Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.
Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.
Last time i worked with an ASIC was way back in the late 1980's. The project was a design for a single chip switching AC synthesized converter controller with 1 percent output distortion and high 90 percent or better efficiency. The single chip was an ASIC. That was so long ago though i hardly remember anything except we had to figure out what parts should be doing what logic functions and what it would take to get them to do that.
But did you mean designing with raw ASIC's as in that example or did you mean taking part in some phase of the making of the ASIC's themselves?
ASIC design is pretty much a software exercise these day. The most significant different areas are analog, RF, or digital design. All work with computer modeling and simulation. (Cadence, Simulink design tools for example). A software design tools seat for IC design can run $20k to $50k per designer.
Most digital designs are verified with FPGA implementation prior to committing to custom silicon.
Small geometry devices (sub-micron) semiconductor design masks are very expensive. A set of masks for a 65 nm chip can cost close to $1M. So screw ups in the design can be expensive and 300 mm wafers of sub-micron design are only for well financed players.
Are we talking the same kind of ASIC? The ones we used back in the 1980's were programmable, such that they programmed similar to how PROM's program, by blowing out tiny "fuses" in the die. Of course you might be talking about making the chips themselves before they are programmed.
The chips came with a bunch of logic and flip flops, and they were connected in various ways. But to get them to function the way we wanted them to, we'd have to figure out what fuses had to be blown out. Once done, the flip flops could be arranged into a string of binary counters or BCD counters or whatever, along with logical gates connected in various ways due to the fuse blow outs.
There were different part numbers which had different overall functionality so one may have more logic gates or flip flops then another. You had to get a part number that had enough flip flops and gates to do the job you needed. But the parts were always the same not custom. They were only custom after being programmed, just like a PROM. So it was never a matter of making an entire custom IC chip.
It's been so long now though i hardly remember any details.
There were different part numbers which had different overall functionality so one may have more logic gates or flip flops then another. You had to get a part number that had enough flip flops and gates to do the job you needed. But the parts were always the same not custom. They were only custom after being programmed, just like a PROM. So it was never a matter of making an entire custom IC chip.
It's been so long now though i hardly remember any details.
I worked on several ASIC prototyping systems in the 1990s for a Japanese company. We had a chip design engineer that would take the description language file from a customer (using a then very cool ULTRA-SPARC workstation). That would be converted to a custom masking markup language in a small Unix mainframe to be feed into a ION beam pattern drawing device (direct-write e-beam tool). By using this we could skip most of the critical photo-mask reticle steps and directly write mask patterns to the photo resist on the wafer so we could make custom chips with under a months turn-around for about $50,000 per device in small batches (a few 6 inch wafers) if the OEM design rules were correct and the machine was in operational (my job).
This was 20 years ago, the modern tools in an ASIC R&D fab like Intels foundry can make custom 22nm designs for anyone with the bucks but it's a very small part of the business. https://www.ebeam.org/docs/ebeam_novel_approach.pdf
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
