A memristor is the lowest energy electronic component. The technology is just at its premisses but it will open a new area of ultra low energy design. 

Background

Invented in 1971 by Leon Chua, the memristor component becomes now accessible for end user. Considered as the fourth electronic missing component, this network mimics synaps plasticity.

Aucune description alternative pour cette image

Due to Covid pandemic, 2020 wasn’t a conventional year wordwide. Instead of stopping its commercial activities , CONCEPTUALISE took the opportunity of the lockdown to start an internal self-funded program around memristors. We believed that memristors will become a key component in near future. After a deep literature research, memristors applications are very wide ( from analog electronic to neuromorphic processor )

The first phase of our program will be to understand as deep as possible this promising component. We decided to split the main program in several agile projects. The first project launch is called SOAP ( Single Operational AmPlifier ). Of course , there is a lot of easy design with classic component to reach the same result, but again the target is to grab experience with this new component to be able to compare apple / apple for the best next steps orientations.

SOAP Board

We identify several programming technique of memristors. To measure & compare these methods we develop simple board. The main board requirement is simple, offering several configuration memristors’s mode and in second step using the programmed memristors into a simple electronic function.

The simple electronic function for SOAP is simple gain amplifier ajustement. Depending on SOAP results a second project will be launched around multistage analog filters.

The different programming methods are

The targeted memristor for this board is a tungsten or Carbon SDC in DIL16 from Knowm inc . Thanks a lot to them to make commercially accessible this kind of components.

The block diagram is the following

SOAP Layout V1

SOAP V1 Results are quite positives : 

  1. Current mirrors topologies are NOT appropriate to program SDC memristors. A voltage-lock in effect was observed each time , which makes a precise programming value very difficult. (Voltage lock-in occurs when the resistance of the device falls below a limit where the current-limited drive circuitry cannot apply a sufficient negative voltage to overcome threshold)
  2. The topology described by “Active programming as described in High precision analogue memristor state tuning DOI:10.1049/el.2012.2295 ” doesn’t fit with our requirements. Negatives pulses generates erase therefore no long period of stable resistor 
  3. The topology described by “Bridge programming as described in A bridge technique for memristor state programming DOI:10.1080/00207217.2019.1692371″ looks promising . We got positive results but due to not precise data sheets of ALDxxx MOSFET we decided to adjust the design.

SOAP Layout V2

New requirements of the design are

  • Four configurations methods:
  • Improved PSU to provide safe power supply to memristor switches (DG445)
  • Antiserial configuration allowed on 4 memristors
  • Precise onboard measurement of Memristor’s Voltage & Current
  • CONF/APP mode by Waveforms interface ( AD2 )

     

     

    In the mean time, in case your are interested by having this board for investigations don’t hesitate to contact us.

    JP Guarrera – Manager & Principal Engineer jp.guarrera@conceptualise.be