[Cryptography] The world's most secure TRNG

Arnold Reinhold agr at me.com
Wed Oct 8 16:57:43 EDT 2014

On Tue, 7 Oct 2014 21:59 Bill Cox wrote:

> I've reduced the BOM for the parts (not board/assembly/test yet) from about
> $7.00 to $2.60.  Unfortunately, my bandwidth dropped from 1MiB/s to maybe
> 25KiB/s.  Also, I was convinced by the argument above that I had to write a
> driver anyway, so why not put the whitener there? ...

Good call. 

> I removed the FPGA from the design and now only have a USB-to-FIFO chip
> acting in bit-bang mode to control the infinite noise multiplier, which is
> much slower.  I think you guys were right to have me focus on cost.  More
> people will copy my $1.10 in parts (without the USB controller) even if it
> generates only 25KiB/s, than ever would copy my $5.50 1MiB/s TRNG.

The Lattice ICE40 FPGA product page http://www.latticesemi.com/Products/FPGAandCPLD/iCE40.aspx says it has a hard I2C core. Is that only for configuring the FPGA or can it be used to output your random bits? If not, how hard would it be to make a second I2C controller from the excess FPGA logic? It sounds like the ICE40 costs about the same as the USB-to-FIFO, so an alternative FPGA-only I2C model should hit the same price point. I2C would open a different market -- embedded -- that might be much larger and really needs a good cheap random source.

> I've got my $1.60 USB interface chip to configure Lattice ICE40 FPGAs,
> which only cost about $1.50 (both in quantities 1,000).  It seems like that
> would be a fun proto-board by itself.  A $5 FPGA USB hacker board might be
> fun... The Lattice tools to configure it runs a free copy of Synplify Pro,
> which looks almost exactly like it did when I stopped working on this tool
> in 1998.  The schematic generator seems to be about the same as I left it,
> though there was a really good guy making amazing improvements for a while
> after I left.  Time seems to have degraded it back to my version.

There might well be a market for hobbyists, e.g. via AdaFruit. FPGA dev boards I've seen start in the $100 range, admittedly much more powerful arrays but how many hobbyists can begin to use that power? You'll need to package up the design software with some documentation and a few simple examples.

Arnold Reinhold

More information about the cryptography mailing list