PSN-L Email List Message

Subject: Re: Rockland filter
From: ChrisAtUpw@.......
Date: Tue, 9 Jan 2007 21:58:00 EST


 
In a message dated 09/01/2007 16:16:40 GMT Standard Time, royb1@...........  
writes:

With the  usual 14 or 16 bit A/D's, is that dynamic range  necessary?
Bob



Hi Bob,
 
    A 16 bit ADC with no noise has +/-1/2 bit  uncertainty. This is about 96 
dB. Each factor of 2 gives 6 dB change. With  +/-10V input 1 bit = 0.305 mV
 
    Quiet opamps may give 0.1 to 10 Hz input noise  levels well below 1 micro 
volt; true chopper amplifiers may be much less than  this. The CAZ opamps 
tend to give 1 to 2 micro volts, but these may give OK  results for long period 
signals, when 1/f drifts become large. It is a good  principle to use odd 
orders of low pass filter with a capacitor across the  feedback resistor of the 
first opamp. Never amplify high frequency signals. This  limits the effect of 
intermodulation distortion and subharmonics
 
    Digital filters tend to give performance in the mid  70 dBs, or less, but 
also suffer from switch transient feed through - 5 mV? -  which may need 
additional analogue filtering before putting it into an  ADC. 
 
    In seismometry, we are seeking very low noise  levels at very low 
frequencies. You just can't afford to throw away +/-4 bits  signal through a poor 
choice of filter, or +/-3 bits by failing to average out  the internal ADC noise. 
While you can increase the amplifier gain to  display small signals over 
internal noise, a reduction of the range by a factor  of 8 or more is very 
undesirable. 
    Your available dynamic range is usually far less  than the maximum range 
of signals that you can receive.
 
    If you use a Lehman or similar long period sensor,  you should set your 
background microseism signal to maybe 200 counts. If you  don't do this, you 
may not be able to sense the long period low amplitude  signals masked by the 
microseism background.
 
    Regards,
 
    Chris Chapman





In a message dated 09/01/2007 16:16:40 GMT Standard Time, royb1@comcast= ..net=20 writes:
<= FONT=20 style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size= =3D2>With the=20 usual 14 or 16 bit A/D's, is that dynamic range=20 necessary?
Bob
Hi Bob,
 
    A 16 bit ADC with no noise has +/-1/2 bit=20 uncertainty. This is about 96 dB. Each factor of 2 gives 6 dB change. With=20 +/-10V input 1 bit =3D 0.305 mV
 
    Quiet opamps may give 0.1 to 10 Hz input noise=20 levels well below 1 micro volt; true chopper amplifiers may be much less tha= n=20 this. The CAZ opamps tend to give 1 to 2 micro volts, but these may give OK=20 results for long period signals, when 1/f drifts become large. It is a good=20 principle to use odd orders of low pass filter with a capacitor across the=20 feedback resistor of the first opamp. Never amplify high frequency signals.=20= This=20 limits the effect of intermodulation distortion and subharmonics
 
    Digital filters tend to give performance in the= mid=20 70 dBs, or less, but also suffer from switch transient feed through - 5 mV?=20= -=20 which may need additional analogue filtering before putting it into an=20 ADC. 
 
    In seismometry, we are seeking very low noise=20 levels at very low frequencies. You just can't afford to throw away +/-4 bit= s=20 signal through a poor choice of filter, or +/-3 bits by failing to average o= ut=20 the internal ADC noise. While you can increase the amplifier gain to=20 display small signals over internal noise, a reduction of the range by a fac= tor=20 of 8 or more is very undesirable.
    Your available dynamic range is usually far=20= less=20 than the maximum range of signals that you can receive.
 
    If you use a Lehman or similar long period sens= or,=20 you should set your background microseism signal to maybe 200 counts. If you= =20 don't do this, you may not be able to sense the long period low amplitude=20 signals masked by the microseism background.
 
    Regards,
 
    Chris Chapman

[ Top ] [ Back ] [ Home Page ]