Open GoogleCodeExporter opened 9 years ago
I note that in the original SALPA paper, all data was presented for fs = 25 kHz
and we just copied the parameters from that paper. Doug Bakkum noted to me,
anecdotally, that it was easy to break SALPA with different sampling rates.
Riley, you have the most insight here. What do you think the issue is? It seems
like SALPA thinks it has a decent fit very prematurely.
Original comment by jonathan...@gmail.com
on 8 Nov 2012 at 3:47
Righto- I'll take a look at this Thursday.
Original comment by RZellerT...@gmail.com
on 13 Nov 2012 at 2:32
Riley, what is the status here?
Original comment by jonathan...@gmail.com
on 10 Dec 2012 at 3:35
still need to look into this on the rig, however:
biggest issue is that the artifact isn't being detected, either through
thresholding or through the 'stim' stream. Whether or not SALPA is fitting
prematurely will be seen after we get blanking working again.
Were their areas of the SALPA trace that were blanked, just not where the
stimuli occurred?
Were stimuli recorded at all during this experiment?
Note that several of the SALPA parameters are based on samples, so changing
sampling rate would very quickly alter their meaning. It would probably be a
good idea to switch to 'ms' as a measure for the curve length, blanking period,
etc.
Original comment by RZellerT...@gmail.com
on 10 Dec 2012 at 7:59
No. (I've previously verified that the stimulation times stay synced to the
expected delivery times to within about 20 us over an hour, so I've stopped
worrying about them).
Yes- agree. I typically use a 15 kHz sampling rate, instead of 25 kHz, and
have always adjusted the default parameters appropriately.
Original comment by jonathan...@gmail.com
on 10 Dec 2012 at 9:02
Using the SALPA parameters you describe (45 sample half width, 0 post-peg, 6
minimum peg, 3 pre peg, 6 min symmetric, right?), SALPA was able to run
correctly on one of our rigs (AD rig).
However, SALPA behaved as you described in your figure if electrical
stimulation input was turned off. Note that this only occurred if I disabled
the 'Stimulation Timing' input in the 'Aux. Input' page of the Hardware
Settings. The Electrical Stim 'real time data stream' does not need to be
activated (on the 'real time' page of Hardware Settings), and the Electrical
Stim data data stream does not need to be recorded (in the Recording Stream
Setup dialog).
SALPA gets a signal from the electrical stim recording set up, which it uses as
way to determine when to blank the signal, IE when the signal is too distorted
by stimulation artifact to recover. When the electrical stim input is not
turned on, SALPA doesn't get that signal and just performs it's usual curve
fitting.
Given that you aren't recording electrical stimulation, that seems like a
probably culprit.
Thanks a bunch and I'm sorry about the extreme delay on this.
Original comment by RZellerT...@gmail.com
on 19 Dec 2012 at 11:06
Original issue reported on code.google.com by
jonathan...@gmail.com
on 7 Nov 2012 at 8:22Attachments: