Capacitors for precision circuits

[tszaboo]

I was designing precision stuff in the past, and there were a few bits, which required that little bit of extra precision for the capacitors. Becuase they affected the signal directly. I'm talking about TrueRMS circuits, RC filters for ppm level measurements, where the leakage currents would skew the output, these kind of applications. For these, my default go to solution was the Panasonic ECPU
Stacked Metallized Plastic Film Chip Capacitor
http://www.farnell.com/datasheets/1733019.pdf
Because it is SMD. And that was important at the time. There wasnt too many choices, but this one worked with the 1GOhm insulation resistance, and the ok AC characteristic up to a few khz.
But damn, these things are expensive if you buy only a few, and you dont care about the package. These will be connected to max 15V so X2 rating is not required.

So my questions are the following:
Do you have a preferred DC film capacitor brand?
Or do you have like a magic NP0 brand and series, which is working?

Thank you for the answers.

[T3sl4co1l]

FYI, I've heard that those parts can leak: the stack isn't sealed on the sides, and is pretty much impossible to wash.

I think this was relevant to a S&H, but similar concerns apply at your ~1Gohm level.

PS was always the old favorite, and Teflon, glass, and other exotica of decreasing production volume and rapidly growing cost; but they're all pretty much unavailable these days.

Afraid I don't have any relevant experience, I haven't worked much with low leakage circuits.

Tim

[daqq]

FYI, I've heard that those parts can leak: the stack isn't sealed on the sides, and is pretty much impossible to wash.

That's a nice surprise

Or do you have like a magic NP0 brand and series, which is working?

Dunno about NP0, but mica capacitors are pretty stable and precise.

[Edwin G. Pettis]

Unfortunately some of the best film capacitor types are no longer with us.  About the best solution is a high quality polypropylene capacitor, you may have to try different manufacturers to find one that works for your application.  If the capacitance isn't too large, go with a non metalized polypropylene.  Depending on your value, you may be able to find some polystyrene NOS, these were discontinued a few years ago but there is still inventory around.  Most of the relatively reasonable priced caps are very similar in performance and most of the higher priced 'exotic' caps do not get you that much better performance for the cost.  You can find polypropylene caps in 50V or 63V ratings.  Stay away from the so-called audiophile caps, you're paying a lot for marketing hype.

Surface mount caps tend to have problems the leaded caps don't and if you need the best performance, forget SMD and no, ceramic caps are not on the approved list for this type of app.

[The Soulman]

Wima fkp and fks series should do well, not small nor cheap tho.

(these days the crappiest website ever.  )

https://www.wima.de/en/our-product-range/parametric-search/

[Kleinstein]

It depends on the capacitance needed. In many cases polypropylene is a good choice - not too expensive but larger volume than polyester. They are still available, e.g. as Wima MKP of FKP series.  If you still get them old NOS polystyrene can be slightly better, but usually only small values.

The linked ones are not really good for a film cap, but with SMD there is not that much choice. Due to soldering temperatures the choice is even more limited here. So film caps are often better in THT. If at all good SMD caps would be G0G / NP0 type ceramics - they are not that bad.

[Andreas]

Hello,

RS-Components still has polystryene capacitors.
not cheap but still in production.

With best regards

Andreas

[tszaboo]

Or do you have like a magic NP0 brand and series, which is working?

Dunno about NP0, but mica capacitors are pretty stable and precise.

Wow, those start above a EUR each. Also they dont go below 100V rating, and up to 47nF. Not a really good deal.

FYI, I've heard that those parts can leak: the stack isn't sealed on the sides, and is pretty much impossible to wash.

I dont think the GOhm was really ever required, if just had to be better than ceramics or elcos. And there arent many technologies, so it was film. I also had a leakage guard trace below, where it was really nessesary. But the Panasonic is molded. In any way, after producing few thousand boards, I never had to measure the leakage of the capacitor.

The linked ones are not really good for a film cap, but with SMD there is not that much choice. Due to soldering temperatures the choice is even more limited here. So film caps are often better in THT. If at all good SMD caps would be G0G / NP0 type ceramics - they are not that bad.

At the time, being SMD was more important. I also think THT is a better choice, I just dont have any experience with them. Is the DC film a "premium part" that I can preatty much choose any brand and series, and it will work as intended?

Surface mount caps tend to have problems the leaded caps don't and if you need the best performance, forget SMD and no, ceramic caps are not on the approved list for this type of app.

I'm not so sure about that. 1206 NP0 capacitors have >10GOhm specified leakage currents. Granted, they will not come in high values, but for example a 10nF 63V capacitor. No excessive board stress, no vibrations for microphonics, used on room temperature... A NP0 might be a better choice.

RS-Components still has polystryene capacitors.
not cheap but still in production.

I was a bit puzzled by the price and performance. They it clicked, that I'm not looking at 500MOhm, but 500 GOhm leakage. That is certainly unique.

[Kleinstein]

There are now surprisingly high value NP0 MLC caps available. If it needs to be SMD, it might be the better alternative to SMD film caps. However large form factor SMD caps (e.g. 1206) are sensitive to board stress.
 
For caps in precision circuits, there are different properties that can be important: TC of capacitance, humidity effects, parallel resistance but also dielectric absorption or loss factor. So it depends on which application. Mica is supposed to have a stable capacitance over time, but are not especially good when it comes to the loss factor.

[Conrad Hoffman]

Silver micas have low losses, though not as good as polystyrene or polypropylene. Unfortunately, they sometimes have much higher dielectric absorption than you'd expect from the loss numbers. NP0/C0G caps are near perfect in every regard, except for limited value. Corning (CGW) made some very good "Spinguard" axial NP0/C0G caps (multilayer of some sort) with values to 0.01 uF and maybe beyond, but they're only available as surplus and only if luck is on your side. They also made non-NP0 parts that look the same, so caveat emptor. Actual DC leakage numbers aren't common and probably have a lot to do with size and how clean the assembly is.

[schmitt trigger]

Unfortunately, excellent dielectrics will be both bulky and expensive.

For sensitive analog circuits, I like polystyrene the best for DC or low frequency applications, as the ones I've seen come in wound format with too much ESL.

Also, polystyrene is easily damaged by solder reflow temps, and I've never seen a SMT part.

[Conrad Hoffman]

Wound capacitors get accused of excessive esl because they look too much like inductors. Remember that the current flows are in parallel opposing paths, so the inductance is far less then you might expect. No doubt a stacked construction is lower, but a lot of that is just due to a smaller loop between the leads- axials tend to be longer. There are radial leaded wound polystyrenes with closer lead spacing, but they don't seem to be as common. Loop area is everything.

[schmitt trigger]

  thanks for the info!

[cellularmitosis]

Thanks for this discussion guys.  I've just taken a risk on some NOS polystyrene's from ebay.  I'm going to use them with an LT1043 -- I'll also try some WIMA's and some regular ceramics and if I can measure the difference, I'll post the result here.

Hmm, if I can figure out Marco Reps mailing address, maybe I'll surprise him with some NOS polystyrenes.  I've been enjoying his electrometer videos lately   

edit: ebay link: http://www.ebay.com/itm/5pcs-Arco-125V-330nF-10-Styroflex-Radial-Capacitor-0-33uF-Polystyrene/112441535089?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649

[Kleinstein]

The LTC1043 charge pump is usually perfectly fine with polyester film caps. As the caps usually keep there voltage, there is no need for low dielectric absorption. An leakage does not have to be much lower than the chip.

[Andreas]

Mhm,

for WIMA MKS02 at 500 Hz and 1uF capacitors I have calculated
typical -1.33 ppm and max. -3.2 ppm error budget
for the capacitor leakage together with the switch resistance.

The leakage of the LTC1043 is about -0.08 ppm maximum.

On one of my dividers I measured with LM399#2 on input:
3430.4833 on high side (between 7V and buffered output of the divider)
3430.4222 on low side (between buffered output and ground)
so all in all -8.9ppm against the 2:1 division factor.

So it would be interesting how better capacitors are behaving.
But I would go for 1uF MKP capacitors with a time constant of 30000 seconds if I would do a redesign of my dividers.

With best regards

Andreas

[cellularmitosis]

I was hoping you would reply Andreas .  I'll keep you posted.

[Smith]

For really low leakage and small capacitance,  use teflon coax. Roughly 1pf per cm.