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

[Don Pearce]

I like your categories. It is possible in concept to have a signal that
is quantized in magnitude and continuous in time, but (unless we resort
to counting electrons) I don't think it's possible in practice.

Yes, I was thinking about that possibility while I was typing, but
since I've never come across such a system I decided it would
complicate things unnecessarily to include it.

d

 

[Jerry Avins]

Because if it could, there would be no need to invent digital which has the
advantage of non-moving parts....................

Actually, I did invent something along those lines, but I was foolish
enough yo leave the plans in my (not yet perfected) time machine, and
they disappeared.

Jerry

 

[Martin Heffels]

Actually, I did invent something along those lines, but I was foolish
enough yo leave the plans in my (not yet perfected) time machine, and
they disappeared.

Shame on you! Radium will be very disappointed now.

-m-

 

[Richard Dobson]

Well, some types of RAM bits are stored as analog voltages
on a MOS gate capacitor. I think old CCD devices could
output some measure of the voltage per bit cell. Or you
could consider the charge digital if you could count the
number of electrons in each well.

These are/were the so-called "bucket-brigade" nominally analog devices
used as delay lines for audio effects such as phasers. Based on storing
audio in a chain of capacitors (typ. NMOS, in VLSI chips). Sort of an
analogue shift register. Limitations: expensive, so delays were very
short (a few msecs) but in their heyday digital was still new and
therefore expensive too); performance - low sample rate; quality -
somewhat noisy. Electronics people loved debating whether such devices
were really digital or analog. At least: digital in concept, analog in
implementation. Suffice it to say, digital is better in all aspects.

Richard Dobson

 

[Richard Dobson]

Radium wrote:


Various compression schemes do that with varying degrees of resulting
quality.

I am talking about:

1. Decreasing the temporal frequency of the video signal without low-
pass filtering or decreasing the playback speed - an example of which
would be decreasing the rate at which a bird [in the movie] flaps its
wings. Hummingbirds flap their wings too fast for the human eye to
see. So the flap-rate of the wings could be decreased until the
flapping is visible to the human eye - without decreasing the playback
speed of the video. This decrease in flap-rate without slowing
playback is visually-analogous to decreasing the pitch of a recorded
sound without decreasing the playback speed. In this case, low-pass
filter would involve attenuating rapidly-changing images while
amplifying slowly-changing images -- I don't want this.

I confess I am jumping into a thread having just discovered it.

There are some mixed metaphors here. There is a video equivalent to
audio pitch shifting. think of the latter represetned in the frequency
domain (spectrum) - the peak correspindsing to the source partial moves
down (or up). the video equivalent is colour cycling or shifting. But
most simply, reds would be shifted to orange, green shifted to blue,
violet to ultra-violet (and hence llost to view). An alternatyive
stratgy is colour rotation using the artists colour wheel, where,
ideally, diametrically opposite colours are complementary. There is no
equivalent that I know of to colour complemenariness in audio.

I ~think~ I get what Radium wants - he wants to be able to modify a
recorded scene the way one can modify a CGI virtual scene, e.g. by
setting a slower wing flapping rate while leaving other parts of the
scene unchanged. As far as I know, computer vision and scene analysis is
nowhere near being able to do this. The only audio parallel I can think
of is wanting to pitch shift just one instrument in a polyphonic
texture, leaving other voices unchanged. With luck, some implementations
of Blind Source Separation can sometimes do this (they need the mixed
sounds to be very distinct - I have seen one example demonstrated at
DaFX); ths difficulties with video I would expect to be order of
magnitude greater.


Richard Dobson

 

[Floyd L. Davidson]

I'm curious to why there are no purely-analog devices which can
record, store, and playback electric audio signals [AC currents at

....

The net result is that an audio CCD is capable of
storing a
decent-quality signal for only a few tens or hundreds of milliseconds,
from input to output.
Another sort of a purely analog signal-storage device,
with no moving
parts other than the electrons which convey the signal, is a simple
length of transmission line (with perhaps some amplifiers mid-way).

....

Come on, Dave, a CCD is a digital device, subject to
aliasing.

CCDs are analog devices, with an analog voltage output.

