Control Transformer,Switching Power Supply Transformer,High Power Switching Power Supply,Dc Negative Ion Generator Huizhou Show-Grand Electronics Co., Ltd. , https://www.sgtransformer.com
It’s a must for analog technology. It’s not a joke.
I want to share my personal growth journey, especially how I improved my technical skills. Let me briefly introduce it. I've been working on audio DAC/ADC, specifically sigma-delta DAC/ADC. In the audio field, DAC is more critical than ADC, so the main focus has always been on DAC. Currently, the performance of DACs has surpassed 100 dB dynamic range, with very low power consumption and a small area. Compared to the company's previous 70 dB design, not only did we achieve over 100 dB, but we also significantly reduced the area and power consumption, surpassing well-known foreign IP suppliers in terms of performance, power, and area.
You might ask: when you buy an MP3 player, is playback or recording more important? That’s why DAC is more crucial than ADC.
I'm not a big name in the industry, but I have some feelings and want to share them. I wrote this, and I laughed while doing it.
When I was in school, I wanted to be a programmer and tried to understand programming (which is actually very important for analog circuit design engineers, as I’ll mention later). After graduation, I became an analog circuit designer, even though my major wasn't computer science or microelectronics. When I first joined the company, I knew only a little about bandgap reference design. I didn’t even understand AC/TRAN simulation, nor did I know that circuits needed TRAN simulation. I had no idea what Miller compensation meant. Looking back now, I laugh at how clueless I was. When I first joined the project team, the supervisor was busy with the project. When a bottleneck arose, he didn’t like to bring people along. I felt useless and was given a task: study Class-D, which was quite popular then. I studied it thoroughly and became an expert—no one had done it before. So I embarked on a sheep-like career for over half a year, during which most of the foundational knowledge was built.
What did I do during those hours? I read Razavi and Gray’s books, found they weren’t related to Class-D, searched for papers online, read nearly 100 articles, but got little clue—there's too much garbage in the literature. Eventually, I found a good Class-D paper from TI, which explained things clearly. I tried to build the circuit and started simulations. Later, I'll talk about some learning methods I found really useful for analog circuit design.
Opportunities are still very important. What matters is your ability to seize them. Life is like that—what time to do what, if you miss it, there won't be another chance. It's very important to cherish the present. I've been working for more than three years now. The IP provider is ChipIdea, which is now acquired by MIPS. Class-D was once hot, but now it's just a common topic. Many people misunderstand that Class-D performance isn't as good as Class-AB, and it's not comparable to Class-A. That’s wrong. Class-D can still be excellent. If designed well, a 96dB dynamic range is easy to achieve, and harmonic distortion is easier to control than Class-AB. ADI has the AD1991, which performs very well, far better than average Class-AB. But I think its architecture could be improved to achieve the same performance with lower power consumption, which would shorten the development cycle. The key is the architecture. In fact, all circuits follow this rule—the key lies in the architecture.
There are four key factors in circuit design:
1. Performance
2. Power consumption
3. Area
4. Development cycle
The trade-off between these four is determined by the architecture.
A solid foundation in simulation is essential. The basics include signal and system theory and circuit theory—these two undergraduate textbooks are the basis for simulation.
Deep understanding of basic analog modules like current mirrors, two-stage Miller OTA, folded-cascode OTA, etc., is essential. It's not enough to just read Allen’s book and understand the formulas; you need to figure out aspects the book doesn’t cover. Even the content in the book can be reinterpreted or corrected. This process requires strict theoretical understanding, intuition development, learning from others, and independent thinking to form your own design method. Real innovation in product development comes from beyond textbooks.
On top of that, building intuition and experience for analog design is extremely valuable. For example, knowing how much Vdsat a current mirror needs, how much area it takes, and the pros and cons of different circuit structures. A deeper understanding helps you achieve required performance with the simplest and most reliable structure, which is very important.
Accumulating engineering experience is also vital. You learn what simulations can't predict in actual processes, which may not be circuit-related but module compatibility issues.
Further understanding of the circuit through problem-solving leads to innovation. In practice, you see many simulations and papers, but when you try to implement them, you find that current mirrors often have Vdsat in tens of millivolts, and differential pairs work in reasonable states, but active loads may not. These are real-world challenges.
