At what point does your circuit become "digital"?

[Beamin]

Vague question but as I learn more I'm wondering when I have crossed this boundary. If you are using an Arduino that's digital with analog passives. But what about a 555 timer circuit? Still analog because you could take the design of the chip duplicate its circuit and make it into all basic analog parts? What about a counter with a 555 and a 4017 counter? Is that digital or hybrid device? IF you are doing electronics and all you care about is getting a result like meeting a  design specification and have to option of making an analog circuit or a digital one isn't it simpler to just make digital circuits and only make analog devices when you have to? Is a switch mode power supply or a class D audio amp considered digital hardware?

Does the definition lie when you are using gates? What about op amps they act like gates don't they? I hvent been doing this hobby for a few years and I have forgot a lot that just use to come to me instantly.

[ataradov]

The easy answer - when you can ignore analog nature of the device.

Even with pure MCU projects with "digital" pins, there is still plenty of "analog" stuff to worry about for high-speed interfaces.

Try to use DDR SDRAM and not do proper impedance matching, and you will quickly learn just how much analog there is in digital stuff.

[vk6zgo]

Vague question but as I learn more I'm wondering when I have crossed this boundary. If you are using an Arduino that's digital with analog passives. But what about a 555 timer circuit? Still analog because you could take the design of the chip duplicate its circuit and make it into all basic analog parts? What about a counter with a 555 and a 4017 counter? Is that digital or hybrid device? IF you are doing electronics and all you care about is getting a result like meeting a  design specification and have to option of making an analog circuit or a digital one isn't it simpler to just make digital circuits and only make analog devices when you have to? Is a switch mode power supply or a class D audio amp considered digital hardware?

Does the definition lie when you are using gates? What about op amps they act like gates don't they? I hvent been doing this hobby for a few years and I have forgot a lot that just use to come to me instantly.

Back in the day, many digital functions were performed using discrete components.
"Flip- Flops" were made using twin triode vacuum tubes, & , if you wished you could duplicate the operation of your 555 with discretes.

Even if  4017 counts to 10, rather than 2 or some whole number power of 2, it still is a "digital" device.

In a "pretend" world, digital devices would only be concerned with "ones & zeroes" , but in the real one, gates are really just "bastardised amplifiers", so things like linearity can, & do, affect your signal integrity.

Frequency domain stuff like stray capacitance etc, transmission line effects, & so on, all affect both digital
& analog circuitry.

In a practical sense, if you have some analog circuitry feeding an ADC, that device would be where the transition is made, but that raises the question of whether an ADC is purely a digital device.
It has an analog input which has to exhibit satisfactory frequency response, linearity, dynamic range, & so on.

[Brumby]

The easy answer - when you can ignore analog nature of the device.

This is pretty much the best way to answer the question - but you have to remember that "digital" is just a special case of the analogue world in which we actually inhabit.

[Tomorokoshi]

Digital: You can analyze the circuit using Karnaugh Maps.

Analog: Everything else.

[IanMacdonald]

If signals only have to fall within a specific voltage or current range to have a given significance, then it's digital.

[KL27x]

IMO, digital means two states. On/off. Hi/lo. 1/0. Or to be more general, something that is digital only occurs is quantum, discrete steps. Analog means 0-X and everything between. Everything can be theoretically replicated in discrete components, including microcontrollers. You would have to slow the clock down because of inductance and capacitance. This has nothing to do with analog vs digital, IMO.

I wouldn't call 555 chip analog. It has digital outputs and inputs on it. Analog describes signals moreso than circuitry.* Even an opamp... is it analog or digital? If you make it into an amplifier, you can call it analog. If you make it into a comparator, it is digital?

*Even better, it describes the signal processing. If you take a curvy signal and only interpret it at a single cutoff point and call it hi or lo, then you have 1 bit digital signal. Take 3 cutoff points, and you have a 2 bit digital signal. If you simply amplify or buffer or invert whatever came in, then that is analog. But notice you can't assign an analog signal a number. You can only compare it to something else. As soon as you label it with a discrete number that doesn't extend to infinity decimal places, you have converted it into digital. You have chopped it into quantum steps. You have discarded the infinite amount of steps between this number and the next higher/lower. That's what an ADC does.

