Precursor’s Custom PCBs
bunniestudios.com
bunniestudios.com
"Out of the box, the FPGA will be configured to work as a 100 MHz, 32-bit RISC-V CPU, but it can be configured to emulate many older processor types. Just keep in mind that it tops out at 100 MHz speeds. Huang says the system will be more powerful than a Palm Pilot or Nintendo DS, but it’s best suited for developers who want to create a single-purpose device completely from scratch…"
That's nice, but it's not clear what it's for. You can make an slow, expensive, tamper-resistant feature phone out of it. Price is around $512. What else?
Summary: ASICs are built by taking existing designs and disabling but not removing stuff because designing new circuits is extremely expensive and error prone. They don't have the money to build an ASIC yet. They also do not trust existing designs. An FPGA both is allows them to iterate towards a final design easier (and with user validation by using it) as well as take advantage of software and cryptography to secure the final SoC design which runs on the FPGA.
The big microprocessors and SoCs that you'd find in a laptop or smartphone have completely escaped anybody's ability to scrutinize the design. Performance is excellent, but there's so much stuff baked into the silicon that nobody actually knows about all of the quirks and hidden misfeatures. You can assume that most chips in this category are backdoored.
At the other end, most "secure" microcontrollers are a joke. Publicly available datasheets just list a pile of checkbox items for lazy compliance people. You need to sign an NDA to have any hope of actually talking to the hardware, and then both the designer and user are expected to trust that the vendor's secret implementation isn't utter trash. I've had to design products that relied on these microcontrollers and I just don't trust them -- This is the main reason I'm not a big fan of security tokens like Yubikey. Given the ease and regularity with which these chips are broken, you can assume that most chips in this category are crap.
So where do we go from here? Desktop performance is nice, but there are lots of applications that don't need it, especially stuff like basic communication and password management -- two areas where we also don't want to trust to an almost-certainly-compromised SoC. Can we build a usable computing environment around a microcontroller? Probably -- you can get beefy microcontrollers with lots of memory and pseudo memory protection that can "run" something like Windows 95. FPGAs in the "reasonably accessible" category have also gotten good enough that you can plop down a few processor cores, peripherals, and bus interconnects. More importantly, the open source tooling and SoC-builder platforms like LiteX have seen enough development that experimenting with this sort of thing is actually practical now.
Also, the funding goal / unit price is high because one of the explicit goals for the campaign is for bunnie and xobs to spend about a year working on the software, an aspect of crowdfunded electronic gadgets that's frequently overlooked. This thing isn't going to be useful if they just crank out the hardware as cheaply as possible and toss it over the fence. They have a history of doing really good work, and the fact that they're interested in actually taking the time to develop this further leads me to believe that they're likely to do a good job. I don't think I would have backed the project if someone else was behind it.
I would consider the project a success if I can eventually store my password database on a precursor, so that I don't need to keep copies on untrusted devices like phones and laptops. There are some usability questions around how this might work, and whether there's a way for the precursor to authenticate on behalf of the device in some cases, but I think it's pretty doable, and I like this approach a lot more than something like a Yubikey.
The order of magnitude lower speed is an expected byproduct of the hardware. If you don't have a use for it, don't use it.
In contrast, even today, sending any level of drawing to a machine shop at the time would have immediately resulted in a phone call to talk about exactly how I wanted something made, and why it couldn't really be made that way. While there are "quick turn" shops that can work directly from files, it's definitely a more recent development.
Even though "Gerbers" were a deeply primitive format, they got the job done.
At my workplace, one of the old-timers showed me the automated wire-wrapping machine, long since retired, on its way to the scrap yard.
Unfortunately, I can't definitively explain why it was better beyond some theoretical platitudes -- in the end, it was literally a process of just making educated guesses and measuring stuff until I found something that worked.
I've tried modelling software in the past, but the results don't match well with measurement, and the paid licenses for modelling software are by far more expensive than the cost of copper tape and a good knife, and then running a few prototypes (and the free stuff I could find wasn't even close to meeting the task). Not to mention the time -- the sims take a long time to run, and building accurate models is also quite time consuming.
Sort of a dissatisfying reality, but that's how the sausage got made.
I'm not sure which changes you made exactly, but I wouldn't be surprised if the end effect was thickening the relevant portions.
Sounds about right. Antenna design is magic.
It is a theme from like...almost 20 years ago. Minimalist, simple. I haven't been able to find a "new" theme that I'm happy with, they have too many features and too much javascript/PHP in them. But yah, I could see how this theme does not play nice with all combinations of mobile+browser.