Are the Glory Days of Analog Engineering Over?

An anonymous reader writes with this article about the future of the analog engineer. Some say technology advancements are obsoleting the need for analog engineers, while others say that good, experienced analog designers will always be needed and currently are in short supply. After years spent encouraging engineering students to focus on software and digital electronics, some people say the day of reckoning appears to be drawing near: Many analog mixed-signal design jobs now stay open longer or are simply going unfilled, say recruiters, with some engineers even unable to retire because they can't find a suitable replacement. On the one hand, some people blame the shift from analog to digital, which produced a generation of engineers who speak the language of code, not circuit schematics. On the other hand, others say that with the advent of systems-on-chip, the easy availability of free circuits, pioneered by companies like TSMC, and software tools to verify designs, there is simply less need for analog designers.

Comments

The world...

[Alioth]

The world is analogue. Someone's going to have to design the analogue front end to your digital system. Even if you have a ready made analogue front end, you still have to understand the analogue world if you ever hope to design high speed digital systems. When it comes to the actual voltage levels on your PCB and signal integrity, the nice clean world of software where you can just expect the hardware to be predictable and just work with no effort goes away, you have to have a little bit of a clue about the analogue side if you want your high speed digital signals to reach their destinations intact. Another example is your (A)DSL line, it might be called "Digital subscriber line" but it required analogue design to get the signal from your modem (and it is a modem - it modulates and demodulates the signal) to the DSLAM in your phone exchange.

You might not need as many analogue engineers as you may have (say) in the 90s, but they'll never go away because the world is analogue, and the analogue world constantly impinges on your digital signals especially once you pass single digit MHz speeds.

Re:The world...

[usuallylost]

The articles headline is a bit missleading. In the body of the article you find that even they admit that analog engineering isn't dead or going anywhere. What is changing is the exact skill sets required. If you are doing traditional circuit design on purely analog equipment you are on hard times because people aren't doing as much of that. If on the other hand you have a foot in both the digital and analog world and can do analog design for digital systems there is a shortage and money is really good. So basically the people having problems are the older analog engineers who haven't kept their skills current. I think you could write that same article about just about any technical field where there has been rapid development in the technology. Some folks end up in dead end specialties that simply aren't in demand anymore. Your options there are retrain, change carriers or compete for the ever shrinking number of jobs. I'd argue that the last one is the worst choice unless you are simply close enough to retirement that the other two are simply unviable. Which actually appears to be the case with most of the guys listed in the article.

Re:The world...

[some old guy]

Here's a news flash: Engineers with actual jobs couldn't care less about quanta. In our world, P.I.D. calculations and control loops are based on the actual phenomena being controlled, said phenomena being analogue in behavior.

If you want to have a mainframe grinding out the quantum equations for a functional room thermostat in near real time, that's your business. For the rest of us in the real world, there's analogue engineering.

Re:The world...

[neilo_1701D]

It would posit that if Heisenberg were still alive, he would disagree with you.

But Heisenberg is in a resolved Schrödinger's cat problem...

Re:The world...

[TheGratefulNet]

Most digital engineers can do analog design as well.
Digital is generally much more complicated than analog design.

uhm, you could not be more wrong!

digital is trivially easy. the tools do the work for you. pcb trace layout, while no one seriously uses autorouters, can be done with little effort and the verification tools ensure signal integrity.

but working with analog is much harder and more of an art than science. you need experience and you don't get that from school.

today's EE's dont' even know how to solder. its pathetic. they run a sim and type on keyboards. some don't even use test gear, like scopes.

no, analog is much harder and still needed. audio and video have a lot of analog nature to them, still, and power supplies, rf systems, antennas, filters (that are not done in dsp), buffers and amplifiers - all analog.

digital has leeway before it fully breaks; but analog has to be done right or performance will suffer.

Analog : Digital :: Embedded : Software Eng.

