THIS IS A LIST OF THE FILES IN THIS DIRECTORY
(ftp://ftp.rush.edu/users/molebio/Bob_Eisenberg/noise/)


1. chnl2.txt    This file contains 100,000 points.  It contains the channel
data without noise AND WITHOUT ANY FILTERING.  There are 812 openings.  The
open channel current amplitude is 1.0.


2. chnl2+noise.txt    This file contains 100,000 points.  IT SHOULD NOT BE
ANALYZED (see below).      It contains the channels
in chnl2 with noise added BUT WITHOUT ANY FILTERING.  The rms
value of the noise is about 3.18.  The open channel current amplitude is
1.0.  Note that the power spectral density (PSD) of the noise has a shape
that is appropriate for more or less best-case recording situations,
provided that it is assumed that the sampling rate is 100 kHz (10
microseconds per point).  ALSO NOTE THAT THIS FILE COULD NOT BE OBTAINED
PHYSICALLY WITHOUT SEVERE ALIASING PROBLEMS.  This is due to the sampling
rate of this data (100 kHz) and the fact that the amplifier has a bandwidth
that extends significantly beyond 50 kHz.  THIS FILE SHOULD NOT BE
ANALYZED.  The noise contains white, f, and f**2 components.  f noise has a
PSD (amp**2/Hz) that rises linearly with increasing frequency; f**2 noise
has a PSD that rises as f**2.


3. fchnl2+noise.txt    This is chnl2+noise that has been passed through a
digital (FIR) Gaussian filter to simulate an anti-aliasing filter.
Assuming that the sampling rate is 100 kHz, the -3 dB bandwidth of this
filter is 15 kHz.  This is an appropriate filter setting for this type of
filter (which closely approximates the 8 pole Bessel filters usually used)
and this sampling rate.  The rms value of the noise in this file is 0.86.
The signal-to-noise ratio (SNR) is 1.16 (SNR is defined as open channel
current amplitude divided by rms noise).  THIS IS THE FILE THAT SHOULD BE
ANALYZED.  It contains 812 openings, with a mean open time of 422
microseconds (42.2 points) and a mean closed time of 808 microseconds (80.8
points).  Of course these values also apply to chnl2 and chnl2+noise.


4. noise.txt    This file contains the noise that was added to create
chnl2+noise.  It contains 100,000 points (like all of the files) and has
NOT been filtered.


5. fnoise.txt    This file contains the same noise passed through a digital
Gaussian filter with a -3 dB bandwidth of 15 kHz.  Thus it is the noise
found in fchnl2+noise.


6. threshold.txt    See explanation below.


Note that half amplitude threshold detection of this data can be
accomplished by filtering the data to achieve a final bandwidth of 3 kHz
(reducing rms noise to 0.127 and increasing SNR to 7.87).

After this filtering, half amplitude threshold detection found 672 openings, with
not more than 1 false alarm.  By allowing a few more false alarms, several more  
true openings can be detected.  The detected mean open time was 506 microseconds
and the detected mean closed time was 980 microseconds.  As expected both
of these are over-estimates due to missing brief events.  Corrections of
the means (taking into account the filter and assuming exponentially
distributed channel lifetimes) is possible but has not been done for this
data.
   
The file "threshold" contains the output of this detection algorithm.