Hi All,
You may remember that there was a discussion a while back concerning the
signal delays produced by the filter circuits used in seismic amplifiers.
After waiting patiently for any figures to be produced, I eventually
gave up and plugged the filter components shown on the circuit diagrams of two
of Larry's amplifiers into experimenter board. I then made some approximate
measurements for the characteristics and the signal delays. I have put approx.
after readings of less that - 30 dB, since the signals had a significant
noise component from the opamp chains.
The capacitors were selected from a 5% polyester range. I used a Wayne
Kerr FG3 Function Generator down to 0.01 Hz and a Philips PM3217 dual beam
oscilloscope. I also checked the frequency calibration of the Function Generator
with a counter timer. The test signals used were continuous wave, not 2 (or
more) cycle sine wave transient signals (a signal which starts on zero,
follows a sine wave pattern for two cycles and finishes on zero, followed by a
zero level pause).
The 'older' type of amplifier circuit used 4 off LF412 opamps in a 6
pole filter circuit, nominally rated at 10 Hz turnover with a Butterworth
characteristic. This circuit is no longer in production, but I am sure that many
are still in use.
The measured signal delay was about 60 milli sec at 1 Hz and increased
to about 70 milli sec at 10 Hz.
The gains were
10 Hz -5.4 dB
15 Hz -20 dB
20 Hz -35 dB approx.
The 'newer' 1 to 3 channel type amplifier uses a TL074 opamp in an 8
pole filter circuit, nominally rated at 5 Hz with a Butterworth characteristic
for the Lehman configuration - see _http://psn.quake.net/eqamp.html_
(http://psn.quake.net/eqamp.html)
The measured delay was 100 milli sec at 1 Hz and it was very nearly
constant out to 5 Hz.
The gains were
5 Hz -2.2dB
7.5 Hz -7.3 dB
10 Hz -14 dB
15 Hz -32 db approx.
The 'newer' type with the geophone configuration is rated at 10 Hz with
a Butterworth characteristic. The component values in the panel on the
circuit diagram give a 10 Hz turnover, not the 20 Hz stated. I note that there is
still considerable signal gain at 20 Hz, which may be relevant when
considering environmental noise pickup at your location.
The measured delay at 1 Hz was 48 milli sec and this appeared to stay
constant out to 10 Hz.
The gains were
10 Hz -2.2 dB
15 Hz -6.8 dB
20 Hz -14 dB
30 Hz -33 dB approx.
I did not find any peaks in the delay characteristics of the TL074 opamp
filter types at about 0.8 x the cut-off frequency, as shown on Texas'
FilterPro design plots for Butterworth Filters. Comparing this filter output
characteristic with the older 6 pole Butterworth filter suggests a more gradual
'roll off' and near constant signal delays over the passband, more
characteristic of Bessel type filters.
I did not test the latest 8 pole version of the 1 to 4 channel amplifier
filter which uses a LF444 opamp, but I would expect them to be similar. The
plots of the characteristics appear to be identical. See
_http://psn.quake.net/serialamp.html_ (http://psn.quake.net/serialamp.html)
If you are digitising several channels with the same A/D conveter, it
may be worth measuring / calculating the delay in between the channel readings.
Even if a conversion only takes ~20 micro seconds, the 'speed' of the
digital link to the computer may need to be considered.
These signal delays are of the same order as the errors which may be
produced in estimating the start time of the signal and are probably not
significant for most 'amateur' work. If, however, you apply a 'real' narrow band
multi pole filter at about 0.5 or 1 Hz to the signal to dig P & S waves out of
the general background, much larger timing errors can be produced. I have no
information on narrow band software filters.
Regards,
Chris Chapman
Hi All,
You may remember that there was a discussion a=20
while back concerning the signal delays produced by the filter circuits used=
in=20
seismic amplifiers.
After waiting patiently for any figures to be=20
produced, I eventually gave up and plugged the filter components shown on th=
e=20
circuit diagrams of two of Larry's amplifiers into experimenter board. I the=
n=20
made some approximate measurements for the characteristics and the signal=20
delays. I have put approx. after readings of less that - 30 dB, since t=
he=20
signals had a significant noise component from the opamp chains.
The capacitors were selected from a 5% polyeste=
r=20
range. I used a Wayne Kerr FG3 Function Generator down to 0.01 Hz and a Phil=
ips=20
PM3217 dual beam oscilloscope. I also checked the frequency calibration of t=
he=20
Function Generator with a counter timer. The test signals used were=20
continuous wave, not 2 (or more) cycle sine wave transient signals (a signal=
=20
which starts on zero, follows a sine wave pattern for two cycles and finishe=
s on=20
zero, followed by a zero level pause).
The 'older' type of amplifier circuit used 4 of=
f=20
LF412 opamps in a 6 pole filter circuit, nominally rated at 10 Hz turnover w=
ith=20
a Butterworth characteristic. This circuit is no longer in production, but I=
am=20
sure that many are still in use.
The measured signal delay was about 60 milli se=
c at=20
1 Hz and increased to about 70 milli sec at 10 Hz.
The gains were
10 Hz -5.4 dB
15 Hz -20 dB
20 Hz -35 dB approx.
The 'newer' 1 to 3 channel type amplifier uses=20=
a=20
TL074 opamp in an 8 pole filter circuit, nominally rated at 5 Hz with a=20
Butterworth characteristic for the Lehman configuration - see http://psn.quake.net/eqamp.html=
The measured delay was 100 milli sec at 1 Hz an=
d it=20
was very nearly constant out to 5 Hz.
The gains were
5 Hz -2.2dB
7.5 Hz -7.3 dB
10 Hz -14 dB
15 Hz -32 db approx.
The 'newer' type with the geophone configuratio=
n is=20
rated at 10 Hz with a Butterworth characteristic. The component values in=20
the panel on the circuit diagram give a 10 Hz turnover, not t=
he=20
20 Hz stated. I note that there is still considerable signal gain at 20=
Hz,=20
which may be relevant when considering environmental noise pickup at your=20
location.
The measured delay at 1 Hz was 48 milli sec and=
=20
this appeared to stay constant out to 10 Hz.
The gains were
10 Hz -2.2 dB
15 Hz -6.8 dB
20 Hz -14 dB
30 Hz -33 dB approx.
I did not find any peaks in the delay=20
characteristics of the TL074 opamp filter types at about 0.8 x the cut-off=20
frequency, as shown on Texas' FilterPro design plots for Butterworth=20
Filters. Comparing this filter output characteristic with the older 6 p=
ole=20
Butterworth filter suggests a more gradual 'roll off' and near constant sign=
al=20
delays over the passband, more characteristic of Bessel type filters.
I did not test the latest 8 pole version of the=
1=20
to 4 channel amplifier filter which uses a LF444 opamp, but I would expect t=
hem=20
to be similar. The plots of the characteristics appear to be identical.=
See=20
http://psn.quake.net/serialamp.=
html
If you are digitising several channels with the=
=20
same A/D conveter, it may be worth measuring / calculating the delay in betw=
een=20
the channel readings. Even if a conversion only takes ~20 micro seconds, the=
=20
'speed' of the digital link to the computer may need to be considered.
These signal delays are of the same order as th=
e=20
errors which may be produced in estimating the start time of the signal=
and=20
are probably not significant for most 'amateur' work. If, however, you apply=
a=20
'real' narrow band multi pole filter at about 0.5 or 1 Hz to the=20
signal to dig P & S waves out of the general background, much larger tim=
ing=20
errors can be produced. I have no information on narrow band software=20
filters.
Regards,
Chris Chapman