PSN-L Email List Message

Subject: RE: Noise
From: "Blair lade" blairl@...............
Date: Fri, 8 Jul 2011 19:54:17 +0930


Geoff, 

 

what is your sample rate and are you averaging the data?

 

How far away from the IA is the seismo?

 

I've had a look at your circuit, 

 

for good common mode noise reduction , the 4 resistors around the 3rd op amp
(summing amplifier) have to be very well matched, typically , they need to
be 0.01% tolerance resistors.  This is quite critical to the performance of
the cancellation circuit. There is an ic made by Analogue Devices, the
Amp03 which is often used for this, it has 4 x  25k laser trimmed resistors
on the chip which are matched typically to 0.01% and saves a lot of effort.
(I use lots of them.)

 

 

The same applies to the  resistors around the 2 buffer amps driving the 3rd
op amp, oviously, if the gains of the input / buffer amps are not exactly
the same then there will be different levels of signals reaching the cmr
circuit.

 

Your use of the IA is a little unconventional, IA's are normally use to
provide a very high impedance (many meg ohms) to the signals so that there
is no loading on the signal  and I2R losses down the cable are also reduced,
that is why the +ve inputs of the op amps are used. 

 

That's not to say that the way you are using it wont work, it's just not the
normal method of using IA's. 

As you are not using the IA the 'normal' way and have the +ve inputs
grounded, ignore the need to have bias resistors to ground in this case..

 

Having long wires from the seismo going to the virtual earth connections  of
the buffer amps could result is all sorts of instabilities within the IA .

 

While it would seem that the connection is a virtual earth, in fact it is
the connection between the feedback resistor and the gain setting resistor
that is the virtual earth, any signal (noise) that gets introduced between
the virtual earth and the -ve input will get amplified a very large amount.
This includes voltages produced from temperature gradients and dissimilar
metals (seebeck effects)..

 

(There were some audio mixing consoles in the 1980's that used  virtual
earth mixing busses and they all suffered from poor noise and cross talk
performance)

 

There is quite a lot written about extending the LF response of seismometers
by applying overdamping using negative resistance, and I guess that's what
you are attempting, using an IA for this is probably not the way to go
unless you are using precision resistors, it's too hard to try and balance
it all up.

 

Have a look at http://www.vaxman.de/publications/teach_gp.pdf for example

 

A single op amp is used for the overdamping followed by lots of gain and
some equalising / filtering

 

blair

 

From: psnlist-request@.............. [mailto:psnlist-request@...............
On Behalf Of Bob McClure
Sent: Thursday, 7 July 2011 1:14 AM
To: psnlist@..............
Subject: Re: Noise

 

Raw data acquisition should always be done at a sample rate more than twice
the upper frequency passed by the amplifier. Sample averaging to a lower
sample rate is not the way to go, as simple sample averaging will reduce,
but not eliminate, aliasing. Instead, low-pass filter the raw data to an
upper frequency less than one half the reduced sample rate. Reduction to the
final rate can then be accomplished by simple decimation, or by sample
averaging.

I have done experiments to verify the above statements. I would recommend
that the digital low-pass filter be of fourth to sixth order.

Bob

On Wed, Jul 6, 2011 at 11:11 AM, Randall Pratt  wrote:

I'm following the discussion of Geoff's noise with interest.  I would like
to ask some questions on alias noise.  

 

If one were to vary the sample rate and sample at say 60sps, 40 sps, 15 sps,
6 sps is there a formula to take the peaks from the various ffts and arrive
at a list of probable noise alias problems?   I have done this and can see
some changes in noise peaks on the spectrums but I have not attempted to
calculate a relationship.

 

Secondly, if the AD runs at 120 sps and averages to 6 sps recording rate
will frequencies between 3 and 60 hz alias into the record?

 

Randy

 

 












Geoff,

 

what is your sample rate and are you averaging the = data?

 

How far away from the IA is the = seismo?

 

I’ve had a look at your circuit, =

 

for good common mode noise reduction , the 4 resistors = around the 3rd op amp (summing amplifier) have to be very well = matched, typically , they need to be 0.01% tolerance resistors.  This is = quite critical to the performance of the cancellation circuit. There is an ic = made by Analogue Devices, the  Amp03 which is often used for this, it has 4 = x  25k laser trimmed resistors on the chip which are matched typically to 0.01% = and saves a lot of effort. (I use lots of them.)

 

 

The same applies to the  resistors around the 2 = buffer amps driving the 3rd op amp, oviously, if the gains of the input / = buffer amps are not exactly the same then there will be different levels of = signals reaching the cmr circuit.

 

Your use of the IA is a little unconventional, IA’s = are normally use to provide a very high impedance (many meg ohms) to the = signals so that there is no loading on the signal  and I2R losses down the = cable are also reduced, that is why the +ve inputs of the op amps are used. =

 

That’s not to say that the way you are using it = wont work, it’s just not the normal method of using IA’s. =

As you are not using the IA the ‘normal’ way = and have the +ve inputs grounded, ignore the need to have bias resistors to = ground in this case..

 

Having long wires from the seismo going to the virtual = earth connections  of the buffer amps could result is all sorts of = instabilities within the IA .

 

While it would seem that the connection is a virtual = earth, in fact it is the connection between the feedback resistor and the gain setting resistor that is the virtual earth, any signal (noise) that gets = introduced between the virtual earth and the –ve input will get amplified a = very large amount. This includes voltages produced from temperature gradients = and dissimilar metals (seebeck effects)..

 

(There were some audio mixing consoles in the = 1980’s that used  virtual earth mixing busses and they all suffered from poor noise = and cross talk performance)

 

There is quite a lot written about extending the LF = response of seismometers by applying overdamping using negative resistance, and I = guess that’s what you are attempting, using an IA for this is probably = not the way to go unless you are using precision resistors, it’s too hard = to try and balance it all up.

 

Have a look at http://www.vaxman= ..de/publications/teach_gp.pdf for example

 

A single op amp is used for the overdamping followed by = lots of gain and some equalising / filtering

 

blair

 

From: psnlist-request@.............. [mailto:psnlist-request@............... On Behalf Of Bob = McClure
Sent: Thursday, 7 July 2011 1:14 AM
To: psnlist@..............
Subject: Re: Noise

 

Raw data acquisition should always be done = at a sample rate more than twice the upper frequency passed by the amplifier. = Sample averaging to a lower sample rate is not the way to go, as simple sample averaging will reduce, but not eliminate, aliasing. Instead, low-pass = filter the raw data to an upper frequency less than one half the reduced sample = rate. Reduction to the final rate can then be accomplished by simple = decimation, or by sample averaging.

I have done experiments to verify the above statements. I would = recommend that the digital low-pass filter be of fourth to sixth order.

Bob

On Wed, Jul 6, 2011 at 11:11 AM, Randall Pratt = <rpratt@.............> = wrote:

I’m following the discussion of Geoff’s noise with interest.  I = would like to ask some questions on alias noise. 

 =

If one were to vary the sample rate and sample at say 60sps, 40 sps, 15 sps, 6 = sps is there a formula to take the peaks from the various ffts and arrive at a = list of probable noise alias problems?   I have done this and can see = some changes in noise peaks on the spectrums but I have not attempted to = calculate a relationship.

 =

Secondly, if the AD runs at 120 sps and averages to 6 sps recording rate will = frequencies between 3 and 60 hz alias into the record?

 =

Randy<= span lang=3DEN-US>

 

 


[ Top ] [ Back ] [ Home Page ]