Bob,

Thank you for cleaning the FUZZ in my brain about filtering.  I don't know 
why, but I have always had a "mental block" when it came to filtering 
as-well-as the FFT usage.

When I finish building my sensors, I'll sit down and STUDY the WinQuake & 
WinSDR documentation more thoroughly.  My trouble in understanding is 
generally terminology, "what is commonly used or normal", and understanding 
what a step or function does; and primarily Old Age!.

Regards,
Jerry

----- Original Message ----- 
From: Bobhelenmcclure@.......
To: psn-l@.............. ; tchannel1@............
Sent: Tuesday, September 04, 2007 1:46 PM
Subject: Filtering


Hi Ted,

I have been looking at your event files, and your filtering does not make 
any sense to me. For example file 070902.011039.tcidf.psn show the following 
filters applied:

1. IIR Filter BW LP Freq: 0.1 Hz  Poles: 2
2. IIR Filter BW HP Freq: 1 Hz  Poles: 2

The first filter attenuates frequencies higher than 0.1 Hz.
The second filter attenuates frequencies lower than 1.0 Hz.

The usual filtering is to apply a lowpass filter to attenuate frequencies 
greater than 1 Hz, and a highpass filter to attenuate frequencies lower 
than, say, 0.05 Hz. The result is a bandpass filter for frequencies over the 
range of 0.05 Hz to 1 Hz. What you have done is to essentially exclude 
almost all frequencies, leaving a tiny response within the 0.1 to 1 Hz 
interval. The very low amplitude of your waveform after filtering confirms 
my belief. Your LP cutoff frequency should be higher, not lower, than your 
HP cutoff.

Please download and inspect my file 070902.0115.rem.psn to see the filtering 
I used for the same event. The filtering for channel z reads:

1. Sensor Period: 4.52s Sensor Q: 0.7 Filtered Period: 40s
2. Butterworth Order 2 HighPass backward Corner Period: 40s
3. Butterworth Order 2 LowPass dblpass-no lag Corner Period: 1s

These filters are my own design, and are available in my utility WinQuake 
filter utility, WQFilter.exe.

The first filter extends the natural period of my vertical sensor from 4.5 
seconds to 40 seconds.

The second filter is a special highpass filter excluding frequencies less 
than 0.025 Hz and executed backward in time.

The third filter is executed both forward in time and backward in time and 
excludes frequencies greater than 1 Hz. The end result of my filtering is a 
passband extending from 0.025 Hz to 1.0 Hz, without any time delay or phase 
shift introduced.

Could you email me an unfiltered version of 070902.011039.tcidf.psn? Also, 
please describe your vertical sensor and its measured or estimated natural 
period and damping. I would like to try my own filtering on it.

Cheers,

Bob





--------------------------------------------------------------------------------
Get a sneak peek of the all-new AOL.com. 







Bob,
 
Thank you for cleaning the =
FUZZ in my=20
brain about filtering.  I don't know why, but I have always had a =
"mental=20
block" when it came to filtering as-well-as the FFT usage.  =

 
When I finish building my =
sensors, I'll=20
sit down and STUDY the WinQuake & WinSDR documentation more=20
thoroughly.  My trouble in understanding is generally terminology, =
"what is=20
commonly used or normal", and understanding what a step or function =
does; and=20
primarily Old Age!.
 
Regards,
Jerry
 
----- Original Message -----=20
From: Bobhelenmcclure@....... =

To: psn-l@.............. ; tchannel1@............ 
Sent: Tuesday, September 04, 2007 1:46 PM
Subject: Filtering


Hi Ted,
 
I have been looking at your event files, and your filtering does =
not make=20
any sense to me. For example file 070902.011039.tcidf.psn show the =
following=20
filters applied:
 
1. IIR Filter BW LP Freq: 0.1 Hz  Poles: 2 
2. IIR Filter =
BW HP=20
Freq: 1 Hz  Poles: 2
 
The first filter attenuates frequencies higher than 0.1 Hz.
The =
second=20
filter attenuates frequencies lower than 1.0 Hz.
 
The usual filtering is to apply a lowpass filter to attenuate =
frequencies=20
greater than 1 Hz, and a highpass filter to attenuate frequencies lower =
than,=20
say, 0.05 Hz. The result is a bandpass filter for frequencies over the =
range of=20
0.05 Hz to 1 Hz. What you have done is to essentially exclude almost all =

frequencies, leaving a tiny response within the 0.1 to 1 Hz interval. =
The very=20
low amplitude of your waveform after filtering confirms my belief. Your =
LP=20
cutoff frequency should be higher, not lower, than your HP cutoff.
 
Please download and inspect my file 070902.0115.rem.psn to see the=20
filtering I used for the same event. The filtering for channel z =
reads:
 
1. Sensor Period: 4.52s Sensor Q: 0.7 Filtered Period: 40s
2.=20
Butterworth Order 2 HighPass backward Corner Period: 40s
3. =
Butterworth Order=20
2 LowPass dblpass-no lag Corner Period: 1s
 
These filters are my own design, and are available in my utility =
WinQuake=20
filter utility, WQFilter.exe.

The first filter extends the natural period of my vertical =
sensor from=20
4.5 seconds to 40 seconds.
 
The second filter is a special highpass filter excluding =
frequencies less=20
than 0.025 Hz and executed backward in time. 
 
The third filter is executed both forward in time and backward in =
time and=20
excludes frequencies greater than 1 Hz. The end result of my filtering =
is a=20
passband extending from 0.025 Hz to 1.0 Hz, without any time delay or =
phase=20
shift introduced.
 
Could you email me an unfiltered version of =
070902.011039.tcidf.psn? Also,=20
please describe your vertical sensor and its measured or estimated =
natural=20
period and damping. I would like to try my own filtering on it.
 
Cheers,
 
Bob





Get a sneak peek of the all-new AOL.com.