In a message dated 2007/06/29, Tangazazen@....... writes:

> Hi Chris,
>       Each of the photo detector quadrants generate about 0.7 microamp which 
> are fed into virtual earth charge sensitive amplifiers.
> There are numerous noise sources but as you point out the shot noise 
> associated with the photo current is dominant. The quadrant detector is 8mm dia, in 
> fact a larger detector means a larger capacitance which in turn increases the 
> noise. 

Hi Martin,

       What sort of seismometer are you using this on? Is it in a feedback 
loop? 8 mm is rather small for use with a Lehman. These tend to have appreciable 
drifts, but +/-1/2" is usually adequate. You can get 10mm long Si photocells 
quite easily.

       You are driving the photo output into a zero impedance maintained at 
zero volts. What I don't follow is where any considerations of capacity arise? 
And why would increased capacity increase the noise? L and C components 
usually have no noise associated with them.

       However, this seems to be a very small photocurrent. What actually 
limits your resolution, if it isn't the shot noise of the photocurrent?

It is the 'lock-in amplifier' approach that controls the noise level. For an 
integration 
> time of a few milliseconds the effective bandwidth at 10KHz is 100 - 200 or 
> so Hz. Lock-in systems can pull out a signal that is substantially less than 
> the noise level.

       Certainly lock-in amplifiers make you immune to 1/f noise and the 
output signal is integrated. I would hope for a bandwidth considerably less than 
200 Hz. I appreciate that lock-in amplifiers can allow you to dig below ambient 
noise levels, but what does this do to the S/N ratio? 

       What measured resolution are you getting?

> The noise level of the LED has not been noticeable. The random conversion 
> to photons is offset by the use of a diffuse encapsulation, a bit like an 
> integrating sphere. The important requirement of the LED is uniformity of the 
> light spot and reasonable linearity when modulated. The feedback from the sum of 
> the quad elements is dynamic but there is a limit to how much the amplifier 
> loop can correct for nonlinearity.

       Which LED and photocell are you using? 
       What is the optical setup? 
       What sort of shutter are you using?
       Can you not use a LED with a flat window and a metal case - no 
integral lens? 
       I would expect a diffuse response to give a reduced resolution? 

>  LED temperature dependence under constant current conditions is non linear 
> but -0.7% over 20C to 80C is an approximate figure for a Gallium Arsenide 
> Phosphide at 670nm. The feedback loop as mentioned overcomes any temperature 
> dependence. Interestingly LED temperature coefficients seems to get smaller at 
> the shorter wavelength but the quantum yield of the LED and the response of 
> the Silicon diode decrease, it's a question of optimisation.  
   
       This is still quite a serious temperature dependence.
       How are you driving the LED? Is it constant on current, variable time, 
or constant on time, variable current? Half sine, triangle or square wave?
       Maybe switch to GaAs photocells?

>  If I had to choose between full capacitive bridge and a quad photo 
> detector I would choose the latter, It's a much more elegant solution.

       Do you have a circuit diagram, please?

       Regards,

       Chris Chapman   
In a me=
ssage dated 2007/06/29, Tangazazen@....... writes:



Hi Chris,

      Each of the photo detector quadrants generate=
 about 0.7 microamp which are fed into virtual earth charge sensitive amplif=
iers.

There are numerous noise sources but as you point out the shot noise associa=
ted with the photo current is dominant. The quadrant detector is 8mm dia, in=
 fact a larger detector means a larger capacitance which in turn increases t=
he noise. 




Hi Martin,



       What sort of seismometer are you using=20=
this on? Is it in a feedback loop? 8 mm is rather small for use with a Lehma=
n. These tend to have appreciable drifts, but +/-1/2" is usually adequate. Y=
ou can get 10mm long Si photocells quite easily.



       You are driving the photo output into a=
 zero impedance maintained at zero volts. What I don't follow is where any c=
onsiderations of capacity arise? And why would increased capacity increase t=
he noise? L and C components usually have no noise associated with them.



       However, this seems to be a very small=20=
photocurrent. What actually limits your resolution, if it isn't the shot noi=
se of the photocurrent?



It is the 'lock-in amplifier' approach that controls the noise level. For an=
 integration 

time of a few milliseconds the=
 effective bandwidth at 10KHz is 100 - 200 or so Hz. Lock-in systems can pul=
l out a signal that is substantially less than the noise level.



       Certainly lock-in amplifiers make you=20=
immune to 1/f noise and the output signal is integrated. I would hope for a=20=
bandwidth considerably less than 200 Hz. I appreciate that lock-in amplifier=
s can allow you to dig below ambient noise levels, but what does this do to=20=
the S/N ratio? 



       What measured resolution are you gettin=
g?



The noise level of the LED has=20=
not been noticeable. The random conversion to photons is offset by the use o=
f a diffuse encapsulation, a bit like an integrating sphere. The important r=
equirement of the LED is uniformity of the light spot and reasonable lineari=
ty when modulated. The feedback from the sum of the quad elements is dynamic=
 but there is a limit to how much the amplifier loop can correct for nonline=
arity.




       Which LED and photocell are you using?=
 

       What is the optical setup? 

       What sort of shutter are you using?

       Can you not use a LED with a flat windo=
w and a metal case - no integral lens? 

       I would expect a diffuse response to gi=
ve a reduced resolution? 



 LED temperature dependence und=
er constant current conditions is non linear but -0.7% over 20C to 80C is an=
 approximate figure for a Gallium Arsenide Phosphide at 670nm. The feedback=20=
loop as mentioned overcomes any temperature dependence. Interestingly LED te=
mperature coefficients seems to get smaller at the shorter wavelength but th=
e quantum yield of the LED and the response of the Silicon diode decrease, i=
t's a question of optimisation.  


   

       This is still quite a serious temperatu=
re dependence.

       How are you driving the LED? Is it cons=
tant on current, variable time, or constant on time, variable current? Half=20=
sine, triangle or square wave?

       Maybe switch to GaAs photocells?=




 If I had to choose between ful=
l capacitive bridge and a quad photo detector I would choose the latter, It'=
s a much more elegant solution.




       Do you have a circuit diagram, please?<=
BR>


       Regards,



       Chris Chapman