[sdiy] $100 Chip fab service

[Florian Teply]

Am Wed, 02 May 2018 16:52:39 +0000
schrieb cheater00 cheater00 <cheater00 at gmail.com>:

> Why would the characteristics differ? They're monolithic on a single
> chip, and chips from different wafers will likely have the exact same
> characteristics. Sure, stuff needs to be characterised, but once you
> have that you're free to go. Bear in mind you get several orders of
> magnitude more transistors than in a normal synth module.. so this
> opens a lot of possibilities.
> 
Well, characteristics can differ somewhat across a wafer and from wafer
to wafer. But thats usually less that what you'd get from
discrete devices. Problem: As far as I can see nobody actually said
which 180nm process that might be, and what the proposed input for
chip fabrication might be. So we have no real chance to infer *anything*
about what could be done easily, and what would need some effort.

Granted, BJTs can always be done, even in a plain digital CMOS process.
But as these would be parasitic devices, their performance tends to be
pretty poor (think current gain in the low tens at best, and pretty
variable). The fact that this probably will be a "digital" CMOS process
(there's nothing in a CMOS process that makes it inherently digital...)
does not preclude analog circuitry. And being a 180nm process does not
mean one would be limited to 1.8V supplies: Nearly always there are
additional devices for higher voltage IO. I'd be pretty surprised if
that technology was 1.8V only, but would rather expect 2.5V and/or 3.3V
IO Cells.

And the minimum feature size also does not preclude the use of bigger
devices. Also in, say, 65nm CMOS some pretty good analog chips can be
made.

But don't expect to be able to do too much in the area they advertize:
Going by the sheer numbers 350 micron by 350 micron would be the
equivalent of 2000x2000 times minimum feature size. That would be
sufficient for maybe 400x400 minimum size transistors of one polarity.
But these minimum size transistors are of little use for analog
circuitry as these can vary quite a lot. For analog applications I'd
expect no more than a few hundred transistors in total on that area,
and passives also take up space. So even IF there was direct
connections available to an external pin, bypassing the common digital
I/O, the available area is not exactly huge.


> On Tue, 1 May 2018 20:25 , <mskala at ansuz.sooke.bc.ca> wrote:
> 
> > On Tue, 1 May 2018, Ben Bradley wrote:  
> > > I'm wondering what you can do in 1.8V "180nm CMOS technology"
> > > that's not digital. Does anyone even have the analog
> > > characteristics of such small N-channel and P-channel mosfets
> > > (and no BJTs, the heart of just about any exponential converter)?
> > > I'm guessing they only have specs for required voltage levels for
> > > 0 and 1, switching times and fanout and such.  
> >
> > That's be my concern.  Sure, any transistor can be used in an
> > analog way, but if the relevant properties are not characterized
> > and controlled, there may not be much point trying.
> >
> > --
> > Matthew Skala
> > mskala at ansuz.sooke.bc.ca                 People before principles.
> > http://ansuz.sooke.bc.ca/
> > _______________________________________________
> > Synth-diy mailing list
> > Synth-diy at synth-diy.org
> > http://synth-diy.org/mailman/listinfo/synth-diy
> >