(Editor's note : This hands-on article is somewhat of a departure from our usual type of technical article, but it is not a “touchy feely” recounting of the emotional trials, tribulations, late nights, weekends, and successes of meeting and overcoming a major challenge.
Instead, it describes how an experienced engineer undertook to teach himself analog IC design, including his planning, the tools, the sequence of events, and the actual IC fabrication process. Whether you are thinking about learning analog IC design yourself, or just want to see how you can use available resources as part of self-paced continuing education regardless of your engineering career stage, you'll find it of interest and with actionable lessons and take-away information you can use.)
We are pleased to present this feature, in three parts:
- Part 1: Goals, sources, curriculum, and “exams”, below
- Part 2: The thesis project and MOSIS, will appear soon
- Part 3: Free tools, Tanner tools, and finishing, will appear soon
Enough of the introduction, let's get on to Stephen's Lafferty's story, which begins here:
Learning analog integrated circuit design may not be as hard as you think. Take advantage of the new software tools and books which make it easy to study at home.
I've enjoyed many years working in electronic design and most of it has been interesting and rewarding. In recent years, I had drifted towards more supervisory design functions, which is natural, I suppose. A few years ago, looking back on my career, I realized that I loved doing analog chip design the most. I decided to make an effort to get back into that.
Many years have passed since I worked at Harris Semiconductor (now Intersil), though, so my IC design skills needed a thorough tune-up. This is the story of how I created and executed a home study program to do just that. The layout of the chip built as the final project, Figure 1 . If you are someone like me, who has good analog-design skills and a strong desire to get into analog-IC design, I hope that reading about my experiences can help you reach your goal.
Figure 1: Final project–a low-noise CMOS op amp with rail-to-rail output, created in Tanner's L-Edit
(Click on image to enlarge)
That brings up the first step: defining your goal. You need to know what kind of analog-IC design you would like to get into. The Web helps a lot there. I looked at job listings, found out that National Semiconductor Corp. has a design center nearby and researched analog ICs in general.
The CMOS analog work in my previous experience had been a brief oddity at the time. Today, CMOS digital processes dominate the landscape, though BiCMOS is not uncommon. So my goal became to be able to do state-of-the-art analog design in digital CMOS processes, with some exposure to BiCMOS as well.
Your goal might be different. I would just urge you to identify actual design jobs that you would like to have, as a model. You might not get exactly that, but at least you will have a target which is based in an actual industry need. Whatever you decide, write down your training goal in a few sentences.
The next step is to define how you are going to get the training you need, along with your time and cost budgets. This has to consider how long you expect to be in your career, what you have saved (or can borrow), what your expenses are, and what effect the career-change will have on your earnings. Based on where I am in my career, I decided that a one- or two-year full-time study program would be reasonable.
As it turned-out, the period was extended. (More on that later.) I am single (sigh) and living costs run about $3500/month to live comfortably. That includes healthcare and some support for a niece in college.
Sources of training
For sources of training, my options included taking courses at nearby Georgia Tech. However, their enrollment for the semester had just closed. Moreover, looking at the roster of IC courses, it was clear that they aren't all given in all semesters. Therefore, you have to get in-sync with the sequence. I did a brief search for online offerings but I just couldn't find on-demand courses which met my needs.
It turns out that it's not so hard to create your own courses from the many excellent textbooks that are out there. In some cases, they will refer you to free or low-cost software which you will need. However, you should also have a budget for tools which you would be better-off buying. Set a reasonable schooling budget in relation to the living expenses which you will have to cover.
My budget was about $10K. Sound like a lot? Not really. It's a bargain compared to tuition costs and it's a lot less than the living expenses. Most importantly, it's small with respect to the anticipated increase in earnings the training would provide. In my case though, getting to do the kind of work I love is more important than the pay. Anyway, my schooling budget was something like this, Table 1 :
(Click on Table to enlarge)
As you will see, the actual chip project is the most important part of the training program, so that justifies the biggest part of the budget. The second chip fab allows a turn in case the first fails. I felt a trade-show trip would be important to get a feel for the industry and to meet people (wrong). (More on the software later.) The IEEE Xplore subscription beautifully takes the place of a library for home schooling. Beats a real library any day. It's well worth the $35/month.
Notice that there is no lab equipment budgeted. Like many analog engineers, I enjoy doing electronic projects and have a decently equipped lab. The fact is though; there isn't much call for lab work in chip design. Since capacitances are so small, there is no way it can help much. Simulation takes the place of the lab. The effort which would have gone into lab skills now goes towards simulation and modeling. I spent a great deal of time validating the device models and working with the simulator software.
The most important tool of all isn't mentioned in the budget: a good personal computer. Many software tools that you will want are Windows-based, by the way. No doubt, you already have a pretty good computer, like I did.
