How do most engineers get information about projects they aren’t working on directly? I often hear it is from technical magazines. I have also been told by my mentors the benefit of keeping up to date on innovation that might eventually become the new standard. Just think, one day however many years ago, they were discussing WiFi in magazines–even though it was not widespread. And now it’s so standard that I’m using it for free in an airport while writing this post! The point is, keeping up on technology is important. But who gathers all this information for us to later digest?

Dr. Sunny Bains is a journalist and editor who covers many different scientific and engineering topics. You’ve probably seen some of her pieces in magazines like EEtimes, The Economist, Wired and many other large scale publications. I first found her site while looking around the internet for more writings on analog topics. She is very interested in both the use of analog information in electrical and biological systems and how these might advance computing power in the future.

Chris Gammell: What kicked off your desire to study these subjects?

Dr. Sunny Bains: Actually, the first thing I fell in love with was holography: I saw my first hologram when I was about 9 when I went to visit my dad in Canada and we took a trip to the Ontario Science Centre. They also had a laser show there, and between the two I got hooked on the technology. Science fiction also helped: I remember being inspired by various shows and movies: Star Trek, Blakes 7, Tron, 2001… Being a girl, I think I was a bit behind my male colleagues in doing actual practical stuff. A number of my male friends got their first computer when we were about 13 (ZX81). My younger brother had computers all through high school. I only got one when I had been at college for a year and decided to start a magazine about holography.

Holography took me towards lasers and optoelectronics, optical computing and signal processing, and then more widely to machine intelligence and vision.

CG: What made you decide to then pursue journalism?

SB: I always knew that writing would be part of my scientific future. I imagined myself sitting in my office doing work and writing articles when people asked me to. I don’t know why… I wasn’t a huge writer when I was in high school. But I actually started applying for science-journalism related jobs when I was still in high school.

CG: I see that you’re a lecturer and researcher at Imperial College in London . What kind of work do you target? Is it still holography?

SB: My favorite subject for some time has been neuromorphic engineering: building analog circuits with brainlike structures. However, I’m still interested in all sorts of things in the area of emerging computing technologies, machine intelligence, optoelectronics and displays.

My PhD (you can see the introductory stuff on my website) was about physical computation and embodied artificial intelligence. Basically, I’m interested in analog information, and using physics rather than digital algorithms to do processing. That theme often comes through in the writing I do these days on my blog.

CG: How about the work you do now?

SB: These days I focus on three things work wise…my company, my teaching (communication skills for engineers) and my writing. Although I will say, it’s hard to write in a recession: advertising budgets are slashed, the number of editorial pages go down, and freelancers like me are cut. Since my PhD I haven’t done any research in the science/engineering sense of actually doing my own work, just in the journalism sense of finding out what others are doing.

CG: Do you think this neuromorphic type work will lead to a singularity, a la Ray Kurzweil?

SB: I hated Ray Kurzweil’s book… you can see my review on my blog.

SB: I think it’s MUCH more important for students to focus on analog electronics side of things. All that math is really hard, but once you’ve mastered it you can do anything. The biology you can pick up by osmosis I think. Anyway, you’re often focusing on some very small system in very great detail, and you’d have to learn all that at the time anyway. In some ways I wish I could be a neuromorphic engineer: I’ve got a fair bit of knowledge and a lot of interest. But I think being an engineer is the most important thing to make progress. That’s not to say that pure biologists don’t play a crucial role… it’s just that they can’t do much to create device: just find things/build models for the engineers to copy and, in some cases, do experiments to determine how well the engineers have done. Of course, I live in an Electrical and Electronic Engineering department, so I would say that…

CG: I’d be more interested in hearing more about your typical day, both as a publisher and a scientist. What is your typical day like?

