PSN-L Email List Message
Subject: Re: Advantages of more ADC bits
From: ChrisAtUpw@.......
Date: Sat, 7 Feb 2009 17:52:20 EST
In a message dated 07/02/2009, rsparks@.......... writes:
I have been considering what advantage the amateur seismologist might
see if making an upgrade to more bits in the A/D converter.
Bits are a basic division of computer technology and are expressed as
powers of two (2^xx). Thus, an 8 bit device will have 2^8 = 256
divisions, whereas a 16 bit device will have 65,536 divisions.
Here is a list of devices/seismometers and the sensitivity/accuracy in bits:
Volksmeter 24 bit
PSN/Cochrane 16 bit
Saum Infiltec 16 bit
AS1 12 bit
Dataq 8/10/12 bit
I get to wondering "Why bother to upgrade?", at least for amateur purposes.
Two reasons come to mind. You will see more quakes with a more
sensitive instrument. You can not really compensate with more
amplification because the dynamic range is smaller with the low bit
count devices. Yes, you can raise the amplification level so that an 8
bit device will respond to the same voltage signal that the 24 bit
device will see, but the 8 bit device will be saturated after only 256
counts while the 24 bit device would have only recorded 256 counts out
of 16.7 million possible counts.
The second reason is to have better fidelity or accuracy. Any FFT taken
of a trace will be more accurate if each individual reading is better
placed in the digital data table. Ultimately, both accuracy and
sample rate must be increased for superior results.
Hi Roger,
Another important question to ask is 'What is the chip background noise
level?' The depends on the integration interval and whether a series of
signals are averaged. Also ask 'What is your seismic background noise level?'
The older 16 bit ADCs used to have 3 to 4 bits of internal digital noise
on them. But with a 20 micro second conversion rate you could take and
average 64 sequential readings. You need to take 4 readings to average out 1 bit
of digital noise, 16 to average out 2 bits etc. You can push signal averaging
to about 18 bit accuracy on a 16 bit ADC for single channel low sample rates.
24 bit ADCs may only give about 19 / 20 bits true resolution, or even
less if you require a rapid sample rate. The 16 bit Sigma Delta ADCs can give
16 bits true resolution for 20 samples / second, but may be less for greater
sample rates. Some 24 bit systems only have +/-2 volts input range, which
requires you to use very low noise / costly amplifiers.
You also need to check the amplifier noise level and the local seismic
noise levels that you detect. The local noise may be mostly due to ~6 second
period microseisems. The noise level for P and S waves at 0.5 to 5 Hz and for
surface waves at periods over 12 seconds may be much lower, so the amplifier
self noise is important. Assuming that you have a 16 bit true ADC with a
+/-10 V input range, 1 bit represents 0.305 milli V. You definitely need the
amplifier to have a lower noise level than this.
12 bits resolution only gives you +/-2048 signal steps, so even if there
is no ADC noise, your practical dynamic range is limited to about 3
Magnitudes. 16 bits gives a more practicable +/-32,768 step range. Remember that
Earthquake Magnitudes have a logarithmic scale. Dynamic range may be a serious
problem if you live close to say, the New Madrid earthquake zone and experience
local quakes. For quake amplitude / distance motions see
_http://jclahr.com/science/psn/magnitude.html_ (http://jclahr.com/science/psn/magnitude.html)
Amplifiers suffer from several sorts of noise, including resistive,
input voltage and current and 1/f noise. The 1/f noise is serious and may be
dominant at periods over ~10 seconds. Chopper amplifiers do not suffer from 1/f
noise, but they are more expensive to construct. The very low noise levels
seen in commercial seismic systems is due to the use of capacitative distance
transducers and chopper detection systems. Velocity measuring systems are
limited by noise at longer periods.
Regards,
Chris Chapman
In a message dated 07/02/2009, rsparks@.......... writes:
<=
FONT=20
style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>I have=20
been considering what advantage the amateur seismologist might
see if=20
making an upgrade to more bits in the A/D converter.
