# EE Approach to Circuit Design

1. Apr 9, 2009

### e-o

Hi everyone,

I have a question for anyone who considers themselves adept at designing circuits. I was wondering, how do you approach the design of complicated circuits? What stages do you go through to get from a concept to a real working circuit?

I myself am not an electrical engineer, nor do I have a great understanding of circuit theory. I'm interested in learning more, but when I look at some complicated circuits (like the interior of an op-amp), they essentially appear random to me, and there doesn't seem to be anything I could learn that would let me figure out how the circuit works.

In my experience, designing a circuit goes something like this:
1) Figure out the objectives of the circuit (what am I trying to build?).
2) Block out the different stages of the circuit (I'll have an oscillator, it connects to a pre-amp, the pre-amp feeds through a filter etc.).
3) Start looking through my textbooks/online for designs for the different stages of my circuit (I'll use an active low-pass filter here, and an astable 555 time there etc.).
4) Sort out all of the actual component values. A large part of my effort here always seems to be trying to match impedances (I'm not sure if that's typical).
5) Make sure the circuit works properly, fix any problems that come up.

However, when I look at something like the interior of an op-amp (it's not just op-amps of course), my approach doesn't seem to work. There are so many unusual connections, which tend to loop-back or overlap one another and it seems to defy my 'linear' (one stage after another) design approach.
I also have problems with my step 3. Some circuits seem simple at first, but their behavior is so odd that I wonder if anyone deliberately came up with the circuit, or if it was an accident haha.

So anyways, if you managed to read through all of that, I thank you. And I'd like to hear your input as to how you go from a concept into a working circuit.

2. Apr 9, 2009

### rrogers

1) Yes, but the first time through is just a sketch/estimate of the objectives. Unless you work for a large company with money to burn; or are upgrading a known application you will be coming back to the specification stage several times, williingly or not.
1-a) You have to interpret the requirements and specification into terms that make sense in terms of current engineering terminology. Hopefully finding regulations and standards.
2) You have to imagine several implementations and probe their weaknesses and strengths. The imaginations are typically loose; the probing should be honest. The block diagrams are mostly useful when the design can be interpreted in terms of signal processing; which is quite often. Let me rephrase that: block diagrams allow you to (and are almost worthless unless you do) define a sequence of interfaces crisply. There are circuit design cases where block diagrams are not relevant. In particular single stage complicated design problems are not amenable unless you generate signal flow charts. They are basically only good for linear systems; although there are also exceptions.
Now If you have a really complicated system you do need to "divide and conquer"; i.e. generate block diagrams. They block diagrams are at the front; the signal flow diagrams (for analog circuits) are done after a topology is selected for evaluation.
3) A good idea, but make sure you don't take somebodies idea without a complete analysis. The applies even to reference designs. "don't trust anybody over thirty" should be "don't trust anybody"; including yourself. Paranoia is a designers friend.
4) Yes but you should also be doing it with respect to detailed analysis and modeling. With contemporary technology modeling can save a lot of physical prototyping. It makes you look better when your design works the first time; and it produces nice output for management review. The truth is I have found computer modeling great for providing insight but in the end creativity is required to apply the insights.
5) Do try to find the limits of the circuit through prototyping. With insight these should be few and identifiable.

Engineering, hardware or software, shouldn't be an experimental science in this day and age.
For analog circuit design remember that over 50% of the performance is in the physical implementation! Thermal aspects, trace impedance control, grounding,
The op-amps are a wonderful case that you should understand. The designs you see are the result of 50 years of experience and brilliant design development. Present designs are very sophisticated. It is worthwhile to analyze the circuits, but start at the beginning. Go back and look at the (now) ancient application notes, design notes, and magazine articles. It might be hard unless you have a good engineering library available.

