Bad Circuit Design | Hacker News Reader

Bad Circuit Design

cmosedu.com

93 pointsnvr21924 comments

Nice. This forum could use a little more circuit discussion IMHO, but I think IRL the ratio of analog-circuit blogs to web-design blogs is close to epsilon.

> This forum could use a little more circuit discussion

Yes, and specifically analog circuit design.

Digital circuits are nice and all, but in the end they're very much like coding than actual electronics.

As long as your tolerances are large enough. And you remember all your pull-ups/pull-downs. And...

the real world always finds ways to interject

To paraphrase a joke I heard before, the only difference between an Analog Engineer and a Digital Engineer, is the analog engineer knows they are working on analog circuits.

I thought it was s/analog circuits/ an antenna/, but yours works too.

My analog mentor once told me: there is no such thing as "digital circuits", everything is analog, it's just a matter of slopes. Which is correct, after spending 5 years as a VLSI digital circuit designer, most of my time was fixing min-delay and slope violations, and track routing fuckups by the auto router. And fitting bug fixes into my allotted area.

That's because a bug in hardware becomes a feature after MP and generates a number of issues or bugs or features in software. Those will eventually bubble up to web-design problems if there's a webpage involved.

Analog circuit design is very very secretive because the semiconductor industry is so cutthroat. I would definitely get fired if I posted about any of my design work on a blog. And to be fair, anything specific enough to be interesting could be copied easily enough. For a website, there is an enormous amount of visibility into the final product. You can even look at the source. For circuits, though, all you get is a tiny piece of plastic and metal with a silicon die inside. Even if you decap it, analyze it, and produce a netlist, the information is still limited and very hard to use.

If anyone has any questions about analog design in general, though, I'd be happy to answer them. Maybe a Q&A blog would be interesting enough?

I've tried a few times to get started learning, but I always hit serious snags. I've tried Practical Electronics for Inventors, but got lost in all the electrons. I got decently far in Art of Digital Design (culminates in making a PDP-8) but hit a part too hard for me and got stuck. Any suggestions?

My canonical answer to this question for the last 2 decades+ has been The Radio Amateur's Handbook. It will take you from zero to basic electronics with a healthy dose of radio theory and it is eminently approachable. I haven't read an edition in years, so I don't know how much it's changed. My recommendation would be the pre 1995 editions. Usually easy to find at libraries, but not these days...

I think I also have this book https://www.amazon.com/Analog-Circuit-Design-Personalities-E...

Or some other similar book from Jim Williams. It's not as much about electronics as it is about the engineers who came up with interesting ideas and designs.

Which brings up another idea: google Analog Devices Analog Dialog. They put out this newsletter for like 50 years and it pretty much tells the history of precision analog design.

Yeah, I'm paid to program, but I'm still an old school analog circuit design pretender at heart :-)

For a practical understanding of electronics, there's no need to go so far deep into the theory that you'd get "lost in all the electrons". I really like HyperPhysics for the basic, basic theory - http://hyperphysics.phy-astr.gsu.edu/hbase/emcon.html#emcon I would go through the Circuit Elements tree first, then the Electric Circuits. If you want to go further, there are tons of college courses with lecture notes available online. You're probably looking for something like an EE201/Electronic Circuits. Here's one I found that seemed to have all the important topics: http://tuttle.merc.iastate.edu/ee201/topics.htm Unfortunately, sometimes the lecture notes are hard to understand since they're not meant to be a standalone tutorial. You may also check MIT or Stanford for their free course versions, which probably have the entire lecture.

From there, to be honest, I'm not sure what to suggest. One thing I usually recommend to people interested in electronics is to pick up a breadboard kit/tutorial like one by Elegoo on Amazon. It's slightly more programming focused, but it covers a lot of other circuit topics and it's a great way to get started with experimentation. I find that being able to actually produce something as you're learning makes it a lot more engaging.

Art of Digital design is a 2nd or 3rd year college level textbook so you might be missing some context there. I would say if you want to go further, look into a transistor-level circuit course (which I took after EE201). That covered both analog and digital transistor circuits, which supported the digital design class I was taking at the same time.