SB: No such thing as an average day I’m afraid. From mid-September to mid-March I’m pretty busy with teaching and spend the most of the rest of my time running my company. If I’m lucky I write the odd piece if I have time. In the summer I try to work on other projects. That could be writing, research, or something for the company. This summer I’m working on a book about technical communication for engineers and physical scientists, and a new project for the company. As such, I don’t really have time to work as a scientist now; I only really did while I was doing my PhD (although that was for a long time). However, I think it’s quite likely that I’ll go back to scientific research at some stage as I tend to go back and forth between my various areas of interest.

CG: How do you first find out about the topics you write about? It seems like you often break stories that are very leading edge.

SB: I get my ideas from three main sources: conferences and other events (like one-off talks, workshops etc.), lab visits (where you just go to a lab and let them show off what they think is cool), and journals. I used to love going to the university libraries in the various cities where I lived and looking at all the new issues… Of course this is all done electronically now (no photocopying, hooray!). I’ve got some great stories by just seeing patterns in papers over a period of time. I do wish I could go back to writing regularly again. I know I could write more on my blog, but it was always supposed to be a byproduct of my freelance work, not a substitute for it.

CG: Do you have any predictions on the future of analog (since your articles are often very forward looking)?

SB: I am convinced that analog is the way to go for applications that are heavy in signal processing, and especially AI. Unlike with symbolic information, the “meaning” of signals does not always translate well into bits. Probably the story in the past that I’ve covered that exemplifies this best concerns Leon Chua’s cellular neural network. But the whole of neuromorphic engineering is built on the same idea: don’t break things into bits if you can just as easily use physics to do the computation for you.

Many thanks to Dr. Sunny Bains for taking the time to talk about her line of work. I was actually surprised when I first learned that she had so much exposure to neuromorphic engineering, a topic someone had previously suggested on the skribit portion of this site. It’s interesting how those futuristic ideas seem to dovetail with much of the analog knowledge of today; that often the most effective signal processing is still done in analog. Hopefully we’ll continue to see this trend and I’ll be able to write about it here.

If you have questions for Dr. Bains, please leave them in the comments!

Sure, it’s something different, but give it a listen and let me know what you think in the comments!

I’ve been thinking about my posts as of late. Moreso, how I can offer something of substance to readers. At least substance that is usable on a daily basis. What do EEs want to read when they get to this site?

I came up with part reviews. I started something similar at my other project, Electricio.us. Spec sheet analysis is a critical skill for any analog engineer. It’s also is time consuming and hard to keep up with. So why not have a few choice parts highlighted on here? I don’t plan on reviewing the newest and hottest parts…but maybe the most useful. So let’s start with the LM741.

A 741? These are parts I joke about, not parts I use on an everyday basis. But like I said, we all have to start somewhere. I’m sure many of us started using this part in electronics labs and at home in hobby projects. It’s the quintessential beginner op amp and it was one of the first ever designed.

LM741 datasheet from National Semiconductor

Pros:

• Cheap — Being made by multiple vendors (necessarily called the LM741 but probably containing the number 741 somewhere in the title) really can help with cost.
• Standard pinout — Sometimes even more valuable than having something cheap in the first place, being able to find a drop-in replacement can be a godsend when in a bind. This has been doubly true with the economic downturn. Sure, the part being out for 42 years doesn’t hurt either. People managed to figure out it’s pretty popular since it first arrived on the scene and duplicated this trailblazing part’s pin assignments.
• Nulling circuit – As we’ll see in the “cons” about this part, the offset voltage is horrible, but the tuneability of the part is a nice feature if you have to use this part. The “offset null” pins give you access to the 1K emitter resistor circuit, basically allowing you to drive the input voltage on one side of the op amp higher by giving a different resistance for the bias current on the input stage. This effectively raises the voltage on one side of the input or the other (inverting or non-inverting). Pretty neat stuff!
• Bandwidth – For an old part, the bandwidth on the 741 is 1.5 MHz. This is set by the internal capacitor, which acts as a pole inside the circuit and limits how fast the circuit can respond. This is extra interesting because the 741 was the first part to ever do this inside of the IC itself; previously you had to set the compensation capacitor externally. At 1.5 MHz, this old geezer of a part can still get up and move like a jackrabbit.

Cons:

• Offset voltage — As you can see above, the nulling circuit is necessary because the offset voltages can get pretty extreme on the inputs of this op amp. “So what?” you say. Well, this causes issues when any kind of precision is required. If you put in a 1 volt signal into an LM741 that is set up in a buffer configuration, the 5 or so mV of offset voltage between the inputs will get passed directly to the output! That’s .5% of your signal right off the bat! And if there is any gain in the circuit (i.e. it being set up in an inverting or non-inverting configuration), then your offset gets multiplied by whichever gain you apply to the circuit! That means for a gain of 10, your 1V input signal now has an output offset of 50 mV! Say goodbye to DC accuracy!
• Offset voltage drift –Even if you decide to “dial in” an offset null resistance on the proper pins, this is only accurate at the temperature you were at while correcting the offset. For every degree change, there is another 15 uV of voltage offset.
• Bias current — We all know that in an ideal model of an op amp, you assume there is no current flowing into the inputs terminals. Well, that’s never really the case, and having any current flowing into the terminals can cause DC offsets once that current flows through resistors (such as in a non-inverting amplifier configuration). The standard bias current spec’d on this part is .1 to 1 uA. That’s nothing to write home about, especially if you care about DC accuracy. The downer on this poor spec is that there isn’t an internal “knob” to help with the bias current; you just have to deal with it and try to design around it.
• PSRR — Though it’s not the worst I’ve ever seen, the PSRR on this part isn’t great either. At a typical value of 77 dB (use 80 for easy math) that means that any noise appearing on the power rails will show up 10^-(80/4) times on the output. So if you see a 1V transient on a 15V supply rail (gah! huge!), then you will have a 10 mV spike on the output. This can cause some serious noise problems down the line and most op amps I’ve been looking at these days have been well north of 100 dB PSRR.

This part is old, no way to overlook that. But it is still relevant part and is the basis for many parts that exist today. That’s pretty impressive some 40 years later. The downsides show the part’s age but are a symptom of the technology and the time this part was developed. The transistors are BJT, compared to the MOSFET based parts of today; many of the specs are guaranteed to be worse on this fact alone (though not the speed nor the power handling ability).

If you have a simple need for an op amp and you know how to properly account for all the shortcomings in this op amp, it’s a fine choice. Like I said above, it’s low cost, easy to use and plentiful in supply. And if history has shown anything, it’s that this part is not going away anytime soon.

So this was my first part review. I think in the future instead of running down the list of a lot of the specs like I did here, I would focus instead on one or two of the spectacular properties of a chip or the extremely underwhelming properties of a chip. Comparing all the middle of the road specs is a waste of everyone’s time. I wanted to make sure I covered a few things here as a baseline for future reviews though. I’d also like to step outside the bread and butter parts of an analog electrical engineer’s part drawer (the op amp) and review other types of components.

If you have any you would like to see reviewed in the future or have any thoughts about this review, be sure to leave a note in the comments.

Thanks to 畢業了吧 for the photo

The number in the parentheses are the number of positions listed online. It’s fair to assume some significant number of those are repeats (Indeed.com is a scraper, not some manual entry site), but we can assume that all the cities listed have a proportionate number of repeat listings. It’s also interesting– but not surprising–to note that certain areas are dense enough with jobs and location (i.e. silicon valley) that three of those cities (3, 5, 9) only show up as one tag.

Now, this isn’t to say these are the best jobs or the easiest to fill nor does it even point out how varied the positions can be! For example, an embedded developer and an analog system engineer might all be under the title “electrical engineer“. If you have experience working on electronics on an oil rig you’re much more likely to get a job in Houston than Fort Meade, regardless of how many jobs are available in either location. But these numbers do  point out where there is a considerable enough chunk of industry to have this many job listings.