BTW: Good circuit designers are sometimes a pain; arrogant, paranoid, and a bunch of other things. Also there are a bunch of designers around that think they know how to design but really don't. Use your own mind carefully all the time. Won't make a lot of friends (doubting them) but your circuits will be better. But never pass up a chance for a design review; the good suggestion statistics are dismal but I have found that every review brings up one or two points that are worth addressing. A reset of the robot on mars should make the designers embarrassed; although the robots have really done well since.
And oh yes don't take circuit design up for a living unless you really enjoy it. You probably won't get rich or really appreciated (outside of select community) so the work should provide sufficient satisfaction. Analog circuit design is really an art form that is not very public:)

Ray

3. Apr 9, 2009

### Bob S

First I will draw a big rectangle. Then I will mark on the perimeter; (left side) all the inputs ; (right side) all the outputs; (and on the bottom) all the power supply requirements. Then I will draw smaller rectangles for the functional components (e.g; oscillator, preamp. mixer, amplifier, etc.) inside doing the same thing, all while minimizing the connections beteeen the smaller boxes.
Then I will draw the detail circuits for all the smaller boxes.

4. Apr 10, 2009

### Staff: Mentor

Welcome to the PF, e-o. I think you'll find it a fun and useful place.

You ask a good question. Are you in about first or second year at a university? (or heck, a very advanced high school student?) There are several aspects to your questions that I'll try to address:

** As said already, the first step in circuit design (or design specification) is to draw a block diagram, and specify the interfaces and performance specs of the blocks and the overall system.

** As a designer, you then go down into each block, and work on the design of that block, either in schematic form, or in Verilog/RTL, or in C, or in C++, or etc.

** In the design of each block's contents, you usually use simulation as part of the design process. For analog circuits, you will use a version of SPICE. For digital circuits, you will use a digital simulator, usually one tied to the digital (Verilog/RTL or other HDL) compiler that you are using. You will write "test benches" and "test vectors" to verify that your digital logic blocks are working correctly, and you will use your SPICE simulations and Monte Carlo simulations to ensure that your design is centered optimally.

** As for understanding why opamps are built the way they are, consider taking an intro analog electronics / IC design class. They will cover the basic building blocks of analog IC design, which will help make a lot more sense out of the "simplified" circuit diagrams of opamps. Understanding fundamental building blocks like matched Diff Amps, Emitter Followers, Cascode stages, push-pull output stages, and feedback, will help you a lot in your quest to understand basic circuit design.

** There's a pretty good intro EE book that I recommend often -- "The Art of Electronics" by Horowitz and Hill. Borrow a copy from the library, or buy a copy, and read it cover-to-cover. It will answer many of the questions that you are asking yourself right now.

Finally, to offer a counterpoint to one piece of advice offered so far in this thread... If you are good at circuit design, have a good fundamental understanding of how things work, are good at applying the fundamentals to unique product design situations, and are willing to work hard at a startup, then you can definitely translate all that hard work into good financial success. The best circuit designers earn their keep, and they are rewarded well, in general.

BTW, if you'd like a copy of the outline of my talk on "Mixed Signal ASIC Design in the Real World", PM me with your e-mail address, and I'll send you a copy. Same goes for others reading this thread.

Last edited: Apr 10, 2009
5. Apr 10, 2009

### waht

Some schematics are like that, where there are cross and overlaping connections drawn all over the place. There was even a contest once, where you had to come up with the most elusive schematic, and there was another contest to come up with the most undecipherable software.

6. Apr 10, 2009

### e-o

Wow fantastic replies, thanks so much!

rrogers - Thanks for such a detailed reply. Looking through older op-amp designs sounds like a great idea, I never even thought of that. I'll definitely be having a look through the library.

Bob S - I have this strange image of a 'fractal circuit' after reading your post haha. I'm glad to hear that block diagrams are a standard approach to beginning a circuit design.

berkeman - I'm actually in my third year of a program called Engineering Physics. But my knowledge of circuits is probably somewhere between a first or second year electrical engineer.
Can you expand upon how you make use of C or C++? I'm familiar with using simulation software (MultiSim), but I've never heard of using C or C++ (is there a particular advantage to programming it yourself?).

waht - Haha, that's awesome.

Again, thanks so much for the replies.

