Ask Hackaday: Why Make Modular Hardware?

In the movies, everything is modular. Some big gun fell off the spaceship when it crashed? Good thing you can just pick it up and fire it as-is (looking at you, Guardians of the Galaxy 2). Hyperdrive dead? No problem, because in the Star Wars universe you can just drop a new one in and be on your way.

Of course, things just aren’t that simple in the real world. Most systems, be they spaceships or cell phones, are enormously complicated and contain hundreds or thousands of interconnected parts. If the camera in my Samsung phone breaks, I can’t exactly steal the one from my girlfriend’s iPhone. They’re simply not interchangeable because the systems were designed differently. Even if we had the same phone and the cameras were interchangeable, they wouldn’t be easy to swap. We’d have to crack open the phones and carefully perform the switch. Speaking of switches, the Nintendo Switch is a good counterexample here. Joycon break? Just buy a new one and pop it on.

What if more products were like the Nintendo Switch? Is its modularity just the tip of the iceberg?

PocKit-Sized Modularity

The PocKit project tackles this question by diving whole-hog into modularity. This impressive platform consists of a central “brain” block that contains an ESP32 and an STM32, as well as ~24 other blocks that can just snap onto it. These blocks include a screen, a keyboard, a speaker, and a wide variety of sensors and other neat bits. The brain module even has a DDR connector that lets you throw a Raspberry Pi Compute Module into the mix for greater functionality.

All of the blocks use the same electrical interface and are held in place by magnets, allowing for near-endless reconfiguration. They have a great demo of the system in action, adding a camera, GPS, and more to a portable battery-powered Linux system:

Some of you may have felt the spark of familiarity here and thought back a couple of years to when Google announced the since-cancelled Project Ara. This was an experimental phone that Google debuted in 2013. It was suppose to be a low-cost, bare-bones phone that the user could easily upgrade themselves by swapping and adding modules onto the frame, including the battery, processor, camera, and display. Unfortunately, Project Ara never made it into our pockets, as Google canned it back in 2016.

Of course, these aren’t the only examples of modular electronics we’ve seen. Some of us even grew up with them.

A Modular Education

A true classic: the Snap Circuits kit

When I was a kid, one of my absolute favorite toys was a Snap Circuits kit. It consisted of a few hundred LEDs, motors, buzzers, and rigid “wires” that all snapped onto a plastic base — sort of like a breadboard. I spent hours connecting the modules in different ways and occasionally came across a configuration that actually did something. The Snap Circuits kit’s modular nature was a perfect introduction to electronics — after all, seven-year-old me wasn’t ready to design and fab a PCB yet. Instead, the large, colorful modules with silkscreened schematic symbols were a great teaching tool.

Nowadays you can still get your hands on a brand-new Snap Circuits kit (and I was very excited to buy one for a friend’s children last Christmas) but as you might expect, there are even more advanced modular kits on the market. Brick ‘R’ Knowledge is one such offering. It’s a modern spin on the classic Snap Circuits that brings advanced components to the table, including an Arduino and networking blocks.

You can tell by looking at these kits and others that modular electronics are a pretty fantastic tool for education. They can be used to introduce complicated concepts in ways that don’t “dumb down” the material, but simply present them in a friendly, approachable manner. Let’s circle back around to consumer electronics though — how much modularity do we need in our day-to-day lives?

Over To You

You know the drill. Time to chime in below and let us know what you think here. Should everything be modular? Presumably, it costs more to manufacture something with such a high degree of customer customizability than it would to just make one or two of the possible configurations. Would you pay more for personalizable products, or are you happy to buy off-the-shelf? If you’re anything like me, you would buy the modular option, find a configuration you like, then never change it — so then, what’s the point of being able to reconfigure your gizmos?

33 thoughts on “Ask Hackaday: Why Make Modular Hardware?

  1. The problems with modularity are size and durability. Phones are hyper-miniaturized and there simply isn’t room in the form factor for the connectors and module walls that would separate the functions. And every one of those connectors is a failure point, especially if they are frequently exercised by actually taking advantage of the modularity. We’ve been building smartphones for quite a while and laptop computers for even longer, and the state of the art is very mature. Modularity sounds like a cool idea but it just doesn’t compete against all-up purposed construction.

    1. Most people have no idea how difficult it is to reliably and cleanly swap parts outside of a proper rated environment. Let alone the average person who has no experience doing so. There is a reason why chip construction has to be very, very clean. Taking apart electronics is certainly possible. Doing so in a normal phone form factor? I agree that is not really possible using current technology.

      That also partially explains why batteries are sealed inside the IP rated enclosure though. It is still possible to be both waterproof and also be fixable or changed out though, right? Anyone?

      1. Samsung has some phones and tablets that are rated for basic immersion and have user-swappable batteries. The backs aren’t even screwed down, it’s just clips and a seal, like _old_ swappable-battery phones. You do need to be a bit careful about getting all the clips clipped before exposing it to water.

        They also detect and show in the UI whether there’s water on the connectors.

        The CAT phones have screws in positions that suggest that they may be reasonable to open up, although I have not done so, as well.

      2. FWIW I have saved more phones by popping the back and soldering the USB port or just replacing the battery than I have lost to immersion.
        Since I need the removeable battery, analog audio phone jack, and an experiance not reporting to another master I am currently stuck between flashing old phones to LineageOS for microG or buying pinephones where the software and user experience is not quite there for the non hackers int he family.

    2. yeah i agree that kind of modularity is a bust for most devices. you want your device to have good components in working condition, you don’t want to be swapping them around willy-nilly. you’d usually rather it break less often than be field-repairable. the trade offs are pretty steep.

      at the same time, there is a huge benefit to modularity when you are experimenting or building, and in fact that advantage is harnessed by all the phone manufacturers. the camera module is modularized to a remarkable extent, even though from the consumer standpoint it’s built into the device. if the module they originally designed it with becomes unobtainium for some reason, it doesn’t seem uncommon to switch the assembly line to a different module, sometimes without even updating the model number. and i’ve seen enough franken laptops to know that a lot of the modules in those highly integrated systems are nonetheless interchangable, even if the physical glue to hold them together winds up becoming a delicate and ugly hack.

      even within a chip..there’s nothing more integrated than an IC, but still an ASIC or SoC will have many modules within it, often coming from different vendors and different development processes.

      modularity is indispensible it’s just not usually user-facing.

    3. But they don’t need to be.

      I used cellphones when they were a lot bulkier than they are today. It’s really not a big deal. The only reason people need them to be so thin is so they can stick them in their back pockets and sit on them. Then they keep using them with big old cracks all over the screen. I don’t get that. I’ll never sit on a hundreds of dollars device with a glass front.

      A fatter phone that is upgradeable and customizable will do much better, thank you. Using some of that extra room for a bigger (and changeable) battery is good too!

      1. It’s not the size, it’s the non-uniformity. They all need to look different, so they all demand slightly different modules. Smartphones basically *already are* modular! They’re just not interchangeably modular.

    4. Yeah, I totally don’t agree. There are buckets of connectors to put things in relatively standard locations on phones/laptops. Phones and laptops aren’t standardized because they all want slightly different form factors so they *look* different. Can’t charge a luxury price for two things that look identical.

      Same thing for desktops: even though there’s a *high* degree of modularity there, standard vendors (Dell, Apple, etc.) all avoid standard form factors for custom wonky things so that their offerings will look more distinct. Really, I think the only reason standardization exists in the desktop market is historical.

      1. In the early PC days there were advantages to using interchangeable assemblies like main board, enclosure, power supply, expansion cards, and socketed chips because it meant that if you were making computers you had multiple options for sources and your customers had options for upgrading. This stayed the case through the 80’s and early 90’s, but in the late 90’s the advantages started to fade because things were changing too fast and by the time you needed a replacement or were ready to upgrade, the state of the art product was no longer compatible with your older assembly. In this case it was technological obsolescence that made the modularity of early PC’s itself obsolete.

        1. I don’t think it was really upgrading that did it – I think it was just the vast number of assemblers. As in, everyone started out making IBM clones, so they all just stole IBM’s power supply connector. Once that connector became too limiting (in both current and voltage), the main company that benefited from the huge number of assemblers (Intel) proposed an improved standard, and hey look, we actually got something reasonable to work with.