It's important that you (and your spouse, if married) make a commitment to the education program. You are going to spend lots of time and money on it and the last thing you want to do is waste it. You have to consider it a fulltime job (assuming you decide to pursue it fulltime). If you can't make a real commitment to it, don't do it. Having made that commitment, you will want to:
- Put family and friends on-notice that you are fully employed during your school hours and are not to be disturbed. You need a separate room for your school work.
- Decide on definite working hours, days and holidays. I chose 8hrs/day, 5-days per week. Don't be too aggressive. Be realistic about the time you need to keep the house up, take care of finances, family and such. You're not a kid in college anymore so you can't afford to drop everything other than studies.
- Organize your school with defined course-names, written syllabi and academic quarters.
- Adopt an academic schedule. Having a formal schedule will answer all questions about when to take breaks and such, without having to feel guilty. I also gave myself two days a month off. Quarters work better than semesters for this kind of study because they are more granular, allowing you to include more variety in the coursework. Also, not having the time-overhead of conventional classes lets you get almost a semester course into a quarter. I adopted the academic schedule of the University of Chicago. After all, they invented the quarter system.
Designing a curriculum
The curriculum is a list of courses that you will need. This should be based on your Training Goal and time budget, taking into consideration the training and experience which you already have. The biggest part of the course-list will be the series which represents your “Thesis Project.” Unfortunately, you cannot start the project until you have completed a good part of the regular coursework. The coursework is supposed to educate you about how to do the design. The project is to carry that on into the real world, with all of its lessons. That's not to say that all of the coursework has to come first, though.
Be aware of the fact that projects take much longer than you think and they tend to expand. (Just like at work! 🙂 In my case, the project became so big that the courses scheduled after it never happened. I call it the project that ate the curriculum.
That's not all bad though. I was able to explore areas of chip design that were not envisioned in the plain-vanilla opamp, originally planned. Many questions were answered that could have been bypassed. In a commercial design, you sometimes work around questions which are murky, rather than taking the time to explore. My rule of engagement for such cases is to pursue if it's something that I would expect an analog IC designer to know. Of course, that has to be tempered by consideration of the schedule. Even academic designs must have a schedule.
This points to the fact that you will need to wear two hats: One for the student and one for the professor. A key responsibility of the professor (the “prof”) is to keep the classes on schedule. That often leads to tough decisions, like leaving-out a feature in a project or skipping a chapter in the book. Remember how your profs used to do that kind of thing back in school? Now you'll get to experience the worry that feels like you let the students down by not making it through that far. But time is limited and you have a course to finish.
Though your choices will probably be quite different, the curriculum that I created originally looked like this (Table 2 ):
(Click on Table to enlarge)
It might seem silly to assign numbers to courses that you are creating for yourself, but it has come in handy. There are countless times that I had to evaluate how the schedule was going or mark a homework paper. It's so much easier when you can refer to the course by a number.
You might ask what a digital course like HDL is doing in an analog curriculum. Much of the analog design work out there is for parts of mixed-signal chips which incorporate both analog and digital designs. So I wanted to have an introduction to modern digital design.
Besides, I've always been fascinated at the power of creating complex digital systems by writing code. It can make high-frequency DSP doable, which is close to analog, in a way. I really enjoyed doing a project with it, combining analog and digital. It was lots of fun. By the way: don't forget to have fun. You learn better when you're having fun.
How to build courses from books
Profs have to make a big effort when they teach a new course. Books have to be chosen, the coursework has to be planned, software acquired, exercises tested, labs worked out and tested and such. It all has to fit neatly into the academic-schedule, too. Unfortunately, the home-student/prof has to do this kind of thing too. Fortunately, there is a big secret to it: Find the right textbook. Some college textbooks are virtually courses-in-a-box.
How do you choose the book? Well, you have an idea of what the course is to cover, from your curriculum. The websites of college courses are a good source for textbook titles. Include “syllabus” in your search engine terms. You can search Amazon for titles related to your subject and will probably find lots of hits. The reviews of textbooks there are very helpful in choosing. Don't buy just one. Get two or three, in case one turns out to be a dud. Also, when something isn't clear in one or you have a question not answered there, you can turn to the others.
About three chapters into the first HDL book I used, I realized that the rest of the book was going to be awful. He was okay in the introductory stuff but used totally bizarre examples, along with other issues. I had to scramble to reorganize the course to continue with a different text. It worked out pretty well, actually. It was scary for a while, though, when I didn't know how well things would mesh. Having to deal with that was another clue that profs earn their cookies.
I highly recommend Jacob Baker's CMOS Circuit Design, Layout, and Simulation (Figure 2 ).
If you want to know why I think it's so great, see my review of it posted on Amazon, titled “It doesn't get any better than this.”, click here.
