Advertisement

Blog

Analog & Digital Are Like Oil & Water

It doesn't matter what product we want to think about. Every product, when put in an environment in which it does something useful, is an analog system. That is because this is an analog world and digital is an artificial construct that our human brains created to attempt to make life simpler. We like binaries, be they for computation purposes or socially defining things, such as male, female, black, white, etc. It is much easier to talk about binaries rather than try and define something in between.

So what makes analog and digital so different? Let's start with analog. Analog entities are continuous in time and value. Things such as sound, heat, and pressure can vary over time, over distance, and can never be defined with exact precision. A sensor can convert between these environmental factors and electrical signals, creating a relationship between them, hopefully with some degree of accuracy. There are equally an infinite number of possible voltages and when formed into a circuit, currents that flow around a circuit. Those current flows create magnetic fields that can induce currents in other signal wires. We rely on this effect to create radios and transformers although within a chip they are usually undesirable effects and referred to as electromagnetic interference or EMI. But they exist and they can influence an analog circuit — a subject we will probably return to several times in the future.

In the digital world there are only three possible values — '1,' '0,' and don't know or don't care. Most of the time we don't even care about what voltage is associated with those logic values. Transitions between logic values are considered to happen at a particular instant in time and we are not interested in what changes they may make except at specific points that we call a sampling point. This is just one of the abstractions that have been applied in the digital domain over the past 20 years. If we consider the analog domain we find that little has changed over that period. Design is still done at the primitive device schematic-level and verified using a SPICE-like simulator.

This has created an ever widening abstraction and complexity gap between the two domains. Those digital abstractions have enabled automated synthesis solutions. Analog remains a manual process.

What happens when the two come together? Most of the time, they may as well be talking different languages and we can blame Gene Amdahl for what happens when they try to work together on a problem. For example, a software engineer may want to simulate for minutes or hours of real time, the digital designer want to simulate for seconds or minutes of real time, while the analog guys might be happy with a millisecond. Digital designers will generally ignore chip to chip variations, temperature, or voltage; instead just rely on their designs being impervious to them within specified limits so long as they follow a number of design rules. For the analog designer, no such things can be ignored and their problems are being made worse with every process improvement made for the digital folks.

But it is important for the two sides to work together on a single chip. Increasing levels of integration are necessary and it is no longer possible to design and verify the analog and digital parts separately. I will be exploring the problems and some solutions to these problems, especially with respect to integrated analog functionality, in the blogs that follow. Watch for further thoughts here on Planet Analog and on the Integration Nation site.

Related posts:

38 comments on “Analog & Digital Are Like Oil & Water

  1. eafpres
    April 29, 2013

    Hi Brian–intersting points about analog.  Just for fun, I'll split hairs a bit.

    EM waves can act like particles and become “digital”.  There are some pretty important things that depend upon that.  

    Although physicists were generating particles and families of particles like crazy in the 1970s, that has slowed down a bit, and we have a set of fundamental particles of fixed (rest) masses.  Those are digital values which underlie the fabric of everything.

    I admit that the vast majority of the “important” phenomema in electronics would fit in your descriptions, but I could argue that all of reality is digital, and analog is an abstraction.

  2. Davidled
    April 29, 2013

    When reviewing in the application circuit, analog and digital is coexisted in the board in the most case. Two worlds cannot be separated in the Electronics world. Simply, system engineer should look at both parties. As simple example, in the system or design level, signal condition circuit include most of analog circuit with Op Amps. Mainly, digital portion includes all memory access with address and data communication/protocol.

  3. Brad Albing
    April 29, 2013

    I don't think Brian is actually suggesting that the two design methodologies can't coexist (since obviously they do coexist in many products). Just that there is a difference in the way designers work on the designs – from their creation thru prototyping and on thru manufacturing and troubleshooting.

  4. Brad Albing
    April 29, 2013

    Uh-oh – I detect another discussion that's about to drift off into the vagueries of the zen of subatomic particle physics.

  5. eafpres
    April 29, 2013

    @Brad–you say that like it is a bad thing…

  6. Vishal Prajapati
    April 30, 2013

    I can point out the Proteus for codesign simulation. It has good capability and has great SPICE tool that can simulate differetn Analog signal sensing using microcontroller with our own written program running on it. I am not sure about the future of Co-design simulation but I can surely say this is a great tool for my co-design simulation.

  7. SunitaT
    April 30, 2013

    ICs with both analog and digital integrated functions faces lot of issues. One such issue is fast-changing digital signals send noise to sensitive analog inputs. One path for this noise is substrate coupling. 

  8. SunitaT
    April 30, 2013

    Most of the time we don't even care about what voltage is associated with those logic values.

    Technology scaling has the great impact on the digital and analog integration. As the technology scales down voltage supply level and depletion width also scale down. For digital volatge scaling will the impact like lower threshold value, lower noise immunity, etc. For analog the available signal range shrinks. At today's 1-V supply level, the signal range may be 0.7 V or less. This lower signal range requires proportionately lower noise levels to maintain the same dynamic range.

  9. Brad Albing
    April 30, 2013

    Those discussions are simultaneously fascinating and maddening due to the cmplexity of the discussion. Generally, I can almost but not quite understand the details.

  10. Brad Albing
    April 30, 2013

    Good to see that you're getting good results with Proteus.

  11. Brad Albing
    April 30, 2013

    Yep – always an issue when the analog and digital circuitry is in close proximity. The related issue (discussed here in several other blogs) concerns grounding and ground return current flow. That's a topic to which we will return.

    Previous ground related blog:

    (Under)standing Your Ground

     

  12. Brad Albing
    April 30, 2013

    That in a nutshell is the summary of the quest for lower power draw and higher speed devices. We operate at lower voltages so that the span of the signals (low to high or V-min to V-max) is small and therefore can be quickly traversed. But by doing that, we lose noise margin or noise immunity.

  13. Scott Elder
    April 30, 2013

    @eafpres

    Since we are splitting hairs here, if you want to argue that everything could be considered digital because of quantum mechanical theory, then you must also concede that while it may be digital, it is also not predictable.  So then we are back to analog vis-a-vis noise.  So mixed mode if you will.  Yes, quantum states.  But different energy levels of states and not always predictable in which state or for how long.  Actually, sounds more like a human.

    Even chairs exhibit 1/f noise.

  14. DEREK.KOONCE
    April 30, 2013

    Another perspective about not digitizing quantum effects, is just add in time and location. Essentially the digital people have to keep breaking things down more and more discrete levels until they reach a divide by infinity; analog folks already have it done from the start.

    Would love to see a digital person represent PI digitally, for the analog  person just says circumfrence over the diameter.

  15. eafpres
    April 30, 2013

    @Scott–I think I have had a few of those noisy chairs.

    You are correct that the uncertainty principle comes into play at this level.  This, of course sets the ultimate limit to things like Moore's law.  Either the irreducible noise caused by trying to extract the signal (analog or digital) will make S/N <=0 or you will reach transistors with single atom doping levels and that's all you will get.

    I haven't done any back of the envelope calculations regarding which of those two limits might come first, or if simply the speed of light in the matrix will be the limit as things are packed infinitely close together but still having physical extent.

    My gut feeling is that the real lower limit will be thermal for practical devices.  Once you have numbers of dopant atoms you can count on your hand enough thermal cycling may kill the device, or migration of other atoms will, or something more catastrophic.  Even with a cryogenically cooled device once you power it up you have heat transfer limits to get the energy out, so you have sub-nanoscale gradients that can get pretty steep.

    And so on, or, as they say–Turtles, all the way down.

     

  16. David Maciel Silva
    April 30, 2013

    When we put digital circuits with analog processing capacity have to take care of various details in a project, do not just connect, but learn to connect to something besides chips and printed circuits, but to follow precise rules, know the final result as “a catastrophe”.

    Some important tips on treating EMI and EMC in electronic circuits:

    http://www.radio-electronics.com/info/circuits/emc-emi/electromagnetic-interference-basics-tutorial.php

  17. Davidled
    April 30, 2013

    -Two different design approach

    I think that they use a different approach/methodology in the design stage. For example, analogy engineer more focus on circuit analysis with worst case which is very challenge tasks with S/N (Signal/Noise) ratio and temperature range. Digital engineer might concern about data speed, communication spec, cross talking in the high speed application. They are more familiar with digital chip such as DSP chips, memory chip related to information capacity. My intention is that integration validation of PCB board consisting of analog and digital could not be separated.

