2013-today : UCSD NVSL Projects

I must be a glutton for punishment.  I am back at the university again on ‘soft’ money.  I am helping so far mainly with two projects for Prof. Steven Swanson.  Steve’s specialty is in computer architecture, nonvolatile memory systems and the software for their optimal use.  This includes work with Flash, MRAM, PCM and similar new technologies.  See NVSL.

However, one of our projects deviates from that thrust.  We have a team designing a system that will allow novices to create simple electronic systems without having to get their hands dirty with schematics, layouts, Gerber plots, bill of materials, etc..  Its well along, but still under wraps.

My other solo responsibility is testing flash memories as in SATA SSDs or PCIe SSDs to make sure that  they erase properly with no residue that can be used to recover sensitive information.

For these two projects I have written test code and designed component libraries, ported very small microprocessor designs from Adafruit, and designed adapters and motor drivers using PADs and Eagle tools.




I interfaced motor drivers to the Qualcomm Snapdragon processor on an Inforce 6410 development board as a side project.  This board can be trimmed and loaded to be an Arduino Uno daughterboard (lower left) or trimmed and loaded to sit vertical (lower right) or horizontal over the Snapdragon, and  uses either I2C or SPI.


I also spent 3 weeks in the summer of 2014 in Telluride CO at the 20th Neuromorphic Engineering Workshop.  (Ironically, I was there at the 1st and 2nd 20 years ago presenting the silicon cortex board.)  While there I studied the NENGO system and programmed a small robot to spar in an olympics and to instantiate part of a model of Verschure’s Distributed Adaptive Control Theory.

It is now well established that spiking neurons with winner-take-all mutual inhibition and spike timing synaptic plasticity can implement both unsupervised Bayesian learning and inference while growing a generative model of their environment (Nessler, et al. ).  If we can get funding support, I plan to to work to combine a Bayesian model of an associative memory system that can learn and make semantic inferences with a platform for experimentation on robots that can do Bayesian inference.  These robots would be semi-autonomous on land, sea or air.  Associative memory is the key to advanced robotics and the future of computer architecture (‘IMHO’).

Memory research is a funny thing.  Some people would pay anything to remember better.  Others are willing to pay to make sure things are forgotten.  DARPA seems to be paying for both kinds of research.  Anyway, what did you have for breakfast on your last birthday?  You can learn something from watching your mental steps to try and answer that.