a simple noise filter, L or C

[DuctDuck]

How could I proceed in selecting a filtering circuit for a DC circuit being measured?

I have a 5pF capacitor being sensed by a jfet (i.e.-2n5459 gen.purp.); meaning the high impedance gate-source is parallel to the capacitor and its Rdischarge. Right now, 1.5kohm is jfet's Rs and is the input load/Rgen to a DC differential amplifier, meant to amplify the tiny DC fluctuations on the capacitor.

I don't want to have a parallel capacitor filter the DC fluctations nor do I want a series inductor too either. Possibly 60Hz, or better yet 6Hz, should be eliminated from the DC measurement signal.

Is there something I'm missing? A compromising point to reconcile with? Please fill me in!

 

[CDRIVE]

A schematic is always helpful.

Chris

 

[DuctDuck]

Yes, I quit agree...check out the thumbnail, please. You will see I have went with an RC integrator as the output for a CD jfet amplifier. Keep in mind this is a low DC voltage measuring circuit and not AC. I predict the RC time constant will have to be much smaller than the previous RC time constant.

p.s.-I would like to deal with the filter issue immediately before anything else.

p.s.s-I'm getting re-aquainted to using jfets again, so please be patient

 

[duke37]

I do not see where the DC signal comes in.

What frequency do you want to eliminate?

 

[DuctDuck]

Oh, sorry. The 5pF cap is an air capacitor I built as an ion chamber (i.e.-volume of air and its ions go to the plates), that is where the DC signal comes in.

The frequencies I want to be rid of are from infernal thermal noise. The S/N calculations point to an extremely low figure; perhaps a low corner frequency Lpf can change it around (e.g.- 60Hz).

 

[duke37]

I do not see how the ions will go to the plates without a potential to attract them.

To get a turn over frequency of about 10Hz, you need an R*C time constant of 0.1sec.
Use C=100nF, R=1M.

I believe that fets have high low frequency flicker noise.

 

[CDRIVE]

Biasing a JFET like a BJT is rather odd. It's putting it nearly outside its linear region.

Chris

 

[DuctDuck]

Yes, Duke37, yes the potential is there! The design of the capacitor uses a coaxial capacitance calculation, and when 5V is there at either plate with respect to ground or better source, there is enough electrostatics for attracting a minute Coulomb population (i.e.- 23.3625x10-12 A worth as maximum).

I'm sorry, but I don't want to put any more capacitance on the input stage; after the buffer amp, Cf can then be very small returning minimal energy. No?

 

[duke37]

Sorry, I just assumed that the voltage supply at the top of the diagram was positive.

Once you have gone through the fet buffer, you have a low impedance source and a load of 1M will not affect it.

 

[DuctDuck]

Great, I will go with your low turnover frequency point of 10Hz, et al. Thanks Duke!

I have noticed some bjt/jfet/mosfet amplifiers using power supply decoupling techniques. Does a CD jfet require this too?

 

[duke37]

Your + and - supplies need to be very well smoothed as any variation will be fed directly to the 5pF capacitor. A large capacitor on the negative divider would help.

 

[DuctDuck]

I want to thank the positive efforts of all towards reducing my thermal noise question thread (e.g.- Vnoise=50uVrms, BW=<4kHz).

I have done some simulations and experiments using your contributions, and some more improvements to analysis and redesigning. This may have castoff your valuable input; sorry Duke, but it was valuable at the beginning of this.

Because I'm using a bjt differential amplifier I have had to reduce the filter's impedance; at the jfet's source branch. This brought me around to using two approaches to solving my problem; AC signal response and DC transient analysis (sorry I don't know plural for analysis...a-nal-i-se-s but how do you spell it???).

Rs=4.7kOhm
Rf=3.3kOhm
(Rt=<8.2kOhm)

Cf=10uF
fc=4.8Hz (and Av=0.01 at f>48Hz...nice)

Ultimately, the results proved very different from all of your valuable input. Not that you're plain wrong, its the contributions I valued and learned from. So, thanks again and keep on contributing!