[ASK] Standard Capacitor ....

[BravoV]

Which capacitor type is considered "good enough" as a standard capacitor ? Also curious what are other types that are used as in metrological standard ?

The words "good enough" here I mean for enthusiast grade people like my self. 

Ok, for me, maybe not a metrological standard grade to be honest, but more like stable and predictable enough for used as hobbyist reference caps. My reason is quite simple, I own few LCR meters, starting from cheap handheld up to benchtop type that can measure and sweep the DUT up to 1 Mhz with adjustable DC bias feature.

As enthusiast, for example sending my benchtop LCR to calibration is way too overkill for my budget also my need.

What I want is to have few caps that are known to be stable in years, once measured and profiled by other calibrated high end LCR meter, these will be my own reference for checking my LCR meters in the future whether they're still in spec, exactly like those voltage standards that are sold for checking DMM.

Any comments regarding long term stable capacitor type and also it's discussion are welcome.

Thanks in advance.

[Vgkid]

Silver mica caps are used for standards. They should be pretty stable, but im not sure on the TC of commercial ones.

[lowimpedance]

Fused silica is a popular dielectric for Metrology grade capacitors.
for example:
http://www.andeen-hagerling.com/ah11a.htm

[quarks]

Fused silica is a popular dielectric for Metrology grade capacitors.
for example:
http://www.andeen-hagerling.com/ah11a.htm

Very interesting
Until now I thought the GenRad 1404 is the very best Standard.
http://www.ietlabs.com/1404.html

before I got the metrology-/voltnut-/ppm-virus, I used this kind of capacitors (see att. pic. 1)
I have them since early to mid 80s and they are still very stable and usefull for hobbiest references (but by no means Standard capacitors)
If I would search for replacements I would look for Glass Capacitors and good NPO (like this blue ones in pic. 2)

[manganin]

Which capacitor type is considered "good enough" as a standard capacitor ? Also curious what are other types that are used as in metrological standard ?

Any comments regarding long term stable capacitor type and also it's discussion are welcome.

Gas dielectric capacitors

The capacitance is proportional to the area of the electrodes and inversely proportional to their distance. That is why mechanical rigidity and low thermal expansion coefficient of the electrodes and their support structures (insulators) are the main design parameters. In the before mentioned General Radio 1404 the electrodes and support screws are made of extremely low expansion alloy Invar (FeNi36). The mechanical rigidity can never be overlooked: when turned upside down, the capacitance of the 1404 slightly changes due to gravity.

The capacitance is also related to the permittivity of the dielectric gas which unfortunately depends on temperature, pressure and moisture. The capacitance element of the 1404 is closed inside a rigid hermetic brass container filled with dry nitrogen. The oxygen and moisture free construction is also important to avoid corrosion. Which basically means that conducting metal electodes slowly turn into insulating oxide which affects the capacitance by reducing the thickness of the electrodes and altering the permittivity of the dielectric.

The supporting insulator material forms a part of the dielectric and also a path for the losses. However in 3-terminal capacitors this problem can be avoided if there are no supporting insulators between the electrodes but only between the electrodes and the guard terminal.

The frequency dependence of the gas dielectric capacitors is very low because the permittivity of gases is relatively constant up to several hundred MHz. In real life the parasitic inductance of the connections sets the limit.





Solid dielectric capacitors

Using solid dielectric is usually the only practical way for making capacitance standard above a few nF. The solid dielectric allows the electrodes and the dielectric to be very thin but still mechanically stable (again the capacitance is proportional to the area of the electrodes and inversely proportional to their distance). The electrodes can even be vaporized on the surface of a thin dielectric. The permittivity of the solid dielectrics is also higher than gases which increases the capacitance. The most useable dielectrics in capacitance standards are mica and fused silica.

The frequency dependence of the solid dielectrics is much worse than gasses and so are the losses.

[BravoV]

before I got the metrology-/voltnut-/ppm-virus, I used this kind of capacitors (see att. pic. 1)
I have them since early to mid 80s and they are still very stable and usefull for hobbiest references (but by no means Standard capacitors)
If I would search for replacements I would look for Glass Capacitors and good NPO (like this blue ones in pic. 2)

Same here, although mine are mostly silver micas.

