Digital, or analog?

[Jerry Avins]

It seems to me important to agree on criteria for deciding whether a
particular circuit or signal is digital or analog

One criterion is intended use; there seems to be general agreement about
that, so I don't address it here. Another criterion is the nature if the
signal or circuit itself, without reference to intentions. That is the
topic of this short essay.

*SIGNALS*
A digital signal consists of a sequence of a fixed number of discrete
states, with no intermediate states are allowed. Outside the realm of
quantum mechanics, a continuous signal is not digital. It can be made
digital by quantizing it, and different quantizers will convert the same
analog signal into different quantized signals. As far as I know,
digital signals exist only as abstractions and in computer circuits. I
would be delighted to learn of exceptions.

*CIRCUITS*
A circuit intended for a digital application has a fixed number of
discrete input and output states. The useful states are limited to those
common to both input and output. For simplicity of design, the number of
states is usually chosen to be 2, although other arrangements are
possible and some have been utilized. States are represented at outputs
as voltage or current ranges, and recognized at inputs in the same way.
The thresholds need not be the same at input or output. For example, the
specification for the 74LS logic family requires a high-level (1) to
equal or exceed 2 volts and a low level (0) to be no more than .8 volts
at the input, and guarantees that a 1 will be at least 2.4 volts and a 0
no more than .5 volts at the output. That specification makes the
devices well suited for digital use, but it defines them as analog
devices by defining voltages which are, as far as states go, ambiguous.

A CMOS CD4011B makes that point better. It is a quad 2-input NAND gate.
With a 15-volt supply, a 1 is 11 volts or greater, and a 0 is 4 volts or
less. The output swings between .05 and 14.95 volts when lightly loaded,
and can sink or source more than 5 ma at 4 and 11 volts. The /intended/
use of a NAND gate makes it digital devices, but these are inherently
analog by construction. Their inherent analog nature seen by connecting
a 1 megohm resistor from output to the inputs tied together. You will
have an analog amplifier with 23dB voltage gain and much higher power
gain. With capacitive coupling, the four gates can be connected as two H
bridges that will deliver 18 milliwatts per channel into 1500 ohms. To
be truly digital, a device must have discrete states, and be incapable
of exhibiting any other state.

That situation is approximated by cross coupling a pair of the gates to
make a set-reset flip-flop. Now there are only two stable states, but
the in-between states still exist as transient states passed through
when the device is in transition. Rise and fall times greater than zero,
and the well known but often ignored metastable state attest to that.
http://www.interfacebus.com/Design_MetaStable.html As far as I know,
digital circuits exist only as abstractions and on schematics. I would
be delighted to learn of exceptions.

Jerry

 

[Eric Jacobsen]

It seems to me important to agree on criteria for deciding whether a
particular circuit or signal is digital or analog

One criterion is intended use; there seems to be general agreement about
that, so I don't address it here. Another criterion is the nature if the
signal or circuit itself, without reference to intentions. That is the
topic of this short essay.

*SIGNALS*
A digital signal consists of a sequence of a fixed number of discrete
states, with no intermediate states are allowed. Outside the realm of
quantum mechanics, a continuous signal is not digital. It can be made
digital by quantizing it, and different quantizers will convert the same
analog signal into different quantized signals. As far as I know,
digital signals exist only as abstractions and in computer circuits. I
would be delighted to learn of exceptions.

*CIRCUITS*
A circuit intended for a digital application has a fixed number of
discrete input and output states. The useful states are limited to those
common to both input and output. For simplicity of design, the number of
states is usually chosen to be 2, although other arrangements are
possible and some have been utilized. States are represented at outputs
as voltage or current ranges, and recognized at inputs in the same way.
The thresholds need not be the same at input or output. For example, the
specification for the 74LS logic family requires a high-level (1) to
equal or exceed 2 volts and a low level (0) to be no more than .8 volts
at the input, and guarantees that a 1 will be at least 2.4 volts and a 0
no more than .5 volts at the output. That specification makes the
devices well suited for digital use, but it defines them as analog
devices by defining voltages which are, as far as states go, ambiguous.

