Precision Capacitor Design

I will start off by saying I don't think I have any answers you haven't already tried. I will put in my two cents if for nothing more than to provoke others to comment with useful information. For one I think I would confirm the calibration bridge to make sure that the frequency spec they state are true. Two, have you tried other types of caps such as tantalum or glass to see if they act the same. The only other thing I might try is finding a cap that gives you the exact value rather than padding them. I would imagine multiple caps would increase the temperature and frequency effects.

Thanks for your input Frank, The bridge is a General Radio 1620 precision capacitance bridge and is good for measurments down to 1 ppm.

The two dielectrics I am using mica and polystyrene show different temperature coefficients well, mica at +20 ppm and polstyrene at -120 (ish) ppm, so mixing the dielectrics I can make a capacitor with almost zero temperature coefficient.

The problem with other types of capacitors are that the tempco can be non-linear and very unpredictable i.e. have a wide tolerance... As for frequency effects I think most data sheets gloss over this aspect of the dielectrics used as 100 ppm is only 0.01%. As for finding a capacitor at exactly the right value... Hmmmm have you seen any capacitors with ultra low temperature coefficients and having a tolerance of +/- 0.02%??

The only ones I have found are the national standards used for calibration they are a special fused silica type which because of its temperature coefficient has to be kept in a very accurately controlled oven!!

Oh! and they cost about £4000 each!! That's why we are looking to make and sell some similar versions, at a reduced price of course..!

John.

Hi,

you should have a look into the data collection from Hippel, this is an old collection of dielectric data.

You should find it in any university bibliotheke.

There in are temperature and frequency dependencies of many many materials.

To my knowledge quartz is much better than mica and the best possible choice combining ultralow temperature coefficient of both mechanical dimensions and dielectric constant, high frequency use, high dielectric strength, high mechanical strength. So which material can be better?

I think that mica has more than 8 times the TC of dielectric constant and the NPO ceramic more than 30 times.

Teflon and crosslinked PE is pretty good but mechanically not particularly stable.

So if you want to do a very good design than I think depositing quartz in vacuum deposition onto aluminum and winding this on a quartz tube will do a good job.

Take a look into shielding and guarding.

Have a look on existing designs from General Radio or DANA (I never had the opportunity to open one of these but I know they are doing a good job).

Ask at NBS if you are in the US or one of the related institutes if elsewhere.

How much money and how much time do you have to spend and how many of these capacitors do you intend to sell at which price?

Have success

RHABE

Thanks for your input Rhabe... Quartz is a very similar material to mica, I've not heard of quartz being used as a dielectric in a capacitor before, maybe because of its piezoelectric properties that make quartz crystals so effective!

I haven't heard of this Hippel research data so I will check that out on Google... Proplem is that most of this research information is only available to members, some isn't though, so its worth a check.

When I found this extreme frequency change with mica capacitors I was surprised, as mica is supposed to be good for high frequency use...

I was hoping someone had some knowledge of the different dielectrics and their characteristics, maybe this Hippel collection has the information I need...

Thanks - John.

Hi Electroman,

there are (I don't know fdrom which supplier) quartz capacitors on the market, small values - pF - only but really the best you can get. These are much better than mica.

Mica is mechanically not good at all but electrically acceptable.

But if you change to quartz glass (not quartz crystal that is piezo-electric), you will have best possible results.

The CMOS gateoxide is pretty thin, down to 5 nanometer, is made by vacuum deposition - I do not have the datails - and is still a very good insulator with high breakdown field strength.

So I would start with vacuum deposition of quartz glass - electron beam evaporation is a good approch, but searching for the procedure that is done in CMOS fabrication is likely to be a better approach.

Look for a used roll-coater that is coating any material onto a foil of any material, either metallic or plastic, if cating onto plastic you have to do quartz glass and aluminum.

If coating onto aluminum then only quartz glass has to be deposited.

Then you can think about best possibilities to contact, to make it rugged mechanically, to add guard and shield surfaces and contacts and to enclose it into a high tightness (helium-tight) vacuum housing to be evacuated and sealed.

I can look at the title of the book written by Hippel if next time in my office but you will not find a material better than quartz glass.

Have success and good luck

RHABE

Silvered mica caps are use for their tight tolerances 1% or better, the do drift with temp but not too much. .5% to 1% depending on their value and the stability of all other factors. Frequecy, voltage, temp drif over all. The drif you descibe make be caused by other factors, not your capacitor network. As frequency goes up so do the problems of stray capacitance. Moisture, suface contamination, (grease from your hands), even the lighting if you use a bench light. I know 2Khz is low frequency in most applications but electrons are funny things. It could be you need to cross reference with another bridge if you have one to hand or can borrow one. I usedto work for Marconi Instruments they used a Wane Kerr bridge, to cross check their own make. In their own instuments they used both silver mica and mainly polystyrene types. The polysterene were made by Sufflex. For special values they bought in very expensive purpose made caps. PTFE was used in micro wave circuits, it behaves well at 10 - 50Ghz. The circuit boards were made of PTFE substate and gold plated copper tracks. The really exotics used thin film techniques with ceramic substate. Microscopic Ceramic chip caps were soldered with special silver loaded low melting point solder. Some even glued in place with silver loaded epoxy resin.

A strip line layout was employed. At lower frequiecies ordinary fiberglass board was considered good enough. For very high voltage ceramic or silver glass were called for. Plenty of plastic film types were used for most other applications. Then there are the electrolytics and the surface mounted types (mostly ceramic) maybe laser trimmed. Resistors posed a bigger problem because of the internal heating they could drif in value in a most alarming way.

Another thing is the leads involved their inductance should also be accounted for.

I have looked high and low for a good stable close tollerance not too expencive type and high vloltage micro wave ceramic seem to fit the bill at the low vallue end. 30ppm and no drift. The oly drawback being their small size. As for the mf side of things you are stuck with polystyrene. 60ppm. Silver mica 100-200ppm. You will have to experiment with some combination where the tollerance and drift are evened out.

I tried but did I succeed? Only you can tell.

Hi Brainwave.... you worked at Marconi Instruments?? So did I!!

I was at Marconi from 1970 until 1979 at St Albans Lonacrea and Fleetville plants in the design labs... Does that coincide with when you were there?

I am using silvered mica because of its ultra high stability and extremely low temperature coefficient of +20 ppm, together with polystyrene for the higher values and to temperature compensate the silver mica with its' -100 ish ppm tempco.

If you calculate it right you can then get a capacitor with almost zero temperature coefficient. So Marconi's used a Wayne Kerr bridge? Do you remember the model number? I'm using a General Radio 1620 capacitor bridge with 0.01% absolute accuracy and a comparison accuracy of better than 1 ppm or 0.0001% - So no chance of comparing against another bridge with any sort of precision. The next highest accuracy bridge I have is a Wayne Kerr 605 which is only 0.05% and doesn't have the resolution to measure changes of 10 ppm etc...

Let me know when and where you worked at Marconi's as we have an annual reunion for all the ex-Marconi people near here, well just outside Hitchen in Herts.

John.

I was still completing my education 1974-5 I then worked at Cam Profiles near city hosp. Then at Elstree air field (fitting and servicing avionics equipment) up until 79, I came to M.I. in late 79 August, sept I think.

So we just missed each other, my father worked there in the fifties, when it was a rather poor set up. He used to describe it as two sheds and a small factory.

They also made medical equipment then Diathermy and small x-ray equipment.

I worked in several depts during my stay mainly in the QA and product planning depts.

I became very adept at running my own businees from their factory. I was doing plenty of repair work as well as my day job. In the eighties it all went down hill rather fast because there were too many people who had no real talent other than consuming as much alcohol as they could get away with. The social club bar made more money than the instruments business two years running. It was sold to an American company and finaly ended its days at stevenage, the long acres site sold off for housing. As new technologies came about, double sided boards surface mounted components etc, more and more work went out to subcontract. They did install two Auto insert, pick and place machines but these wasted more components than they ever fitted. Radial did the caps and Tr's the axial the R's and diodes.

In the mid eighties when GEC owned it, when lord Winestock had over site it did well when he went it all started to stuggle. We did pelican crossing equip for GEC traffic Borham wood. The trouble was M.I. was too small to compete. The mobile phone test set market then the auto test equipment (based at Fleetville) began ok then gave out under pressure from the big American europian compediters. I got made redundant in 1990 and collected all of £2400 for my troubles.

The reference is von Hippel, Dielectric Materials and Applications, 1954 MIT Press, huge, detailed, accurate. Old.

I excerpted 7-8 pages for "Capacitive Sensors," Larry K. Baxter, IEEE Press/Wiley 1997.

It shows AlO2, Steatite ceramic, mica, and quartz, for instance, with flat dielectric constants from 100Hz to 10GHz. Paper, glass, PTFE, polystyrene etc. have a few percent change, usually kicking in at a few hundred MHz. Polystyrene is flat to 10MHz I think NP0 ceramic is flat, high-K ceramic is bent. See also

www.ttiinc.com/object/tech_seminars_052505muratapacket

Look once more into Hippel's book at the TCs,

quartz is much better than the other good materials.

If anybody needs a particular page copied for personal use, please send me an email-request.

RHABE