{149} revision 12 modified: 01-03-2012 03:23 gmt

IEEE-01258173 (pdf) Wireless implantable microsystems: high-density electronic interfaces to the nervous system - January 2004.

  • very impressive!
  • based on the old / well established beam-lead technology (see the image of the paper at the bottom of that page).
    • required 20 years of development to create an etching process with sufficient yield, though. Microprobes have been in development since 1966.
    • Silicon is slowly attacked by saline; however, the use of a boron etch-stop to define the substrate virtually eliminates such erosion.
    • Silicon dioxide is known to slowly hydrate in water, but this can be mitigated by CVD of silicon nitride / silicon oxide stacks. Polysilicon can be used too, since it forms a tight bond with silicon oxide, keeping water out.
      • Why don't they just seal it with a known impermeable plastic/epoxy/whatever? (They do, later) Utah probe is sealed in parylene.
    • Shunt capacitance is negligible compared to site capacitance; heavy substrate doping minimizes electrical or optically induced noise & virtually eliminates crosstalk.
    • (Of course) Silicon allows amplifiers and circuitry to be formed at/near the electrode, eliminating the need for (some) interconnects.
    • Silicon ribbon connectors cannot be made much longer than a few centimeters. 4um thick silicon cables are 100x more flexible than a 25um gold wire (!!) - but that is out-of-plane; they are relatively weak for in-plane stress.
  • Gold has a maximum charge delivery of 20uC/cm^2 ; platinum, 75 uC/cm^2 ; iridium oxide, 3000 uC/cm^2.
  • Glass can be hermetically bonded to silicon if both flat clean surfaces are put in opposition with a high voltage (1500V) placed across the interface at an elevated temperature (400C). These packages have been shown to be stable and inert in guinea pigs.
    • Silicon nitride, thin metal films, and metal films over polymers are all attractive coatings for probes (with no mention of biocompatibility); they last decades in salt water, and eventually succumb to pinholes.
  • Silicon probes outperform microwire arrays by a factor of (up to) 50 in terms of volume of tissue displaced / recording site. Michigan probes are typically 15um thick x 60um in cross section.
  • they tend to use many more recording sites than recording channels, hence, have a low expected yield. e.g. they have a 1024 site electrode (see the cool figures!), and can record from the best 128 of those. good idea, reasonable strategy, I guess.
    • they demonstrate that it is not too hard to remove the artifact of multiplexing on their systems - the multiplexing noise is below electrode noise.
  • talk about spongifying their iridium electrodes using current pulses in a PBS solution to (apparently) lower electrode impedance.
  • talk about drug delivery too
  • describe the exact manufacturing procedures that the Michigan arrays are created, including the critical back-etch (which i don't exactly understand).
  • describe the circuits used to amplify low-level neural signals.
  • Their charge-redistribution ADC is okay - 250ksps, 9b resolution, 1.4mW from a 3v source. Commercial ADCS are better - AD7467 is 0.6mw, 200ksps, 10bits. (though it scales up to 1.3mW @ 3V, 200ksps ; 0.36mW @1.8V - so the devices are comparable.)
  • some of the (very tiny) electrodes have 'holes' (!)
  • also have wireless microstimulators.
  • demonstrate long-term recording: 383days post implant in a rat & there are still many cells (though the figure is not that great, it is easy to understand) (this might be an exemplar)
  • associated website?
  • (quote:) "For ranges of a few centimeters, the high attenuation of RF signals in biological tissue dictates carrier frequencies below 10Mhz." Their solution is to use FSK with octave jumps in freqency & modulation rates up to 60% that of the carrier frequency.


WISE, K.D. and ANDERSON, D.J. and HETKE, J.F. and KIPKE, D.R. and NAJAFI, K. Wireless implantable microsystems: high-density electronic interfaces to the nervous system Proceedings of the IEEE 92 1 76 - 97 (2004)