[fractoid]

Analog circuits are always going to be faster, more accurate per area of silicon, and less deterministic than digital circuits. They're also always going to be harder to understand than digital circuits for anyone who isn't a wizard. There's less need for analog circuit wizards than there is for digital circuit designers just the same way there's less need for deep embedded software wizards than there is for your garden-variety software engineers. It hurts to say it but technology is advancing to the point where it's less important to get 100% out of our current technology than it is to get 25% out of it in a manner that mere mortals can understand.

There'll always be a place for analog design but it will be confined to an ever-shrinking niche on the cutting edge where, as bogglingly capable as it is, our digital technology just isn't quite up to the task.

Re:Analog : Digital :: Embedded : Software Eng.

[nerdbert]

No. And I say that as an analog designer. I've been doing this now for 25 years and I can tell you that analog circuits are typically limited to 8-bit accuracy without fancy digital techniques behind them.

And that statement alone should tell you why mixed signal is really where the action is for accuracy. Take the example of delta-sigma ADCs. You need the best comparator/DAC you can design, but you follow that by massive oversampling to get your 15+ ENOB accuracy by putting the noise out of band. Similarly, all the fast electronics in your o-scope these days uses massively parallel oversampled designs.

So no, analog circuits aren't going to be faster and more accurate per area of silicon. A good design that uses an appropriate mixture of both analog and digital is really where the best (smallest/lowest cost) solution is. There are times when you pretty much have to go pure analog (LDOs after your switched regulator in a phone, for example), but in general the best solution for nearly all problems these days is a mixture of analog and digital.

Yes, analog circuit design is "wizardry" to some people, but I personally put it as a deeply specialized niche that's extremely difficult to master and as such it's no different than the equivalent specialization in other fields. When we get a new MS grad in here in our chip shop try to start analog design I tell them flat-out that what they learned in school is less than 5% of the knowledge they really need to make a product and not to take it personally when they are closely supervised for 5 years as they learn what's really needed. You thought circuits were hard in school? You ain't seen nothin' until you've actually tried to make a mixed signal chip in a deep submicron technology (although strangely enough, the latest FinFET processes are relatively more analog friendly than the planar stuff we were dealing with before).

To me the real issue is what's happening in the chip industry. SoCs have huge economies that are driving their use in things like phones. But an SoC takes a huge company to make since you have to supply an incredible amount of IP and by far the bulk of that IP is digital. The problem that creates is cultural. Analog guys have hugely different needs that get ignored by digitally-oriented SoC companies, and without enough analog guys they tend to wave off what the analog guys need to do their jobs as too hard and too specialized for their support teams to bother with. That leaves the analog guys in those big companies generally supplying inferior solutions, which means that analog guys don't want to work for those big companies, which means the big guys don't get the best analog guys, etc. until you have a death spiral. So what you're seeing in the chip industry these days are big digital IP companies and smaller, specialized analog companies and that increasing segregation is roiling the traditionally very secure and stable analog design positions and making it appear analog design is going downhill.

Digital is only digital if analog is right

[Big_Breaker]

Part of "Digital", the lowest level of digital, is a contract concerning how signalling between transistors occurs. This includes timing, rise and hold times, voltage thresholds and current. I'll include avoidance of race conditions, clock distribution, refresh cycles on DRAM and temperature effects as a side car. These are all design constraints that make sure the 1s and 0s working properly. It's only when you have a 99.99999999% solid digital contract that you can begin the digital side of the design.

All of this digital design is solidly analog and will NEVER go away.

I could make another whole post about the absurdity of traditional "analog" going away. All these mobile devices have some amazing RF design going on from the antenna down to the mixed signal SoC. Analog is everywhere and at the core of every electronic gadget.

Some engineers even unable to retire?

[zennling]

Just because there is no suitable replacement should not be grounds for an engineer to work forever! Which companies are saying that you arent allowed to retire because of no suitable replacements? Name and shame them!