Full Width 
Shortcuts 
Sign Up 
Log In 
Articles 
Reading List More from Evan Hahn (dot com)
In short: I use iOS Shortcuts to run a Mastodon bot. I run a Mastodon bot that posts random quotes from The Legend of Zelda: Link’s Awakening. (The writing in that game is so charming!) I used to run it off a server, but for the past few days, it’s used an iOS Shortcut on my phone. How it works The bot is called Ulrira, named after a beloved Link’s Awakening character. Ulrira (the bot, not the character) samples a line from the text dump, then makes an HTTP request to post it. Here’s a screenshot of the full thing: As you can see, everything here is hard-coded, from the server credentials to the file name. (It probably wouldn’t be hard to make this more generic, but I didn’t bother.) This used to be a short, hand-written Bash script. I used the new AI-powered Shortcut creator and prompted this into existence. I said something like, “Here’s a Bash script. Turn this into a Shortcut.” Like a lot of LLM output, it wasn’t quite right, but it was useful for getting started. Shortcuts doesn’t seem to have an elegant way to express something like “run this every 6 hours”, so I created four additional scheduler Shortcuts that call the main one. They run at 12am, 6am, 12pm, and 6pm. That means there are five shortcuts in total: one to actually make the post, and four to schedule the main Shortcut at different times. Why do it this way? I didn’t want to pay for my server anymore, nor did I want to maintain it. When I went looking for alternatives, they either (1) cost money (2) had some significant limitation (for example, GitHub Actions requires me to update the repo every 60 days). This solution runs on a device with high uptime and is effectively free. If you’re a pro software genius, you might complain that this is not reliable! It doesn’t work if my phone is offline, for example. Or what if I change timezones? But in the words of Link’s Awakening: You’re good! You’re a pro, aren’t you? ………… Well, beat it! Pros aren’t allowed in here! Follow the bot here.
Here are the Vim forks I know about: Neovim: “a Vim-based text editor engineered for extensibility and usability, to encourage new applications and contributions.” Vim Classic: “a stable, long-term support fork of the ubiquitous text editor, maintained without the assistance of generative AI tools.” EVi: “a hard-fork of Vim v9.1.0 (Jan 2024) before AI was used in the project.” MacVim: “a macOS version of the Vim text editor, providing a graphical user interface for Vim, while tightly integrating with macOS and providing features specific to the platform.” This list doesn’t include Vim-inspired editors like Kakoune, Vim-style plugins for other editors like IdeaVim, or Vim “distributions” like LazyVim. Contact me if you know of other forks!
In short: from my experience, people like map and filter, but not reduce. I use functions like map and filter all the time. When I put that code up for review, my peers rarely complain. I get plenty of feedback about other decisions, but not about my use of map and filter. I cannot say the same for reduce. Often, when I’ve submitted a patch with reduce inside, I get a comment like, “this part is hard to read.” And I see reduce way less than map, filter, some, and so on. Anecdotally, I have come to believe that programmers don’t like reduce as much. I don’t know why, but I have a few theories: reduce is harder to read. reduce is less familiar. reduce can have worse performance compared to other options. reduce is less elegant in languages I use, like JavaScript, Python, and Swift. In my blissful stint as a Clojure developer, I did not get this feedback. I’m wrong, and I’m seeing a trend that’s not real. I usually just change reduce to something else and move on. Even though I prefer it, I don’t usually care much. But it’s a little social phenomenon I’ve observed, and I thought I’d document it. I’ve also noticed this less recently, possibly because code review is less thorough nowadays. Do you notice this? Do you like reduce? Please tell me.
In short: Vim has a joke autocmd called UserGettingBored that doesn’t do anything. Vim’s automatic commands feature, usually shortened to “autocmd”, lets you run code when various events occur. For example, you could implement an auto-save feature by binding the TextChanged event to the :w command. Vim has over 100 events, from “buffer was created” to “file was saved”. But one of them sticks out to me: UserGettingBored. Here’s the documentation: UserGettingBored: When the user presses the same key 42 times. Just kidding! :-) When I saw this, I was busy doing something else and it completely derailed me. “I must know more,” I thought. Here’s what I found: Unfortunately, it doesn’t do anything. It only exists in the documentation (and some tests). If you try to use it with somethig like autocmd UserGettingBored ..., you’ll get a “no such group or event” error. It’s present in Vim, Neovim, and Vim Classic. It was first added by Bram Moolenaar in July 2000, over a year before Vim 6.0 was released. The original description was, “When the user hits CTRL-C. Just kidding!” And it didn’t do anything back then, so I don’t think it’s ever been real. In August 2001, he added the smiley face to the documentation. It then read, “When the user hits CTRL-C. Just kidding! :-)” Twelve years later, in 2013, the description changed to its current iteration: “When the user presses the same key 42 times. Just kidding! :-)” In 2022, developer Mike Smith created an unofficial plugin inspired by this joke autocmd. If you press the same key 42 times in Insert mode, a picture of Samuel L. Jackson appears. 22 years later, it’s finally real.
