Raven goes to fabrication, hooray!

I had an unfortunate problem three months ago when I submitted the Raven project to X-Fab for the May shuttle run, and discovered that the shuttle run does not support the 6-metal backend process that we implemented on the efabless platform as EFXH018C. Instead, they support a 4-metal stack (way too few metal layers to route the RISC-V core) and a 6-metal stack with thick top metal.

Between May and August, I rushed to implement a new process variant EFXH018D which is the 6-metal stack with the thick top metal that is compatible with the shuttle run. This required substantial re-work to the Raven chip, the qrouter tool, and the Open Galaxy platform techfiles.

The new tapeout date was August 3, and although there was a bit of last-minute shuffling around of forms, I am happy to report that as of today I have approved the mask data for the shuttle run, and it is actually happening!

The schedule on the X-Fab (public) website shows that this shuttle run will be done on December 28 (yes, it’s dreadfully slow), so once it goes through packaging we should have parts to test early in the new year.


Raven was returned from fabrication over the holiday and is currently in packaging; a test board has been prepared, and everything should get assembled, delivered, and be ready to test some time mid-January.

This morning I successfully validated operation of the Raven chip: First time silicon success with open source tools! The test board has a few errors and a few quirks, and in retrospect (of course) a few things could have been implemented better, but the Raven chip is running code. The processor speed is 100MHz (12.5MHz crystal on the circuit board and an 8X clock multiplier PLL on the chip), and I have tested and validated the 16 bits of general-purpose digital I/O (GPIO) and the UART, as well as the 3.3V housekeeping SPI. All of the analog functions (other than those supporting the processor, such as the POR, PLL, and 1.8V LDO) remain to be tested.

I cannot overstate how happy I am about this project. First time silicon success is a BIG goal in the industry, and to have first time success with a project as big as a microprocessor is an important validation of the open source tools used to create it. Not to mention the short time from concept to execution, and the very, very short list of people directly involved in the project.


Congratulations - this is a delight to read!