Voltage noise resulting from a buffered precharged capacitor?

[Echo88]

Assume we have a 1µF capacitor, that is charged beforehand to 5V and then buffer it with a JFET-OP: what is the resulting 0.1-10Hz voltage noise coming from the output of the OP?

Realistic specs for cap and OP would be: 1µF foil, PP, 10pA leakage current / ADA4625 or OPA140

I couldnt really make something out of the definition from https://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise#Thermal_noise_on_capacitors and apparently LTSpice needs a Vin source for noise analysis.

[Atomillo]

Might be I misunderstand how noise "works", but my understanding would be that when you disconnect the capacitor from the charging source (with it's equivalent output impedance and thus it's Johnson noise), the voltage it is charged to is that of the source *plus* the instantaneous value of the noise.

Thus if you perform many times the reading you would see that the Johnson noise of the charging voltage still remains, even if the capacitor itself has no noise: so you can not escape Johnson!

Is this what you were referring? I hope to have been of assistance!

[Echo88]

I was interested to evaluate how a charged and puffered cap would behave as a voltage reference, assuming its leakage and OP-bias current would be compensated and its all used in a temperature controlled environment.
If the cap-"reset noise" really only acts as an offset voltage to the charge voltage during disconnection of the charge-source, then the only remaing noise sources are the OP-voltage noise and the ESR of the capacitance times OP current noise?

[mawyatt]

Awhile back there was an IC voltage reference ( Intersil ) that used trapped charge on an internal cap Floating Gate by means of Fowler-Nordheim tunneling, the same Floating Gate technique that Flash memory utilizes. The inventors gave a presentation at the IEEE ISSCC we attended, very interesting and clever technique utilizing an old Flash Floating Gate memory process.

Edit: See,

https://patents.justia.com/patent/7759727

https://patents.google.com/patent/US7759727

Best,

[razvan784]

If the cap-"reset noise" really only acts as an offset voltage to the charge voltage during disconnection of the charge-source, then the only remaing noise sources are the OP-voltage noise and the ESR of the capacitance times OP current noise?

That is my understanding as well.
With the cap you get a constant kT/C error, only opamp noise is then relevant.
Of course, the hard part is keeping the temperature constant (to keep C constant) and managing leakage and input bias current.

[Atomillo]

I was interested to evaluate how a charged and puffered cap would behave as a voltage reference, assuming its leakage and OP-bias current would be compensated and its all used in a temperature controlled environment.
If the cap-"reset noise" really only acts as an offset voltage to the charge voltage during disconnection of the charge-source, then the only remaing noise sources are the OP-voltage noise and the ESR of the capacitance times OP current noise?

That's my understanding as well, that it only adds a random offset.

[Echo88]

Yeah, the floating gate technique is interesting.
Its really remarkable how the isolation of the gate is apparently so good that it achieves comparably low drift and "only" 30µVpp 0.1-10Hz despite using a mosfet, that must have biascurrent in the sub fA or aA-range, assuming the gate cap is in the nF-region (?).
Would be nice if user Noopy could open one up for die-shots

https://www.eevblog.com/forum/metrology/intersil-fga-(tm)-voltage-references/
https://www.renesas.com/us/en/document/dst/x60003-datasheet?r=481586

[Andreas]

Yeah, the floating gate technique is interesting.

Yes but at least the ISL21009 (3uVpp noise) has a "feature" (turn on time of serveral days until ppm level) which makes it completely useless for volt-nutting.

https://www.eevblog.com/forum/metrology/intersil-fga-(tm)-voltage-references/msg1229037/#msg1229037

with best regards

Andreas

[iMo]

@Andreas: I've been observing the same "feature" with my 2xLT1021 ref - 60ppm after 1 year off (3+ weeks to get back), 7ppm after 12 hours off (hours to get back). Btw, my TC is aprox 0.2ppm/C. I think it is related to that epoxy package (you had too)..

Capacitors have got noise too - ie wiki/Johnson–Nyquist_noise says 1nF has got 2uV rms "without considering R" as the R drops out of the equation.. Interesting..

[Andreas]

@Andreas: I've been observing the same "feature" with my 2xLT1021 ref - 60ppm after 1 year off (3+ weeks to get back), 7ppm after 12 hours off (hours to get back). Btw, my TC is aprox 0.2ppm/C. I think it is related to that epoxy package (you had too)..

Hello,

the ISL21009 went from 4.99979 V to ~4.99960 V after less than 1 minute "off time"
After 100 minutes still 60 uV = 12 ppm too low. (at same temperature).
So no humidity effect in my case.

with best regards

Andreas

[Kleinstein]

The kt/C reset noise only happens during switching off. So for the time after the capacitor is connected the reset noise is like part of the noise of the 5 V source to charge the capacitor.

Over long time the current noise of the OP will become relevent.
The capacitor loss (including leakage, but also other parts like dielectric absorbtion)  will also contribute to the noise, just like having an extra large. With a good capacitor this should not be real issue for the 0.1 - 10 Hz noise.  The leakage current may show excess noise. So like some resistors with applied voltage that can show extra 1/f type noise the capacitor loss can also be causing more low frequency noise than just the johnson noise, when the capacitor is charged. Expect such an effect especially for electrolytic capacitors. The Johnson noise is only for the loss part, not the ideal capacitance.
With the usually rather low loss factors of foil capacitors I would not worry too much about this noise part.

The capacitor can also convert variation in the temperature and mechnical stress to a voltage signal. The capacitance TC can be in the 50-100 ppm/K and thus relatively large compared to a good voltage reference.  So the capacitor would still like a stable temperature for sub ppm level noise.

[Echo88]

Maybe some sort of charge collection takes place on the inside of the ISL21009 and when it looses power, the charge equalization takes time?
Like the dielectric absorption that leads to a long settling effect on glass reed relays, since the charges dissipate only slowly over the glass surface.

Regarding the capacitor-voltage reference:
To keep the voltage on the buffer cap constant, i assume it would need to be used in conjunction with a normal voltage reference.
The voltage difference between both references would be detected and via a PID-feedback loop and an isolated bipolar pA-current source produce a corresponding tiny current that compensates the capacitor leakage and OP bias current.
A photoisolator could be used as the current source and could be mounted inside the heated enclosure without problems.

[Vtile]

Don't underestimate the mechanical part as Kleinstein pointed out, some condensators can be awfull "microphones" as I personally did find out with my non-scientific^tm poking with needle meter.

[openloop]

Capacitors are weird. I played with a few of different, mostly film types (in 1uF and up range) when learning to use my SR560. If I remember correctly, quietest were PP (didn't have a large polystyrene at the time). PPS were next. Acrylics, regular electrolytics and tantalums were fairly noisy and, let's say, they "behaved". (talking .1-10Hz here). Also, noise appeared to be proportional to voltage. And all that with me being careful about temperature - caps have fairly bad tempco: change in capacity, by definition, with constant charge, causes proportional (to change in capacity) change in voltage on terminals.

It seems that my experience is in agreement with Jim Williams's (AN 124). He had to resort to using crazy expensive (to this day) wet tantalum...

[Wolfgang]

The kt/C reset noise only happens during switching off. So for the time after the capacitor is connected the reset noise is like part of the noise of the 5 V source to charge the capacitor.

Over long time the current noise of the OP will become relevent.
The capacitor loss (including leakage, but also other parts like dielectric absorbtion)  will also contribute to the noise, just like having an extra large. With a good capacitor this should not be real issue for the 0.1 - 10 Hz noise.  The leakage current may show excess noise. So like some resistors with applied voltage that can show extra 1/f type noise the capacitor loss can also be causing more low frequency noise than just the johnson noise, when the capacitor is charged. Expect such an effect especially for electrolytic capacitors. The Johnson noise is only for the loss part, not the ideal capacitance.
With the usually rather low loss factors of foil capacitors I would not worry too much about this noise part.

The capacitor can also convert variation in the temperature and mechnical stress to a voltage signal. The capacitance TC can be in the 50-100 ppm/K and thus relatively large compared to a good voltage reference.  So the capacitor would still like a stable temperature for sub ppm level noise.

I agree. A phenomen that should not be ignored, however, is dielectric soakage. This can lead to a "tail" of charge and discharge currents and associated flicker noise. Old styroflex are good in this rspect, but they are also large, unfortunately.

[Kleinstein]

......
I agree. A phenomen that should not be ignored, however, is dielectric soakage. This can lead to a "tail" of charge and discharge currents and associated flicker noise. Old styroflex are good in this rspect, but they are also large, unfortunately.

For low dielectric absorbtion a new option are C0G capacitors. In my tests I found most of them caparable to the classic PP or PS types and a TDK brand about a factor of 5 better. Up to some 10 nF the C0G caps are not even expensive.

[Wolfgang]

......
I agree. A phenomen that should not be ignored, however, is dielectric soakage. This can lead to a "tail" of charge and discharge currents and associated flicker noise. Old styroflex are good in this rspect, but they are also large, unfortunately.

For low dielectric absorbtion a new option are C0G capacitors. In my tests I found most of them caparable to the classic PP or PS types and a TDK brand about a factor of 5 better. Up to some 10 nF the C0G caps are not even expensive.

I had a range of up to microfarads in mind. Are there better options available today for those as well ?