On Tue, 2012-10-09 at 22:28 -0400, Brett Nordgren wrote:
> Hi Bob,
> 
> The AD706 is quite a low-noise device.  Typically 15E-9 V/sqrt-Hz 
> with 1/f corner at 4Hz and 50E-15 A/sqrt-Hz with 1/f corner at 8Hz.
> 
> At lowest frequencies the LT1007A looks like it typically has 12x 
> lower voltage noise and 400x higher current noise than the 706.  I 
> need to look at how it would work in some of the higher-impedance 
> locations of our circuits.

Hi Brett,

One way to look at this - each op-amp has a "characteristic resistance"
above which its current noise dominates and below which its voltage
noise dominates.  As an example, take the voltage (en) and current (in)
noise density at say 1 Hz for these op-amps.  At 1 Hz for the AD706 en =
40 nV/sqrt(Hz) and in = 105 fA/sqrt(Hz).  At 1 Hz for the LT1007 en = 4
nV/sqrt(Hz) and in = 4.7 pA/sqrt(Hz).  The characteristic resistance
(en / in) for the AD part is 380K,  The characteristic resistance for
the LT part is 850 Ohms.

Say the circuit of interest is at a 50K resistance level.  The AD part
would contribute noise mainly from the en source (since 50K < 380K) or
on the order of 40 nV / sqrt(Hz).

The LT part would contribute noise mainly from the in source (since 50K
> 850 Ohms) or on the order of 235 nV / sqrt(Hz) (50K * in).

If the circuit of interest is at a 1K resistance level the AD part would
still show 40 nV / sqrt(Hz) but the LT part would show something like 6
nV / sqrt(Hz) (rms sum of the voltage and current noise sources).

The geophone is to be overdamped, so the circuit it will load into needs
a negative resistor somewhat smaller in magnitude than the Rc component
in your model - a few hundred Ohms.  The negative resistance circuit in
my experimental circuit uses a 499 Ohm resistor.  So, for my circuit at
least, the LT is the better choice.

My circuit can be downloaded from my Dropbox:

  https://www.dropbox.com/sh/icm65xdz8ga7w9y/Eh382QBfD2

The file is "overdamped.pdf."  It is overdamped to get a response down
to 1 Hz for the 10 Hz phone.  I am pleased with the performance of this
circuit as it appears to be dominated by local seismic background noise.
I probably need to find a quieter place to put it.  My next experiment
will try to extend the lower corner some.

>   Also Dave has generally used dual op amps 
> to simplify the board and keep its size down.  Though I suspect that 
> in seeking the best voltage noise performance one finds that dual 
> devices suffer from some degradation due to thermal interaction 
> between the two circuits.
> 
> At this point the broadband designs generally have lower noise than 
> the local background, though at Dave's excellent site that may not be 
> entirely true, below about 20 mHz.  What I really like to know is 
> which parts are generating how much noise and why.  So far the 
> position sensor has resisted noise modeling.  Measurements are in order.

The program CCICAP has a Noise Table feature that can tell you exactly
that.  See the noise table for the design I am experimenting with:

  http://c-c-i.com/2.95171615931122/ntbl0000.ac.txt

For example the entry for 0.1 Hz:

Frequency, total spot noise, total integrated noise: 
     1.0E-01  4.3E-05  2.1E-05
 Component integrated noise table:
   A2        1.503E-05   A1        1.399E-05   Rl        1.137E-06
   Rc        9.946E-07   R1        9.141E-07   R2        9.141E-07
   R3        6.153E-07   Rint      6.114E-07   R4        4.846E-08
   Reqv      1.324E-08

The spot noise is 4.3E-5 V/sqrt(Hz) (remember this is at the output with
substantial gain in the circuit) and the total noise is 21 uV.  The
elements are listed in order of decreasing noise contribution - A2 and
A1 dominate, with Rl, Rc, .. contributing lesser amounts of noise.



__________________________________________________________

Public Seismic Network Mailing List (PSNLIST)

To leave this list email PSNLIST-REQUEST@.............. with 
the body of the message (first line only): unsubscribe
See http://www.seismicnet.com/maillist.html for more information.