In short: I prefer strict tables in SQLite because they avoid some datatype problems, such as putting text in number columns. SQLite has a feature that I think is underrated: strict tables. Strict tables help enforce rigid typing, preventing mistakes like putting text into integer columns. I like them, and wrote this post to promote their use! To make a strict table, add STRICT to the end of its definition. Like this: -CREATE TABLE people (name TEXT); +CREATE TABLE people (name TEXT) STRICT; That’s it! But what does it do? Advantages of strict tables Broadly, strict tables help enforce rigid types, like other SQL engines do. Prevents type mismatches on insert/update Most significantly, strict tables keep you from inserting the wrong type into a column. For example, SQLite normally lets you put text into an INTEGER column, but not with strict tables. -- Non-strict tables let you put anything anywhere. CREATE TABLE people_nonstrict (age INTEGER); INSERT INTO people_nonstrict (age) VALUES ('garbage'); -- => works fine -- Strict tables don't allow that, which I prefer. CREATE TABLE people_strict (age INTEGER) STRICT; INSERT INTO people_strict (age) VALUES ('garbage'); -- => error: cannot store TEXT value in INTEGER column Personally, I think it’s a mistake to try to put text in an integer column, or vice-versa. I don’t want SQLite to let me make this error! The same validation happens for UPDATEs, too. Notably, if a value can be losslessly converted, it will still be accepted. For example, the string '123' can be perfectly converted to an integer, so it’s allowed. These two lines are equivalent, even for a strict table: INSERT INTO people_strict (age) VALUES ('123'); INSERT INTO people_strict (age) VALUES (123); Prevents bogus column types on table creation By default, you can create columns with bogus types. For example, all of these work even though they aren’t valid SQLite datatypes: -- SQLite doesn't support these types, but this is all accepted. CREATE TABLE tbl (name GARBAGE); CREATE TABLE tbl (name DATETIME); CREATE TABLE tbl (name JSON); CREATE TABLE tbl (name UUID); CREATE TABLE tbl (name BLOBB); I think these aren’t what the developer intended. Some of these are typos, some of them are misunderstandings of which datatypes SQLite supports, and some are egregious mistakes. Appending STRICT to any of these statements makes them error. In my opinion, that’s the correct behavior! -- All of these give errors, which I prefer. CREATE TABLE tbl (name GARBAGE) STRICT; CREATE TABLE tbl (name DATETIME) STRICT; CREATE TABLE tbl (name JSON) STRICT; CREATE TABLE tbl (name UUID) STRICT; CREATE TABLE tbl (name BLOBB) STRICT; Only INT, INTEGER, REAL, TEXT, BLOB, and ANY are allowed. Strict tables also require a column type, so you can’t do CREATE TABLE tbl (name). Still allows flexibility with ANY If you still need a column to be flexible, you can use the ANY datatype. As the name suggests, it allows anything—even in a strict table. CREATE TABLE tbl (value ANY) STRICT; -- All of these are valid because the column is ANY: INSERT INTO tbl (value) VALUES (123); INSERT INTO tbl (value) VALUES ('text'); INSERT INTO tbl (value) VALUES (12.34); INSERT INTO tbl (value) VALUES (X'8647'); I haven’t found a use for this, but maybe you will! Disadvantages of strict tables I prefer strict tables but I must share a few cons. Not everything is better! Can’t strict-ify an existing table I think it’s best to use strictness from the start, but that’s not always possible. Unfortunately, I don’t think there’s a way to ALTER a table to make it strict. I think you have to copy the data out of the non-strict table into the strict one. Something like this: -- 1. Create a new strict table with the same schema CREATE TABLE new_people (name TEXT) STRICT; -- 2. Copy data (risky if types are wrong!) INSERT INTO new_people SELECT * FROM people; -- 3. Replace the old table DROP TABLE people; ALTER TABLE new_people RENAME TO people; Note that this could be tricky if the non-strict table has invalid data! For example, if the old data accidentally contains text in an integer column, you’ll get errors when doing the migration. You’ll probably need to clean the data or cast it. You could make a rule for your codebase that all new tables are strict. That might be useful—at least some of your tables are valid! But it might also mean you have inconsistent validation across your tables, which might be more surprising than having weak validation on all tables. It’s up to you to decide whether this is a good fit for you. The SQLite developers disagree with me SQLite has a whole page called “The Advantages Of Flexible Typing”, where they argue that SQLite’s flexible behavior is good, actually. I hesitate to wade into the controversy of static-versus-dynamic, but I disagree in most cases. I’ve personally encountered many bugs where an unexpected data type caused subtle headaches. I’d much rather these mistakes explode loudly. But it’s worth noting that SQLite’s developers seem not to share my preference for strict tables! They point out a few good uses for flexible tables, such as “a pure key-value store” or “a place to store miscellaneous attributes” of different types. They also mention that you might want to keep the invalid data in some cases, like if you’re directly importing a messy CSV and don’t want to lose any data. I still prefer strict tables, but acknowledge there are some reasonable cases for non-strict ones. (There’s also at least one comment in the SQLite