Nine levels of analog integrated circuit design

Nine levels of analog integrated circuit design

One paragraph: You just started to enter this line. You only have a general understanding of what PMOS/NMOS/BJT is. You are not sure about the characteristics of various devices. You don’t have any idea what kind of circuit you design. The circuit diagram mainly looks at the articles in domestic magazines, or according to the ready-made circuits in the textbooks, you always feel that they are justified. The circuit you are working on is mainly a small-scale module, doing a point differential op amp, or a simulation of a bandgap reference, and you are calculating the article, for fear that the paper will not be enough. In general, basically see the op amp or the hairpin. You think that spice is a very difficult and eccentric thing.

The second paragraph: You start to know what is circuit design, every day holding this textbook on the draft paper mad. You also often start to mention some technical parameters, such as Vdsat, lamda, early voltage, GWB, ft and the like. I always feel that sometimes the circuit and the hand are almost the same, and sometimes the difference is quite big. You are also beginning to care about voltage, temperature and process variations. For example, low voltage, low power systems or something. Or what is the super high-speed and high-precision stuff, from time to time also come to the last two sentences. When you design the circuit, you start planning to tape out, although the tape out looks quite far away. At this stage, you think that the spice is very powerful, but it is often a big headache because the AC simulation results are wrong.

Three paragraphs: You have been fighting PVT for some time, but in general there are basically few successful design experiences. You think it's really hard to design a circuit that can really be used. You are anxious to build your confidence, but you don't know what to do. You start reading some JSSC or doctoral thesis, but you think they are talking about the same thing, the real chip is not the same thing. What do you think Vdsat's metrics are not accurate enough, the default settings of the emulator are not enough to meet your requirements, so you try to adjust the parameters of the emulator, or try to change the emulator, but the results they give are still Sometimes it is sometimes not allowed. You go to the forum and hope to get the guidance of the master. But they are also unclear, and sometimes things are sometimes wrong. At this stage, you think that the spice is very good, but the help manual is too unclear.

Four paragraphs: You have had a relatively large failure experience. You know that to be a good circuit, you need to be better, and you need to carefully check every detail. You find that there are many problems that you have never imagined in the design process. To complete the circuit, you need to fully grasp every aspect. So you begin to systematically relearn the textbooks that have been sold at the time of college graduation. You have carefully looked at the relevant information that you can find, hoping to find some more inspiring ideas. You already know the circuit specifications and performance you need to achieve, and you know that circuit design essentially requires a lot of reasonable trade-offs. But you can't figure out how this "reasonable" is determined, and how to choose between the different indicators is good. You think it's really hard to design a proper working circuit. You don't believe that in this world, someone can do what they claim to be so good, because smart people feel that they are helpless in the face of such a complicated choice. How can they do it? At this stage, you feel that the spice function is still too limited, and often stunned by the "TIme step too small" error message, and occasionally you will create huge simulation files to crash everyone and the computer.

Five paragraphs: You think that many competitors' things are just that. You start to have a familiar design approach. But you don't know how to optimize your tools at hand. You have used some scripting languages ​​compiled by others, but you can't think of awk or perl when you encounter a lot of problems. You start to take a lot of simulation time on the server. You believe that after a lot of simulation, you can clearly adjust the module you designed to the right look. Sometimes you think that circuit design is too boring. If you can't do it, you are considering whether to give up. At this stage, you think that the spice is good, but it is far worse than the fast spice series of emulators; you don't believe in AC emulation, but instead a lot of transient simulations.

Six paragraphs: You begin to understand that there are only the most suitable designs in this world, and there is no best design. You start to have a set of design methods that are truly your own. You will prefer one or two simulation tools and be able to use them to evaluate your design. You start thinking about PVT changes in your design. You know the evolution of a circuit from start to now and can tailor them to different applications. You start to pay attention to power consumption and area, and your tape out chip starts to meet some of the product requirements. But sometimes you still can't fully understand the design of some complex systems, and make some stupid mistakes and lead to catastrophic consequences. When you start reading JSSC, you should not only pick one or two articles, but perhaps using JSSC as a toilet reading is a good choice for you. At this stage, you think that spice is a great tool, you know how to simulate the accuracy and speed in the spice, and make the most suitable choice at any time.

Seven paragraphs: You begin to truly understand the nature of analog circuit design, both for high-precision systems and high-speed systems have their own unique views and experience. You can trade off different module metrics at the system level for the best performance. You will understand a potential market and start your own product definition, and you know that as long as the method is correct, the products you design will be very competitive. You can carry out the function and index division of the whole circuit from head to toe. You know every technical detail and how their compromise will affect your product. You start to pay attention to the reliability of the design. At this stage, you think that spice is a very useful tool and likes Monte Carlo simulation, but you often complain that the server is too slow, although you often run the simulation in the middle of the night.

Eight paragraphs: At this time, it is a common practice for you to make a chip. Just like a driver driving a car, you will stop at the red light and the green light will do. The design of a product is almost unconscious for you. You don't need to adjust the parameters and optimizations constantly to the simulation results. More often, you need a small amount of simulation to end the design of a module. You can clearly feel that the circuit module of a certain indicator is technically possible or impossible. You don't have to care about the noise figure or signal-to-noise ratio or distortion of a particular module. You just need to know that it can be designed, and more detailed technical indicators are meaningless to you. You start to think that the things on JSSC are actually making up the number. Sometimes JSSC is considered unqualified even if it is used as toilet paper (too thin and too brittle). You think that the spice is used very often, but it is not reliable. When you look at the work point, it is almost enough.

Nine paragraphs: At this time, you are familiar with many circuits, you can predict the development direction of many technologies in the next round. You only run a few simulations a year, or a simulation is a few years. When you rarely draw circuit diagrams, most of the time you are playing golf or fishing on a small island in the Pacific Ocean. Except for occasional fun on the ISSCC, you never talk about circuit matters with others, because you know no one can understand.