# Digital circuits are made of analog parts

I came across this here.

Assuming that this is a true statement and not an experts' small joke, what makes circuits with digits (say with 0,1 bit machine codes) "analog"?

• When I was in college, I made a four-bit counter out of 2N3906 transistors, a few cheap diodes, and some resistors. (and some LEDs for indicators.) I wanted to make a 4-bit full adder, but my prof told me to stop wasting my time. Mar 30, 2022 at 22:24
• Suggestion for anyone who has the time: write an answer about post-2000 spinning hard drives, PRML, etc. Basically the bits are read and written so fast that you never get a voltage swing into what would normally be considered "0" or "1" territory, the controller just looks at the wiggles and goes "well, given the previous bits, and the response of the media, and the response of the head, this much wiggle probably means a 1, in context". Someone used to good chunky digital tech, seeing that the first time, will generally think "how the crap does that even work?" Mar 31, 2022 at 13:43
• All the answers are wrong! There's no such thing as analog or digital :-)) It's all Human Brain ;-)) Mar 31, 2022 at 17:13

Counter question: Can you name any circuit, that is NOT made from analog parts? And further, what is a digital part?

The basic components of any (integrated) circuit are transistors. While they can be configured to act in a digital way (On/Off), they are analog, they have characteristic curves, transient behaviours like rise times and such.

One could say "digital" is just a quantized way of looking at "analog".

Nothing in the physical world is digital.

As far as signals are concerned, digital signals only exist inside a computer (or on a piece of paper perhaps). In order to actually interact with or experience a digital signal it MUST be turned into something physical, i.e. an analog signal.

Example. A spoken word starts analog an analog signal in your glottis. It can be digitized into a digital signal and than processed, transmitted, received, stored, etc. Interestingly enough, each of these steps require something physical and so they are all executed in the analog domain. The "digitality" of the signal is just an abstraction layer that makes the analog circuitry easier since it only needs to distinguish between two different states and doesn't need to know that it's a spoken word or something completely different.

In order to actually USE the digital signal, it must be made analog again: as light from a screen, as air motion from the cone of a loudspeaker, voltage at an electrode that stimulates a neuron, etc.

what makes circuits with digits (say with 0,1 bit machine codes) "analog"?

The fact they use a physical quantity. For example the input and output of a NAND gate (which is one of the most basic building block of a CPU) are physical voltages. These voltages can have any value between 0V and the supply voltage. It's considered to part of a "digital circuit" for two reasons: a) we only care if the voltage is above or below a threshold value and not the exact voltage an b) it's only used at certain points in time and the voltage in between doesn't matter.

"Digital Circuit" does NOT mean that the actual parts are digital (they are not), but that the circuit only looks at above/below and only at certain points in time.

• The nerves controlling your glottis could be said to be digital. Mar 29, 2022 at 18:48
• @user253751: because they have activation functions with sharp transitions. But one thing this answer omitted is the analog nature of time (in addition to signal amplitude) -- the input and output of a NAND gate are time-varying physical voltages, which aren't quite limited by the power rails (transient overshoot is a thing), mainly voltage ramps, and the time delay between ramps and slope of each ramp are analog quantities with very analog effects, even under a digital interpretation you get "glitches". Ramp timing and propagation delay matters for nerves, too. Mar 29, 2022 at 19:38
• Nerve signals are not digital. They are sort of binary in terms of amplitude but they are not discrete in time: they can occur at any time Mar 30, 2022 at 16:27
• The opening sentence is misleading. "Digital" means "discrete", most generally, and there certainly are discrete phenomena in the physical world. The entire world of quantum mechanics opens the possibility for many types of quantized electronic systems which can operate on single electrons, etc. There is no fundamental "analog" nature to such a system - either an electron is present or it is not, so the device always exists in one of two available states.
– J...
Mar 30, 2022 at 18:30
• Single electrons have some non-zero probability of not being where you think they are, or when you think they are. Mar 30, 2022 at 20:30

The diagram below shows the different voltage thresholds for TTL (transistor-transistor logic) logic 0 and logic 1 values. As you can see, the $$y$$-axis has an analog voltage, and the valid "low" (logic 0) values are below 0.8V while the valid "high" (logic 1) values are above 2.0V.

The devices we use (whether TTL or other form of logic gates) are inherently in the world and are, thus, analog. It's only the interpretation we put on specific analog values that make them digital.

The diagram comes from here.

Digital machines are made using analog components. And the analog world tends to be discrete if you view it in sufficient detail.

This matters if you buy an expensive record player based on a flawed philosophical idea about «resolution» and «stair cases». For more reality oriented applications it is more about «what can I use this technology for?», and «how should I think about this tool in order to facilitate comprehending it».

Thinking about the voltages inside a cpu or dsp in terms of abstract digital ideals allows us to make and control a machine and behaviour that is vastly more complex than we ever could if we needed to care about clock flank rise time when programming a web page.

There is no such thing as a digital circuit in the real world, except by probabilistic approximation.

All electrical current depends on the movement of charged particles (electrons or ions) which, according to quantum theory, have some non-zero probability of being both outside and inside some circuit component (capacitor or transistor junction, etc.) at any given time. So a "1" or "0" is only a statement of probability.

If you add large enough guard bands, in both voltage levels and measurement time windows, then you can get the error rate low enough that you can pretend or assume a circuit (some probabilistic collection of electrons in motion) is digital, and hopefully not be wrong for as long as the circuit is needed to function.

But any transition through or probability of being inside a guard band region (transition voltage or clock edge, etc.), and thus outside the fully legal "1" or "0" state, is an indication that the circuit is actually analog.

• As an exercise for the student, take a clocked digital bit, violate its setup and hold time by varying amounts, and see what happens to the assumed "digital" output. Then look up meta-stability. Mar 30, 2022 at 20:29

Digital circuits are a human construct which has become widespread because it is so useful. My definition would be that a circuit is digital if information is represented exclusively as discrete (usually binary) signal levels. A key characteristic is that the underlying implementation uses amplifying non-linear elements that settle into one of the allowed discrete levels. This means information can be replicated with (in theory) complete faithfulness even in the presence of appreciable real world noise.

Digital is an abstraction, '1' and '0' are simpler to think about then voltages, currents, depletion regions, doping and bandgaps, but then, so is analogue, we speak of voltages and currents because that is a simpler thing to reason about then fields and waves and Maxwell, which is in simpler to reason about then QM.

All models lie, some are useful.

When engineering you need to be asking if you are working at the correct level of abstraction all the time, does the 'digital model' apply in this case or do I need to go down and look at the eye diagram, but also, can I go up and view this as software instead of digital logic?

And yes, you can trivially build 'digital' things out of analogue components, digital just means we have defined a range of (analogue) values that we consider logic '1', a range that we consider logic '0' and consider everything else to be indeterminate.