So I ask you to respond in the shiny new comments section: are these really the only areas employers are hiring these days? Is there a significant long tail that I’m not seeing on Indeed? (i.e. 30 more cities with 250 listings each?) Are there any obviously booming spots that are left off the map? What about outside the good ol’ U S of A? I know there are a couple of readers, writers and witty commenters from outside my home country. Looking forward to your responses!

I imagine if a doctor was diagnosing the medical condition of the embedded community, he would walk into the tiny exam room, take one look at the embedded community sitting there in its socks and underwear on the crinkly disposable exam table cover and say:

“Yup, still fragmented.”

What do I mean by this? It means that even with my posturing about the need for community AND my lack of expertise in the topic, there are some undeniable rifts in the embedded community. And they will always be there. Why?

1. Too many vendors with their own pieces of silicon
   • Guess what? Companies like making money! Amazing, right? I can name at least 5 monstrous companies that produce independent silicon chips, almost always with similar cores that rhyme with “schlARM”. They have their niche areas and peripherals that are used in that segment; examples areas that a vendor might try to target are motor control, display processing, low cost, low power or RF. But in the end, the very things that distinguish them from their competitors and therefore allow them make money, also drives the community apart.
2. Too many closed doors
   • Another problem on the vendor side can be the amount of information provided to the people working on their chips. Without open access to the information, users are forced into the “camps” of the vendors in order to access features buried within the silicon. Less mobility between chips means more fragmentation.
3. Too much software
   • Well what about abstraction? For those out there that are more on the analog side of things, abstraction is writing code that isn’t controlling something directly. Think about it like you’re a teacher. You care a lot about turning the lights off in your classroom and want to teach your kids about why it’s important in order to save energy. In a non-abstracted case, you would tell each of the kids to turn the lights off when it’s their turn. Perhaps Wednesday it’s Johnny and Thursday it’s Susie. So you tell them directly. Abstraction in the simplest sense would be assigning Bobby to remember whose turn it is each day of the week. That way, you only have to tell Bobby to have someone turn off the lights; it’s the same every time. Bringing it back to processors and the embedded community, if things were abstracted, you could always tell “Bobby” to do the same thing and he would have close to the same response each time. Well there is such a thing that even the layman such as me is familiar with: operating systems. But this isn’t like the PC world where the choices have been culled down to a select two or three. There are embedded versions of larger OSes (think Win CE or Embedded Linux) and RTOS (Real Time Operating Systems) which are an even lighter version of their half cousins named previously. Beyond that there are superloops and other small implementations. The point is, there are a lo00000t of choices for software for a looooot  of different processors. It’s fragmented. But why all the trouble? Why do we need so many choices?
4. Too many market segments
   • It’s true. That’s why embedded has been growing steadily for the past 20 years and will likely continue to keep growing. There are a lot of  different needs! I guarantee you that engineers working on high-reliability industrial controls don’t care that much about Android. Sure, it could work, but it’s a new OS with lots of potential bugs and doesn’t really fit the needs. Similarly, handset makers don’t want to use reliable code from 10 years ago because all the reliability in the world doesn’t make a flashy new interface for mobile, web-enabled handsets. Chip vendors pick and choose to play to specific segments, as do the software vendors, creating hundreds of potential combinations; it’s much more likely that whatever current developers are working on though is a much smaller combinatorial subset. And so the fragmentation continues.

I know that analog is my niche and that there are some very compelling cases for using it in different areas of electronics. But I’m not stupid; there was a reason I took interest in the embedded space and why you should do the same. Everyone will continue to expect more from their devices, whether scientific, consumer or somewhere in between; if you’re on the internet reading this post, you’re likely used to the benefits of Moore’s Law and will continue to be.

What I’m trying to say is that there is value in learning about embedded systems; learning about some component of embedded computing is better than ignoring it. As software continues ascending into further levels of abstraction (think Python instead of C), there will be fewer people around that know how to reach down into silicon and flip a bit. Knowing how to do so not only will help you in your day to day tasks, but could make you a very employable engineer/programmer.

And who knows, perhaps embedded design will be the next black magic, much like analog is considered today!