The fact that they are commonly used as the sensor in
digital cameras results in the output of a CCD virtually
always going directly (well, after a bit of signal
processing for things such as white balance, ISO gain,
etc.) to an analog-to-digital converter that digitizes
the analog signal.

The charges represent the signal at a
particular instant of its average over a particular
interval. (My CCD digital camera can take time
exposures.) A CCD's content may not be quantized in
amount, but it is quantized in time. In a camera, where
the charges pertain to individual pixels, the result is
also quantized in space.

But none of that quantization changes the fact that the
device itself has an analog output.

 

[glen herrmannsfeldt]

Ron N. wrote: (snip)

These are/were the so-called "bucket-brigade" nominally analog devices
used as delay lines for audio effects such as phasers. Based on storing
audio in a chain of capacitors (typ. NMOS, in VLSI chips). Sort of an
analogue shift register.

(snip)

I believe there is a device more like an analog RAM used for
sound recording in toys. One can record up to about a
minute of voice and replay it many times.

-- glen

 

[glen herrmannsfeldt]

Dave Platt wrote:

(snip)

As I believe the term "digital" is usually meant, it implies a
two-state (on/off) storage representation. It's not just that the
signal amplitude is quantized, but that the quantization uses a
power-of-two representation and storage system of some sort.

It means discrete states, but the base does not have to be two.

Many of the early computers were decimal based, and not
necessarily BCD.

The Fortran standard still allows for any base greater
than one to be used for representing values.

-- glen

 

[Floyd L. Davidson]

That describes a binary digital system. Not all digital systems
are binary. What is called M-ary is very common, with multiple
states.

It doesn't require a power of two representation, though that
certainly makes a lot of other functionality much easier. The
key is "discrete states" from a "finite set". That makes it
digital.

My reading of the possible systems goes like this.

analogue - a continuous representation of the original signal
sampled - a representation of the signal at discrete time points

Note that discrete time points does not make a signal digital,
if the value of the signal can still be varied infinitely.

quantized - a sampled signal, but with the possible levels constrained
to a limited set of values

That is by definition a digital siganl. As soon as the possible values
are "constrained to a limited set", it is by definition digital data.

digital - a quantized signal, with the individual levels represented
by numbers

It makes no difference how the levels are represented.

Aliasing is going to happen as soon as you move beyond the first line
of that list.

Your definitions are pretty good! The significant
points are that analog is continuous with an infinite
set of values, while digital has a discrete number of
values from a finite set.

The standard definitions of analog data and digital data
(these are milspec and Federal Standard 1037C
definitions) are:

analog data:

Data represented by a physical quantity that is
considered to be continuously variable and has a
magnitude directly proportional to the data or to a
suitable function of the data.

digital data:

1. Data represented by discrete values or
conditions, as opposed to analog data.

2. Discrete representations of quantized values of
variables, e.g. , the representation of numbers by
digits, perhaps with special characters and the
"space" character.

See http://glossary.its.bldrdoc.gov/fs-1037/

 

[Floyd L. Davidson]

I like your categories. It is possible in concept to
have a signal that is quantized in magnitude and
continuous in time, but (unless we resort to counting
electrons) I don't think it's possible in practice.

If you quantize the magnitude, it is digital. That is
by definition.

 

[glen herrmannsfeldt]

Floyd L. Davidson wrote:

(snip)

In a digital channel you cannot pass frequencies higher
1/2 the Nyquist rate, which in theory is a very sharp
cutoff but in practice it becomes very similar to the
gradual analog cutoff.

If you read Nyquist's paper, that is pretty much it.

He was figuring out for fast he could send pulses
through a band limited channel and separate them out at
the other end. Electronic communication was digital
before it was analog.

-- glen

 

[Don Pearce]

If you quantize the magnitude, it is digital. That is
by definition.

No it isn't. It isn't digital until you assign numerical values to
those quantized levels. Until then it is simply a quantized analogue
signal.

d

 

[Andor]

Yes, I was thinking about that possibility while I was typing, but
since I've never come across such a system I decided it would
complicate things unnecessarily to include it.

