Re: [buildcheapeeg] I published current design at SourceForge

From: Jim Peters (jim_at_uazu.net)
Date: 2001-11-30 12:20:44


A fantastic bunch of links, Doug !! (I'm going to follow up some of
those wavelet links). How about putting those links on the web-site
as `research material' or something like that ?

Just a couple of things I picked up on:

Doug Sutherland wrote:
> 3) Jim has achieved results far superior to all
> other low-cost EEGs I have seen, band width
> is 0.1-100 HZ with sampling at 120 Hz. This
> is better than all of the low-cost EEGs I
> have seen including commercial ones like
> WaveRider, Brainmaster, POD, etc.

A response of 0.1-100Hz is not possible when sampling at 120Hz. In
theory the maximum is 60Hz, but even then detection of a 59Hz signal
would be rather unstable (going in and out of phase with the sampling
frequency). I like the recommendation you quoted from the "IFCN
Standards" of using a higher sampling rate, say 200Hz.

> 5) The 555-timer based design is interesting, but
> my concern is that timing/interrupts are a big
> issue (must be DOS only to disable interrupts).
> I am wondering what the band width of this
> design is (frequency response in Hz) and what
> the CMRR looks like. Anybody know?

If we have a signal sampling frequency of 256Hz, and we require 12-bit
data, then we need to represent 4096 possible voltage levels through
the pulse width. This requires an accuracy of 1/(256*4096) == 1us
(i.e. a 1MHz clock reference in the computer). If we settle for
8-bit, this only requires 1/(256*256) == 15us (i.e. a 65536Hz clock).

The overhead of monitoring voltage levels or interrupts could be
bypassed by feeding the pulse-width-modulated signal to a soundcard
input which will buffer it for us. Processing would then be very
quick. However, we are then limited to 44 or 48kHz sampling which
would give us a maximum of 8-bit data and 188Hz sampling of our
signal.

This is not so good as other solutions, such as serial cables, where
12-bit data and 256Hz sampling only requires 3000-4000 bps per
channel, and gives lower processor overhead than either of the
alternatives above.

Using a soundcard might still be possible by modulating the signal
with a sine-wave or square-wave carrier (amplitude modulation). This
doesn't have the same problems as pulse-width modulation, and the
carrier effectively sets the sampling rate. However, there are
probably other problems with this approach -- would the signal have to
be offset to differentiate +ve and -ve peaks ? (AM radio must have
solved this.)

Anyway, I think you've all already discussed this. It would limit
most people to two channels anyway (unless you multiplex the
signals ...), and it doesn't give you a two-way communication.

Hopefully I should be able to use the finished design whatever you
decide to go with. I have two Linux-based laptops, neither with IrDA.
One has audio and serial inputs, and the other has only USB ports.
Hopefully I should be able to pick up a cheap USB<->serial or
USB<->IrDA adapter (there are a few listed as supported in Linux).

I'm really interested to see where this project is heading.

Jim-P (not the other Jim)

-- 
Jim Peters (_)/=\~/_(_) Uazú
(_) /=\ ~/_ (_)
jim@ (_) /=\ ~/_ (_) www.
uazu.net (_) ____ /=\ ____ ~/_ ____ (_) uazu.net


This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:32 BST