7. Apr 10, 2009

### Staff: Mentor

I was mostly referring to how you might implement a block as a microcontroller (uC), as opposed to digital or analog electronics. For example, with a circuit called a Programmable System on a Chip (PSoC), you can integrate analog, digital and uC blocks on a single chip:

http://www.cypress.com/psoc2/?id=1353

And when you are architecting a device or system, you will look at different ways to partition the functions in your block diagram, to find the optimum cost & performance mix. That may entail pulling some digital functions into a uC, and even some analog functions into a uC (in the form of DSP), instead of implementing them in analog electronics.

8. Apr 10, 2009

### Bob S

In a sense you are right. The objective of designing subcircuits is in part to divide the whole circuit into parts that logically are minimally interdependent units, so each subcircuit may have only one or two inputs and one or two outputs. Also, compartmentalizing the overall design into small boxes is absolutely necessary. An early circuit I designed (at HP, ~1955) was a very low noise 420 MHz receiver amplifier with a voltage gain of about 2 x 106 and with a down-conversion to 30 MHz. Each amplifier stage was a grounded-grid amplifier (6AM4) with the middle of the tube sockets literally being divided (cathode and filament side, plate side) by the aluminum sides of the subcircuit boxes to keep the thing from oscillating. So in a sense, it is fractal.

9. Apr 10, 2009

### waht

BobS, HP was the ultimate test equipment company at the time. At school, we studied alot of HP schematics that were freely available, of various measuring equipment, and analyzed why they were designed that way. Some of the designs are really incredible considering the time they were conceived.

10. Apr 10, 2009

### rrogers

I agree, coming in at a later time 70's I found the designs and techniques to be quite instructive and inspirational.

Ray

11. Apr 10, 2009

### Bob S

I worked at HP from 1954 to 58. For a while I worked on their first oscilliscopes, including the model 150. It was 100% vacuum tubes on printed circuit boards. My first glimpse of a Tektronix scope was in the HP development lab, all disassembled on the work bench.

12. Apr 10, 2009

### Corneo

Agilent still has some great designs but everyone I talk to tells me its no longer like the way HP was.

13. Apr 10, 2009

### waht

HP entered the oscilloscope business decades after Tektronix, so it would make sense they would try to reverse engineer it. However as Ray pointed out, designs coming out since the 70s are of interest. The company produced one of the highest quality equipment. Even today, you will find a 30 year old HP analyzer on benches in labs that work flawlessly without any maintenance.

I hear the same, ever since they took over the test equipment division, they killed the "HP way" which spawned many innovations, plus they outsourced everything.

14. Apr 11, 2009

### rrogers

I wish I could remember the name (or even the function of the machine); but there was one instrument that ran around sampling internal errors, saving them on capacitors, and then using the error voltages to correct subsequent operations. It looked incredibly complicated (switches flying all over) but when you delved down it was a brilliant way to economically build a very precise instrument. I got a chance to use it on a touch pad; really worked great. You applied voltages every which way to the pad then sorted out the saved (on capacitors) voltages to figure out where the fingers were. Worked great as long as you were carefull about leakages and such.
I sort of despaired when HP started treating the instrument division as an unwanted step-child to the information processing (accounting in my opinion) division. But they turned into Agilent who has introduced some instruments that are impressive. I am quite impressed by the current generation of RF/spread spectrum/CDMA (and so forth) instruments. These are quite difficult things to do. Those and the scopes that go over 10Ghz. When I started at Beckman in 1965, simple measurements over 1GHz was quite an effort. I remember we figured out how to do automatic arithmetic reference oscillator frequency compensation for over 1GHz frequency measurements one Friday; we came in on Monday to find that HP had just written up the technique in a magazine. Glum time for a while.

Ray

15. Apr 11, 2009

### waht

HP pioneered sampling oscilloscope technology, and rolled out a world's first 18 GHz oscilloscope in early 70s or late 60s. We also looked at an RF deck removed from a spectrum analyzer. The design was more of an art then mere mixers and meandering semi-rigid plumbing.

16. Apr 12, 2009

### e-o

Thanks berkeman, that makes more sense. I thought you meant using C to help solve the equations relating to how the circuit operated. Thanks for the link too, I've never heard of PSoC's before.

Wow, you guys have some really interesting backgrounds.