          I mean, if you look at tablets a few years back, for instance, it’s easy to say “oh, they’re all just bespoke designs” but the ultra-cheap ones were all just dead-clones of the manufacturer’s reference board, so they were practically all completely compatible with each other. But because the cases and connectors weren’t, they weren’t interchangeable. So the only reason they weren’t totally “modular” is just because the manufacturers made the plastic different.

        2. I find it fairly amazing how standard PC components still are though. I am old enough to remember the “AT” power supplies that were supplanted by ATX PSUs though not not old enough to remember any other types, so that’s two standards in my lifetime, which seems incredible given the incentives to lock customers in to a bespoke design. Granted the maximum current permitted has gone from 5 V heavy to 12 V heavy, the main board connector has gained a few extra positions and now there are SATA power connectors and those 8-way connectors for powering GPUs (of which I now need three!), but I can still use a modern ATX PSU to power my 486 and I can still power it up myself by grounding the green wire.

    5. I think two different ideas are being conflated in the article. You have modularity in terms of configurability where devices are built out of almost literal building block components that are a standard form factor and interface both physical and electrical.

      This is something that has been tried many times and never works. One of those things that sounds good but is dumb. They are complicated, expensive, heavy, the connectors wear out as you noted. The list goes on.

      The other end of the modularity, the one you actually see depicted in media is probably better thought of as a very high level of reparability. In star wars you can replace a hyperdrive so easily not because they are building blocks, but because the republic and presumably the empire that came afterwards were very pro right to repair. Corellian tech is perfectly happy to talk to mon cal ships because everyone is using standardized electrical signaling and its well documented. Nobody is using some weird voltage or phase and there is no drm to prevent their usage.

      I imagine a good in universe explanation about why this always seems to be the case in Sci-Fi is that nobody wants to die in the cold vacuum of space because space Apple decided you couldn’t replace your own life support system.

      In some ways the modular configuration paradigm can actually work against you because the interface itself could be subject to IP law restrictions and licensing fees. Meaning that while your cool thing-a-majig can be reconfigured you can only use authorized space acme® modules.

      But if everything is well documented and has to conform to some basic low level standards then clever characters or hackers can jury rig all sorts of things without too much work.

      1. An issue with the former point that few people pick up on: Modularity for performance-enhancing upgrades is a fools errand.

        There was a laptop I saw recently where the USB/thunderbolt ports on the side were simple blocks that slid in and out so you could choose what you wanted. There are the usual arguments against it like wasted space, material, and product packaging.

        But, more relevant to the message: this is actually not upgradable at all.

        What are the chances someone will change those blocks after they’re in place? They have to plug in over thunderbolt, or PCIe 3.0, or what have you. All future modules that could be made have to work with the limitations of those legacy interfaces. This becomes even worse when you move from peripherals and storage to core system components which are nesisarily tightly coupled because they have to be for performance due to the laws of physics. A DDR3 ram chip will not run faster than a DDR2 chip on an SoC built for DDR2, if it can even be made to work at all.

        No, when a device is made its maximum performance is largely locked-in.

        You can swap parts with replacements when they break to keep it running, but you’ll never turn a 486 into an i7 by adding more RAM. The focus should instead be placed on limiting the explosive growth of software bloat which increases faster than moore’s law. The postmarketOS people provide a good example of this, basing off of a lightweight distro like alpine so that old phones can work just as well as new ones.

        extra: for an extreem example of the last point, you might look at suckless.org

  2. Working with industrial robots… they’re pretty much modular. When you’ve got a production line down that’s costing 1000’s of or even 10’s of thousands per hour… swapping out prebuilt subassemblies is important!

    1. It’s certainly possible to much more easily do with larger, more designed for software types of applications such as industrial robots and adding this modular functionality to it that way. Much more difficult to do with cell phones or smaller hardware is all.

      Though still curious about efforts to actually robotic applications waterproof which seems very helpful to do with industrial robots of most types, even if the water application is intermittent. Industrial robots are not cheap, especially if they are halfway decent in what they can do.

  3. The connectors, loose wires/solder joints etc in those modules does can be additional points of failure.
    So going excessive in modularity makes your project unreliable.