Bits are a ba=
sic=20
division of computer technology and are expressed as
powers of two=20
(2^xx). Thus, an 8 bit device will have 2^8 =3D 256
divisions, w=
hereas=20
a 16 bit device will have 65,536 divisions.
Here is a list of=20
devices/seismometers and the sensitivity/accuracy in=20
bits:
Volksmeter 24=20
bit
PSN/Cochrane 16 bit
Saum Infiltec  =
; 16=20
bit
AS1 &n=
bsp;=20
12 bit
Dataq &nbs=
p;=20
8/10/12 bit
I get to wondering "Why bother to upgrade?", at l=
east=20
for amateur purposes.
Two reasons come to mind. You will see=
=20
more quakes with a more
sensitive instrument. You can not really=
=20
compensate with more
amplification because the dynamic range is smalle=
r=20
with the low bit
count devices. Yes, you can raise the amplifica=
tion=20
level so that an 8
bit device will respond to the same voltage signal=20=
that=20
the 24 bit
device will see, but the 8 bit device will be saturat=
ed=20
after only 256
counts while the 24 bit device would have only recorded=
256=20
counts out
of 16.7 million possible counts.
The second reason i=
s to=20
have better fidelity or accuracy. Any FFT taken
of a trace will=20=
be=20
more accurate if each individual reading is better
placed in the digit=
al=20
data table. Ultimately, both accuracy and
sample rate must be increase=
d=20
for superior results.
Hi Roger,
Another important question to ask is 'What=
is=20
the chip background noise level?' The depends on the integration interv=
al=20
and whether a series of signals are averaged. Also ask 'What is your=20
seismic background noise level?'
The older 16 bit ADCs used to have 3 to 4 bits=20=
of=20
internal digital noise on them. But with a 20 micro second conversion rate y=
ou=20
could take and average 64 sequential readings. You need to take 4 readi=
ngs=20
to average out 1 bit of digital noise, 16 to average out 2 bits etc. Yo=
u=20
can push signal averaging to about 18 bit accuracy on a 16 bit ADC for=20
single channel low sample rates.
24 bit ADCs may only give about 19 / 20 bits tr=
ue=20
resolution, or even less if you require a rapid sample rate. The 16 bit Sigm=
a=20
Delta ADCs can give 16 bits true resolution for 20 samples / second, but may=
be=20
less for greater sample rates. Some 24 bit systems only have +/-2 volts inpu=
t=20
range, which requires you to use very low noise / costly=20
amplifiers.
You also need to check the amplifier noise leve=
l=20
and the local seismic noise levels that you detect. The local noise may be=20
mostly due to ~6 second period microseisems. The noise level for P and S wav=
es=20
at 0.5 to 5 Hz and for surface waves at periods over 12 seconds ma=
y be=20
much lower, so the amplifier self noise is important. Assuming that you=
=20
have a 16 bit true ADC with a +/-10 V input range, 1 bit represents 0.305 mi=
lli=20
V. You definitely need the amplifier to have a lower noise level than=20
this.
12 bits resolution only gives you +/-2048 signa=
l=20
steps, so even if there is no ADC noise, your practical dynamic range is lim=
ited=20
to about 3 Magnitudes. 16 bits gives a more practicable +/-32,768 step=20
range. Remember that Earthquake Magnitudes have a logarithmic=20
scale. Dynamic range may be a serious problem if you live close to say,=
the=20
New Madrid earthquake zone and experience local quakes. For quake amplitude=20=
/=20
distance motions see
http://jclahr.com/scie=
nce/psn/magnitude.html
Amplifiers suffer from several sorts of noise,=20
including resistive, input voltage and current and 1/f noise. The 1/f n=
oise=20
is serious and may be dominant at periods over ~10 seconds. Chopper amplifie=
rs=20
do not suffer from 1/f noise, but they are more expensive to construct. The=20=
very=20
low noise levels seen in commercial seismic systems is due to the use of=20
capacitative distance transducers and chopper detection systems. Velocity=20
measuring systems are limited by noise at longer periods.
Regards,
Chris Chapman
[ Top ]
[ Back ]
[ Home Page ]