Robotic Application Processing Cores. We are building advanced motion and motor controls on FPGA and ASIC, fully open source from bits to gates.
by SJ Kelly
Welcome to the June monthly update on RAPcores. Development in the month of June has been focused on continued support for SoC deployments and eFPGA systems.
If you are just hearing about RAPcores, welcome! We are developing Robotic Application Processing cores, with the goal of enhanced robotic proprioception through motor control. We are leveraging hardware and software co-design to create tightly coupled motor control and sensing solutions for robotics and advanced manufacturing.
In June I finished the preliminary work on register mapping. This yielded a net 20% reduction in resources and has given interfaces far more access and control to internal registers inside RAPcores. For context only about 50% of internal registers were covered in the original design. Now we can access all of them for fewer resources. A big win. There are still some final integrations to do for fixed point. I suspect we are on target to net 40% resource savings on the system, with a far more maintainable and simple design. For context the register mapping work deleted over one hundred lines of source from our state machine, and greatly streamlined adding new capabilities. Once the fixed point work is merged in we can begin exploring different control modes and more telemetry.
The last week of June was entirely dedicated to bringing up the Quicklogic EOS S3 chip and Symbiflow. The EOS S3 is a cool little chip with a Cortex M4F and eFPGA. Internally there is a system bus between the ARM core and eFPGA, allowing us to evaluate how these types of systems will work together in physical hardware. Unfortunately the EOS S3 eFPGA fabric is somewhat under powered for multi axis control, and the Symbiflow tool chain does not seem as polished as yosys+nextpnr. I have started an early port for the EOS S3 in the branch. So far, the port has been relatively easy. In part, our efforts to improve the parametrics of RAPcores has allowed us to scale down certain precision’s to fit on these smaller architectures.
Once our sensing story is more complete, I have a strong feeling RAPcores in IoT applications could be a far more important market than even robotics. One of our goals is to improve the overall efficiency of motor control with better control methods. Our parametric core designs we should be able to support a wide range of applications in motor control.
QuickLogic is the first publicly traded FPGA vendor to commit to a fully open source stack. There seems to be many interesting chips in their pipeline, so they will be one to follow.
I presented our work at the FOSSi Foundation Dial-up on June 15th. The video can be seen here.
The past few months we have been designing some hardware examples to the evaluate costs and form factors. One of these was a minimal and cost effective four axis control board, designed to be paired with a RasPi. Unfortunately we designed the board around the Ice40 HX8K, which is somewhat difficult to obtain these days and priced out by the ECP5 25k part. Moreover, the availability of FPGAs is low right now with the semiconductor shortages. The coming months will likely be a holding pattern on hardware design until we get news on availability and future products.
June was mostly about some big projects wrapping up, and planning for new ones. July will be full of more experiments and open source developments, so stay tuned!
tags: