"logic is encoded using the phase of oscillatory signals, rather than voltage levels. Such phase encoding has long been used in radio communication for its superior noise immunity; we show that it can also be used for computation, using self-sustaining nonlinear oscillators as underlying logic elements."
Reminds me a little of the two labs I've worked in in the course of my gravitational wave interferometry research in the LIGO collaboration. Analog electronics are still (usually) far faster and far less noisy than digital as a way to implement feedback loops and frequency modulation/demodulation, and most of the best op-amps are 70s or 80s designs (e.g. TL074, OP27, LT1028, etc.). Our labs are full of them, and, they work at least as well as modern equivalents, so why not keep using them?
The LF411 is more than 30years old and can be understood as an upgrade to the jellybean 741, but that one is quite shoddy by today's standards. With the JFET inputs and corresponding low bias current, it comes much closer to the ideal OpAmp than a 741, but there's little else speaking for it (well, it's cheap too). LT1028 OTOH is still the OpAmp with the lowest voltage noise afaik (which it buys with fairly high input currents and input current noise), it's still available, but it ain't cheap.
There has been progress in other regards (lower supply voltages, lower power consumption, rail-to-rail operation, smaller packages), but clearly progress in the analog realm over the last forty years has been much more modest than in the digital.
Class D amplifiers have benefited a lot from better FETs [1].
But note that class D amplifiers are mainly applicable to relatively low frequencies. Perfect for audio, but not adequate above this range, see e.g. [2].
The TL07x series op-amps along with quite a few others have been in continuous production for over 40 years. They are still manufactured by the million. When is the obsolescence you refer to planned for?
I love looking at old processes like this and seeing how they influence modern methodologies. The tape out in particular is really interesting, and reminds me of the old school photo and pasteup techniques[1] that graphic designers used to use.
It makes me a bit sad that it's getting harder to find places to get hands-on with these old manual processes, because it's a great way to wrap your head around the fundamental concepts of the thing you're learning. I'm younger than the author, but well into the internet age we were still taught manual pasteup and letterpress on old Heidelberg platen presses in our high school graphic design courses and I draw on some of that process knowledge even today.
It also feels like those experiences were really helpful for learning to think systematically and predicting how the consequence of the design decision I make now will affect me down the line. Today if I screwed up a plate it would take me a few minutes to fix; back then it would've taken literally hours and, if we were running business, cost a ton of money in lost production time.
I love looking at old processes only to think "Boy...we've lost the ability to make mechanical things well because almost everything (barring servos, actuators) is now software controlled". In electronics, it is dumbed down to reading a 870 page datasheet about an all-in-one-mother-of-all-peripherals SOC. No more discrete circuitry, no more logic chips, none of that.
Note that I am not saying those solutions were better, but there was a lot of hardwork, talent and intense discipline in making stuff back in the day. Drawings were hand drawn! Some guy was bending down for 3 days straight to make that schematic!
It's all too viscerally physical and mechanical. I want a time machine to go to the 70's for a bit, and furiously document and interview people. Lot of this knowledge and expertise is lost because we don't need this stuff anymore.
Remember things when they had toggle switches and knobs? What happened? We have touch screens that ask us to scan the QR code of the water filter before dispensing water from the Fridge. Oh, but it shows today's weather on the door. So great.
On Hackernews, we are such an intense bit of people - we should bring some of the old tech back and simplify our world. If you're in software, build flat, simple and fast apps that are maintainable. Make them composable and modular. If you're in hardware, build high quality stuff for twice the price but then convince the user through honest & truthful marketing that this product is better for you in the long run. Instead, the rush these days is to market products with abhorrent psychological tricks and backstab users with all kinds of dark patterns. Shame on you guys. When you're on your death bed, none of this will matter. Do good things and build great stuff. Atleast I will reflect back that I made the users, you know the ones that provided me with a living wage, happy.
