To: psnlist@..............
Sent: Thu, December 27, 2012 3:33:06 PM
Subject: Re: Instrumentation Question
Hi Bob,
I have tried the real-time period extending filter in WinSdr with very =
good results . I am using it to extend the period of broadband force =
balance seismometers from 50 or 120 seconds to 1000 seconds I know =
that was not the original intent but it really works well. I have =
compared its performance with an analog "inverse filter" with =
essentially the same transfer function .The resulting waveforms are =
essentially identical. =20
I have not yet tried the WinQuake version but I will do so
My objective is to provide a system with real-time capability to observe =
earth normal modes following a large event and analyze the spectrum with =
WinQuake. The increase in noise at very long periods is clearly =
evident but that is expected and OK. The noise source is mostly the =
instrument self noise enhanced by the period extension filter since the =
instrument noise is significantly above the low noise model beyond 100 =
seconds. The very long period waveforms from side by side broadband =
instruments are incoherent but they will be coherent following a large =
teleseismic event upon excitation of the Normal modes. The period =
extension process could be either in the data acquisition as I am now =
doing , or in post processing using WinQuake. The pros do it in post =
processing but they don't get so excited watching waveforms in real time =
like the amateurs.=20
Just one comment on the relative merits of the inverse filtering =
techniques vs. the negative impedance loading for a geophone. When =
filtering a critically damped geophone the geophone dynamics are =
unchanged and the clipping level is unchanged. When the geophone is =
loaded with a negative impedance the dynamics of the seismic mass are =
changed dramatically. The mass is heavily over-damped and moves much =
less in response to large ground displacements. One of the problems in =
seismic instrumentation for public safety is clipping. The over-damped =
geophone has a much larger dynamic range as well as period extension.
Thanks for the heads up on WinQuake and sorry for the overly long =
response. This is really interesting stuff and it gets me going :-}=20
Best Regards,
Dave =20
Hi Barry,
The latest version of WinSdr =
(4.6.7) has=20
the period extension capability which was intended for geophones and =
other non=20
feedback sensors. It also has "derived Channels" which are additional =
output=20
channels which can use one of the standard input channels as a =
source but=20
process the data differently.
The period extension is a new filter =
under the=20
integrate / filter heading. To set it up first download the =
new=20
version and start it -- If you have an unused channel on the A/D just =
connect=20
the seismometer to that channel in addition to the normal channel. =
If you=20
need to add a new channel then go to the system settings/more options =
and select=20
"additional channels" chose the number channels you wish to add. =
Then -- go to settings and =
select the=20
new channel and then select the source channel which will be one of the =
original=20
8. Then select integrate /filter control and enable the =
period extension filter and set pendulum frequency to .02 =
Hz =20
(50 seconds for Inyo and Yuma) leave Q at 0.707 the set High =
pass=20
filter to .002 or .001 Hz and Q=3D 0.707. Do NOT check the =
integrate=20
data box. Now --- two options -- you can use the "post =
digital=20
filter" or the normal event detection filter. If you use the post =
digital=20
filter the data will be recorded as filtered and you have no =
options to=20
change the filter on replay or transfer to WinQuake. If you use the =
normal=20
Event Detection filter the data will be recorded after the period=20
extension process but before the second filter . The settings in either =
case are=20
..002 Hz 1 pole High Pass and I usually use .07 Hz low pass 6 pole. =
WinSdr will=20
not accept a high pass at longer than 500 seconds (.002 Hz). I will ask =
Larry if=20
it is possible to extend that filter to 1000 seconds.
When you start this channel it will =
take as much as=20
an hour to settle and then the trace will be dominated by very long =
period noise=20
.. This is the actual seismometer self noise which is due to a lot =
of things=20
such as 1/f noise , and dominated by pressure / temperature rate =
effects.=20
You will need to set the Y axis gain to 4 or more to reduce the =
noise to=20
an acceptable level. If the instrument is not very well insulated =
from=20
temperature change the very long period noise could be excessive =
but you=20
have seen earth normal modes before without a =
problem.
NOW -- we wait for a BIG =
event that=20
generates the earth modes as you have observed =
before. The=20
difference is the signal the normal modes should be much =
larger as=20
the long roll off of the instrument has de-convolved by the=20
filter. I have yet to see an =
event so it is=20
still unproven but the very long period noise I see is what I would=20
expect. I am using this technique on the Trillium Compact =
and my=20
Napa sensors. I am waiting for delivery of a new =
Nanometrics Trillium=20
120 which should have significantly reduced noise beyond 100 seconds. My =
own=20
sensors were equal to or better than the Trillium Compact so I =
need a=20
better reference.
This capability will add a new =
dimension to our=20
amateur seismic observations. THANKS TO LARRY and BOB McCLURE=20
!!!!
Cheers, Dave
Sent: Saturday, December 29, 2012 2:48 PM
Subject: Re: Instrumentation Question
Dave
The=20
"real-time period extending filter in =
WinSdr". Is that the high pass/ low pass digital filter located =
with each=20
channel or is there something else in the program that =
I'm not aware of ?=20
Regards
Barry
www.seismicvault.com
From: Dave Nelson <davefnelson@.......>
To: psnlist@..............
<=
SPAN=20
style=3D"FONT-WEIGHT: bold">Sent: Thu, December 27, 2012 =
3:33:06=20
PM
Subject: Re: =
Instrumentation=20
Question
Hi Bob,
I have tried the real-time period =
extending=20
filter in WinSdr with very good results . I am using it to extend =
the=20
period of broadband force balance seismometers from 50 or 120 seconds to =
1000=20
seconds I know that was not the original intent but it =
really works=20
well. I have compared its performance with an analog "inverse =
filter" with=20
essentially the same transfer function .The resulting waveforms are =
essentially=20
identical.
I have not yet tried the WinQuake =
version but I will do so
My objective is to provide a =
system with=20
real-time capability to observe earth normal modes following a large =
event and=20
analyze the spectrum with WinQuake. The increase in noise at =
very=20
long periods is clearly evident but that is expected and OK. The noise =
source=20
is mostly the instrument self noise enhanced by the period =
extension=20
filter since the instrument noise is significantly above the low noise =
model=20
beyond 100 seconds. The very long period waveforms from side by side =
broadband=20
instruments are incoherent but they will be coherent =
following a=20
large teleseismic event upon excitation of the Normal modes. The period=20
extension process could be either in the data acquisition as I am now =
doing , or=20
in post processing using WinQuake. The pros do it in post =
processing=20
but they don't get so excited watching waveforms in real time like the=20
amateurs.
Just one comment on the relative merits =
of the=20
inverse filtering techniques vs. the negative impedance loading =
for a=20
geophone. When filtering a critically damped geophone the geophone =
dynamics are=20
unchanged and the clipping level is unchanged. When the geophone is =
loaded with=20
a negative impedance the dynamics of the seismic mass are changed =
dramatically.=20
The mass is heavily over-damped and moves much less in response to =
large=20
ground displacements. One of the problems in seismic instrumentation for =
public=20
safety is clipping. The over-damped geophone has a much larger =
dynamic=20
range as well as period extension.
Thanks for the heads up on WinQuake and =
sorry for=20
the overly long response. This is really interesting stuff and it gets =
me going=20
:-}
Best Regards,
Dave