source that calls non-strict tables “legacy”, but I trust that less than the official documentation.) Only in SQLite 3.37.0+ SQLite introduced strict tables in version 3.37.0, released November 2021. If you’re on an older version of SQLite, you can’t use strict tables. It’s worth noting that old versions of SQLite can’t read databases with strict tables. For example, if you create a strict table in the newest version of SQLite and then try to read that database in SQLite 3.36.0 (before strict tables were added), you’ll get an error—even if the strict table is already in the database. Performance maybe? Strict tables are theoretically slower because they have to do a little extra work. For example, they check datatypes when doing an insert or update. But in practice, I don’t think this is an issue. I wrote a hacky script that inserted millions of rows into a table with 100 columns, and there was no obvious difference on multiple machines I tried. The file size on disk was also the same. I didn’t test this thoroughly, so maybe there’s something I missed, but I don’t think strict tables present a performance problem. In fact, one might expect better performance because you won’t be accidentally mismatching SQLite’s column affinities. But again, I haven’t tested this. Conclusion: I like strict tables! Personally, I think the pros of strict tables outweigh the cons. I generally prefer when types are rigidly enforced. It squashes a class of mistakes, and help enforce good data integrity. They’re not a panacea, but they’re usually easy to add and go a long way. If there’s a SQLite feature you think is underrated, please tell me.
More in technology
There are few things worse than a dead AirTag when you need it. That's why we designed the The 4-Year Battery Keychain for AirTag. It replaces the low power CR2032 coin cell with an included Toshiba CR2477 battery for over 4x more capacity. And does so in a minimal form factor. Assembles in seconds. Split key ring + braided cord included. Easy to open with included tool. Perfect on a keychain. And great for backpacks or a purse. Also available on Amazon.
In the corner of the rear Floodgap lab sits a pile of classic game consoles. It used to look like this: Sega Dreamcast all the consoles here only came with NTSC RF output and date from the pre-flat panel TV era. That includes multiple Atari consoles (including the inevitable VCS, here a Darth Vader 2600), several Pong machines — including two Commodore units — with various chipsets, an Intellivision (as the rebadged Tandyvision One, my first) and a Bally Astrocade with lots of upgrades. It seems wrong to use them with a flat panel, and CRTs lag less. Plus, games with a light gun (here the Dreamcast and one of the Pong consoles) often rely on the CRT's rapidly scanning beam to know where they're pointed. So what do you do when the CRT TV goes bad? I should note that this isn't all the old school gaming fun in the house. In the commons next to the pinball machines are an Xbox 360 Slim and a PlayStation 2, and in the home theatre connected to the 3D projector is a PlayStation 3 since it has some stereoscopic titles. These might appear in future articles. But most of the rest are here. I bought this Panasonic Omnivision PV-M2035 from a friend at church in 1999 for $100 [about $200 in 2026 dollars]. He had no use for it, and the top of the plastic shell had been heat-warped by sitting in his car too long, but it worked, still had the remote, and was a suitably cheap TV for my unsuitably cheap studio apartment. I used it for local news and Friends with a set of rabbit ears, VHS movies, and speakers for playing music by connecting a CD player to the single rear input. When I moved to a less cheap 1-bedroom, it came along, and its flat top was flat enough to become a shelf for a DVD player. I got the Dreamcast from Fry's in 2001 after Sega discontinued it and managed to share the composite input between them. That apartment was my crashpad for about a decade before moving to the current orbiting HQ, whereupon I bought a passive-3D flat panel TV and moved the Omnivision to its current location in the back of the house. It shared the RF input between the Inty, and later the Atari 2600, and later still the Astrocade and various Pongs. At any one time there were at least four consoles always connected to it via switches and switchboxes, and I wired up an accessory port for more. plastic was broken and on these TVs that's pretty much the load-bearing part. When it arrived at the mailbox and I unpacked it, I could see the lower half had sandwiched right through the VCR slot, with visible PCB fragments and broken buttons; aside from the tube itself (ironically), the thing was a total loss. The seller nevertheless made good on the refund, to their credit, and I got to schlep it to the recycler. Overall, however, it had occurred to me that there were very few TVs sufficiently contemporary with the M2035 out there, which might be a hint about their general longevity. So, if I was going to be forced to buy a different Panasonic, then it might as well be a more recent one, and it should be one of the "triple play" units (i.e., TV, VCR, DVD) so that I could at least get rid of the external DVD player. Mission: Impossible '88 and '89 DVD sets, where the "in stereo" animation was even burned into the video stream.) not recording and thus impossible to watch many DVDs, for which the Sima CopyMaster was the cheapest solution: it simply blots out that portion