A CMOS CD4011B makes that point better. It is a quad 2-input NAND gate.
With a 15-volt supply, a 1 is 11 volts or greater, and a 0 is 4 volts or
less. The output swings between .05 and 14.95 volts when lightly loaded,
and can sink or source more than 5 ma at 4 and 11 volts. The /intended/
use of a NAND gate makes it digital devices, but these are inherently
analog by construction. Their inherent analog nature seen by connecting
a 1 megohm resistor from output to the inputs tied together. You will
have an analog amplifier with 23dB voltage gain and much higher power
gain. With capacitive coupling, the four gates can be connected as two H
bridges that will deliver 18 milliwatts per channel into 1500 ohms. To
be truly digital, a device must have discrete states, and be incapable
of exhibiting any other state.

That situation is approximated by cross coupling a pair of the gates to
make a set-reset flip-flop. Now there are only two stable states, but
the in-between states still exist as transient states passed through
when the device is in transition. Rise and fall times greater than zero,
and the well known but often ignored metastable state attest to that.
http://www.interfacebus.com/Design_MetaStable.html As far as I know,
digital circuits exist only as abstractions and on schematics. I would
be delighted to learn of exceptions.

Jerry

Jerry, that's good work on a rational foundation for the terms.

I think, however, that the significant quibbling comes in only on
boundary cases where the distinction gets blurred, and it's just
always going to be tough to avoid semantic arguments popping up there.
For example, even with "digital" devices, by your definitions, at both
ends of a circuit board trace, the trace may still need carefully
designed termination and interference isolation which are often best
treated as "analog" phenomena.

Although your definitions generally hold, there's still plenty of room
for argument and confusion at the boundaries. Because of that, I
don't think the problem of whether certain things are really "digital"
or "analog" is going to be solved with world-class
unanimously-approved definitions.

Eric Jacobsen
Minister of Algorithms, Intel Corp.
My opinions may not be Intel's opinions.
http://www.ericjacobsen.org

 

[Ron N.]

It seems to me important to agree on criteria for deciding whether a
particular circuit or signal is digital or analog ....
As far as I know, digital circuits exist only as abstractions ...

I remember an old-time engineer once telling me "There
is no such thing as digital". We merely interpret certain
behaviors of analog circuits as transitions, noise, or
metastability, etc. and then attempt to ignore those for
our convenience, or at our own peril.

Of course, he was talking about the real world. We are
free to make abstract models, but should not confuse
a model with some engineering reality, especially if
a circuit is working near its limits of behaving similar
to our first-order models.

So I would agree with you that digital circuits exist only
as abstractions.

However, there is also no such thing as analog. It is
merely a model where we assume that some continuous
curve or function is associated with our measurements
of what are actually numbers discrete quantum events
(the measurements are usually far too course to
notice the error).

If the quantum events are below some noise floor and
you are already ignoring this noise floor, then a continuous
model might be the computationally or cognitively more
efficient tool of abstraction.

So, as to the question about whether some circuit or
signal is digital or analog, I would say the answer
depends on which model best serves your specific
purpose or question regarding the given circuit or signal.

For some things DSP engineers do, the answer might
well be both.


IMHO. YMMV.

 

[Adrian Spilca]

It seems to me important to agree on criteria for deciding whether a
particular circuit or signal is digital or analog

One criterion is intended use; there seems to be general agreement about
that, so I don't address it here. Another criterion is the nature if the
signal or circuit itself, without reference to intentions. That is the
topic of this short essay.

*SIGNALS*
A digital signal consists of a sequence of a fixed number of discrete
states, with no intermediate states are allowed. Outside the realm of
quantum mechanics, a continuous signal is not digital. It can be made
digital by quantizing it, and different quantizers will convert the same
analog signal into different quantized signals. As far as I know,
digital signals exist only as abstractions and in computer circuits. I
would be delighted to learn of exceptions.