[Brumby]

IMO, digital means two states. On/off. Hi/lo. 1/0.

That is a bit limiting.  Binary is just one of a number of possible implementations of a digital system.

Or to be more general, something that is digital only occurs is quantum, discrete steps.

That's better.

I wouldn't call 555 chip analog. It has digital outputs and inputs on it.

Can't agree.  The "inputs" could hardly be called digital.  The flip-flop output I could give you.

Analog describes signals moreso than circuitry.* Even an opamp... is it analog or digital? If you make it into an amplifier, you can call it analog. If you make it into a comparator, it is digital?

These are better questions - ones that show digital is just a special case of analogue.

*Even better, it describes the signal processing. If you take a curvy signal and only interpret it at a single cutoff point and call it hi or lo, then you have 1 bit digital signal. Take 3 cutoff points, and you have a 2 bit digital signal. If you simply amplify or buffer or invert whatever came in, then that is analog. But notice you can't assign an analog signal a number. You can only compare it to something else. As soon as you label it with a discrete number that doesn't extend to infinity decimal places, you have converted it into digital. You have chopped it into quantum steps. You have discarded the infinite amount of steps between this number and the next higher/lower. That's what an ADC does.

A good progression in the argument - but, even so, all the stages within even the most complex ADC, MCU, FPGA or whatever, still have to follow analogue rules of electron flow.

[TheUnnamedNewbie]

No circuit is digital. Digital is a abstract conceptual thing used to design systems with.

It just so happens that we can make a lot of real-life things behave very similar like these conceptual building blocks. The most basic example of this must be a PMOS on top of an NMOS with their gates tied to together - the inverter.
However, building digital-like circuits is not limited to electronics: Mechanical logic can also be made (in fact the first computer can be argued to be mechanical, if you look at the machines thought up by Charles Babbage. Designing these blocks is then a matter of optimizing the behavior to be digital-like enough for our application - every gate has leakage, noise sensitivity, delay, rise-time and fall-time. (granted, this is often part of digital design too, but even then we tend to use very "digital" models with loading and unit-size inverters).

In truth, where you draw this line is very subjective I think. To a analog design engineer, things like latches and SRAM are "digital". However, a processor architecture designer might think of the design of SRAM as very "analog"-like design, since you are dealing with differential pairs and gain etc.

You could perhaps also consider how the design is done: what techniques are used? Digital design will often use simplified delay-based models, since they can ignore a lot of effects like noise, cross-coupling, risetime (they simplify risetime and bandwidth to propagation delay). Analog design usually uses SPICE or even lower level RF simulation tools.

[tggzzz]

Analogue signals convey energy from the transmitter/source to the receiver/destination.

Digital signals convey information from the transmitter/source to the receiver/destination.

Consider an RS232 (or whatever) link connecting one computer to another. The digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the voltages and currents in the wires.

Consider storing programs on a domestic tape recorder (yes, it used to happen). Again the digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the two different frequencies stored on the taps.

Consider a dial-up modem or many modern comms systems. Again the digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the phases and amplitudes on the wire.

To corrupt a digital signal, change the 0s and 1s.
To corrupt an analogue signal, corrupt the amplitudes, frequencies, phases, as appropriate.

[TheUnnamedNewbie]

Analogue signals convey energy from the transmitter/source to the receiver/destination.

Digital signals convey information from the transmitter/source to the receiver/destination.

What about digital data stored in QAM signals? What about the relationship between SNR and the amount of information that can be transported along the channel? I feel like this is a very poor distinction. A very substantial part of analog design is precisely conveying digital data - not energy (the energy is just needed in order to do so). The main reason of an RF link is to convey information, and not energy. I would hardly call that digital though: You can convey analog information such as speech without at any point using digital "gates".

[tggzzz]

Analogue signals convey energy from the transmitter/source to the receiver/destination.

Digital signals convey information from the transmitter/source to the receiver/destination.

What about digital data stored in QAM signals?

As I wrote, and you snipped, "Consider a dial-up modem or many modern comms systems. Again the digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the phases and amplitudes on the wire."

Digits are not "stored" in QAM/(C)OFDM/GMSK/FM/etc signals. See below w.r.t. interpreting energy.