Remember that preface that we never read in school? It's actually a goldmine for nuggets about how to layout a course. Right up front, you can decide what parts of the book fit well with your needs, as it often tells you how the chapters are organized and generally how to build courses around them. For example, Baker writes, “Several courses can be taught using this book including VLSI or physical IC design (chapters 1-7 and 10-15), analog IC design (chapters 8-9 and 20-24), memory circuit design (16-19), and advanced analog IC design, (25-29).”
For my course, M501, “Basic BiCMOS Analog IC Design,” I chose Chapters 1-9 and 20-24. I felt that the physical design aspects are important and didn't need the memory design portion. You can see the syllabus here. To fit all this into a 10-week quarter, I covered the physical-design chapters at two or three per week, which was very ambitious. The others were about a week per chapter. I ran the course as 3-sessions per week.
Of course, there is no lecture. Reading the chapter is something a student would normally do. How could I provide the reinforcement which a lecture would provide? My solution was to read each assignment as if I were going to teach it. This meant understanding it completely, with every significant point, equation and figure completely understood.
For me, that meant a LOT of writing in the book. It was to be treated as a workbook. Nothing was taken for granted. I argued or restated lots of things! But Baker usually won in my imagined arguments, often after I had read a few pages more. I caught a lot of errors (which were already documented on Baker's website and corrected in a new printing). It was great fun actually; made it much more interesting, though time-consuming. Covering a chapter a week that way, in addition to the other course and the term paper was tough, at times.
Each week had written homework problem assignments and the results were reviewed with my professor hat “on”. (That guy was pretty tough, too :). Baker and other authors post solutions to the problems. If access is restricted, you can usually contact the author, explain why you need them and get access. No peeking before completing assignments!
These days, software is a big part of engineering courses. This poses a conundrum for professors and particularly home schoolers: How do you find suitable software tools when they are normally purchased in a corporate environment for big bucks? That is where books like Baker (and their websites) can help a great deal.
For example, he introduces free and low-cost tools which might not be easy to use but get the job done, at least for teaching purposes. As you will see later, though, some such tools are not sufficient for a serious project. Some universities are fortunate enough to partake in special arrangements with companies like Cadence, which make expensive engineering tools available to institutions with affordable terms. Such solutions don't help the home schooler. I was able to make a special arrangement with Tanner Tools, in a critical case (discussed in detail below).
Exams, projects and papers
Everyone asks me what I did about exams. If you want exams for your course, the textbook authors often provide those too. (I don't hear a clamor out there.) In a regular school, exams motivate students to study the material and give them feedback on how they are doing. For self-schooling, it is pretty much inherent that you are going to study the material and are motivated. Otherwise, you don't make progress through the book. You have a schedule laid out and you want to stick to it.
After all, there is the carrot of the break between quarters. Man, those 7-to-10 day breaks are great! The academics really know how to live! If you run late in your schedule, it will eat into your break. In case you might feel guilty for taking the quarter breaks, remember that it is important that you stay motivated and avoid burnout. Cramming semesters into quarters is challenging. Adding in papers and projects makes for long days and nights. Having a break every 10-weeks or so really helps.
Another reason for tests is to prove that you know the material, so you can get credit for doing the work. Unfortunately for the home schooler, self-administered tests just wouldn't have much credibility, so that purpose is defeated. The proof I will use of this education program consists of:
- The Thesis Project, with fabricated chip. That is supported by various documents, including a design review package and 500 pages of typed and illustrated notes.
- Electronic documentation of the curriculum with all the supporting documents, term papers, project papers, homework assignments and notes.
So there isn't much of a role left for exams. To fill the void, I substituted projects and term papers. Believe me, exams would have been much quicker and easier! The project reports and term papers provide good evidence of your work. They also add a lot to the curriculum.
I wanted to include BiCMOS in my M501 course. Unfortunately, neither Baker, Gray nor Allen addresses this much. So the term paper became, “An Overview of BiCMOS Analog IC Design.” It really gave me an opportunity to research the BiCMOS area.
Alas, it was too much of an opportunity. I suppose an experienced professor might have counseled me not to choose such a broad subject or might have helped me rein it in. At almost 40-pages, the paper became quite the little thesis. I did learn a lot, but it really did blow the academic schedule. I report this, so you can be forewarned not to make the same mistake.
Make sure that the scope of papers and projects is much smaller than what you think you can complete in the period. Don't let scope grow. They simply take far longer than you imagine. You can see the M501 term paper here.
(End of Part 1)
About the author
Steve Lafferty holds an MSEE degree and has been working in electronic design for over 30-years. He was previously a principal engineer at Wegener Communications. Having just finished a two-year educational program in CMOS analog design, he is now available for employment in that or a similar field. He can be reached at firstname.lastname@example.org, 770-664-6192.