The mechanical things are still there, and it is even easier to make them. Unfortunately, it turns out that they are expensive, so they have been relegated to either very expensive devices or to beginner's devices. Take a look at high-end scope or ultrasound machine control panel -- they are full of beautifully designed knobs and switches. And on the opposite end of spectrum, if you look at instructables/hackaday, there are tons of project to interface anything mechanical.
Same goes for the discrete circuity. Get one of those cheap plastic solar lights -- you may find there is no controller in sight, just a bunch of old-style transistors. And on the other side, no SOC will have a 1GHz AFE or 20 amp output.
Toggle switches and mechanical interaction in general was abandoned because it costs more: not just the parts, but having a human assemble them into the chassis.
>Shame on you guys. When you're on your death bed, none of this will matter. Do good things and build great stuff. Atleast I will reflect back that I made the users, you know the ones that provided me with a living wage, happy.
Users are welcome to pay the real price of the product if they don't want it to be spyware.
At any rate, if you're thinking about work on your deathbed you have more things to worry about than how happy users are with the junk you made while living.
> Users are welcome to pay the real price of the product if they don't want it to be spyware.
Once upon a time there were many profitable business models for selling software.
Then software got big. You weren't selling a thousand copies anymore, now it's a billion.
The economics of software is that the price you "need" to charge goes down when the number of users goes up. If it costs ten million to develop and you have a thousand users, they each have to pay at least $10,000. If you have a million users they only have to pay $10.
If you have a billion users they each only have to pay $0.01.
You can get more than the $0.01 from each user through advertising. What you can't do is get an amount on that scale through payment processing, because the transaction overhead is larger than that.
So now you have a user that the advertiser is willing to pay $0.03 to advertise to. The users themselves are willing to pay $0.05 to not be spied on, but you haven't got any efficient way to get $0.05 from them. It'd have to be much more than that to justify the payment processing overhead, and the user isn't willing to pay that much.
So it's not a matter of the users being unwilling to pay, it's that we don't have an efficient way for them to pay that small of an amount.
Maybe if it wasn't locked to a small area, and had people willing to back it long term, it wouldn't fail?
A common Google problem is that for all that they are a global company, they can't launch shit globally. I was really interested in using Google Contributor, both sides, but limited launch meant that I could only look from far away and never ever encounter a site using it. Hell, I think some of the engineers who worked on it weren't in the very small group allowed to use it.
The real price of software includes the price of delivery.
Users would rather not pay that, so we are stuck with craptastic spyware that is selling their data to google, facebook, the NSA, the Chinese communist party and that organ harvester from Moldova.
> The real price of software includes the price of delivery.
The real price of delivery of software over the Internet is zero.
(Well, it's slightly above zero, because all the networking equipment uses electricity, but you pay for your part of that in your power and Internet bills.)
AWS, which charges ridiculously high prices for bulk bandwidth, charges ~$0.15/GB, and half that if you have high volumes. If your software is 50MB, that's less than a penny. Meanwhile, if even this is too much for you, BitTorrent exists.
Point is: it’s important to keep the grand picture in mind and once in a while step back and think - “Shit...what am I doing?” even though you can’t do much about it due to family, responsibilities and other hard realities of life. When building a company, I don’t want to just build something that deceives people, deploy shady tactics and dark patterns. I want to run an honest business that I can reflect back in life and say “yep, it was a struggle but I did the right things.”
This reminds me of my time spent working as an engineer in BBC television in the late 80s. There was a lot of kit designed and built by the BBC design department. It Was just so well designed, built and documented.
In the 90’s it all started to go digital and less fun to maintain
Yes, I remember them building one of the first digital frame stores at Kingswood Warren, must've been early to mid-80's - huge 14" disk packs with the electronics to digitise/recreate the analogue broadcast-quality video signals ("ah, the smell of a burning disk pack in the morning" - after a head crash).
Eventually installed at Television Centre, together with our PDP-11 driven motorised rostrum camera.
Wow, these photos are beautiful and I really enjoyed reading through this. They capture a rich piece of technical history that most people wouldn't normally get to appreciate.
Also, I didn't understand why Google Photos was in the title at first, but I actually really like this presentation format. It's kind of like a super clean blog post.
