worrying about intelligibiilty.

It can be measured, but it does *not* cause appreciable
effects.

No? It imparts a very characteristic sound quality to normal voice! It is
often called "Telephone-like".

So, as I've stated... your example shows that the
filters are placed at 0.95 times the maximum possible
frequency that can traverse the channel. And those
filters are on the *input* to the channel.

The fact that the power amp, the last part active of the
system, actually has a greater bandwidth (even then,
only 2x), is insignificant.

So what?

What about the bandwidth of
the output stages of the D to A converter...

Usually the output buffers of a D/A have very wide bandwidth. 100s of KHz or
more.

That is
where it makes the most difference in regard to
conversion of digital to quasi-analog to analog without
appreciable artifacts.
Wrong.

But secondarily, the speakers
are also an important part of the overall channel... and
ultimately the point at which a digital signal is
absolutely converted to analog (perhaps artifacts and all).

Wrong again. The input to typical speakers is analog, pure and simple.

You said it was simple. Apparently you should read up
on the topic.

You're talking trash.

The brick wall filtering is on the channel *input*, to
prevent aliasing. It does not limit what could show up
on the output, as if it were not there the frequencies
that it removes would be folded at the output anyway.
The absolute upper limit (at which point it generates
100 percent distortion) is Fs/2, for input bandwidth.

Okay... if you want to limit this to separating the ADC
from the DAC, the DAC doesn't have the brick wall
filter, the ADC does.

You quote me without proper attribution, presumably to avoid admitting your
rather gross errors?

How do you get an image out of a DAC, if it wasn't
produced at the ADC?

The image is produced by the convolution of the signal with the clock. If
you read the samples out of digital storage or calculate them with a
computer they never have to come anyplace near an ADC. However, they will
still produce images.

Improperly filtered channel output might have artifacts
from the sampling though, but no HP output filtering is
necessary to remove any signal resulting from the input
signal.
Semi-gibberish.

They are both part of a "digital channel", and *that* is
what I am talking about.

You don't seem to be able to distinguish the start from the finish.

Nothing else makes sense in this context.

Very little of what you say makes sense, unless you quote or paraphrase
someone who knows what they are talking about.

 

[Arny Krueger]

That has no relevance to whether it represents a discrete set
of values.


It could also be a digital signal. You simply cannot
tell from looking at a scope.


Oh, are we back to the idea that the NTIA had never heard of
PCM when they came up with those definitions?

This response is irrelevant to what I said.

I get it. You can't read.