of the vertical blank along with the pulses, which also has the side effect of sometimes stripping closed captioning (in the vertical blank as well), though DVDs with proper subtitles are of course unaffected. Later Sima devices could regenerate portions of the vertical blank for better quality, and Macrovision actually sued them over their products in 2005, but later settled. The mixer on the right was purchased literal decades ago at Radio Shack (more specifically a "Video Enhancer Stereo Audio Mixer") that I think we used for a Halloween display exactly once and then forgot about. Here it was used first as a cutoff switch, and second as a way to match the audio volume from the DVD player with the audio volume from the Dreamcast, which was also directly connected to the second rear Y-cable. This TV is mono, so the stereo channels from the audio mixer and the DC get merged using their own Y-connectors (also Radio Shack) before going into the Y-cable to the single line audio input. The video function isn't used here at all; it was just a audio mixer on the shelf I wasn't otherwise using. The DVD player is under the mixer and the CopyMaster, a cheapie Sony special that will only eject discs if it's in the mood or summarily threatened with violence. You already met this system when we adjusted the laser pickup and refurbished its rechargeable battery. you met the Tandyvision One (otherwise identical to the Mattel Intellivision Master Component, and manufactured by them as an OEM) when we installed a composite video modification and a front power-cutoff reset. I later upgraded it with an original Mattel Intellivoice speech synthesizer (Tandy never rebadged these) and a CuttleCart. But you haven't met the Bally on the bottom, officially introduced in 1978 as a game console and simple home computer, sold under various names such as the Bally Professional Arcade seen on early units, later the Bally Computer System badging here, and now collectively referred to under its final incarnation as the Astrocade. These machines were based around a 1.789MHz Z80 (half NTSC colourburst) combined with the same graphics hardware Dave Nutting Associates used for Midway's arcade games Gorf and Wizard of Wor, among others. Unfortunately, much of its theoretical graphics power went unrealized because it only has half the video memory of the arcade units (i.e., 4K) — and that's all the memory it has, so it has to interleave bits in the video memory to store BASIC programs, meaning only game ROMs can take full advantage. Still, that was enough to be solid competition for the likes of the Atari VCS and Fairchild Channel F, and it holds its own well against the slightly later Intellivision. does have is a doozy: the cooling vents are on the underside. These machines are very prone to overheating and can be easily damaged by it, which is why I use two foam struts to keep it off the carpet. The yellow label is for Astrovision, a separate enterprise that had been attempting to develop their own console. The (then) Bally Professional Arcade was selling poorly, though part of this was due to a tardy launch and Bally's weak retail presence, and Bally decided to sell off the consumer products division in 1979. Astrovision, unable to get their unit to market quickly, bought it in 1981 along with a limited license to the Bally name and reintroduced it at Winter CES 1982. Astrovision units are otherwise the same except for the badging and are the most common Astrocades today, though the company became another victim of the 1983 video game crash and went bankrupt shortly afterwards. Simultaneously, an Astrocade derivative merged with Bally's unfinished ZGRASS-100 expansion chassis was reworked into the Datamax UV-1 graphics generator, sold in small numbers until around 1982. the wall warts and power strips, and do a little spot vacuuming with the Roomba. out in spring. Gremlin's Blockade; and a simple calculator and screen paint toy ("Scribbling"). The display is a bit noisier than the Inty/Tandyvision, probably because of the difference in shielding, but still quite good. Videostak Video Game Center, which allegedly could accomodate a VCS, Intellivision or ColecoVision. The front decal is somewhat worn and the woodgrain veneer doesn't match the Vader VCS nearly as well, but it's still a fun-looking setup. Intellivision System Changer which has Atari 2600 hardware in an expansion box for the Intellivision II (useable on the original Inty with modification). It might be fun to explore the reputed irregularities between what is believed to be a VTI clone TIA in the System Changer and the slightly earlier Coleco Expansion Module 1 (and the related Coleco Gemini), and see how they differ in practice. Now for the Pong machines. These attach to the "accessory port" on the input switcher. The current Pong menagerie are mostly MOS 7600-based, MOS' own Pong-on-a-chip: two from Commodore themselves but PAL only using the MOS 7601 (the TV Game 2000K and TV Game 3000H), plus two MOS 7600 (NTSC) systems, a Coleco Telstar Gemini, not to be confused with the Coleco Gemini or Coleco Telstar, and an Allied Leisure Name of the Game II, possibly the earliest use of the chip in NTSC regions. We'll play the Name of the Game II and Telstar Gemini below. These will be consoles for another day, but I also have a boxed Atari Video Pinball, an "honorary" Pong machine though it plays several other games, and a boxed Atari Ultra Pong Doubles, which we had before we got the Tandyvision, but I don't ever recall it working fully and it later failed to power up and got thrown away, so this is a replacement. My least favourite of the Pongs here (I really only picked it up because it was cheap) is a 1977 Magnavox Odyssey 3000. The 3000 is basically an upmarket Odyssey 2000 with proper handheld controllers, but otherwise the same me-too AY-3-8500 Pong-on-a-chip generating the same black-and-white display. Allied Leisure Name of the Game II is a two-player Pong based on the MOS 7600-001. This chip notionally contains three field games (a tennis game, a "goal" game usually billed as either soccer or hockey, and a racquetball game) and a skeet shooter, but the racquetball game can be put into a single player mode by setting it to "four" players and the NotG2's switch triggers this mode in place of the skeet shooter. It was noisy on the other TV and it's noisy here, possibly because it has its own internal RF switchbox, but it plays fine. Coleco Telstar Gemini, an MOS 7600-004 system. This has two "pinball" variants (with flippers and moveable Pong paddles, so it's still a Pong of sorts) and two shooting gallery variants involving shooting a deformed bear sprite. The light gun connector is the same as the Coleco Telstar, so I borrowed the gun from that. sing along with The Offspring. But the output looks crisp through the rear jacks, the sound is proper stereo, and we now have front inputs I can use for something else without rigging a whole bunch of Y-adapters again. Plus, if the Telstar Gemini's gun works, it's all but guaranteed the Dreamcast's does. am likely to play some Video Pinball, and in its box that's easy to move, so that can go on top of the 2600 and the NotG and Telstar Gemini can sit next to it. The old TV is out in the garage and as I have time I'll extract its circuit boards and go over them more thoroughly, since I had a lot of good times with that television and it would be nice to fix. If I do, I'll write it up here, since it's an actual old VCR. Meanwhile, I have some possible projects in mind for the Tandyvision and Astrocade, and maybe some mods to go in the VCS, and you can read about those in future articles too. In the end, with the new TV all set up, the 1980s can live again in my rear office. Or, as my wife likes to point out, they never actually died there either.
By a mile, the iPhone Duo has garnered more questions about what I think about it than any other phone I’ve ever had. It’s clearly a phone that has people intrigued, so I thought I’d jot down some loose, unorganized thoughts about what I think about it after about ten days. If I could sum all my thoughts up into one word, it’d simply be “fun”. I think for the vast majority of people, myself included, the iPhone 18 Pro is a more practical phone that through scientific analysis you’d likely even find a better phone. Even if they were priced the same, its package is more traditionally comfortable and ergonomic, with a better quality main camera, a helpful telephoto camera, a screen that’s more practical to use more of the time, a better biometric system (Face ID versus Touch ID) and a more established software experience. But damn if the iPhone Duo isn’t fun. Phones have been downright boring for so long. Incredible technology, lovely apps, but the hardware has been a bit slow moving since the iPhone X. Maybe that’s why I jumped onto the iPhone Air so readily. Even that slight departure from “black slab of glass” to “a dramatically thinner black slab of glass” was exciting versus what we’ve grown accustomed to. So I pick up the Duo, and it’s kinda just exciting to hold a phone for the first time in ages. Maybe that’s a honeymoon period and after a few months I’ll be over it. Maybe everyone who tries a folding phone for the first time experiences this. And to be clear, it’s not just all raw novelty. The fun aspect is fueled by having an absolutely incredible inner screen that fundamentally transforms what a phone is and can be. In addition, support the Apple Pencil has turned out to be phenomenally handy for me to be able to jot down quick thoughts and notes. For downsides, Touch ID is objectively worse than Face ID, but honestly, I don’t mind it as much as I expected I would. The smaller camera sensor than the Pro series phones is still decidedly great (rumors were it was a 17e size which was not correct, it’s the size of the normal 17’s). Software support gets better by the day, though iOS on the Duo is still decidedly a bit buggy at times. Other random thoughts Touch ID is objectively worse than Face ID, but honestly, I don’t mind it as much as I expected I would, previewing notifications is where I miss it the most After getting used to having a single speaker in the iPhone Air for a year, having stereo (not to mention great sounding) speakers is amazing The FaceTime camera on the inside is indeed not great for selfies/photography but honestly perfectly acceptable for FaceTime, and I’d definitely prefer this solution over a cutout on the inside so I’m happy! It definitely has a heavy feel to it, but it’s nothing honestly that bothersome A lot of the reviews said something to the effect of “the crease is practically invisible” and while I don’t really mind it or anything I definitely would disagree with that statement, you can see it a fair bit even with light content The hinge is a bit harder to open than I expected, I’ve found the secret is don’t pry from the middle of the side, instead put a thumb near the top and bottom (also helps prevent pressing Camera Control instead which I was doing all the time). I imagine this is necessary to have it be able to stop at arbitrary angles once open (which I do like) and hopefully it loosens up a bit with more use, but playing with Android foldables at Best Buy I did find their hinge easier to open I am really split on the inner, nano texture screen. It definitely helps reflections, has a neat look to it, and is super nice to draw on with the pencil, but it’s significantly less vibrant than the outer display which to my eyes looks a fair bit nicer. I think they made the right decision ultimately but it’s definitely not a zero-compromise solution. I was considering nano texture on my next MacBook but after this I think I’ll stick with glossy. My glass half full take with the Duo for now though is that we get two really cool screens to play with The outer screen is awesome for taking photos, unlike taller iPhones, your viewfinder fills the screen completely so it’s the largest feeling viewfinder I’ve used on a phone I miss the action button, nice little launcher for ChatGPT/Claude depending on which month it is Battery life is downright phenomenal versus my iPhone Air and 15 Pro (two devices admittedly not known for having great battery life) So, I don’t think the Duo is for most people who either don’t care much about phones, or care a lot about having the perfect phone, but for a decent amount of people in the middle I think the Duo represents a fundamentally different experience, that’s a ton of fun, and I’m excited to be excited about a phone again. (But, Apple, please figure out a way to add a telephoto lens.)
Going through a collection of old parts, CuriousMarc encountered some ICs that he couldn't identify. He opened up one of these ICs, and I studied it under the microscope. After some reverse engineering, I determined that it was a logic chip that implemented four NAND gates using old-fashioned diode-transistor logic (DTL).1 The chip turned out to be the ST680A, a member of the Signetics 600 series of chips from 1965. In this article, I discuss the circuitry in this chip and how diode-transistor logic works. The die of the ST680A under the microscope. Integrated circuits can be packaged in many ways. One of the most popular is the plastic DIP (Dual In-line Package), a black rectangle of epoxy with two rows of metal pins. This type of package was invented by Signetics: "Introduced in late 1965, the 600 series was the first integrated circuit family to be offered in a solid, plastic, dual in-line package (DIP)." Earlier integrated circuits were packaged in round metal cans or in small flat-packs. In 1965, engineers at Fairchild invented the DIP, which had two rows of pins on 0.1" spacing. Unlike the surface-mounted flat-pack, the DIP had pins that were inserted into holes, making manual or automated assembly quick and easy. Because the pins of a DIP were on a regular grid, rather than the circular pattern of a metal can, circuit board design and manufacturing were also easier (details). The first DIP chips had ceramic packages. The plastic DIP, introduced by Signetics, was much cheaper and became very popular. An advertisement for the Signetics 600 series. From Electronics, March 21, 1966. The plastic DIP is less popular with me, however, because it is harder to open. With a ceramic package, I can simply tap it with a chisel and pop it open, but that doesn't work with epoxy packages. An epoxy package can be dissolved with nitric or sulfuric acid, but I try to avoid dangerous acids. Instead, CuriousMarc used Antoine's hot air decapping method. Unfortunately, the hot air technique doesn't always work. In this case, tiny bits of epoxy were stuck to the die, obscuring the circuitry. I used a few drops of boiling sulfuric acid to clean the die. (I'm okay with sulfuric acid in tiny quantities.) The die on its metal frame. The photo above shows the silicon die on its metal frame, the lead frame. The golden blobs are where the bond wires were attached to the die. You can see the doped silicon regions that form diodes and bipolar NPN transistors, as well as 16 resistors in the center, forming serpentine paths. Most chips have a passivation layer—a thin layer of silicon dioxide—over the chip. Inconveniently, this old chip didn't have a passivation layer, so the acid also stripped off the metal layer. (A few remnants are visible in the upper left.) Without the metal layer, I had to guess at how the components were connected to form logic gates. The die has very faint color differences where the silicon was protected by the metal, so that helped. Eventually, I came up with the diagram below, showing how the metal layer was probably configured. This wiring reveals that the chip implements four DTL NAND gates, which I'll describe in more detail below. One puzzle was that the chip has two unused diodes, indicated with question marks below. The metal layer for a quad NAND chip. I desaturated the image to make the labels clearer. Transistors, diodes, and resistors In this section, I'll explain how the patterns in the image correspond to components of the circuit. The silicon die is doped with various impurities to form regions with different electrical characteristics. These regions appear red or green under the microscope due to thin-film effects. As a result, the chip's transistors, diodes, and resistors are visible. The chip is made from bipolar NPN transistors, which have a completely different construction