Can anyone recommend a good intro to analogue circuit design? I've dabbled enough in digital circuit design to have a rough intuition for how to do interesting things with logic gates. But I don't have the faintest clue how even the most basic analogue circuits work, eg. a variable-pitch tone generator. I have no intuition for how you could build an oscilloscope with a fully analogue signal chain.
I’ve recently enjoyed a three part course on analog electronics on Coursera. It assumes no background in electronics but you do need some calculus (it’s adapted from a class at GeorgiaTech, apparently). The courses move from linear DC circuits to linear AC circuits and finally to diodes and transistors.
Look for the instructor „Bonnie Ferri“ on Coursera.
Read as much as you can from Robert A. Pease. He was the guru to contact when it came to analog design. The column he wrote for Electronic Design can still be accessed on line. If your luck you can find one of his books used. Look him up on Amazon.
I switched my major from EE to CS, but I did take a few courses in EE. It seems analogue circuits are more about complex signals (the type you’d analyze and describe with Fourier and Laplace transforms) and less about simple voltage switching (though op-amps can get you voltage switching if you need it as an output).
Physics is the problem that nature has thrown at us. Technology problems are something that we create to solve real problems. Unfortunately, we lost the real problem solving part somewhere and the biggest technological problem we have now is how to keep eyeballs throughout the globe glued to your app or service.
Of course, it will feel crappy in solving that problem.
I mean it is insane to me on a rational level that an embedded engineer/programmer working for a company designing avionics software/train control software in the UK/France or Germany earns less (by half) what someone in SV does working for some start-ups.
I get the whole "they are paid what the market determines they are worth thing" but intuitively I find it weird.
If I could do my life over again I'd have gone to Uni and done electronics and then moved into that kind of field, not for the money - it's about the same but because it is interesting and you get to point at a train or plane and say "it's safer because I did my job".
Humans are amazing, the whole planet full of life is amazing. And instead of exploring everything the universe has provided for us, for itself, we have created a shitty prison for everyone involved.
I sat on the beach yesterday watching a 4 and half billion year fire that is burning 8 light minutes away, overhearing two ladies in their mid 50s talk about how much money they were making flipping houses and https://www.youtube.com/watch?v=Wf4cea5oObY
The universe doesn't care if you contemplate it or not, it doesn't care if a couple of 50 year olds make their money selling pieces of dirt with some bricks on for more than they paid.
Enjoy your life, do the things that make you content (and I say content not happy carefully here, happiness is fleeting, contentedness lasts).
We had colleagues, photocopiers and telephones. A good lab had a large library and many shelf-feet of schematics. Search was difficult, sure, but copy pasta was just as fundamental to a young engineer then as it is now. The consequences for bad judgement were however a bit smokier (is that why my memories are hazy?)
The “trust what worked and copy” mentality and potentially lack of access to the original designer sometimes can lead young engineers read too much into the intent of designs and be unwilling to consider changes. Sometimes this helps keep the “magic smoke” in, and sometimes the original designs truly are elegant, but other times it can be an impediment to progress. “But this is the founders’ work! How could we possibly improve on that!” I’ve been in some design reviews like that. Old designs sometimes become legendary and thus hard to convince others to change, even when needed.
The Chibitronics kits are great for kids or really anyone who wants to play with laying out circuits and components by hand. I'm not sure there's anything else in the space, but they're pretty reasonably priced: https://chibitronics.com/lovetocode/
> Chibitronics is an evolution of Jie Qi’s passion for combining technology and art through making electronics using paper craft. The circuit stickers were developed as part of her PhD research at the MIT Media Lab. Together with Andrew “bunnie” Huang and Patricia Ng, Jie’s research has evolved into the Chibitronics toolkits.
I never did the sticky-tape and photoresist circuit cards. But Radio Shack sold some circuit transfer-decals that you pressed onto your cleaned copper board, prior to dunking it in the etchant solution. The trick was to remember that the order of the pins was reversed, since you were laying out the underside.