    The worse case is modularity at per chip level and you build everything on a breadboard so that each of them and the loose wires can be plug and unplug easily. :P

  4. > Presumably, it costs more to manufacture something with such a high degree of customer customizability than it would to just make one or two of the possible configurations

    you’ve got more SKUs, more money spent on connectors, more money on injection molds (or whatever) as each component requires (waterproof!) housing, etc.

    there are trade offs to everything, an unfortunate aspect of modern discourse is that we can’t acknowledge there are costs to any of our proposals.

    > Unfortunately, Project Ara never made it into our pockets, as Google canned it back in 2016.

    even if it hadn’t been google, everyone knew Ara was never going to work/be cost effective.

    how many consumers ever upgrade/replace the aspects of their devices which _can_ be upgraded? plenty of (non-apple) laptops can still upgrade the memory, drive, and probably the wifi for just a couple of screws. not many people bother (and then they all hop on hackaday or hacker news and assume everyone does such things).

  5. LG tried for limited modularity with their LG ‘Friends’ modules for I think the G5. It was a total bust. They didnt release most of the modules they intended to and the whole thing was just a bit disappointing. Now most users expect IP68 environmental on their phones so I doubt we’ll ever see that again.

  6. I think there’s an unseen assumption that BRANDS being intermodular is the topic at hand. I would argue that motorcycles are to a large degree modular. Of course when there are “No user-serviceable parts inside” modularity is irrelevant.

  7. Modularity is a marketer’s word. It implies savings and obsolescence avoidance, but in reality is a colossal waste of time and money for anything beyond a child’s learning electronic circuits kits. Instead front-loading expense for modularity is a fools recipe for project failure.

    1. USB and modular gun rails are doing fine. It’s just things that try to be all UNIXy and fully general that don’t work, they don’t solve a problem or a class of problems, they just encapsulate a logical idea and expect you to do everything yourself in a nonstandard ad hoc way.

      Smartphone screens, CPU, and RAM are all very much good enough except OLED that wears out. A good modular device would assume you will be keeping those for 25 years, but allow swapping the radio, battery, and adding accessories like flashlights and projectors and Ethernet connectors.

      Unopinionated modular is not modular, it’s just a pile of parts playing off classical buzzwords like “creativity” and “education”.

      And things like Grove aren’t really modular either, they are meant from the start for education and playing around and totally lack things like supplying 6-12v to motors and lights.

      A real modular grove-type system would probably standardize the voltage at 3.3v, none of this figure it out yourself crap, and have a second 5-pin connector for a separate unregulated Vin, an unregulated Vbus, regulated 12v, and regulated 5v, all with well defined rules requiring never supplying power without reverse protection, defining that chargers must charge the internal bat from Vin and supply power to Vbus if able, etc.

      It would probably also have a 6-pin connector for SPI/UART/GPIO/etc.

      If you list out 50 common use cases, you can design a standard that covers all of them cheaply. But instead they seem to try to be as general as possible and only include the bare minimum in the actual standard, which means when you go to actually do anything, you have to do it ad-hoc or define your own extensions.

      Ditch the UNIX and modular is awesome!

  8. I would prefer at least some modularity in household appliances. Standardized power supplies, motors and drivers (f.e washing machines) to simplify the repair and reduce the necessary amount of spare parts to keep things going.

  9. I really would love to see battery for power tools to be normalised.
    I mean every manufacturer have his own format, but in the end, inside, it’s mostly the same LiIon battery and voltages.
    It would be super easy to make a universal battery connector, and would cause less headache when you are looking for a charger or even a new tool (I love thi one, but I have all the batteries in the other brand. What should I do?)

  10. Of course modularity in a spaceship the aize of a house is much easier than in a mobile phone and fwiw houses are very modular. Need a light switch? Take it out of your house and put it into mine and it will work. Similarly ypu can argue that the router for your internet connection is modular, quite easily exchangeable between houses.
    I am sure when technology improves modularity will be possible to achieve in smaller hardware and it will be taken advantage of.
    Maybe the key to modular phones is a miniaturized wireless very short range protocol for data and power? The there are no exposed connectors and reliability will be improved.

Leave a Reply to PatCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.