Yannis Tsividis once asked in comp.dsp what signal processing
practitioners thought of his continuous-time signal processing
(filtering) scheme. As I remember, it didn't go down well with the
crowd. After reading a paper from him explaining the concept I thought
that the scheme had at least educational merit. There are some
references on his webpage:

http://www.ee.columbia.edu/fac-bios/tsividis/faculty.html

Regards,
Andor

 

[Floyd L. Davidson]

No it isn't. It isn't digital until you assign numerical values to
those quantized levels. Until then it is simply a quantized analogue
signal.

If you quantize it, you *have* assigned a value to it,
and that value is not from a continuous set, but from a
discrete finite set, and therefore it is digital.

A "quantized analogue signal" is digital by definition.

(Emphasis added)

QUANTIZATION:
A process in which the continuous range of values
of an analog signal is sampled and divided into
nonoverlapping (but not necessarily equal)
subranges, and *a* *discrete*, *unique* *value* *is*
*assigned* to each subrange.

http://ntia.its.bldrdoc.gov/fs-1037/

 

[Don Pearce]

If you quantize it, you *have* assigned a value to it,
and that value is not from a continuous set, but from a
discrete finite set, and therefore it is digital.

A "quantized analogue signal" is digital by definition.

No, you haven't. You merely have a signal at a set of discrete levels.
You need an analogue to digital converter to take each of those
quantized levels and convert it into a digital word (of 1s and 0s).

Digital means "represented by digits", not "in discrete voltage
steps".

d

 

[Don Stauffer in Minnesota]

Is it true that unlike the-frequency-of-audio, the-frequency-of-video
has two components -- temporal and spatial?

Kind of. This gets into some pretty involved engineering and math and
was originally used to get into how to analyze images when designers
were first trying to develop television systems. It involves what is
known as linear systems analysis, which originally was for one
dimensional signals such as audio. In this type of analysis any
arbitrary shape/waveform can be broken down into a collection of many
sine waves of different frequency. For images this was extended to
work as a two-dimensional array, with duplication of the signal by
considering two sets of so-called "spatial frequencies", at right
angles to each other.

This was extended beyond TV engineering when optical engineers
developed the Modulation Transfer Function by borrowing EE ideas of
linear systems to predict and measure performance of optical systems.
It involves things like Fourier transforms.

AFAIK, the-frequency-of-audio only has a temporal component. Do I
guess right?
Right

II. Digital vs. Analog

Sample-rate is a digital entity. In a digital audio device, the sample-
rate must be at least 2x the highest intended frequency of the digital
audio signal. What is the analog-equivalent of sample-rate? In an
analog audio device, does this equivalent need to be at least 2x the
highest intended frequency of the analog audio signal? If not, then
what is the minimum frequency that the analog-equivalent-of-sample-
rate must be in relation to the analog audio signal?

The analog equivalent is, loosely, the bandpass or cutoff frequency of
an analog filtering circuit. Any electrical network designed to
reproduce faithfully the analog signal must have a bandpass such that
the high frequency cutoff is equal to or higher than the highest
frequency in the analog signal.

 

[Doug McDonald]

Well, yes and no. That's true for what is called PCM, used
on the Compact Disc and MPEG. It is sort of true for Delta-Sigma
coding, but the for the actual useful sampling rate limit, its
not really true. D-S modulation is used for the Super Audio CD.

There is no analog-equivalent of sample-rate? Then what the limits the
highest frequency an analog audio device can encode?

The circuits used. All circuits have a low-pass filtering action
of some sort. For example, 78 RPM records went up to maybe
10-12 kHz usefully, while 33s actually could go up to 40 kHZ if
pushed (e.g. discrete quad.) Many high quality audio power amps will
happily go to 100 kHZ or even a megahertz. This may be intrinsic
with the circuits, or, far more common, a simple resistor-capacitor
filter circuit.

What determines the highest frequency signal an analog solid-state
audio device can input without distortion?

The nature of the transistors is the ultimate limit. Because
at this limit nonlinearities of a rather terrible nature occur, the
circuits they are used in usually limit the frequency with the RC
filter mentioned above, or equivalent.