This is my favorite circuit book:

https://www.amazon.com/Analog-Circuit-Design-Personalities-E...

It is written from zen-level experts in a very ELI5 way. The examples start off with a naive approach, and step by step explain how they were improved, in some cases over decades.

Plus it focuses on how analog really is an art. Circuits have so many competing design parameters, and the canvas is so wide open, that it really takes a different kind of brain to approach the problem. This book gives a peak into that process in a comical, light-hearted way.

“Look at how offensively huge this oscillator is!”

— http://www.wastedtalent.ca/comic/matter-scale

Great webcomics feel like one of victims of our new internet...

Some chump has run the data lines right through the power supply! Amateur hour! I've got tears in my eyes!

I am a giddy goat.

The main circuit confuses me, but apparently that's because I haven't studied CMOS design much.

What's going on is that CMOS chips have accurate capacitors, but inaccurate resistors. So the "switched capacitor" converts capacitors into resistors by oscillating the switches Phi1 and Phi2 at a certain rate.

However, the switched capacitors prevent the feedback loop from properly forming on the OpAmp.

http://cmosedu.com/cmos1/bad_design/bad_design.jpg

If anyone else has issues understanding the designs... remember that this seems to be tailored for CMOS circuits, not ordinary circuits that you'd make on a breadboard.

I guess its a "leaky abstraction". The switched-capacitors are supposed to be "thought of as resistors", but because it cuts off the op-amp feedback, you cannot use them directly as resistors in this case.

The full discussion of this circuit is here: http://cmosedu.com/cmos1/bad_design/bad_design6/bad_design_6...

Very enlightening for sure. But I'm not entirely sure how one would even get into CMOS design as a hobby. So this sort of stuff remains the realm of textbooks and theory for me. Just gonna stick with simpler breadboards for my own hobbies...

To add a bit of context for anyone interested, there are some interesting reasons why switched capacitor circuits are powerful in integrated CMOS design. In terms of matching, the choice between resistors and capacitors is due to size and matching. Large values of resistance require large resistors or can be implemented with smaller capacitors, switches, and a clock (https://en.wikipedia.org/wiki/Switched_capacitor#The_switche...). Resistance values which would be prohibitively large to get with a resistor in modern processes can be achieved with switched capacitors. Additionally (and more importantly), capacitors can have better matching due to their geometries. If we want a gain of 10 on-chip, switched capacitor circuits are very well suited. Switched capacitor circuits obviously can't be used everywhere (need a clock, not continuous time), but they provide some good advantages.

Considering a resistor as a sheet with a length and width, increasing the length or decreasing the width will increase the resistance. A large resistance means a long and thin material. For a capacitor, we just care about the overall area of the plates and normally use squares. There are variations to the dimensions generated for both of these shapes. The long, thin resistor is more sensitive to these absolute changes in its width and thus the best matching is reduced. The absolute values of both resistors and capacitors are poor, but the matching of capacitors can be as good as 0.1% (enough for 10-bits of accuracy). http://ims.unipv.it/Courses/download/AIC/Layout03.pdf

There are a few switched-capacitor ICs, though they're more for boutique applications at this point. https://www.analog.com/media/en/technical-documentation/data...

This website assumes you've studied CMOS circuit design for analog circuitry (amplifiers, analog-digital converters), and have the corresponding course book. It's a good list of when the linear model breaks down and you actually have to do the analysis.

I really like #3, placing a capacitor on a high-impedance node.

javascript:[...document.querySelectorAll("*")].map(x=>x.style.fontFamily="Helvetica")

https://designblog.rietveldacademie.nl/wp-content/uploads/20...

Anyway, the blog was about "bad circuit design", not "bad design".

Meanwhile, on every page maintained by the OpenBSD devs: https://3.bp.blogspot.com/-Y2lMqbfEn8E/T-aJSC76_6I/AAAAAAAJC...

Apparently I don't have Comic Sans installed on this machine, though. I should probably fix that.

ehy, leave comic sans alone!

Reply on news.ycombinator.com