Tobi Delbruck
INI WWW
Home Wiki Motivation History People Projects Publications Resources Fun Contact

The Physiologist's Friend Chip

Downloads Availability Significance Users Links

The Phyiologist's Friend Chip electronically emulates cells in the visual system and responds realistically to visual stimuli.

New: complete design available

If you are a physiologist working on the properties of the early visual system, you could use this chip as a simple substitute animal.  You could use it to train students to run your rig. Or you could use it as a set of realistic and known receptive fields with spiking or analog responses for developing new spike-triggered averaging techniques. Or when debugging a new setup, you can use it simply to make sure everything runs as expected. The idea is that experimenters can test nearly the complete loop -- which generally involves stimulus generation, recording, and analysis, without using actual animals. Bugs in the large portions of the setup can be found before any animals are used..

If you teach about how the brain works, the chip is a captivating way to involve the students in demonstrations of a number of basic properties of the visual system: adaptation, complementary (ON/OFF) coding, spikes, complementary push-pull input to cortical cells. Using an overhead projector, you can conduct experiments on this model visual system.

The chip outputs an ON and OFF retinal ganglion cell and two orientation-tuned cortical simple cells. Analog membrane potentials are also output and can be used to check spike discrimininators.

A software simulation that models more types of cells is the Physio Friend program.

Physiologist's friend system

The Physiologist's Friend Chip

Downloads

A user guide shows how to use the chip for demonstrations and in the lab Download PDF  
Slides that we have used in conjunction with the chip Download PDF Download PowerPoint
T. Delbruck, S.C. Liu. (2004). A silicon visual system as a model animal. Vision Research, vol. 44, issue 17, pp. 2083-2089 (preprint; copyright restrictions prevent putting journal version here.)   PDF of vision research article
NEW Physiologist's Friend Chip design See this page  

Availability

As of August 2009 we are in the process of building up the final lot of approximately 15 of the Physiologist Friend Chip systems based on the friend23 chip; the process in which this chip was designed has been discontinued and no more silicon can be produced.   Please contact Tobi Delbruck (tobi@ini.phys.ethz.ch) to inquire about availability. The most recent shipment of 3 systems went to the NatLab at the Freie Uni Berlin for hands-on use in teaching children and undergraduates about the nervous system:

Significance

Why do I think this chip is a significant advance in neuromorphic engineering? It is true that it is a very simple chip that uses circuits and architectures that for the most part have been around for a long time. So why do I think it is important?

  1. A chip doesn't have to be complicated to be useful. It only needs to do something useful.
  2. The chip really works well enough to sell to naive users and they actually use it. This is the first example I know of a neuromorphic chip that models real neural behavior that has been sold and routinely used outside the creator's lab.
  3. There are no adjustments on the chip; it is completely parameter-free. That means it is just like an opamp or ADC; you turn it on and it works. The dozen internal bias currents, which span 6 decades, are generated on-chip by a bias generator. Every chip acts the same; there is no calibration necessary. To appreciate this, you need to understand that in the field of neuromorphic chip design people always assumed that this part was trivial and had no qualms about throwing a mole of parameters onto the chip, which would sometimes take months to optimally tune up. Well, it wasn't as trivial as Andre and I thought, and it wasn't something that the traditional analog design community had figured out either.

Users

The Physiologist's Friend Chip is used by about 15 labs around the world.

David van Essen Washington Univ.
Gilles Laurant Caltech
Kevan Martin Inst. of Neuroinformatics, ETH/Univ. Zurich
Daniel Kiper Inst. of Neuroinformatics, ETH/Univ. Zurich
Matteo Carrandini Smith Kettelwell
Rich van Wezel Univ Utrecht
Florentin Worgotter Univ. Stirling
Matthias Henning Univ. Stirling
Natalie Hempel de Ibarra Freie Universitaet, Berlin
Rudolf Menzel Freie Universitaet, Berlin
Lyle Borg-Graham Neurophysique et Physiologie du Systeme Moteur CNRS
Ulf Eysel Ruhr Univ. Bochum
Pam Reinigal UC San Diego
Wyeth Bair Univ. of Oxford
Bernd Porr Univ. of Glasgow
Henry Markram EPFL, Lausanne, Switzerland
Lauriston Kellaway University of Cape Town
Sebastian Seung MIT
Prakash Kara Univ. of S. Carolina
Petra Skiebe-Corrette Freie Universitaet, Berlin
Sorinel Oprisan College of Charleston, South Carolina

News

August 2009: New boards are being built! This will be the last lot because MOSIS has shut down the 1.6u AMI process in which this design has been fabricated.

Nov. 2006 & Oct 2007: Daniel Kiper, Stephan Roth, and Tobi Delbruck used 4 Physio Friend chips in a biology block course first for understanding receptive fields and later for quantitative measurements of photoreceptor response properties.

Students learn to plot receptive fields of retinal and simple cells. Dan Kiper is in the right photo.

Links

Matthias Hennig and Bernd Porr at the University of Stirling have used the chip for demonstrations and instruction. They have developed a linux-based USB data acquisition device.

Return to PhysioFriend Home page. Go to PhysioFriend Program.


August 21, 2009
Home Motivation History People Projects Publications Resources Fun Contact