from the MOS transistors in modern ICs. In an NPN transistor, the current from the collector to the emitter is controlled by the current through the base. The collector is a large region of N-type silicon, silicon that is doped with atoms such as phosphorus or arsenic. Since these elements have one more valence electron than silicon, they give the silicon more negative charge carriers, so it is called N for negative. The N-type silicon happens to appear reddish. On top of the collector, the base is a thin region of P-type silicon, which appears green. This silicon has been doped with atoms such as boron or gallium, which have one fewer valence electron than silicon. The missing electrons form "holes", which act as positive charge carriers. Finally, the emitter is N-type silicon, on top of the base layer. These layers work together through semiconductor magic, amplifying the base current. Two NPN transistors on the chip. The second transistor is a larger output transistor. The chip's output transistors are slightly larger so they can sink more current. In the transistor on the right above, note that the collector contact wraps around three sides of the transistor. The emitter is also larger, with an L-shaped contact. According to the datasheet, each output transistor can sink 2.5 mA of current. The chip uses many diodes since it is diode-transistor logic. The structure of a diode is similar to a transistor, but without the emitter. In a diode, current flows from the cathode (P-type silicon) to the anode (N-type). A diode and a resistor on the chip. The chip uses 16 resistors, located in the middle of the die. A resistor is constructed from a long, thin region of doped silicon. Since silicon has moderately high resistance, a long strip of silicon acts as a resistor. Most of the resistors have a serpentine layout to fit a long resistor into a smaller area. These resistors consist of P-type silicon (green) in a "tub" of N-type silicon (red). The tub is connected to the positive voltage supply; since current doesn't flow from N-type silicon to P-type silicon, this isolates the resistors from each other. Operation of a DTL gate Logic gates can be created in many ways. The earliest integrated circuits (such as the ones in the Apollo Guidance Computer) used Resistor-Transistor Logic (RTL), where logic gates were constructed from resistors and transistors. In March 1962, Signetics announced the first family of commercial Diode-Transistor Logic (DTL) integrated circuits (details). By using diodes on the inputs instead of resistors, DTL provided better performance than RTL. However, DTL soon became overshadowed by TTL (Transistor-Transistor Logic), which was faster and used less power. (If you built digital circuits a few decades ago, you're probably familiar with the popular 7400 series of TTL chips.) The diagram below shows how the DTL NAND gate works. If both inputs are high, the current from the first resistor flows through the base of the first transistor, turning it on. The transistor amplifies this current, passing a much larger current from the transistor's collector to the transistor's emitter. This current goes through the base of the second transistor, which turns on the second transistor. The second transistor passes an even larger current from the output pin to ground, pulling the output pin low. The DTL NAND gate with both inputs high. The schematic is from the datasheet. This circuit is considerably more complicated than the DTL gate on Wikipedia. But if one of the input pins is pulled low, the current from the resistor goes through the diode instead of the transistor. The first transistor turns off, and without a base current, the second transistor turns off. The third resistor pulls the output high. Thus, the output is low only if both inputs are high: a NAND gate. The DTL NAND gate with an input low. I'll point out a few features of this circuit. The diodes on the inputs ensure that one input doesn't affect the other input. The diode in the middle raises the voltage on the first transistor's emitter, so it requires more voltage to turn on, reducing the sensitivity to noise. The output uses a resistor to pull the output high; more advanced circuits use another transistor. When the output is low, the third resistor is still "trying" to pull the output high, which is wasting power. Moreover, the resistor limits the output current, so the high output is fairly weak. In some DTL families, you can get gates with different values for the output resistor, so you can select the tradeoff between wasting power and having an output that is too weak. If you've used TTL integrated circuits from the 1970s, you know that chips usually have power and ground on diagonal pins: ground on pin 7 and power on pin 14. (Putting power and ground diagonally opposite both helps die layout and circuit board layout.) Much to my surprise, I found that the Signetics chips predate this standard and have ground on pin 1 and power on pin 8: still diagonal, but the wrong diagonal! The diagram below shows how the four NAND gates are wired in the chip. Pinout of the 680A quad NAND gate. From the databook. Three chips from one die When I examined the die under the microscope, I figured out the circuit, but I was puzzled by two extra diodes, as if two NAND gates had a third input that wasn't used. The explanation is that Signetics used the same silicon layout for at least three different chips, applying a different metal layer for different