[onerng talk] Are zero-NRE low-cost mixed-signal ASICs interesting?

[Paul Campbell]

On Wed, 01 Oct 2014 17:02:01 Bill Cox wrote:
> I am trying to find out if there is any need for board-level designers
> out there to be able to create small mixed-signal ASICs.  I'm not
> talking about an iPod-Nano on a chip, but simple arrays of capacitors,
> resistors, transistors, a few logic gates, and maybe some amplifiers -
> not coincidentally a good platform for building Infinite Noise
> Multipliers  The die would be tiny, and each would have the same
> components.
> Designs would be configured with custom routing.  The minimum order
> might be 1,000.

> So, for example, a chip you could design using say 100 0.1pF caps, 300
> 6K Ohm resistors, maybe 50 analog N and P mosfets configured for
> analog (wide gates), maybe 20-ish T-gates, and 20-ish logic gates
> (NAND/NOR/INV), a couple of op-amps, and maybe 16 pads, and come in
> some tiny 16-pin surface mount package.  It might even have 1K-ish
> gates of real logic, and 128 FLOPs, and even a small block of SRAM, if
> people think it should.  The resistors would not be very accurate, but
> they would match well.  Same thing for the other components.  It would
> come with free design tools, likely based on existing open-source
> tools.
> 
> Something like this I think can be done for under $1/chip, in
> quantities of 1,000. 

I haven't built chips for a decade but $1 in those low volumes seems 
impossible

 (you're likely going to need CAD tools in the ~$100k range to design/lay out 
the thing and that adds $100 to each chip in those volumes right there) 

NRE for tapeout, packaging, tester time are all going to add up - from memory 
last time I built (much bigger) chips silicon,  tester time, packaging cost 
very roughly 1/3 each of our finished silicon cost 

I suspect you need to have a volume in the millions to get to $1 - however as 
I said it's been more than a decade since I was building silicon ,things may 
have changed - if I can knock out 1000 small chips like that for ~$1k I'll 
stop looking at fpgas ....

> I am trying to figure out if this is a good fit
> for helping enable the Internet of Things.  It might be useful for
> simple sensor interfaces, for example, or reducing part-counts and
> size.  Of course, you guys here know as well as I that the Internet of
> Things has a major security problem...

Since I nominally do IoT stuff in my day job I have to say that mixed-signal 
stuff is probably iffy, depends on the process, low, really micro power is 
everything - if you can't get a 3 year battery life out of your 'thing' it's 
going to be a burden to its owner

	Paul