Bought few of this NPO recently, going to use it as reference cap, with tempco of 30ppm and virtually doesn't age as Kemet claimed. 

What made me interested at this cap is apart from its a NPO, it has relatively big capacitance (100nF) compared to most silver micas I have.

[BravoV]

Solid dielectric capacitors

Using solid dielectric is usually the only practical way for making capacitance standard above a few nF. The solid dielectric allows the electrodes and the dielectric to be very thin but still mechanically stable (again the capacitance is proportional to the area of the electrodes and inversely proportional to their distance). The electrodes can even be vaporized on the surface of a thin dielectric. The permittivity of the solid dielectrics is also higher than gases which increases the capacitance. The most useable dielectrics in capacitance standards are mica and fused silica.

@Manganin, thanks for the insight of the big boys standards, its amazing to see how it was build, to my eyes its more like a piece of fine art.

As you pointed above on solid dielectric, fused silica is definitely out of question, compared to mica, how about the 100nF NPO at my previous post ? Is it good enough as a hobbyist reference cap ? Also reading Kemet's literatures, the NPO's tempco is quite linear and predictable.

It has relatively huge capacitance compared to common mica caps which usually only max out at few thousands pF.

[manganin]

how about the 100nF NPO at my previous post ? Is it good enough as a hobbyist reference cap ? Also reading Kemet's literatures, the NPO's tempco is quite linear and predictable. It has relatively huge capacitance compared to common mica caps which usually only max out at few thousands pF.

Capacitance standards constructed from NP0 SMD ceramic capacitors have succesfully been used even in primary impedance metrology. But as far as I know for short term transfer only. There are application like high frequency and large capacitance where the stability requirements are not as high. But I am still a little suspicious if there are any types stable enough for other than transfer use. On the other hand there has been a lot of development in the materials and manufacturing of the ceramic capacitors recently...

The NP0 capacitors have many good properties. The dissipation factor equal or lower than mica capacitors. And the high frequency behaviour is good, mostly because of the small size. The temperature coefficient is too high as-is, but it is easy to make an oven for a small chip.

The transfer standards that I have seen were constructed from a large number of a few nF range capacitors. Usually higher capacitance in the same size chip means more compromises.

[TimFox]

In the old General Radio standard capacitors and decade capacitors, silver mica was recommended for high frequencies, but polystyrene for lower frequencies.  The manuals had examples of change in effective capacitance for mica at low frequencies, which they blamed on polarization effects (presumably of the water of crystallization in the mica).  Polystyrene has a higher tempco, however.

[BravoV]

When in old General Radio's era, apart from the proven mica, just wonder if the development of NPO material at that time already matured as now ?

[ManateeMafia]

There have been several different Russian NOS silver mica capacitors for sale.
Search for the SSG mica capacitors on ebay. They appear to be hermetically sealed and many are sold in lots.

[TimFox]

General Radio and Cardwell made air-spaced standard capacitors up to about 1000 pF, with low-expansion glass mounting plate and spacers.  They were contained in coffee-mug sized cylinders with two male banana plugs (one to the case).

[BravoV]

There have been several different Russian NOS silver mica capacitors for sale.
Search for the SSG mica capacitors on ebay. They appear to be hermetically sealed and many are sold in lots.

Cool, didn't aware of those thingy, I will be watching ebay on those caps, thanks.

General Radio and Cardwell made air-spaced standard capacitors up to about 1000 pF, with low-expansion glass mounting plate and spacers.  They were contained in coffee-mug sized cylinders with two male banana plugs (one to the case).

Yep, now I have an idea on how to hermetically seal these reference caps, thanks.

[amspire]

I have a pile of Russian 50,000pF 0.5% SSG capacitors. I see you can get them on eBay for about US$10 for 4 capacitors.

They are enclosed in a metal body and they appear to be hermetically sealed with glass seals around the leads. Mechanically very stable - squeezing the body causes less then 1ppm change. I will try and do some tests including a comparison with a GenRad type 1409 Silver Mica standard capacitor. I suspect they will perform very well.

I also have some Russian plastic case 360pF 5% Mica caps that I will test as well.

I will have to set up a little shielded box for 3 wire measurement since measuring changes in a 360pF capacitor down to 1ppm means the bridge is sensitive to capacitances of 360 atto Farads (360 x 10^-18 Farads). Without a shielded box, walking into the room will change the capacitance.