Many people use complex structures for simple tasks, ignoring area and efficiency.
When I first studied Class-D, I couldn't grasp the principle. How can a square wave represent a sine wave with low harmonic distortion? Starting with an open-loop structure, using the ideal PWM generator, it's still hard to get a clean waveform. What about analog non-idealities? Some literature said it was time to do it. I thought, how is that possible? Won't that create nonlinearity? Without dead time, power consumption is high, and EMI is hard to manage. It seemed contradictory and impossible to compromise.
Switching circuits simulate slowly. Simulating one cycle can take half an hour, and accuracy is low. If I had an idea, I'd spend at least half an hour verifying it. In reality, it took longer due to convergence issues in HSPICE, which often failed to converge. It was exhausting. Later, I found Spectre had better convergence and used it going forward. But convergence problems still occurred. I studied a book on Spectre simulation (written by the developer) to understand how to make the simulator converge. This is my style—when I don’t understand a problem, I go to professional books and study them seriously, even if it takes a lot of time. Understanding is the key. After that, I rarely had convergence issues, and even if I did, I could solve them quickly.
The contradiction in Class-D wasn't resolved then. I eventually realized my idea was wrong. I kept thinking about the problem from the time domain. Then, I found the direction from literature: solving problems in the frequency domain and focusing on noise shaping.
So I started studying signals and systems. At the time, I felt that many people saw signal theory as important, but few truly learned it. Most read through it in 3–7 days without applying it. Signal and systems were too theoretical, and I struggled to apply it. I thought Class-D was an analog filter, so I studied filter books and used Laplace transforms to derive transfer functions. Slowly, the concepts of signals and systems began to form.
But I still didn’t understand noise shaping in Class-D. I learned that the key was noise shaping, which I discovered through sigma-delta modulators. Sigma-delta is a mature theory, and few people study it now. Why not start with sigma-delta and then move to Class-D? I looked up information and found a colleague who collected analog books. He recommended “Delta-Sigma Data Converters – Theory, Design and Simulation†by ADI. I call this the "sigma-delta bible." Most of my sigma-delta knowledge came from this book. Another important book is “Understanding Delta Sigma Data Converters,†also by ADI.
If someone asks me, what’s the best way to learn analog circuits? I’d say starting with sigma-delta ADC/DAC is best because it teaches you about signals, transfer functions, modeling, and noise—core elements of analog circuits. If you want to learn sigma-delta, I recommend these two books. The reason I succeeded in this field is because of these two books.
But it’s not enough to just finish these two books. Absorbing the knowledge is difficult. Moreover, circuits aren’t textbook perfect. Reality is often unclear. Papers have errors, and you may face unreported issues. You question whether your derivation is correct. Many wonder after completing a circuit: will it work? Will it meet requirements? Are there bugs? Have I missed something?
The biggest difference between analog and digital circuits is the uncertainty in simulation. Software runs, FPGA logic works, but analog circuits don’t always behave as expected. Simulations are slow, and you can’t simulate real situations. It’s hard to have 100% confidence before tapeout. Confidence in analog comes from judging which simulation results are credible.
Real improvement comes from debugging existing chips. At the time, the company had a chip under debug with many issues. One was a problem with the sigma-delta DAC. I noticed the output had a triangle wave—a serious issue. I studied the design document, modeled the structure, and used MATLAB for modeling. I spent a golden week reading “Digital Signal Processing Practice Method,†which helped me understand key concepts. Combining MATLAB made it especially suitable for sigma-delta modeling.
This field is strange. Analog designers think it's digital, and digital designers are too busy to care. I had time to learn and found it interesting. With MATLAB as a tool, learning was fast. I wrote code, optimized algorithms, and studied the book until I saw the phenomenon described in it. Modeling showed the triangle wave could be reproduced. I demonstrated the model to leadership, saying the problem was likely here, but I needed time to find the root cause.
What matters isn’t strength, but investment. My most rewarding experiences were times when I invested the most. The return was disproportionate—80% effort might give 20%, but adding 20% more could yield 80%. As the saying goes, the hunter gets half the ninety.
