Hi Chris, you are correct. I had a little free time today and took out my
scope and my very old Preston Mod-135 Waveform Source Generator and created
some test files. I'll email you and Larry the files off-line and anybody
else that would like to see it. I generated test data at 1, 3, 5, 7 and 9-Hz
at voltage levels of .5 .2 and 1 volt PP for each Hz range. I'm using
Winquake to view the resulting datasets and with the Windquake X-scale set
at one, can see that CH6 the unfiltered channel leads the filtered ATN
channel by six clicks on a scale of 1 to 14 clicks which I estimate to be
57% of the scale range which should directly convert to 57MS if my
calculations are correct...

I guess my first question is what am I measuring? This is not just the
filter, the circuit also includes, a 3-poll filter, an inverting stage with
decoupling cap and then a final gain/leveling stage and whatever processing
time the micro processor takes up. The three op-amps are op-27's and the A/D
is Larry's 8-channel serial data collection system which is housed in
another box connected via short 12-inch jumpers. Please take a look at the
data and see what you think.

Steve
  -----Original Message-----
  From: psn-l-request@.............. [mailto:psn-l-request@.................
Behalf Of ChrisAtUpw@.......
  Sent: Tuesday, May 17, 2005 4:33 AM
  To: psn-l@..............
  Subject: Re: Time/Filtering


  In a message dated 17/05/2005, shammon1@............. writes:
    Hi, I was thinking about simple ways to measure this and was able to use
one
    of the unused channels on Larry's A/D board by splitting the unfiltered
data
    from one of my sensors into it before it was filtered. After capturing
some
    sample data,  I picked a nice large spike from one of the datasets and
was
    able to see an approximate 60ms delay when comparing the filtered and
    unfiltered datasets. I'm using the Pete Rowe filter design Amp/filter
found
    at http://pw2.netcom.com/~shammon1/equip.htm#Electronics and this is
    consistent with what I expected to see.
    Regards, Steve Hammond  PSN San Jose
    Aptos, CA

  Hi Steve,

      The component values on Pete Rowe's circuit diagrams on your Website
are unfortunately almost unreadable on my computer.
      A three pole 10 Hz Butterworth filter should give about 30 milli sec
delay, peaking to about 40 milli sec at 8 Hz.
      If you have a Butterworth filter and wait for a spike, you will
probably be measuring the delay near the peak, rather than the low frequency
delay which applies to P & S waves.
      Regards,

      Chris Chapman





Hi Chris, you are correct. I had a =
little=20
free time today and took out my scope and my very old Preston Mod-135 =
Waveform=20
Source Generator and created some test files. I'll email you and Larry =
the files=20
off-line and anybody else that would like to see it. I generated test =
data at 1,=20
3, 5, 7 and 9-Hz at voltage levels of .5 .2 and 1 volt PP for each Hz =
range. I'm=20
using Winquake to view the resulting datasets and with the Windquake =
X-scale set=20
at one, can see that CH6 the unfiltered channel leads the filtered ATN =
channel=20
by six clicks on a scale of 1 to 14 clicks which I estimate to be  =
57% of=20
the scale range which should directly convert to 57MS if my =
calculations=20
are correct... 
 
I guess my first question is =
what am=20
I measuring? This is not just the filter, the =
circuit also=20
includes, a 3-poll filter, an inverting stage with decoupling cap and =
then a=20
final gain/leveling stage and whatever processing time the micro =
processor takes=20
up. The three op-amps are op-27's and the A/D is Larry's 8-channel =
serial data=20
collection system which is housed in another box connected via =
short=20
12-inch jumpers. Please take a look at the data and see what you=20
think.
 
Steve

  
-----Original Message-----
From:=20
  psn-l-request@.............. =
[mailto:psn-l-request@...............On Behalf=20
  Of ChrisAtUpw@.......
Sent: Tuesday, May 17, 2005 4:33=20
  AM
To: psn-l@..............
Subject: Re:=20
  Time/Filtering


  

  
In a message dated 17/05/2005, shammon1@............. =
writes:
  Hi, I was =
thinking about=20
    simple ways to measure this and was able to use one
of the unused =

    channels on Larry's A/D board by splitting the unfiltered =
data
from one=20
    of my sensors into it before it was filtered. After capturing =
some
sample=20
    data,  I picked a nice large spike from one of the datasets and =

    was
able to see an approximate 60ms delay when comparing the =
filtered=20
    and
unfiltered datasets. I'm using the Pete Rowe filter design =
Amp/filter=20
    found
at http://pw2.netcom.com/~shammon1/equip.htm#Electronics =
and this=20
    is
consistent with what I expected to see.
Regards, Steve=20
    Hammond  PSN San Jose
Aptos, =
CA

  

  
Hi Steve,
  
 
  
    The component values on Pete Rowe's =
circuit=20
  diagrams on your Website are unfortunately almost unreadable on my=20
  computer.
  
    A three pole 10 Hz Butterworth filter =
should give=20
  about 30 milli sec delay, peaking to about 40 milli sec at 8 Hz.
  
    If you have a Butterworth filter and wait =
for a=20
  spike, you will probably be measuring the delay near the peak, rather =
than the=20
  low frequency delay which applies to P & S waves.
  
    Regards,
  
 
  
    Chris=20
Chapman