*CIRCUITS*
A circuit intended for a digital application has a fixed number of
discrete input and output states. The useful states are limited to those
common to both input and output. For simplicity of design, the number of
states is usually chosen to be 2, although other arrangements are
possible and some have been utilized. States are represented at outputs
as voltage or current ranges, and recognized at inputs in the same way.
The thresholds need not be the same at input or output. For example, the
specification for the 74LS logic family requires a high-level (1) to
equal or exceed 2 volts and a low level (0) to be no more than .8 volts
at the input, and guarantees that a 1 will be at least 2.4 volts and a 0
no more than .5 volts at the output. That specification makes the
devices well suited for digital use, but it defines them as analog
devices by defining voltages which are, as far as states go, ambiguous.

A CMOS CD4011B makes that point better. It is a quad 2-input NAND gate.
With a 15-volt supply, a 1 is 11 volts or greater, and a 0 is 4 volts or
less. The output swings between .05 and 14.95 volts when lightly loaded,
and can sink or source more than 5 ma at 4 and 11 volts. The /intended/
use of a NAND gate makes it digital devices, but these are inherently
analog by construction. Their inherent analog nature seen by connecting
a 1 megohm resistor from output to the inputs tied together. You will
have an analog amplifier with 23dB voltage gain and much higher power
gain. With capacitive coupling, the four gates can be connected as two H
bridges that will deliver 18 milliwatts per channel into 1500 ohms. To
be truly digital, a device must have discrete states, and be incapable
of exhibiting any other state.

That situation is approximated by cross coupling a pair of the gates to
make a set-reset flip-flop. Now there are only two stable states, but
the in-between states still exist as transient states passed through
when the device is in transition. Rise and fall times greater than zero,
and the well known but often ignored metastable state attest to that.
http://www.interfacebus.com/Design_MetaStable.html As far as I know,
digital circuits exist only as abstractions and on schematics. I would
be delighted to learn of exceptions.

Jerry

Yes, our world is analog, isn't it?
I agree that the digital world is an abstraction, therefore not real.

But what is our world or what is real? (keeping the technological
perspective rather than philosophical). Isn't it because our sensors
integrate incoming signals so that they appear continuous? (I prefer
continuous rather than analog, as opposed to digital). Because our world,
the real world, is what we perceive using our natural sensors. But this is
the macroscopic world, in the microscopic one (as you mentioned quantum
mechanics) it could be the other way around. The continuous signals might
be the abstraction.

Adrian

 

[Jerry Avins]

[snip much]
As far as I know,
digital circuits exist only as abstractions and on schematics. I would
be delighted to learn of exceptions.

electo-mechanical relays

'nuff said

Better say more. Make before break, or break before make? We're looking
at transitions here. Even with SPST, contacts bounce

Jerry

 

[Ron N.]

[snip much]
As far as I know,
digital circuits exist only as abstractions and on schematics. I would
be delighted to learn of exceptions.

electo-mechanical relays

'nuff said

The first time I tried to make a toggle flip flop out of
relays (in junior high, I think), it went into metastable
oscillation, dependent on the power supply voltage.

Another DPDT relay (there were scavenged discards)
had one bad (randomly high resistance) contact.

Try again.


IMHO. YMMV.

 

[Ron N.]

The contacts have *ONLY* 2 states.

No matter how much bounce, the contacts are *EITHER* "open" or "closed"

Said by somebody who's never put a VOM across a batch
of old rusted relays (scavenged from dead pinball machine
parts found in a outdoor scrap heap I think).

Your abstraction does work a bit better with new ones.


IMHO. YMMV.

 

[Jerry Avins]

Jerry Avins wrote:

[snip much]

As far as I know,

digital circuits exist only as abstractions and on schematics. I
would be delighted to learn of exceptions.


electo-mechanical relays

'nuff said

Better say more. Make before break, or break before make? We're
looking at transitions here. Even with SPST, contacts bounce

Jerry

The contacts have *ONLY* 2 states.

No matter how much bounce, the contacts are *EITHER* "open" or "closed"

ain't no other option!

Arcing contacts? Zero risetime?

Jerry

 

[Ron N.]

irregardless of "contact resistance" they are *either* open/closed

So what resistance value, and for what duration, would you
call closed?

Do you think that all "open" relays have zero resistance?


IMHO. YMMV.

 

[Floyd L. Davidson]

[snip much]
As far as I know,
digital circuits exist only as abstractions and on
schematics. I would be delighted to learn of exceptions.

electo-mechanical relays

'nuff said

I don't think it is possible to say enough to convince some
folks.

I'd say a light switch is another excellent example of something
clearly digital. That can be compared with a large variable
resistor (a "light dimmer") which is analog.

With a switch, the light is either on or off. With the dimmer,
the light can be set half way between any two other points.

One is discrete, the other is continuous. Digital, and analog.

 

[Ron N.]

"closed" is CLOSED "open" is OPEN


NO *OPEN* relays have INFINITE resistance

That's what I get for looking at the wrong pin of a DPDT
relay. :^)

actually i'm not cheating fair as i knew what you meant

My previous experience was in industry that wished to "KNOW" within 1 ms
when a contact closed with 10's to 100's of ms bounce

In my case, it was what voltage, frequency and duration
of an oscillating relay pair would close the next (rusty)
relay downstream. There was also the amplitude of
whacking the side of the breadboard. Analog.

 

[Jerry Avins]

BULL!
irregardless of "contact resistance" they are *either* open/closed

You could say the same about a door, but degree matters, especially to
the obese.

Jerry

 

[Ron N.]

Richard Owlett wrote:

Arcing contacts? Zero risetime?

I can't remember where, but I seem to recall some
ultra high-speed photography of the formation of an
arc. It didn't look like it was happening instantly.

Or does Richard count that first electron at "closed"?


IMHO. YMMV.

 

[Bob Myers]

It seems to me important to agree on criteria for deciding whether a
particular circuit or signal is digital or analog

Circuits and signals themselves are neither digital nor
analog, even though we unfortunately do tend to classify
them as such.

The terms "digital" and "analog" properly refer only to two
different means of encoding information, in this context
on to an electrical signal. The signal itself remains just
electricity no matter what we call the encoding, and all
signals and circuits obey the same basic physical laws, etc..
The true distinction between these two arises SOLELY in
how we interpret the signal (or similarly, how it was
intended to be interpreted).

In an "analog" encoding, some parameter of the electrical
signal (generally, either voltage or current, but there are
other possibilities) is manipulated such that it varies in a
manner directly related to the variations seen in the original
information source - for instance, a voltage which is caused
to vary in the same manner as a sound wave, in the case of
an "analog audio signal." And hence the name itself - the
voltage is varying ANALOGOUSLY to the original, hence
"analog" encoding.

Similarly, in a "digital" information encoding or transmission
system, various states of the transmitted signal correspond
to numeric values - or more generally, symbols - and must
be interpreted accordingly. Again, the name says it all -
we're not sending something that directly represents another
thing, but instead are sending symbols or "digits."

Neither term NECESSARILY implies a lot of what are
commonly thought of as the distinguishing features of
either; for instance, "analog" does not necessarily imply
that the system is either continuous or linear, even though
many common analog systems are both. Similarly, "digital"
does not necessarily imply a discrete or sampled representation,
and certainly is not limited to a straight binary encoding -
although again the vast majority of "digital" systems exhibit
these characteristics.

From this perspective, arguments as to whether the world
itself, or basic natural phenomena, etc., are "digital" or
"analog" are meaningless; the world is what the world is,
and these terms only refer to methods for encoding
information which describe some real, original thing.