  18. jimfordbroadcom
    May 1, 2013

    @eafpres, as an analog guy with a fair amount of digital experience (I don't know how you could avoid it these days), I immediately thought you meant S/N <= 0 dB, or signal power divided by noise power <= 1.  S/N (or SNR, as I prefer to call it) < 0 dB happens all the time; error-correcting codes and processing gain, both application of digital technology to the effects in an analog world, allow us to extract signals out of much larger noise.  I'm not even mentioning quantum computing, in large part because I know next to nothing about it.  The point is that we think we can extrapolate to the end of Moore's law, but there's "so much room at the bottom" to paraphrase Feynman, that we may see some interesting devices pop up soon.  Rumors of Moore's law's demise have been greatly exaggerated, to paraphrase another of my favorites, Samuel Clemens, AKA Mark Twain.

  19. Brad Albing
    May 2, 2013

    Makes sense – except for the last part – who says, “Turtles, all the way down”?

  20. Brad Albing
    May 2, 2013

    OK – thanks for clarifying your point of view.

  21. eafpres
    May 2, 2013

    Hi Brad–my refernce to Turtles all the way down is a possibly misguided attempt at humor.  In the parts of physics dealing with cosmology, this is a well known if vaguely documented quote.  If you go to Wikipedia there is a lot on this; I cannot vouch for what is there but I did read Hawking's book and that is probably where I first read this.

    The reference is various forms is to the difficulty in explaining any sub-system without any external reference.  You can think about it in various ways.  In the cosmology case, it deals with the problems like “what came before the big bang” (one of Hawking's more controversial topic areas).  Historically, it is said to represent a cosmology where, say, you take the Earth, and explain what holds it in place–“it is on the back of a turtle”; then “what is that turtle standing on” and you get trapped into “it is turtles all the way down”.

    In my case I was referring to the problem if you suppose continuous miniaturization of electronic devices with the commensurate continuous increase in functional density (say, transistors per unit volume).  My point was that Moore's “law” (which he did not claim was a law, of course) will break down when we get to quantum limits or before.  If you don't believe that, it is tantamount to believing “It's turtles all the way down”.

  22. goafrit2
    May 2, 2013

    >>  but I could argue that all of reality is digital, and analog is an abstraction.

    I do not understand this comment because it is analog that is reality since the world is analog in nature; digital is the abstraction created by lazy people somewhere in the 19th century.

  23. goafrit2
    May 2, 2013

    >> Just that there is a difference in the way designers work on the designs – from their creation thru prototyping and on thru manufacturing and troubleshooting.

    Absolutely correct. Analog design and design designs are two different paradigms. The designs, the tools, the processes and everything indeed are different. The only constant is that the analog guy is more respected because the digital guy kind of works for him.

  24. goafrit2
    May 2, 2013

    >> ICs with both analog and digital integrated functions faces lot of issues

    Interesting there are few commercial core IC products that do not have both analog and digital. In Intel processors, we have plenty of analog despite the obvious digital systems. We need to understand that ADC and DAC are there for a purpose because the world is analog which most times the processing power is digital (now we have spiking processing from the neuromorphics guys)

  25. eafpres
    May 2, 2013

    Hi goafrit2–My point, at least half seriously, was that electromagnetic energy acts like particles and comes in “quanta” so at the quantum level, it is in discrete packets, therefore inherently “digital”.  Same goes for fundamental particles–there is a finite set, they have fixed (rest) masses, so the construction of all matter is from discrete pieces of mass and therefore also inherently digital.  Space you could argue is analog, and time.  So perhaps I should have said the universe is inherenty mixed signal at the fundamental level.

  26. Scott Elder
    May 2, 2013

    @eafpres – Enough with the quantum stuff….let's go back over to the Linear versus not Linear blog discussions.:-)

  27. eafpres
    May 2, 2013

    @Scott–LOL.  Agreed!

  28. GarethBaron
    May 3, 2013

    In the digital world Pi=Circumference/Diameter.  It's all about precision & what's needed to perform the calculation.