The problem’s phenomenon differed from the book, even contradicting it. Through modeling, I guessed the possible cause and simulated the circuit. I studied Verilog-A, DFT algorithms, and methods to improve simulation accuracy and speed. Sigma-delta simulations were slow—sometimes taking a day to simulate one case. I almost gave up after two months of simulation, but finally reproduced the issue. The performance didn’t match tests, and I thought I hadn’t found the root cause.
New projects started, and this module was chosen by no one, but I was eager. I wanted to find the problem before designing. It was risky, but I persisted. Finally, I saw the waveform and identified the issue. From a signal and system perspective, it matched the test results. I built a new model, and simulation confirmed the consistency. Adjustments in SPICE fixed the issue. I achieved full consistency between theory, modeling, testing, and simulation.
The most exciting part was confirming the root cause. I was confident, though worried about undiscovered issues. Analog is fuzzy, and simulations have blind spots. Fortunately, the modified circuit worked. The problem turned out to be a small mistake—a single line adjustment. But it led to significant gains. The key in sigma-delta is noise. During this time, I gathered data, refined simulation methods, and improved noise calculation confidence.
Low power means understanding noise. Knowing noise lets you simplify circuits, reduce non-critical power consumption, and plan critical parts. There are many innovations possible, including structural ones. After debugging, I worked on the DAC, making comprehensive modifications to reduce noise and plan power consumption. Despite tight timelines, the circuit was completed quickly, and the project met its goals.
Though the project failed for other reasons, I felt confident. Others doubted me, but I believed in my design. Looking back, I was surprised—it was my first circuit, yet I felt like an expert. No experience, but I vowed my circuit had no issues and could reach a high level.
I didn’t expect the post to add value. I’m stubborn, always wanting to learn more. I dislike half-understanding, so I invest time. This slow approach taught me more than others. After finishing the project, I returned to a Class-D project. I defined the top structure, calculated signal and noise transfer functions, built a Veriloga model, studied non-idealities, added a spread spectrum algorithm, and verified it. This shifted the design difficulty to the system level, similar to transferring analog to digital.
Class-D turned out to be simple, achieving low power. I lost interest in it, seeing it as a fully proven topic. Later, the circuit was handed to colleagues, and it performed well. In contrast, TI’s Class-D was only average. To sum up, Class-D uses noise shaping, simpler than sigma-delta ADC/DAC, with no real digital-to-analog conversion, making it easy to implement.
Original post address: http://bbs.eetop.cn/thread-141642-1-1.html
Finished address: http://bbs.eetop.cn/thread-313574-1-1.html
Watch EETOP, enter **one hundred** chest backstage view Recommended Reading:
Analog digital product development process
A brief history of SPICE
Dry goods sharing: measuring your own ECG (from theory to circuit)
Good analog IC engineers should have the literacy
Classic in the field of analog IC design
Do you need post-simulation simulation?
I see the pole zero
Six classic analog IC books wonderful comments and summary
100 Bibles designed in simulation
Suggestions for getting started with analog circuits
Analog IC film experience sharing
Simulated IC annual salary of hundreds of thousands of brothers' model learning experience
What should I do to become a simulated ic designer during my undergraduate degree?
Nine-level advancement of analog circuit design
Where is AnalogIC difficult, structure? parameter? territory? system?
Analog IC Design Part 2: Transfer Function, Zero-Pole Formation and Time Domain Response
I still don't want to continue to learn analog design - a question of inter-disciplinary master
How to learn analog IC design
Simulating the big cow to talk about the price and development direction of the simulation engineer
Analog circuit should know 200 questions
Several practices and advantages and disadvantages of post-simulation simulation
Very interesting explanation about analog land and digital ground
Dry goods! In-depth explanation of the classic lecture notes on RF analog circuit design
The Road of the Modem--The Analog Circuit Learning Feelings
Is analog IC design really hard to learn?
Analog IC forward design process summary
Simulation learning sentiment and summary
Analog filter design handout
Lazard Analog CMOS Integrated Circuit Design Lecture Part3
Lazard Analog CMOS Integrated Circuit Design Lecture Part 2
Simulation learning sentiment and summary