I actually started one level lower than that: in high school, those cost too much, but I discovered that a Sharpie made good, cheap etch resist :-)
Later at my first job, we made our own prototypes in the basement. We had a giant Agfa stat camera that would take the 2x artwork and reduce it 50% onto clear film that we would use to expose photosensitive resist coated PCB and then etch.
First job the draftsman would layout PCB's with tape and pads on mylar sheets And then we'd send those out to be reduced. Then to the FAB house who'd run some prototypes. All told the whole process seemed to take about 3 months from scratch.
The couple of times I delt with homemade PCB's... janky.
If I had to prototype something I'd buy vector board with plated through holes and then solder wire wrap wire to sockets. Quicker and almost as reliable as a professional PCB.
Using sharpies for etch resistant drawing was common method described to me in electronics books from 1970s or so (German series by Dieter Nuhrman, unfortunately I don't know the original title, was sold in Poland as "Elektronika łatwiejsza niż przypuszczasz", literally "Electronics (is) easier than you thought")
I used the transfer decal method extensively for ICs/pads along with fine tip water resistant markers for traces.
The trick that helped me a lot in avoiding errors was to redraw the circuit mirrored before writing it on the pcb, that is, with ICs having pin 1 on the upper right instead on the upper left. With time I learned to think the circuit mirrored and routed, so all my diagrams were drawn on paper very close to how they would appear on the pcb.
Of course I mean a bunch of TTL/CMOS ICs, or one-two LSI max, and some transistors. Anything more complex would have been a nightmare to make this way, at least for me.
Edit:
The transparent sheets with pcb traces transfered on them are called "acetate sheets", made of the same material used in transparency projectors slides. They're used to transfer the drawing onto copper boards previously treated with photosensitive material via exposition to UV light, then the boards can be etched by the usual FeCl solution.
I have a very strong interest about how folks used to make develop and build technology back on the time. Sometimes I think we are losing A LOT by not keep certain design principles in our minds, and that applies to maths, software development, electronics, engineering in general and many other fields.
I believe that we could get so much more out of our available, modern tools if we started to pay more attention to the history of our fields.
So, I've seen circuit boards with ground planes before, but what I don't quite understand is: why are the holes that are connected to the ground plane connected by two small traces (i.e., are surrounded by two C-shaped cut-outs), instead of just being drilled directly into a solid, uninterrupted ground plane?
It is to make the soldering process easier and faster by increasing the thermal resistance between the pin and the large metal of the gnd plane. If the gnd pins were sunk into the ground plane directly, the assembler would have to leave the solder iron in place longer, which would heat up more area, and possibly cause damage to the nearby parts.
Just look. Sources in near-perfect order. All build artifacts attached and stored (including the punch tape for the drilling CNC machine). Evidence in the form of photos from an oscilloscope screen attached. The results worked for 40+ years.
Huge respect to the engineering culture of these people.
Wow I'm glad this person saw the value and wrote this! I hope it becomes a series of posts. I'm on the software side, but have always romanticized the hardware side -- gotta stop wasting time and start learning and tinkering with hardware...
If anybody happens to be looking for a nice electronics project idea that they can blog about: try building something like a Wacom stylus tablet device. I'm curious how far you could get with just a hobby budget.
I am really looking forward to read the blog post "design software like it is 20s" where someone found my commits from this year and (hopefully) comes to the same conclusion.
The circuitry is incredible, but man, seeing these examples of PCB design makes me shudder.
I'd never want to design a printed circuit board using a ruler and a few sheets of paper. It's suddenly wildly apparent to me why CERN funded the creation of KiCAD.
However, there is a lab at Berkeley working on the topic, with really interesting work being done: http://people.eecs.berkeley.edu/~tianshi/research.html
"logic is encoded using the phase of oscillatory signals, rather than voltage levels. Such phase encoding has long been used in radio communication for its superior noise immunity; we show that it can also be used for computation, using self-sustaining nonlinear oscillators as underlying logic elements."
Harmonic injection computing, anyone?