[BravoV]

I have a pile of Russian 50,000pF 0.5% SSG capacitors. I see you can get them on eBay for about US$10 for 4 capacitors.

They are enclosed in a metal body and they appear to be hermetically sealed with glass seals around the leads. Mechanically very stable - squeezing the body causes less then 1ppm change. I will try and do some tests including a comparison with a GenRad type 1409 Silver Mica standard capacitor. I suspect they will perform very well.

I also have some Russian plastic case 360pF 5% Mica caps that I will test as well.

I will have to set up a little shielded box for 3 wire measurement since measuring changes in a 360pF capacitor down to 1ppm means the bridge is sensitive to capacitances of 360 atto Farads (360 x 10^-18 Farads). Without a shielded box, walking into the room will change the capacitance.

Hi Richard, welcome back, its been a while since your last post ! 

Looking forward to see that little box of yours and how it performs. 

[acbern]

So how do you verify a 1ppm change? The IET1689 digibridge, although one of the standards in metrology industry, will not resolve this.
Just a 1616 bridge I guess or an andeen hagerling. If you have one of these, I am jaleous.

[Vgkid]

The genrad 1620/1615 gives 1ppm resolution as well. Though if i remember correctly you can use the 1689 to give the fifference in  values.

[amspire]

Something similar to the Genrad transformer bridge. I have an ESI 707b monster that is also a transformer bridge, but it is a different design to the Genrad. The Genrad needs a self calibration step to calibrate all the range capacitors from the inbuilt standard capacitor. The ESI references everything back to a single 1nF Invar steel dry nitrogen dielectric hermetic standard capacitor. The bridge main range goes to 5 places, but it has a second Difference section that goes to 1ppm. It can resolve down to 10 atto Farads. Made about 1967.

[acbern]

This one (just as the 1615 and 1616 bridges) need a high voltage oscillator and a selective microvoltmeter. For curiosity, what do you use for these? I would think, if you are not going into a range lower than say 10pF, a 10V generator may actualy even do it.

[amspire]

I actually have the ESI 701 Capacitance Measuring System that is a hardly liftable box including the 707B bridge and the 861A AC Generator Detector. This generates 20hz to 20kHz at up to 240v RMS, and has a null Detector with a tracking bandpass filter, and a 0.25 microvolt noise level. All valves, but it still works great.

[acbern]

So from your experience, is it as good as it looks on paper?

[amspire]

I will get some photo's when I do the tests - I have just been making up some leads and making a little shielded box out of brass shim sheet. I am going to do some simple temperature tests on the Genrad capacitor and the Russian capacitor. All Genrad/IET say is that type 1409 capacitors at 23deg +/- 5 deg C, they should be within specifications. For the Genrad caps, that was often 0.05% and the newer IET 1409 capacitors, you can get 0.02%. No temperature coefficient curves are published, so I am curious.

I think the specifications are pretty conservative. It came from the factory with the standard capacitor adjusted to 0.001% and probably has never been adjusted since. In spite of this, I am pretty sure it is still within the 0.01% accuracy.

The quality is fantastic. The custom made switches have a light quality feel and are designed so they should never need maintenance. They still seem to be perfect after 50 years. The last two digits is a vernier dial connected to a custom 1.4 ohm resistive slider across transformer winding and they are constructed to be accurate to 1%. It is quite good having the last two digits as a dial as it is better then switches for tweaking. The bridge can get pretty clear nulls even down to the 1pp level. Readings seem very stable and repetative.

The Genrad bridges are masterpieces of design and you would probably still go for one first, but the ESI is very impressive.

The huge problem with all these transformer bridges is just they are so incredibly big and heavy. It sits in the garage most of the time, and occasionally comes out. I do love the concept of precision transformer bridges though. Basically the transformer cannot lose accuracy. The oscillator does not have enough power to ever damage the bridge windings, so as long as the switches are good, all I need is one accurate reference capacitor and I can calibrate the bridge with one adjustment. There is a small hex screw on the front panel that allows the standard capacitor to be adjusted +/- 0.03%. Strait after calibration, it is probably accurate to near 0.001%.

Concerning the calibration, I am trying to come up with a plan for a quadrature bridge, so I can precisely calibrate a capacitor from a resistor. It should be possible to make a quadrature bridge with a 10ppm precision. I need two sinewaves of exactly the same amplitude (within a few ppm) and exactly 90 degrees phase difference - I think I have worked out a plan to do that. Unfortunately, I may be forced to use a FPGA to realize my concept. I hope to find a way to avoid that.

[acbern]

Very interesting. I have ordered a 707 bridge, do not have it yet. I still need a generator and a detector, I plan to use a 8904A DDS generator and a lock in amplifier as detector. These I have, so no additional spendings. l have a set of Genrad capacitors from 0.1pF to 10 nF, a homemade 100nF, and an ESI SC1000 hermetic standard cap which I will use as my CAL standard. That will be calibrated next week and will be the basis for deriving all the rest. The concept of decade switchable transformers, not needing any calibration, is great. That way really everything can be derived from one capacitor. I had done some initial tests building up a capacitive bridge based on a Dekatran 7 digit ratio tranformer, and got pretty good results. A finished bridge though has the advantage that it has a defined accuracy as specified by the manufatcurer, versus a self-built bridge which needs to be validated (so you need the set of calibrated capacitors, not cost efficient).

Interestingly enough, I had also planned to build a quadrature bridge. I started out with a precise reference capacitor, a precise coaxial resistor and again the dekatran bridge. My results were not so good, from what I can say, not knowing exactly the value of the standard I used at that time. That will change now with the SC1000 comming back from calibration. As generators I used an Agilent 16 bit DDS (81150A) and I measured (verified) the phase between the two outputs with an Agilent dual channel counter, which is very precise. Levels with a 3458A, pretty precise too (SYNC), especially if you just need to know the voltage relation of the two channels, rather than the absolute value. So I believe that the signal quality/accuracy were not the problem, but rather the setup. I worked with 10kHz, and I think going to 1kHz is the first change to be implemented. Also the overall coaxial setup needs improvement, so I had ordered isolation transformers. Quadrature bridges as used in national metrology labs are pretty complex, so I started out with a much simpler approach (see below, last page).
http://home.mit.bme.hu/~kollar/IMEKO-procfiles-for-web/tc4/TC4-11st-Lisbon-2001/Imeko%202001%20Full%20Papers%20Directory/imeko-105.pdf
What configuration do you use?

[amspire]

You need lots of voltage for the ESI 707B bridge, but the easy way to do that is with a transformer - if you can find something suitable.

When you are measuring 1uF, the voltage across the 1uF is 1000 times less then the voltage across the standard capacitor. So to even get 0.2v across the 1uF cap, you need 200V across the standard capacitor. Also to measure at full resolution on the 10pF range, you need 200V across the measured cap to get 1V across the standard capacitor. The bridge works perfectly well at lower voltages, but the minimum for the 707B is 100V. The maximum voltage is 350V I think.  The oscillator in my bridge it only 1W power but it can output 200V at 10K output impedance, 20V at 100 ohms impedance, 2V at 1ohm and 0.2v at 0.01 ohms. Distortion is 2% max. The transformer output helps in meeting the isolation requirements needed by the bridge. It is important that the bridge is earthed at one point only - the input to the detector.

I assume you have the manual? I have the PDF if you need.

[amspire]

My quadrature bridge is much simpler - I have two very stable sinewave generators that I first null at a 1:1 resistive divider. It is dead easy getting a 1:1 divider that is as accurate as you need. This gives two digitally generated sinewaves that are exactly the same amplitude and exactly 180 degrees out of phase. I then exactly shift one oscillator by exactly 90 degrees (easy if it is digitally generated) and apply it to the reference C-R pair. With a few extra refinements I can make sure the load on the sinewaves is identical in both the calibrate and measure stages, and handling capacitor dissipation factor, etc can also be managed. Distortion of the sinewave may not be critical, but it has to have a very stable short term amplitude. It doesn't have to be a general purpose tool, as its main job is to be able to calibrate at least one capacitor value very accurately from a resistor. I have ways that even a simple 8 bit microcontroller can possibly be generating the oscillator source waveforms, but I can do better with a FPGA. Most high end oscillators/synthesizers are unlikely to have the stability and phase accuracy I need.

Then all I need is a way to get my resistor standards calibrated.