Ask HN: Why isn't there an open-source FPGA? | Hacker News Reader

Ask HN: Why isn't there an open-source FPGA?

17 pointsorndorffgrant10 comments

I'm asking about the FPGA itself - not boards or IP cores or toolchain.

The only thing I've found while searching is https://git.slaanesh.org/kfpga/kfpga which is still very new and small, but inspiring.

What are the major hurdles that have prevented people from creating an open source FPGA design?

View on HN

CPUs you can implement on FPGAs or something else whereas FPGAs are really only useful if you can get it made. Getting the design rules from a foundry usually involves an NDA and you will need (proprietary) EDA tools.

Also the utility of open source CPU cores is much higher than an open source FPGA since you're not going to be able to customize it anyway because of mask costs.

I hear that repeated over and over. And while it is true now, it could change if concepts like

[1] https://www.yokogawa.com/yjp/solutions/solutions/minimal-fab...

catch up a little technology-wise, and take hold.

edit:

[2] http://pmt-minimalfoundry.businesscatalyst.com/minimal-found...

I see they have a branch office in Moscow now

[3] https://minimalfab.eu/

This FPGA is not open source by design, but this project seems fairly complete: http://www.clifford.at/icestorm/ "Project IceStorm aims at documenting the bitstream format of Lattice iCE40 FPGAs and providing simple tools for analyzing and creating bitstream files."

Lattice (the FPGA's manufacturer) seems open-minded regarding such efforts: https://twitter.com/latticesemi/status/1269115302140231682

I like the idea. I do a lot of FPGA work and have often thought the same thing. The link is very interesting. I think part of the challenge would be how do you get the foundry capacity to actually produce it. And how to pay for that, who will buy it etc. etc. Those questions are beyond me, but I do think that if it were to get momentum there could be a lot of interest.

Xilinx has like 2,000 FPGA patents, proudly displayed on plaques on their cafeteria wall.

So you would have to navigate around those.

They were founded 36 years ago. How long do they last? How many of them are meanwhile void?

It's a valid question.

The current duration is 20 years.

But companies with sophisticated IP lawyers, like FPGA vendors, can create "evergreen" patents by tweaking existing ones to last 100 years or more.

Also, small companies generally cannot afford patent litigation, so most fold after just receiving a letter.

Hollywood is located in LA because East Coast companies didn't want to pay Edison patent royalties. Moving to another state is less likely to work these days.

The laser patent killed commercial laser products until it expired, since nobody wanted to pay the 4% or 5% royalty. (I'm guessing that it wasn't that one royalty that was the concern, but the likely additional demands that would come out of the woodwork.)

Same thing with the Wright Brothers airplane patents. It took war for the Wright Brothers and Curtis to work together, which is why Europe got so far ahead of the US in aviation at the time.

> Moving to another state is less likely to work these days

Would these patents apply in the EU?

The short answer is not automatically, but there is a streamlined process for filing patents in the US and then the EU.

https://www.epo.org/

I suspect that US evergreen patents would be annoying to the EU if pointed out, though.

I don't really get the desire to make your own FPGAs though. They will probably end up being slower and more expensive. It makes more sense to me to work on toolchain software.

That would certainly be a concern. I'm unfamiliar with the details of how patents work, but there are other companies producing FPGAs aside from Xilinx, so it must be possible to create an FPGA without violating the patents, right?

Reply on news.ycombinator.com