Analog solid-state audio device = a purely analog electronic device
that can record, store, playback, and process audio signals without
needing any moving parts.

The above device inputs the electrical signals generated by an
attached microphone. These electric signals are AC and represent the
sound in "electronic" form. Sound with a higher-frequency will
generate a faster-alternating current than sound with a lower-
frequency. A louder sound will generate an alternating-current with a
bigger peak-to-peak wattage than a softer soft.

What mathematically determines the highest-frequency electric signal
such a device can intake without distortion?

The overall design. Such things as you describe are rare, very, very,
very rare. It's very hard to STORE signals purely analog without
moving parts. In fact, I had a hard time thinking of any such
device that is or was purely analog. However, the old analog
storage oscilloscopes would meet your criteria if you don't
include electrons in a vacuum as moving parts. There the limit to the
frequency response is the size of the focus spot .... i.e.
the quality of the lenses! (Such device of course uses analog
electron lenses). If you don't intend to store forever, there
were things like analog mercury delay lines which stored signals
as sound waves travelling through mercury.

Doug McDonald

 

[Arny Krueger]

On Aug 19, 8:54 pm, [email protected] (Dave Platt) wrote:

Thus introducing an important concept - sampled, non-digital signals.

Sampling and digitizing are somewhat independent. The necessary connection
comes when you realize that you have to sample something to digitize it.
OTOH, you don't have to digitize it when you sample it.

Is CCD a form of analog non-volatile RAM?
Yes.

Interestingly enough, CCDs are widely used for video. Reason being that
their dynamic range is as you say poor for audio, but its OK for video.

I wonder how a PC would perform if it used CCDs in place of digital
storage devices. Lots of errors.
Exactly.

Only if you have a fairly liberal idea of "decent-quality".

What is the highest frequency an audio CCD can input and output? My
guess is 0.5x the clock rate.

Well, a scosh less. Nyquist rules.

Ancient computers used quartz delay lines as storage devices. Case in point
was the IBM 2848 video display controller. There was one delay line per
attached CRTs.

Where is the "storage" in this device?

The delay line.

Where is the signal being stored?

It was stored in whatever made up the delay line. It could be a rotating
disk of magnetic material, a piece of quartz or glass, a bunch of coils and
capacitors, whatever. All of these were used up until RAM became an
economical solution.

 

[Jerry Avins]

I'm curious to why there are no purely-analog devices which can
record, store, and playback electric audio signals [AC currents at

...

The net result is that an audio CCD is capable of
storing a
decent-quality signal for only a few tens or hundreds of milliseconds,
from input to output.
Another sort of a purely analog signal-storage device,
with no moving
parts other than the electrons which convey the signal, is a simple
length of transmission line (with perhaps some amplifiers mid-way).

...

Come on, Dave, a CCD is a digital device, subject to
aliasing.

CCDs are analog devices, with an analog voltage output.

The fact that they are commonly used as the sensor in
digital cameras results in the output of a CCD virtually
always going directly (well, after a bit of signal
processing for things such as white balance, ISO gain,
etc.) to an analog-to-digital converter that digitizes
the analog signal.

The charges represent the signal at a
particular instant of its average over a particular
interval. (My CCD digital camera can take time
exposures.) A CCD's content may not be quantized in
amount, but it is quantized in time. In a camera, where
the charges pertain to individual pixels, the result is
also quantized in space.

But none of that quantization changes the fact that the
device itself has an analog output.

We agree on the facts. We disagree about how to classify borderline
cases. Is that important enough to warrant further discussion?

Jerry

 

[Jerry Avins]

Dave Platt wrote:

(snip)


It means discrete states, but the base does not have to be two.

Many of the early computers were decimal based, and not
necessarily BCD.

The Fortran standard still allows for any base greater
than one to be used for representing values.

Glenn,

I believe that's also a borderline area where definitions become
smudged. I know that the Russians built a computer with trinary logic,
but all the decimal systems I know, whether BCD, excess-three, or
something more exotic, encode the numbers on sets of four wires that
carry two-state signals. Making a case that that isn't binary opens the
door to claiming that hexadecimal is distinct from binary.

Jerry