We still refer to circuits themselves as "digital" or "analog"
primarily because of the optimization of that class of circuits
for dealing with that sort of information, but again the ciruits
themselves ALL behave according to the same laws of
physics. And there are certainly types of circuits which
don't fall into either category - power systems being the
most obvious example. (The transmission of power does
not involve the transmission of information, so to speak of
power engineering as dealing with either "analog" or
"digital" is just silly.)

Bob M.

 

[Floyd L. Davidson]

You could say the same about a door, but degree matters,
especially to the obese.

Then obesity is clearly analog. But the described switch contacts
are not...

 

[Floyd L. Davidson]

Neither term NECESSARILY implies a lot of what are
commonly thought of as the distinguishing features of
either; for instance, "analog" does not necessarily imply
that the system is either continuous or linear, even though
many common analog systems are both.

By definition, it implys that it is continous. Linear, no, but
continuous is essential.

Similarly, "digital"
does not necessarily imply a discrete

Discrete, yes, again by definition.

or sampled representation,
and certainly is not limited to a straight binary encoding -
although again the vast majority of "digital" systems exhibit
these characteristics.

I don't believe that necessarily most are binary nor sampled.

From this perspective, arguments as to whether the world

From this perspective, if you want to make up your own
definitions, and so does everyone else, what point is there to
discussing *anything*?

You are denying that the standardized definitions are correct,
and that is an absurd stance to take, which makes *nothing* you
say worth discussing.

 

[Ron N.]

You are denying that the standardized definitions are correct,
and that is an absurd stance to take,

I disagree. Many standardized definitions are committee
compromises, and have sometimes what comes out of a
committee has little to do with what real people try to mean
or understand when using those words (especially if a
committee has gotten politicized or influenced by marketing
agendas).

(in general. not expressing an opinion on digital/analog).
IMHO. YMMV.

 

[Floyd L. Davidson]

I disagree. Many standardized definitions are committee
compromises,

I was not referencing just *any* definitions. I say *those*
definitions. They are not in dispute by any reputable source.

You stake *your* reputation against the entire industry when
you deny they are correct, and that is simply absurd!

and have sometimes what comes out of a
committee has little to do with what real people try to mean
or understand when using those words (especially if a
committee has gotten politicized or influenced by marketing
agendas).

(in general. not expressing an opinion on digital/analog).
IMHO. YMMV.

So we are going to ignore indisputably good standardized
definitions that the industry (not to mention the Federal
government) uses, and you think that is going to in any way
assist in discussing this topic???

 

[Floyd L. Davidson]

I've seen a 7400 series inverter used as an 'op amp'
Don't recall why it was done, suspect was just to prove it could be ;/

Of course to make it work as an analog circuit... components had
to be added. Which means that was *not* the same "circuit"
which functions as a digital circuit, even if it used the same
IC.

Circuits may or may not be analog or digital, though *clearly*
some are one or the other. Signals *are* one or the other, by
definition.

 

[John E. Hadstate]

This is a general prerequisite for any kind of
communication, digital or analog ;-)

Circuits and signals themselves are neither digital nor
analog, even though we unfortunately do tend to classify
them as such.

That is demonstrably incorrect. All circuits that have a
physical instantiation are analog, as are the signals that
they process.

[snip fairly lucid description of analog encoding]

Similarly, in a "digital" information encoding or
transmission
system, various states of the transmitted signal
correspond
to numeric values - or more generally, symbols - and must
be interpreted accordingly. Again, the name says it all -
we're not sending something that directly represents
another
thing, but instead are sending symbols or "digits."

In and of itself, a digital state has no representation in
the physical world. To instantiate or process a digital
state, it must be encoded into a voltage/current analog.
People have been known to group binary states into a
multi-state logic and represent the combined states with a
single value of voltage/current analog. If you group enough
binary states together the analog of the digital value
becomes indistinguishable from continuous.