On Jul 10, 2021, at 8:58 AM, Rodney Brown via cctalk
<cctalk at classiccmp.org> wrote:
Alan Mantooth, Carl-Mikael Zetterling and Ana Rusu (28 April 2021)
"The Radio We Could Send to Hell: Silicon carbide radio circuits can take the
volcanic heat of Venus" IEEE Spectrum
https://spectrum.ieee.org/semiconductors/materials/the-radio-we-could-send-…
"The average surface temperature on Venus is 464 ?C, the atmosphere is dense with
highly corrosive droplets of sulfuric acid, and the atmospheric pressure at the surface is
about 90 times that of Earth."
(Teflon melts at 327 ?C)
Decomposes, actually.
I saw some articles about high temperature semiconductors a couple of years ago. Those
were just ring oscillators, a proof of concept that SSI was possible.
The following thesis describes the Vulcan II chip
mentioned a bit more
Benavides Herrera, Maria Raquel, "An RS-485 Transceiver in a Silicon Carbide CMOS
Process" (2018).Theses and Dissertations 3067
"The RS?485 was designed in a 1.2 ?m SiC CMOS process technology developed by
Raytheon Systems Limited (UK) called High Temperature Silicon Carbide (HiTSiC?).
The components available in this process include: NMOS and PMOS devices,
on-chip resistors, diodes, and capacitors.
The process key features are given below:
? 4H-SiC process
? N-type substrate
? Supply voltage of 15 V
5? Single metal layer, two layers of polysilicon (one being high sheet resistance poly)
? Operating temperatures greater than 300?C" ...
Curious that they called it RS-485, which is a serial link transmission standard.
"Greater than 300 C" is quite a lot below 464 C, though.
---
In 1988's 1.2?m CMOS process the MIPS R3010 floating-point coprocessor was about
75,000 transistors on an ~8mm x 8mm die.
Are there markets for SiC CMOS devices with large transistor counts?
Doesn't seem likely. SiC is great for heat tolerance, so it's used in high power
devices. But that's not normally a consideration for VLSI.
Watching Curious Marc's video mentioning Triton
missile/Saturn V bit-serial
computer implementations, reminded me of:
Olof Kindgren (2019) Bit by bit - How to fit 8 RISC-V cores in a $38 FPGA board
https://riscv.org/wp-content/uploads/2019/06/12.15-SERV-Zurich-Copy.pdf
https://github.com/olofk/serv "SERV is an award-winning bit-serial RISC-V core"
(RV32I)
(Not an engineer - guessing)
If process limits mean large SiC memories are unlikely, what other technologies
would work in the 400?500?C temperature range? Magnetic bubble memory? Twistors
if threading cores automatically remains infeasible?
Magnetic memory -- bubble, core, or whatever, including core rope ROM -- would require a
core material with a Curie temperature well above 460 C. A quick look says that
"ferrite" has a Curie temperature of 450 C but several other magnetic materials
go much higher. That statement is puzzling though, because there isn't such a thing
as "ferrite", rather there are a large number of ferrite type materials with
different magnetic characteristics.
For cameras could you build vacuum tube sensors
containing SiC devices if useful?
Vacuum tube sensors could be vidicons or the like, but the question is what the leakage
characteristics of photo-emission targets look like at that temperature. If the leakage
is too great the scan rate would have to go way up to catch the picture before the charges
leak away.
That suggests something else: if indeed leakage at those temperatures is reasonable,
Williams tube or Selectron tube memories could be used. Now *that* would be a classic
computer fan's delight. For that matter, if SiC integrated circuits can't handle
Venus, there are always tubes. Those can be made quite small -- some of us are old enough
to remember Nuvistors and acorn tubes. And it would be possible to build integrated
circuits -- tiny versions of the Loewe NF3, with a bunch of tube electrode systems as well
as the needed passive component all inside one vacuum enclosure.
paul