What about the relationship between SNR and the amount of information that can be transported along the channel?

See Claude Shannon's seminal paper which completely defines that. See any communications theory textbook; they all refer to it and examine its implications in depth.

The effectiveness of digital communications systems are often defined by how close a single transmission approaches the Shannon limit. For a cellular system, also consider the capacity defined by bits/s/Hz/m².

I feel like this is a very poor distinction.

Nope; that's because you don't understand comms theory. Think about it.

A very substantial part of analog design is precisely conveying digital data - not energy (the energy is just needed in order to do so).

Irrelevant. Energy is conveyed. If you choose to interpret the energy (and changes in the energy) as representing digital signals, that's no different to choosing to interpret the energy as sound.

The main reason of an RF link is to convey information, and not energy. I would hardly call that digital though: You can convey analog information such as speech without at any point using digital "gates".

True, but the non-trivial parts of that statement completely misses the point.

In practice there are two points at which analogue signals become digital: photon counting (astronomy, optical TDR), and femtoamp currents.

[vk6zgo]

Analogue signals convey energy from the transmitter/source to the receiver/destination.

Digital signals convey information from the transmitter/source to the receiver/destination.

Consider an RS232 (or whatever) link connecting one computer to another. The digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the voltages and currents in the wires.

Consider storing programs on a domestic tape recorder (yes, it used to happen). Again the digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the two different frequencies stored on the tape.

Consider a dial-up modem or many modern comms systems. Again the digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the phases and amplitudes on the wire.

To corrupt a digital signal, change the 0s and 1s.
To corrupt an analogue signal, corrupt the amplitudes, frequencies, phases, as appropriate.

This is a fair enough  distinction if you are looking at  a digital system, but it gets a bit blurry in a purely analogue system, where the term "analogue" signal  can mean several things.

(1) In a simple example, the original signal could be that appearing at the microphone output caused by a person saying "Hello world".

(2) That signal after amplification &, perhaps, frequency response reshaping, or whatever.

(3)A signal modulated by (2).

All these things carry the original spoken information "Hello world" in some form, with some characteristic which varies in proportion to the original spoken words.

Not a digital signal to be seen, but information is conveyed from the acoustic input to the microphone, &  ultimately to an audio transducer of some kind.

[tggzzz]

Analogue signals convey energy from the transmitter/source to the receiver/destination.

Digital signals convey information from the transmitter/source to the receiver/destination.

Consider an RS232 (or whatever) link connecting one computer to another. The digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the voltages and currents in the wires.

Consider storing programs on a domestic tape recorder (yes, it used to happen). Again the digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the two different frequencies stored on the tape.

Consider a dial-up modem or many modern comms systems. Again the digital signal might be "hello world" converted to a sequence of 0s and 1s. The analogue signal is the phases and amplitudes on the wire.

To corrupt a digital signal, change the 0s and 1s.
To corrupt an analogue signal, corrupt the amplitudes, frequencies, phases, as appropriate.

This is a fair enough  distinction if you are looking at  a digital system, but it gets a bit blurry in a purely analogue system, where the term "analogue" signal  can mean several things.

(1) In a simple example, the original signal could be that appearing at the microphone output caused by a person saying "Hello world".

(2) That signal after amplification &, perhaps, frequency response reshaping, or whatever.

(3)A signal modulated by (2).

All these things carry the original spoken information "Hello world" in some form, with some characteristic which varies in proportion to the original spoken words.

Not a digital signal to be seen, but information is conveyed from the acoustic input to the microphone, &  ultimately to an audio transducer of some kind.

Agreed.

OTOH, I didn't say that analogue systems don't carry information! I deliberately omitted that possibility to force people to consider the essential differences without getting confused by a grey area.

As far as I concerned there are only two types of fundamentally digital electronics: photon counting and femtoamp circuits. Everything else is a matter of interpretation  

I've never done femtoamp electronics, but if I could have found a source of cryogenic temperatures usable in a trench by the side of a road, I would have done photon counting back in 1980.

