Why Is Analog Design Harder Than Digital Design?

Digital is On or Off, where Analog is continuous.  Because of this difference, designing Analog circuits becomes much more difficult, as there are so many considerations that need to taken into account.  Temperature, noise, power signals can all effect an analog circuit.  In an Analog circuit, all of these forces need to be understood and compensations need to be designed in to the circuit to prevent them from influencing the intended functionality.  A simple opAmp circuit is relatively easy to understand, but how to prevent outside forces from effecting this circuit is much more complex.  I remember a simple saying, "amplifiers oscillate, oscillators don't"  that hints at how difficult it can be to achieve the functionality you want to occur, without having unintended results. 

I think digital is more of an illusion as signals are analog, and that fact will at some point have practical implications even if one tries to ignore that. I also think not that many are into analog because it requires a lot of math and physics, which is something I don't think many enjoy, specially when you get into differential equations, modelling and statistics.

This post reminded me of Bob Widlar, one of the pioneers of analog ICs who had had some pretty strong (and offensive), but kinda funny opinions on digital design:

I would challenge the statement "Analog Design Harder Than Digital Design". Define harder for me?

Either discipline taken into the real word (i.e practical application) is hard. What cushions the application of the discipline is experience. You could argue knowledge is also a cushion but I have found it only works well after the fact

I tell the story of an RF engineer designing a radio transmission/receiver system. The individual was an iron ring engineer and I was a technologist. The site we were working on, had reception issues. The replacement design was suppose to resolve the issue. The engineer design used the latest technology. He used the engineer specifications of the replacement equipment to maximize the RF energy available.

After reviewing his design I asked him some questions regarding theory applied. I agreed the specification indicates the RF connections has .5dB  loss. If that connection was made by an installer in -35 Celsius temperature and the cable was rolled out in a snow storm would he still feel confident in his design. My comments were to help him understand what happens in the real world.

I continue to discover, even after retirements, what works in the my lab doesn't work the way I think it should when it meets the real world. I thought this pic was a great visual of applied theory.

I am of an age when studying a BSc Electrical & Electronic Engineering was still mostly analogue (one of my first year courses was on designing with valves - the vacuum type). Then I had a course on digital circuits. Wow! Amazing! It was just so easy and it worked (virtually) every time. It was several years later that I actually really understood that even digital circuits were really analogue circuits. Still, by then microprocessors were abounding and I could even stop worrying about digital stuff. I still also remember a lecture or course I did on communication theory (possibly when I was trying to understand one of Alan Turing's paper on computing) when I realised that digital values mean whatever I want them to mean and were not absolute values. Good days.

Dubbie

I find analog design harder than digital design because of a number of hurdles:

• The maths - analog design is very math heavy and I found that quite challenging to get my head around.
• The "non-idealities" of reality - while it's easy to simplify components and think of them as "ideal" resistors/capacitors/inductors, in real life, the non-idealities can easily come to bite you in the butt - so what may work in theory doesn't work (or doesn't work well) in practice.
• Problems with simulation - I've had a good amount of fun with Pspice/LTspice, but it's not easy to simulate complex circuits realistically - again, shortcuts we take with the models here often yield simulation results that don't match real-life performance.
• Cost - unlike digital circuits which can often be easily digitally reconfigured (as software), analog circuits often result in needing specific components (which means having a good stash of them) and occasionally letting out blue magic smoke.
• The unforgiving nature of the signal - analog signals are exactly that. There is no "leeway" in interpretation - so whatever inadequacies are in the circuit will degrade the output (e.g. noise). The design has to be done much better and with more care than with digital signals that have a wide threshold between values and are tolerant of imperfections, especially when error-correcting codes are involved.
• The "dark arts" - some specific analog fields, such as RF, are still often referred to as dark arts as they can be difficult to prototype without getting stray inductances/capacitances dominating. To troubleshoot these designs takes a lot of knowing-rules-of-thumb and luck.
• More difficult debugging - a demanding analog design is not particularly easy to debug as the probes will often change the value of the signal due to loading or not read the expected signal/introduce corruption into the signal.

As a result, I tend to like digital - it's simpler even though technically everything is "analog" to begin with, digital methods merely interpret the analog as one of several states, making it more tolerant and allowing us (most of the time) to abstract the analog complexity away to just the first few blocks of a circuit that deal with input and output. I suppose a good illustration of this is the explosion of Software Defined Radio (SDR) in even consumer-grade products.

- Gough

Until digital becomes analog (high frequency) it is a bit like a cookbook - just follow the datasheet.

With analog, the datasheet is just the start of an adventure.

Douglas,

I can still remember in one of my microwave lecturers being informed that at higher frequencies that if the copper tracks on a PCB were at too sharp an angle the signal just kept going in the original direction. I was a lad from the countryside, we didn't have weird stuff like that.

Dubbie

Digital is a subset of Analog.

The boundaries of digital design are fairly easy to establish.

In Analog, you have a more open set of criteria that make a good circuit more challenging.

DAB

Interesting perspective. I used to work with a bunch of physicists who bragged that chemistry was a subset of physics, but that didn't mean they were knowledgeable about chemistry.

If you are a physicist, then you need a working knowledge of Chemistry if you are involved with elements.

If you are a chemist, you need a working knowledge of physics, depending upon what depth of elements and compounds you need to work with.

Digital engineers study circuits that have timing and decision making involved.

Analog engineers study circuits in which a much wider range of possible events within its solution.

I agree that as digital has gotten much higher speeds, the analog effects are introducing a wider range of issues than they did back in the early days of digital IC circuits.

Analog however still remains a wide open domain requiring a very specific set of skills, usually much deeper than those needed to do basic digital circuits.

I would never ask a digital engineer to do an analog circuit and vice versa.

They are two different domains requiring two different mind and skill sets.

DAB