FemtoFarad capacitors

[Tiflo]

Hi,

I am working on a charge sensitive preamplifier (CSPA/FBFA) for a detector.
To achieve a suffisant gain, it require feedback capacitor as low as 10 femtoFarads (0.01 pF).
The lowest value that I can find on Digikey &Co is 0.1pF, which is way too high to obtain the desired gain.

I found an article where they make them themself, but i'm not convinced ...
"built with a 4 mm teflon piece with two wires in the edges in order to minimize the dielectric noise of the capacitors"
Article DOI : 10.1016/S0168-9002(02)01806-5  (sorry, can't share legaly the pdf ...)

Is interdigital capacitor a solution and if yes how to calculate them ?

Thanks for your help,

Tiflo

[jmelson]

This will never work.  You simply can't get that much gain out of a single amplifier stage.  Probably the parasitic capacitance of the device package is a whole pF (or more).

You need a very quiet first stage amplifier with a sane amount of gain, and then another simple voltage amplifier after it.

Jon

[sdouble]

silicon detector ?
x-ray detection ? what energy ?

[Tiflo]

This will never work.  You simply can't get that much gain out of a single amplifier stage.  Probably the parasitic capacitance of the device package is a whole pF (or more).
You need a very quiet first stage amplifier with a sane amount of gain, and then another simple voltage amplifier after it.

According to the first link : https://www.researchgate.net/publication/229038928_X-Ray_Spectroscopy_with_PIN_diodes (Fig11 & 15)
It seems to work pretty well (FET input configuration).

silicon detector ?
x-ray detection ? what energy ?

Si-PIN photodiode for ~10keV gamma detection

[IanB]

Capacitance doesn't only live inside capacitors, it lives everywhere in your circuit. If you get down to small enough capacitance values you need to control the capacitance with the physical design and layout of your circuit and not just rely on an off the shelf part.

[jmelson]

Capacitance doesn't only live inside capacitors, it lives everywhere in your circuit. If you get down to small enough capacitance values you need to control the capacitance with the physical design and layout of your circuit and not just rely on an off the shelf part.

Right, a PiN preamp I did some years ago needed about a 1 pF feedback cap, I ended up using 0.5 pF as the stray capacitance of the board provided the remaining capacitance.
Must have mostly been in the feedback R.

Circuit layout is REALLY critical in these sorts of things.

Jon

[mikeselectricstuff]

If you have to ask where to but a 0.01pf cap, you have bigger problems than buying a 0.01pf cap.

[Tiflo]

If you have to ask where to but a 0.01pf cap, you have bigger problems than buying a 0.01pf cap.

No need to be unpleasant, as you can guess, I'm not EE.
Sorry to try to rely on peer reviewed article, next time I should trust random people on internet (do not take it for yourself, I like your teardown )
I was surprised when I saw this cap value in the schematic and I come here for more informations.

Circuit layout is REALLY critical in these sorts of things.

If you have some references ?
Here, they only talk about guarding

[james_s]

You saw a published schematic with a capacitor that small? I'm not sure how it's even possible, you can easily have several pF between two traces with no actual capacitor at all. Any time you have two conductors separated by some sort of insulator you get a capacitor.

[mikeselectricstuff]

Circuit layout is REALLY critical in these sorts of things.

If you have some references ?
Here, they only talk about guarding

Guarding is about high impedance, reducing leakage currents by reducing the potential difference between sensitive nodes and anything else.
Reducing capacitance is much harder as it's simply the proximity of the parts to each other, combined with the relative areas of those parts, so it's a balancing act between keeping things small, yet not too close to each other.
Dielectrics will also be significant - what's between the parts - air, PCB, package material
It's a pretty unusual requirement, so there may not be much good reference material out there. It's made harder by the fact that you basically can't probe anything to test, as any probe will have lots more capacitance than the circuit.
You may find some info in the high-frequency RF field, though they tend to use much lower impedances to reduce the effects of capacitance. 

[mikeselectricstuff]

You'll notice the PCB shown in that paper has the JFET sticking up vertically - this will be to reduce stray capacitance and leakage. The caps in the schematic are probably just the stray capacitance of this arrangement.
 

