Tessel 2, A $35 Linux Computer That’s Truly Open Source

We’ve seen the first version of the Tessel a few years ago, and it’s still an interesting board: an ARM Cortex-M3 running at 180MHz, WiFi, 32 Megs of both Flash and RAM, and something that can be programmed entirely in JavaScript or Node.js. Since then, the company behind Tessel, Technical Machines, has started work on the Tessel 2, a board that’s continuing in the long tradition of taking chips from WiFi routers and making a dev board out of them. The Tessel 2 features a MediaTek MT7620 running Linux built on OpenWRT, Ethernet, 802.11bgn WiFi, an Atmel SAMD21 serving as a real-time I/O coprocessor, two USB ports, and everything can still be controlled through JavaScript, Node, with support for Rust and other languages in the works.

Instead of going the usual route and determining the future of Tessel through market research and the apparent pragmatism of whoever happens to be in charge, this week Technical Machines did something wonderful: the ownership and direction of the Tessel Project is now independent of Technical Machine. This makes Tessel a completely open source and community driven platform for I0T, robots, and whatever else would benefit from an open source community disconnected from hardware.

The Tessel project is completely disconnected from manufacturers, something the Arduino project has been struggling with for the last few years, unbeknownst to most of the founders for most of that time. It’s a boon for the open source community, and something that should see an incredible uptake in the next few months.

22 thoughts on “Tessel 2, A $35 Linux Computer That’s Truly Open Source

  1. At minimum I’d expect a “truly open source” board to use a SoC that has detailed documentation available to everyone.
    You know, like the BeagleBone with its 4973 page AM335x technical reference manual or the WandBoard with its 5817 page i.MX6D/Q reference manual (and these still skip things like the GPU).

    1. That’s a strong point.

      I went looking for the sheets for the MediaTek MT7620n. It turns out this WiFi SoC was designed by Ralink and released in 2012. Then Ralink got bought out by MediaTek and this chip must’ve been one of the less important projects.

      The only mention of the chip on MediaTek’s site is a press release — no PDFs, no spec sheets. The closest thing I found was a samizdat site — http://www.anz.ru/files/mediatek/MT7620_Datasheet.pdf

      I wonder how Tessel is getting a supply of these chips, and when that would run out. No one needs another Betamax.

      Bummer. I was excited about this diving board for IO that runs Linux as an RTOS. They’ve got modules, they’ve got precompiled tools for popular languages, they’ve got all sorts of ways for things to plug in and talk to other things. It’s like a weather station: you could spend a little while doing maintenance, but you really just want the raw data to hand to the database on the bigger box.

      Instead I’m back to the suffering of… oh wait, not suffering.

    2. true open source doesn’t need documentation, it’s just nice to have. I don’t believe written anywhere in any open source license like GNU GPL requires some form of documentation

      1. No, the GPL doesn’t require documentation, but we are talking about hardware and using a basically undocumented SoC is like linking a binary blob into your open source application.

      2. Open source needs some serious documentation. As long as the programmers still expect the end users to compile or figure what stuff from the source code or documentation or website info out of date, they would never be good enough for the average person. There are some very well documented projects, but they are the exceptions not the rule.

        Without proper hardware documentation, you can’t really program the low level hardware correctly. Yes you can try, but never quite sure what the bits does or have any special sequences or interdependences for initialization.

      3. Are you also going to say that real men don’t eat quiche? Will you then reject the scrambled eggs and bacon in a pie crust so long as I don’t use the ‘q’ word?

        The whole point of open source anything is that we have the SOURCE. We have the original code, we have the specs, we know which circuits start where and go where.

        Otherwise we’ve only hacked a black box. This isn’t a bad thing — we do it all the time. However we also accept with that hacking glory that we tossed the warranty, can’t ask the creator any questions without making their lawyers get on alert, run the risk of some DCMA or other nonsense but still legal violation.

      4. good luck getting people to contribute if they have no idea what to do with the chip.

        or do you think just because there is some rudimentary code in a repo, people will go out of their way to buy a new device and then magic elves will make it work for you?

        1. ‘or do you think just because there is some rudimentary code in a repo, people will go out of their way to buy a new device and then magic elves will make it work for you?’

          Yes, isn’t that what happened with the Raspberry Pi? If you market to the right people, they will do just as you said.

      5. Asking the user to read the source code is like asking an average person to do file corporate taxes for a complicate case. Hey all the tax rules are right there, you can read it, understand, apply and get around it, right?

        The self centered programmer centric view has to be changed to an end users if you want open source to go more main stream. Otherwise you are only writing code for fellow programmers.

        1. true.
          For most programmers open source is a means to an end, not a goal in itself. As a development model it works very well, given the ubiquity of its poster child Linux. Most users of those instances are not asked to read the source, Luke. Instead they just use their router or android phone with minimal instructions. You might have fallen into the self-centred trap by assuming the open source user is just like you.
          I share the concern that a badly documented processor is not ideal, but if you consider what happened with the esp8266 I think the community(someone) will work around these obstacles.

  2. At first I didn’t understand: it’s slower than an RPi or BBB, so why should I be fascinated?

    Instead it’s not a tiny computer; rather, it’s a great diving board for IO that runs Linux as an RTOS. They’ve got modules, they’ve got precompiled tools for popular languages, they’ve got all sorts of ways for things to plug in and talk to other things. It’s like a weather station: you could spend a little while doing maintenance, but you really just want the raw data to hand to the database on the bigger box.

    I get it. If I ever have free time again, I may really like it.

    1. … not linux. No adhoc wifi, no Ethernet, no full-fledged TCP/IP stack, no USB devices, and only so-so support for PWM (this board has uC for that too). You could go on. But sure, if you want wifi cheaply and something that is capable of some simple logic esp8266 is the best thing there is for your money (at least until you run into problems with I/O, but then I guess you would add additional microcontroller).

  3. I wish we could have an OpenRISC, or RISC-V board that was open all the way down to the silicon. But I wonder, even if we did have one, how would we know if the CPU was actually following the design? If the fab were to slip in a secret backdoor, would third parties be able to spot the deviations with an electron microscope, or something else? Sorry if this is slightly off-topic, but I’ve been wondering for a while…

    1. The HDL side can still be open sourced and allow others to contribute/modify. Can’t do much about the closed source tools and IP under it. There are huge investments that the vendors poured into the chips which they make it possible even for individuals to implement custom logic at tens of dollars. Essentially every of their paying customers pooled together to make this possible.

      You won’t know enough nor have the tools nor $$$ to contract a fab to make your own ASIC. So that’s the trade off – freedom to make your chip that is powered by closed source or go on a high horse and do nothing but whine.

Leave a Reply to jelleCancel 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.