functionality. This saved money since it's much cheaper to produce a single optical mask for the metal layer than a complete set of new masks. Signetics sold semi-custom ICs to customers, using an approach called variFEBS (Variable Functional Electronics Blocks). This approach used a custom metal layer on top of fixed silicon, providing custom circuits without the cost of a full mask set. These circuits didn't sell well, but Signetics used the same approach for standard ICs. Pinout of the 670A triple NAND gate and the 616A quad NAND gate. All three chips have some similarities, such as using pins 2 and 14 as outputs. From the databook. One IC with the same silicon as the quad NAND gate is the 670A chip, which has three NAND gates, each with three inputs (above left). The diagram below shows how the silicon can be wired to provide this functionality. The hypothesized metal layer for the 670A triple-NAND gate. The layout for the triple chip above is a bit tricky because the top gate has inputs on both sides of the power (Vcc) pin. This creates a layout problem because the power and ground metal runs vertically to the resistors, blocking metal wiring from one side to the other. The solution is a "crossunder", a region of silicon that allows a signal to pass under the metal layer, connecting the inputs on the left to the rest of the circuitry on the right. This crossunder is visible near the top of the chip. Another chip with the same layout is the 616A dual NAND gate, with five inputs on each gate. Since this chip has 10 inputs in total (compared to 8 on the quad gate), this chip uses the two "extra" diodes that I spotted. If that isn't enough inputs, each NAND gate has an expansion input; by connecting external diodes to the expansion input, the gate can be made as large as desired. The hypothesized metal layer for the 616A dual-NAND chip. Some history The history of Signetics is closely tied to Fairchild. The story of Fairchild's founding is well known: in 1957, eight engineers—the Traitorous Eight—left Shockley Semiconductor and created Fairchild Semiconductor. Hundreds of companies (the "Fairchildren") were started by engineers who left Fairchild. Signetics was one of those companies, formed in 1961 and said to be the first company to exclusively design and manufacture integrated circuits.2 Signetics announced its first product in 1962, a NAND gate that cost $125 (over $1000 in current dollars).3 Not surprisingly, sales were weak. Signetics soon ran short of money, so in 1962, Corning Glass invested $1.7 million for 51% ownership of Signetics. Corning Glass may seem like a strange investor, but Corning was more of a high-tech company than you'd expect, making components such as resistors and glass capacitors, and later creating fiber optics. Moreover, in the early 1960s, integrated circuits were the hot area (much like AI now), and everyone wanted to get involved. In 1963, the Department of Defense pushed for contractors to use integrated circuits in new products. Riding this wave, Signetics became the leading producer of integrated circuits.4 However, Fairchild soon introduced a series of DTL chips that competed with Signetics. The Fairchild 930 series was faster than the Signetics chips and less than half the price, so Fairchild took much of the market from Signetics and dominated the IC market. Signetics also produced analog integrated circuits. Signetics is probably best known for creating the 555 timer chip in 1970, said to be the world's most popular IC, with billions sold.5 Signetics was acquired by Philips in 1975, and then Philips Semiconductors was spun off as NXP (Next exPerience) in 2006. NXP is currently the #10 semiconductor manufacturer worldwide, so the heritage of Signetics lives on. For more, follow me on Bluesky (@righto.com), Mastodon (@[email protected]), or RSS. Thanks to CuriousMarc for providing the IC and thanks to TubeTime for finding the databook. AI statement: Despite the presence of the em dash, no AI was used in the writing of this article (details). Notes and references After I had reverse-engineered the chip, TubeTime found datasheets for it, making my reverse engineering unnecessary. (See the Signetics Condensed Catalog 1969, Utilogic II Databook 1971, and Utilogic 2 Databook 1972.) We must have been unlucky with our original web searches since we didn't find anything. But this way you get to see what's inside the chip :-) The Signetics ST680A IC in its package. The large circular depression marks pin 1. From the date code, this chip is from 1971. ↩ For more Signetics history, see Signetics Founding History, written by Lionel Kattner, a co-founder of Signetics. He states that the company name was inspired by the words "signals" and "networks", as well as the high-tech suffix "-netics". ↩ See History of Semiconductor Engineering, page 156. ↩ The book Making Silicon Valley describes the history of Signetics. It has a negative view of the Corning investment, saying that Signetics executives "Van Poppelen and Kattner dissented on the grounds that Corning was not a good fit for an integrated circuit company. In their view, Corning was a staid Eastern firm, ignorant about semiconductors and antithetic to the ways and culture of the electronics industry on the West Coast. Their view later proved to be correct." ↩ The 555 timer was created by analog IC wizard Hans Camenzind. If you're interested in analog circuits, you should read his book Designing Analog Chips, which you can download for free. ↩