[Mechatrommer]

You saw a published schematic with a capacitor that small? I'm not sure how it's even possible, you can easily have several pF between two traces with no actual capacitor at all. Any time you have two conductors separated by some sort of insulator you get a capacitor.

solder the components up in the air.

[ArdWar]

solder the components up in the air.

I still not convinced even this measure is enough in this case.

You only need a few millimeters of flying leads separated by 100 mils to get a few tens of femtofarad capacitance

[amspire]

It is actually surprisingly easy measuring down to 1 femtofarad and in this case it is a two stage amplifier. With shielding between the stages, you could possibly arrange a controlled 10 femptofarad of capacitance.

The trick to measuring is the two plates have to be inside a shielded container connected to ground. Connect one plate to a 10Khz oscillator with as much voltage as you can manage. 20V would probably ok, but 100V is better. The other plate goes to an opamp inverting input with a 1M feedback resistor. The other opamp input is grounded. have the output going to an oscilloscope so you can see if you are picking up 10KHz or just noise. Without the shield, the plate going to the opamp is also an antenna picking signal from other places like the oscillator leads and it makes it hard to measure anything accurately. With a bit of trial and error, you will quickly work out the dimensions and spacing to get the capacitance.

It may be you can get away with the same shielded capacitor in this circuit. I do not know the fet gate capacitance but if you ground the shield of the shielded capacitor, and keep the grounded shield to plate capacitance much lower then the gate capacitance, it could work.

[Kleinstein]

Usual sray capacitance can reach the 1 pF range. It takes extra effort to reduce the parasitic capacitance to below 0.1 pf. It also makes a big difference if the small capacitance to ground or only a coupling capacitance between two wires with much more capacitance towards ground allowed.
There are FEM based program packages to calculate fields and capacitance - some of the very expensive layout programs may include this.  Its possible but still a little cumbersome to transfer the layout data.

With discrete parts there is plenty of parasitic capacitance inside the parts too.
On a small scale chip, one might have to consider a 100 fF cap as an extra element that need considerable space.

[Zero999]

If you have to ask where to but a 0.01pf cap, you have bigger problems than buying a 0.01pf cap.

No need to be unpleasant, as you can guess, I'm not EE.
Sorry to try to rely on peer reviewed article, next time I should trust random people on internet (do not take it for yourself, I like your teardown )
I was surprised when I saw this cap value in the schematic and I come here for more informations.

Circuit layout is REALLY critical in these sorts of things.

If you have some references ?
Here, they only talk about guarding

Don't blindly  trust random people on the Internet. Independently verify what they're telling you.

Look up the formula for calculating the capacitance of two conductors and you'll see, things need to get very small before the capacitance is as small as 10fF. It may be possible on an integrated circuit, but nothing that can be built from discrete components.
https://en.wikipedia.org/wiki/Capacitance#Mutual_capacitance

[tszaboo]

A 0.2mm wide PCB trace on a two layer PCB above a GND plane has 42fF capacitance to ground. So your maximum length of a PCB trace is less than a mm. Clearly, you cannot build a circuit with traces that short. The fF is relevant to IC designs, not PCB designs.
It is not possible to use capacitors this size to make amplifier circuits.

For precision measurements, IDK, detecting single electrons or something like that. If your application is something like that, and you have a lot of budget to build this (something like 100KEUR, maybe more), get in contact with an analog IC designer firm, like Linear Technology or ON semi.
Alternatively, I think CCD cameras had to deal with small capacitance, so there might be some knowledge from companies making those.

[mikeselectricstuff]

I find it slightly suspect that the paper shows no pictures of the complete assembly including the photodiode. Considering the critical layout, why would they not include that?
 The schematic looks a lot like it's from a simulator - I'm sure you can get it to work in simulation where stray capacitance is ignored.
I do wonder if this was actually ever built as described and worked.

[amspire]

A 0.2mm wide PCB trace on a two layer PCB above a GND plane has 42fF capacitance to ground. So your maximum length of a PCB trace is less than a mm. Clearly, you cannot build a circuit with traces that short. The fF is relevant to IC designs, not PCB designs.
It is not possible to use capacitors this size to make amplifier circuits.

Not true - a 10fF cap does not have to be small - it can be big. You can make it meters long if you want

For example:
Two 4x4mm plates separated by 14mm is 10fF.
Two 10x10mm plates separated by 88mm is 10fF.
Two 0.1Mx0.1M plates separated by 0.88 meters is 10fF.
Two 1Km x  1Km plates separated by 880 Km is 10fF. OK that one may be a bit hard to do.

If you put them in a shield, there is no inteference from anything else.

Ok, but what about all the plate to shield capacitance?

Well if the shield is grounded, the opamp output doesn't care less about a a few pF capacitance to ground.
On the input, the important thing is that the capacitance to ground is significantly smaller then the fet gate capacitance. If the FET has a 2pF gate capacitance, you will try and make the input to shield capacitance, say, 0.2pF.

It may mean a big shield. If you shield the whole FET input stage, there is no stray capacitive leakage from the opamp out.

I have an old capacitance bridge that can measure 10fF capacitors to 1% accuracy. As long as it is in a shield, it is not a big problem. The way the capacitance bridges work is that when balanced, the non-driven end of the capacitor is exactly at shield potential, and so the capacitance to the shield is totally cancelled out.

With a good enough measurement circuit, you can make and accurately measure a 1 attoFarad capacitor - two 1mm x 1mm plates separated by 8.8 meters.

[Zero999]

A 0.2mm wide PCB trace on a two layer PCB above a GND plane has 42fF capacitance to ground. So your maximum length of a PCB trace is less than a mm. Clearly, you cannot build a circuit with traces that short. The fF is relevant to IC designs, not PCB designs.
It is not possible to use capacitors this size to make amplifier circuits.

Not true - a 10fF cap does not have to be small - it can be big. You can make it meters long if you want

For example:
Two 4x4mm plates separated by 14mm is 10fF.
Two 10x10mm plates separated by 88mm is 10fF.
Two 0.1Mx0.1M plates separated by 0.88 meters is 10fF.
Two 1Km x  1Km plates separated by 880 Km is 10fF. OK that one may be a bit hard to do.

All good in theory, until you find that the capacitance of the leads to the plates dominates the capacitance of the plates very quickly.

[amspire]

A 0.2mm wide PCB trace on a two layer PCB above a GND plane has 42fF capacitance to ground. So your maximum length of a PCB trace is less than a mm. Clearly, you cannot build a circuit with traces that short. The fF is relevant to IC designs, not PCB designs.
It is not possible to use capacitors this size to make amplifier circuits.

Not true - a 10fF cap does not have to be small - it can be big. You can make it meters long if you want

For example:
Two 4x4mm plates separated by 14mm is 10fF.
Two 10x10mm plates separated by 88mm is 10fF.
Two 0.1Mx0.1M plates separated by 0.88 meters is 10fF.
Two 1Km x  1Km plates separated by 880 Km is 10fF. OK that one may be a bit hard to do.

All good in theory, until you find that the capacitance of the leads to the plates dominates the capacitance of the plates very quickly.

For a shielded capacitor, it is not hard to get essentially zero stray coupling to the opamp output and the connection to the capacitor to the opamp output is trivually easy.

At the other end, you could have the fet inside the shield right near to the plate if you want.

I admit the reality of getting a circuit to work that depended on a 0.1fF feedback capacitor would be scary. How big is the FET gate to drain capacitance? Then there is the capacitance of the wiring to the PIN diode and the capacitance of the PIN diode to the 65V supply. The capacitance to ground of the plate to FET connection is probably the least of the problems.

To me, this is one of those circuits that gives a false reading if you just take a breath in the next room.

[Damianos]

It seems as an experiment started a few decades ago: https://www.osti.gov/scitech/biblio/10158886
At the right side of the above site, you can download the full text.

[Tiflo]

Thanks all for the interesting discussion.

I have done some test today with a cheap PIN diode and a charge preamp dedicated to proportional counter (gain ~1V/pC  -> ~1pF feedback capacitor)
The result was highly noisy due to the crappy setup (poor shielding & long cable), but the gain should be sufficient for my application (25keV was detectable, 14keV could be with beter setup). So I will stay with the "standard" design.

An other article showing the faisability with 0.5pF feedback capacitor : https://link.springer.com/article/10.3938/jkps.69.1587 using a cooled FET preamplifier.

I do wonder if this was actually ever built as described and worked.

Maybe ...

It seems as an experiment started a few decades ago: https://www.osti.gov/scitech/biblio/10158886

Thanks for the theory.