     

    Now the benefit of Digital is that there is very little noise due to the fact that the original signal is represented with a very high SNR in analog terms.  (ie bits).  However you sacrifice silicon/circuit space for bits.  Analog computing is excellent for some things but it does lose precision very quickly when chained (Noise gets amplified etc).  The best solutions these days are do as much in the digital realm and then convert back into Analog when necessary.  Both are tools and both are necessary.  I know there are necessary things like sampling that have to be done in the analog domain, but if there is complex math required or some form of extreme precision needed then digital is the way to go.  For speed, analog will always be faster.  I'm a fan of both design spaces I guess!

    Now as far as Quantum computing is concerned, the qbit is going to be interesting.  We won't be dealing with time or frequency domain issues but will be dealing with probabilistic domain issues.  Quantum computing is still in it's infancy so who knows what's going to happen in the space.  Care to take a bet 😉

     

    And as far as Quantum level signals being digital, that's all to be discovered.  That's how we see them at the moment but as newer theories come in we may start seeing an analog nature of sub atomic particles.  I think it is too early to make the judgement that they are analog or digital in nature.  We have a model at this point that fits the boundary conditions.  What's in the box ???

  29. goafrit2
    May 4, 2013

    >> So perhaps I should have said the universe is inherenty mixed signal at the fundamental level.

    Absolutely, outside the domain of the Newtonian Physics to the quatum mechaical world, everything changes. The good thing is that the mixed domain is seamlessly entwined than oil and water are.

  30. goafrit2
    May 4, 2013

    >> Enough with the quantum stuff….let's go back over to the Linear versus not Linear blog discussions.:-)

    Why must we not discuss it if we care? I am curious. What is bad about discussing quantum stuffs? If we think that nanotechnology will redesign the constructs of the ways we design analog and digital systems, it makes sense to discuss quantum effects.

  31. goafrit2
    May 4, 2013

    >> The best solutions these days are do as much in the digital realm and then convert back into Analog when necessary.  Both are tools and both are necessary. 

    Yes, both work in harmony. Except for power consumption, that strategy of doing digital and then getting back to analog makes sense. But when you work in low-power systems like in biomedical systems, doing more in the analog realm becomes a huge strategy.

  32. Brad Albing
    May 6, 2013

    Hmm… the zen of analog and digital. Maybe, maybe not. You'll have better luck (get better results) discussing the zen of particle theory + wave theory. Which neatly ties in to the earlier portion of this thread. So that would be the zen of our replies….

  33. Brad Albing
    May 6, 2013

    I was hoping you were going to explain how a further discussion of Moore's law is a good walk spoiled.

  34. Brad Albing
    May 6, 2013

    EafPres – that actually makes surprisingly good sense. As cosmological explanations go, I like it.

  35. eafpres
    May 6, 2013

    @Brad–as a Chemical Engineer turned into Quality Engineer, then turned into RF Engineer, and finally going over to the dark side of Corporate, and now (!) Consultant, I'm happy to hold up the Cosmology part of the discussions, since it is obvious I have no aversion to jabbering about things I know little about. 🙂

  36. goafrit2
    May 8, 2013

    >> You'll have better luck (get better results) discussing the zen of particle theory + wave theory.

    It is very likely that few and fewer are spending time understanding that even in schools. I know that in EE, everyone is doing something on Arduino, connectivity and then apps/web APIs. It is getting harder to see students pushing into the gray areas in EE at least based on the ones we have as interns.

  37. goafrit2
    May 8, 2013

    >> I was hoping you were going to explain how a further discussion of Moore's law is a good walk spoiled.

    Moore's Law is going to be like Faraday which was law until it was no more law. Why? There is nothing law about either, just that authors became media personalities and gave names to observations.

  38. goafrit2
    May 8, 2013

    >> @Brad–as a Chemical Engineer turned into Quality Engineer, then turned into RF Engineer,

    It will be good to know if you run any consulting outfit on career planning. I have people here that struggle to redesign their careers after some bumps. If a chemical engineer can move to RF Engineer which is scary for even EE majors, I mean anyone needs not give excuses to change.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.