In a message dated 2008/02/16, Brett3mr@............. writes:

> If carried down to low enough frequency, integral feedback can go a long 
> way toward resisting instrument drift.  This is 'just' a matter of making 
> the instrument force / acceleration response approach zero at very low 
> frequencies.
> 
> Unfortunately, there is dirty little secret about using R-C integral 
> feedback to resist drift error forces, and that is evident when the 
> integrator is 'straining' to cancel a fairly strong unbalance force, in 
> which event there will be a substantial voltage across the integrating 
> capacitor.  The cap. has a temperature coefficient of C which is of the 
> same order of magnitude as the Temp. Coeff. of a steel spring i.e. pretty 
> large.  Since in a feedback integrator the charge, Q in the cap. changes 
> relatively slowly and can be considered to be constant as a first 
> approximation, if its capacitance goes up with temperature, its voltage 
> goes down in proportion because of Q=CV thus introducing its own rather 
> large drift effect.  It will only work as expected if the system is already 
> reasonably well balanced mechanically making the voltage across the cap. 
> not too large. 

Hi Brett,

       You have four drifts here. The change in the magnet strength with 
temperature, the changes in the coil with temperature and the change in the 
capacitor with temperature. The magnet strength is decreasing, the coil area and 
resistance are increasing and the capacitance is decreasing with increasing T. I 
don't know to what extent these can be chosen to cancel? However, you only 
have to put the low pass frequency below the minimum response frequency, but this 
could give problems with 1000 second instruments.

       Regards,

       Chris Chapman
   
In a me=
ssage dated 2008/02/16, Brett3mr@............. writes:



If carried down to low enough f=
requency, integral feedback can go a long 

way toward resisting instrument drift.  This is 'just' a matter of maki=
ng 

the instrument force / acceleration response approach zero at very low 

frequencies.



Unfortunately, there is dirty little secret about using R-C integral 

feedback to resist drift error forces, and that is evident when the 

integrator is 'straining' to cancel a fairly strong unbalance force, in 

which event there will be a substantial voltage across the integrating 

capacitor.  The cap. has a temperature coefficient of C which is of the=
 

same order of magnitude as the Temp. Coeff. of a steel spring i.e. pretty 
large.  Since in a feedback integrator the charge, Q in the cap. change=
s 

relatively slowly and can be considered to be constant as a first 

approximation, if its capacitance goes up with temperature, its voltage 

goes down in proportion because of Q=3DCV thus introducing its own rather 
large drift effect.  It will only work as expected if the system is alr=
eady 

reasonably well balanced mechanically making the voltage across the cap. 
not too large. 




Hi Brett,



       You have four drifts here. The change i=
n the magnet strength with temperature, the changes in the coil with tempera=
ture and the change in the capacitor with temperature. The magnet strength i=
s decreasing, the coil area and resistance are increasing and the capacitanc=
e is decreasing with increasing T. I don't know to what extent these can be=20=
chosen to cancel? However, you only have to put the low pass frequency below=
 the minimum response frequency, but this could give problems with 1000 seco=
nd instruments.



       Regards,



       Chris Chapman