PSN-L Email List Message
Subject: Re: Volksmeter sensor and the AD774x CDC chip
From: ChrisAtUpw@.......
Date: Thu, 20 Apr 2006 22:51:13 EDT
In a message dated 20/04/2006, lcochrane@.............. writes:
> Hi Larry,
> What period compensation are you claiming for this sensor?
> Do the plots show a compensated velocity / compensated displacement
> / uncompensated signal, or what?
> How are you performing the velocity feedback / damping necessary to
> stabilise the response?
There is no period compensation in either the SG sensor or the Volksmeter
(VM). Both use simple damping. The SG sensor has a feedback loop but it's just
there to damp the pendulum. The VM sensor uses eddy-current damping and has
no feedback system.
Randall will be sending a response to this shorty.
The plots are from the raw data right out of the sensor, but with a 60
second 2 pole
high-pass and 2 Hz 2 pole low-pass filter applied to the data before making
the GIF
image. The event files are the raw data from the AD7746 chip.
Hi Larry,
The original S/G circuit used a 62.5 sec AC coupled integrator with a
gain of 10, a 1 sec low pass filter and critical damping. This was to give it
an output linear with velocity.
> I note that in your SG EMail reference
(_http://psn.quake.net/freqtest.html_ (http://psn.quake.net/freqtest.html) ), you claim a flat response
> from 50 sec to 3 Hz, but remarked that you had to substantially increase
> the damping to get this. I note that in your latest circuit diagram, you
> seem to have greatly reduced the itegration time and wondered if the
> circuit values were correct?
Like the VM channel (LCTST) my LC8 channel is pretty much the raw data from
the
pickup except there are two 60 second high-pass filters in the signal path.
My LC3
channel, the integrated output, should have a velocity response from about 1
second, the period of the pendulum, to about 50 seconds. I have not made any
changes to my SG sensor for many years so the value of the integrator parts
should be the same.
The circuit at _http://psn.quake.net/sg-schm.gif_
(http://psn.quake.net/sg-schm.gif) shows a 100 K Ohm input resistor, a 470 K Ohm feedback resistor and
a 2 mu F parallel capacitor. This RC combination rolls off at a period of
only 5.9 seconds - a factor of 10 less than we want. This is why I asked if the
published circuit values are correct??
Do you still use a 4.7 M Ohm feedback resistor, like in your previous
circuit at _http://psn.quake.net/sgproc.gif_ (http://psn.quake.net/sgproc.gif) ?
Moreover, there does not seem to be a 1 sec low pass circuit to produce
a velocity response?
It is possible to ''squash'' this sort of humped response by heavy
overdamping and a considerable increase in the amplifier gain, but you tend to run
into serious noise problems.
The original circuit used an input capacitor to the integrator, which
should effectively limit the VLF 1/f noise. A 100 mu F non polar capacitor?
Note that it is usually possible to lengthen the period using digital
processing - if there is sufficient signal resolution, as seems likely with the
VM 24 bit ADC. However, doing this to a displacement signal is likely to
give rather large amplitudes at the low frequencies. Maybe the VM displacement
signal could be converted into a velocity signal?
Regards,
Chris Chapman
In a message dated 20/04/2006, lcochrane@.............. writes:
<=
FONT=20
style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>> Hi=20
Larry,
> What period compensation are you claimin=
g=20
for this sensor?
> Do the plots show a compensate=
d=20
velocity / compensated displacement
> / uncompensated signal, or=20
what?
> How are you performing the velocity feedb=
ack=20
/ damping necessary to
> stabilise the response?
There is no=
=20
period compensation in either the SG sensor or the Volksmeter (VM). Both u=
se=20
simple damping. The SG sensor has a feedback loop but it's just there to d=
amp=20
the pendulum. The VM sensor uses eddy-current damping and has no feedback=20
system.
Randall will be sending a response to this shorty.
The=20
plots are from the raw data right out of the sensor, but with a 60 second=20=
2=20
pole
high-pass and 2 Hz 2 pole low-pass filter applied to the data bef=
ore=20
making the GIF
image. The event files are the raw data from the AD7746=
=20
chip.
Hi Larry,
The original S/G circuit used a 62.5 sec AC cou=
pled=20
integrator with a gain of 10, a 1 sec low pass filter and critical damp=
ing.=20
This was to give it an output linear with velocity.
<=
FONT=20
style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000=20
size=3D2>> I note that in your SG EMail reference (<=
A=20
href=3D"http://psn.quake.net/freqtest.html">http://psn.quake.net/freqtest.=
html),=20
you claim a flat response
> from 50 sec to 3 Hz, but remarked that=20=
you=20
had to substantially increase
> the damping to get this. I note tha=
t in=20
your latest circuit diagram, you
> seem to have greatly reduced the=
=20
itegration time and wondered if the
> circuit values were=20
correct?
Like the VM channel (LCTST) my LC8 channel is pretty much=20=
the=20
raw data from the
pickup except there are two 60 second high-pass filt=
ers=20
in the signal path. My LC3
channel, the integrated output, should have=
a=20
velocity response from about 1 second, the period of the pendulum, to abou=
t 50=20
seconds. I have not made any changes to my SG sensor for many years so the=
=20
value of the integrator parts should be the same.
Moreover, there does not seem to be a 1 sec low=
=20
pass circuit to produce a velocity response?
It is possible to ''squash'' this sort of humpe=
d=20
response by heavy overdamping and a considerable increase in the amplif=
ier=20
gain, but you tend to run into serious noise problems.
The original circuit used an input capacitor to=
the=20
integrator, which should effectively limit the VLF 1/f noise. A 100 mu=20=
F=20
non polar capacitor?
Note that it is usually possible to length=
en=20
the period using digital processing - if there is sufficient signal=20
resolution, as seems likely with the VM 24 bit ADC. However, doing this=
to=20
a displacement signal is likely to give rather large amplitudes at the low=20
frequencies. Maybe the VM displacement signal could be converted into a velo=
city=20
signal?
Regards,
Chris Chapman
[ Top ]
[ Back ]
[ Home Page ]