[bson]

It's a bit like asking at what point do lines on a sheet of paper become letters.  Letters are to words what digits are to numbers.  Digital = numeric.  A binary digit can be represented by a single logic level, and in general if there's two or more bits it's digital, while a single bit is merely a logic signal.  For digits other than binary obviously more than two voltages or currents is needed.  Our use of binary voltage levels is purely a convenient convention; other schemes have been used, and it's not hard for example to envision the receiver maintaining a voltage and the transmitter sending by switching a load (which the receiver detects as a current)...  I mention this because it's a trap to think of binary logic as something not analog - it's very much analog and not treating it as such is going to result in suboptimal performance and reliability.  There is no point where analog physical reality can be ignored.

[Beamin]

How about asking this way: You run a company and like most managers you have no idea what you are talking about or how your products work. You have two departments an analoge department and a digital department with analog engineers and digital engineers. In order to make the company work efficiently you must assign things to the right department. How do you know what to assign to each department? IE you need to make an Arduino clone you would send it to the digital dept. If you needed a replica transistor radio it goes to the analog dept. If the president of the company who is not an EE asks you what type of projects go to which group, what do you tell him without going into a philosophical explanation?

[tggzzz]

How about asking this way: You run a company and like most managers you have no idea what you are talking about or how your products work. You have two departments an analoge department and a digital department with analog engineers and digital engineers. In order to make the company work efficiently you must assign things to the right department. How do you know what to assign to each department? IE you need to make an Arduino clone you would send it to the digital dept. If you needed a replica transistor radio it goes to the analog dept. If the president of the company who is not an EE asks you what type of projects go to which group, what do you tell him without going into a philosophical explanation?

The only sensible response would be to resign from such an idiotic dysfunctional company. To all intents and purposes, I have done that! (Cryptically, that specific company was in Browns Lane in the S of England, and was infamously run by accountants).

At an interview once, the HR-droid listened to my background and experience, and then asked "now, are you really a hardware or software engineer?". It was instantly obvious that such "listening" was a euphemism - I made my excuses and left - and got a job at a decent company (HPLabs).

The fundamental principal is that attempts to pigeon-hole people leads to sub-optimum solutions - famously expressed as "organisations are constrained to produce products that mirror their organisational structure".

[KL27x]

Can't agree.  The "inputs" could hardly be called digital.  The flip-flop output I could give you.

Well, the timing cap input is something you can call analog. (OTOH, you could call the voltage level analogue, but the processing of it you might consider as digital, because the response to this input is only 1 of 2 outcomes like a comparator. A comparator is essentially a 1 bit ADC.)  But if you set it up as a monostable multiivibrator, for instance, the trigger input can hardly be described as analog.

I suppose you would call the timing circuit on the 555 analog. It's based on filling a cap. Compared to a delay done by say microcontroller, which is counting edges coming out of a clock (which could be an LC circuit, too, so there's a subtle difference, I suppose.)

[innkeeper]

Zen of Digital:
Digital exists as long as you think of it in terms of 1 and 0.
The closer you look at something you consider digital, the more analog it becomes.
Digital is just a frame of mind, the reality is that everything is analog.

[T3sl4co1l]

Digital is a subset of analog.

Digital logic follows three rules:
1. Gates exist.  That is, logic elements with well-defined input and output states.
2. The gates operate with a Boolean algebra, utilizing a certain number of signal states.
3. Specified min/max propagation delays.  (Propagation is a delay between an input state change and a corresponding output state change.)

Anywhere these characteristics arise in an analog circuit, it can be considered digital as well.

Conversely, it often happens that a circuit, intended to be digital, violates these rules.  If the interconnects between gates are not simple propagation delays, but say they exhibit ringing: now the input to the next gate does not match the output from the proceeding gate.  This is why so much effort is spent on maintaining good signal quality.

Explanation:

(1) is necessary, to show that the circuit is not reciprocal.  That is, a signal applied to one port does not necessarily come out another port.  A passive filter network is reciprocal, i.e., it exhibits the same gain in either direction.  An amplifier is usually non-reciprocal (but not necessarily digital).  Digital must be non-reciprocal, in order for sequential logic to be implemented.  (It's noteworthy that, as an algebra, Boolean algebra needn't be non-reciprocal: indeed, that's one of the special powers of the equals operator.  But it's rather difficult to implement and design circuits that are reciprocal.  This is an important theoretical aspect of quantum computing, by the way: irreversible logic operations are equivalent to lost information and increased entropy.  Quantum computers can only achieve their power by harnessing this behavior!)

(2) allows for more than strict 0/1 Boolean algebra.  IEEE 1164 is an example of such a system.  In short, the system is extended to allow for outputs driving outputs, as long as they do so in a well-behaved manner.  This allows complex bus systems to be built with ease, while allowing the compiler to detect conflicts.

(3) is simply reality: the speed of light is finite, and no physical logic can be truly instantaneous.  In analog systems, the lesson again is well learned: there is delay everywhere, and its low-frequency approximations, inductance and capacitance.

Tim

[Beamin]

Zen of Digital:
Digital exists as long as you think of it in terms of 1 and 0.
The closer you look at something you consider digital, the more analog it becomes.
Digital is just a frame of mind, the reality is that everything is analog.

But really since we all live in the matrix our analog systems are just computer simulation which are digital. Or we could be a boltzman brain.

[tggzzz]

Digital is a subset of analog.

Thus spake an analogue circuit designer!

I agree that digital electronic circuits can be considered a subset of analogue electronic circuits. But in the wider context, digital systems as embodied in hardware and software, are not a subset. Information theory is an important extra dimension, e.g. the concepts of entropy, error detection and correction, and state spaces don't have an analogue equivalent.

[T3sl4co1l]

Well, at the absolute lowest level, you have quantum information theory, where interactions can be tracked in a graph structure.  Each node being a discrete state of any given particle, and each edge being an interaction causing a state transition.

Physicists are more than happy to integrate over a few quadrillion state transitions, taking the sum as if they were a continuum, at which point everything is analog again (statistical mechanics, the averaged properties of large populations of particles).  This includes most semiconductors (which are still quantum systems, but we owe their average characteristics to the statistics).

Meanwhile, if you're working entirely in QM (QIT is more of an adjunct to it, as Feynman diagrams are to QED, to write down which particle interactions to sum over), you have the "digital" property of wave quantization*, and the analog property of waves themselves (continuous amplitude, position and time).

*I put this in quotes because energy levels are discrete, but not usually Boolean.  Systems that exhibit useful qubits can be considered digital without quotes.

Tim

[T3sl4co1l]

Oh, sick!

So at a glance, this seems right:

Information theory is an important extra dimension, e.g. the concepts of entropy, error detection and correction, and state spaces don't have an analogue equivalent.

I was just contemplating these and considering counterpoints.  Offhand, entropy and state space are perfectly comfortable in the analog domain -- statistical mechanics seems an adequate, very low level example of the former, while state space control theory is perfectly comfortable with real-valued (or complex, I suppose) variables in the state vector (hey, it's just an array of differential equations, nothing digital needed).

But offhand, I didn't know of any examples of the emphasized case.

First hit on Google.  The headline and abstract seem to be on the money.  I will read this with great interest.
http://preserve.lehigh.edu/cgi/viewcontent.cgi?article=2035&context=etd

Thanks for inspiring this idea

Tim

[EEVblog]

Vague question but as I learn more I'm wondering when I have crossed this boundary. If you are using an Arduino that's digital with analog passives. But what about a 555 timer circuit? Still analog because you could take the design of the chip duplicate its circuit and make it into all basic analog parts? What about a counter with a 555 and a 4017 counter? Is that digital or hybrid device? IF you are doing electronics and all you care about is getting a result like meeting a  design specification and have to option of making an analog circuit or a digital one isn't it simpler to just make digital circuits and only make analog devices when you have to? Is a switch mode power supply or a class D audio amp considered digital hardware?
Does the definition lie when you are using gates? What about op amps they act like gates don't they? I hvent been doing this hobby for a few years and I have forgot a lot that just use to come to me instantly.

If you are talking in terms of not being able to duplicate with passive components, then you could argue that CMOS basically invented the "digital" era, because CMOS chips only use MOSFET transistors acting as on/off switches, nothing else, no resistors or anything else.
e.g. a CMOS NAND gate:



A TTL NAND gate on the other hand uses resistors: