Capacitors have memory like us - handle them carefully

I could talk for hours about the 'humble' capacitor and its properties as well as its failure modes...

I certainly wouldn't like to sit here for a day typing it all out! Unless maybe it was for a book on the subject.

John.

I think it will be a great idea to have a book specially on capacitors. Effect of temperature, voltage and time based history are the one ignored. How do I estimate the life of a capacitor? How long a capacitor will survive and which capacitor will hang around longer. In embedded amplifiers, I prefer non-polar capacitors and of NPO/COG or X7R types. I am also now seriously considering air and Mica capacitors. I did use polyester and paper capacitors other than normal ceramic capacitors.

I worked in a television repair shop in high school. We bought and sold lots of used sets. We would scrap some and save parts to repair others. We had a room full of CRT's. I have been shocked from CRT's that had been sitting on the shelf for months. Uh-huh-huh, yeah, it was cool, uh-huh-huh.

The "electret" effect is of course the basis for millions of microphones in current use. The electric field is "frozen" into a polymer that is used as a permanently charged dielectric element in the microphone.

Shyam,

I would say that A capacitor is a capacitor, forever:... is a fairly true statement. Having 35 years as an electronic technician, and "seat of the pants" engineer, designing, building, and testing electronic equipment, analog Audio, digital, power supplies, etc., I have found that most capacitors (except electrolytics) come about as close to ideal components, as anything else. Various plastics, oil or waxed paper, ceramics, oil, and many other materials have excellent dielectric properties. Among these; polycarbonate, mica, glass, and ceramics, are quite stable materials, dielectric properties do not degrade, possibly for well over a century, and they work well in lots of environments and over a wide temperature range. There are also quite a few good materials for the plates. Tin is one of the most popular and deservedly so. Most "bypolar" (meaning non-electrolytic) capacitors do not exhibit a "memory" effect, (other than holding a charge, which may then be discharged).

Electrolytic capacitors have improved dramatically over the last 50 years. The electrolytes last much longer, are less corrosive to the internal foils, and the outer cases are less susceptible to springing leaks. They can exhibit a "memory" effect because there are ions that form and physically drift, in the electrolyte forming layers,....but the effect is so much less than the memory effect exhibited by Batteries, on a lesser order of magnitude, as to be almost insignificant. If your statement were true, Shyam about a capacitor "getting used to" passing a certain level of signal, and then refusing to "open-up" to a higher voltage excursion, then wouldn't it be the case that in audio applications, just passing through several capacitors would result in limiting transient response. This is not the case. In spite of deliciously poetic prosey statements to the contrary from some capacitor manufacturers, of audio caps, the differences in performance between many types of caps with the same measured capacity, and voltage, operating within design parameters are less than single parts per million, and repeatably and measurably so.

As to air dielectric capacitors as a better choice, I say NO. Not only is the user of air gap capcitors subject to atmospheric moisture content which vaires enourmously, (could be +-20% from one day to the next), but one is also subject to the foils or plates attracting charged particles like smoke which can accumulate bridging the gap between the foils or plates, and shorting the capacitor. The old air gap tuning caps, in radios (40 years ago) were infamous for this type of failure.

Also the usual failure mode for caps is: punched through dielectric; which results in arcing through, at low voltage: not at all like a fuse.

Electrolytics also have high failure rates in switching power supplies because they are cycled (charged and discharged (100,000 plus times per sec) at high rates) near their operational limits.

Geoffrey Reed

what makes capacitors to come down in capacitor value with increased voltage? Assuming that there is no charge leak, then what makes the capacitor of say ceramic type to become less than 50% value for voltage from 100V to 600V? This does not happen in air capacitors.

I looked around and found documents characterizing the effect but they did not indicate a cause. As I think about it though it seems to parallel in many ways the saturation of magnetic materials. Perhaps there is a limit to the contribution that a dielectric can make to capacitance as the voltage increases which is a nice complement to the limit in increase a magnetic core material can make to inductance as current increases. Also, there is no "saturation" when the "core" is air.

In that case capacitor should recover fully. Capacitors show memory effect. We often ignore this part. However it is of great importance in measurement of signals of different level. After loading too much charge on the capacitor and even after the discharge cycle of the capacitor, it will show residual charge. We have come across this problem even in Keithley charge integrators and current amplifiers.

Not exactly. Except for air core inductors, they also display memory. There is a residual magnetism after current is removed.

I don't think there is any trick to defeat the residual charge memory. You may have to find a way to use a capacitor with air or even a vacuum dielectric. One other thought would be to try a capacitor with a much higher voltage rating than your operating voltage. That would minimize the stress on the dielectric material and perhaps reduce the undesirable effect.

Another problem I consider a serious one is in use of ceramic capacitors as filter for reference voltages. I have 10K pot that feeds signal to the comparator and has 0.1uF ceramic capacitor as filter for reference. I am sure these capacitors are good temperature sensors so they will change easily with temperature to large amount. Now that will also make the reference to fluctuate in same way as you find the electret microphone do. This sure will make my reference of worst kind and comparator will easily see it.

Yes virtually all passives are microphonic to some extent. If you have that much vibration perhaps you can minimize the effect by isolating the instrument from vibration. Also you may want to test capacitors of different dielectric materials to determine which one has the lowest response to vibration. You may find that the use of several capacitors in parallel yet with different physical orientation would ameliorate the effect of any single axis vibration.

Not the vibrations but temperature variations from air flow causes lot change in capacitor value. They are like thermo-capacitor temperature sensors and are very sensitive like thermistors. I did not take seriously and used to placed them along with 10uF Aluminum dielectric capacitors in parallel for noise reduction. I am bit scared of them and will use only X7R in future designs even though X7R in 0.1uF 50V range going to cost me a fortune when I am going to buy million numbers.

I don't understand unless the temperature change is very rapid. How high is the impedance sourcing current to the filter? Any slow change in capacitance would be made up for from the supply and would hardly be noticeable unless the source current is extremely low. If that were the case then perhaps you could reduce the effect by using a smaller capacitor and still have adequate filtering. Alternatively, perhaps you could insulate the circuit to slow down the change in temperature and the consequent rate in change of charge.

Shyam, you're using a ceramic capacitor for noise reduction??

Now I've heard everything.... ceramic should be avoided at all costs due to its piezo electric effect, they make excellent noise generators!!

Also the X7R and above are useless as far as temperature coefficient goes, you will be lucky to get a tempco of less than 1000 ppm per degree Celsius.

What with your noisy power supply because you're using out of date noisy series pass regulators (78XX and 79XX)...

I do begin to wonder about some 'engineers' on here.

John.

Dear John

I tried to convince the user who pay $4000 for amplifier but do not want to pay for power supply and are happy to get free those 78XX and 79XX series regulated version.

After putting 10 ohms 10uF|| 0.1uF ceramic, the noise came down to +/-5mV at the output of the amplifiers so they have good noise rejection from power supply. I am afraid the device is not doing enough good. In some places I can not use large capacitors as filters. I was using Polystyrene for DPMs. I have Teflon capacitors also but they are big and must have greater memory effect due to their electret properties. I use them only for charge integration.

Now I have 5V signal range so you can say that 10 bit dynamic range possible.

I do want to go down to 14 bit and that requires noise to go down an order of magnitude below.

I have seen CERN people using MICA and other capacitors and I am still using X7R for signal isolation.

There are two problems now. Signal isolation from high voltage and noise filters. India is a bad country for electronics parts. To test one capacitor I need to order 10,000 numbers as no one wants to ship small quantity. I also end up paying huge shipment cost and customs duty.

I want to shift my research facilities in USA and UK. I am moving out of he country from October month and will somehow look into this aspect more closely.

When is your book on capacitor coming to market? Book one for me.

Oooowww dear Geoffrey, I have to disagree with some of your comments...

Capacitors are far far from ideal components.... It depends what you want to use each one for as to the choice of the dielectric...

I have an excellent example of very poor design on my workbench at the moment, it was designed and built in the USA and its crammed with tantalum capacitors used as supply bypass capacitors...!!! I'm amazed it ever worked at all, those tantalum bead capacitors have a very poor dv/dt rating, as shown by this dead instrument, as tantalums often fail short circuit, they are sods to find when someone has designed tens of them decoupling the supply rails. So I'm having to replace every single one with a decent aluminium electrolytic and a ceramic capacitor to get this instrument working again.

What is this you write about .... dielectrics do not degrade? They certainly do!

Also what is this about most dielectrics don't show a memory effect? Every single one has a memory effect!! whether its large enough to concern a particular design or not is up to the designer to evaluate. If you don't believe me just try getting a nice CRT, discharge the anode and then touch the anode!! the glass dielectric is good but still has enough memory to throw you across the room!!!

Air dielectric is one of the better ones for the right applications... If you have a problem with dust contamination then put them in a box... the relative humidity of the air only has a tiny effect on the capacitance of 2 ppm per 1% RH, enough to ignore in normal use.... "The old radios of 40 years ago were infamous for this failure mode...." Really?!! Why are old radio sets still working after 40 years then? Surely if there was a real problem with dust / smoke build up all that is necessary is a spring clean using a brush, as I've done on numerous occasions on elderly 50 year old sets!!?

"The usual failure mode is punch through of the dielectric.....", this is only one of many failure modes and is why the capacitor designers made self-healing capacitor types, which are a metal deposited on the dielectric so that when punch through did occur the metal was evaporated around the locality of the hole to 'heal' the capacitor to enable it to continue working....

John.

Excuse me?????? By definition a capacitor holds a charge!!!!!!! The fundamental operating principle of a capacitor is that it has CAPACITY. That means it takes time to charge (with a finite supply of current), and will hold the charge indefinitely under perfect conditions (when taken out of circuit and the dielectric does not "leak". Of course you can store charge with a capacitor!!!!!!, that is what it is for!!!!!!!

If you are calling that a memory effect, then EXCUSE ME!!!!!!! Any cap that cannot hold a charge is not a capacitor. The term "memory effect" is a different term and is used to describe how batteries for instance will aclimate to a certain voltage storage level, and resist fully charging, after many cycles of partial charging. Having to do with ion implantation in the plates.

I said very clearly that unless you are talking about the ability for a capacitor to hold a charge, then, they do not have a "memory effect", except to a very slight degree for electrolytics. Notice that I did not mention Tantalums, Tantalums, are the very furthest from ideal, 1st they ARE electrolytic, and among electrolytics they have the poorest back leakage spec's, they are filled with electrolyte and are like miniature batteries, charging and discharging, The leakage accross the plates has a real "memory effect" because the plates are actually forming and ionizing into the electrolyte, as the cap is charged. They were popular because it was possible to make them extremely small before the advent of monolythic film caps, which are also small and have a number of drawbacks, like the leads, break off easily and they have a finite measurable resistance to charging discharging.

And if you think an air-gap capacitor cannot hold a charge, BOY are you ever wrong!!!!!!!!!!!! The first caps were called Leyden jars (Air gap capacitors). In the early 18th century, various electrical experimenters including Volta, and Benjamin Franklin used to charge one of these, or several connected in series with some of the early electric machines, (static generators), and then shock a group of people to their amusement and chagrin, as the first person on one side of the circle grabbed one pole and a person on the other side of the circle grabbed the other pole. The first evidence of cosmic rays was evidenced by the dischage of a leyden jar after many days, due to the cosmic rays providing a discharge pathe for the charge on the capacitor. True they were made of glass but the foil was open to the air. I'm wondering why you guys are equating the "imperfection" of a capacitor's dielectric with the capacitor's ability to stay charged. The more perfect the dielectric the better is the ability to hold a charge when disconnected from a circuit. Most dielectrics convey charge better than air does, so the plates can be placed closer together, giving the capacitor higher capacity, but given some large plates (a few sq feet of aluminum, and maybe a few glass beads to hold them appart, I'm sure you could build a capacitor of .1 mfd at say 600V, and it will hold a charge for several hours (cosmic rays and ionized particles in the air and moisture will discharge it eventually. I know because I did this experiment, about 30 years ago.

Shyam, I have found that a 600 volt .05 polyprop is about the same size as a 100V .33 . Is this what you are saying about capacity changing with voltage??. Because if you are saying that a 600 V .05 mfd changes capacity (increasing) at 100 volts you are just plain wrong!!!!! Unless perhaps you meant by .00001%. There are all kinds of electronic measurement instruments used to filter specific frequencies from a wide spectra over a huge voltage range (like micro volts to many volts) that are based on a specific capacitive reactance, and are accurate to one part in 100,000, and most of the inaccuracies are caused by other components.

Again, there are many "near" perfect caps, that behave within parts per million of theoretical values. Low voltage, thin ceramic disc caps are the measurably worst they are slightly microphonic, other than electrolytics and tantalums. (if you have a voltage gain of 10,000 sometimes (especially thin disc caps) may represent one part in that 10,000 of acoustic sensitivity. If this were not the case then they would cause big problems in electric guitar amps, ( they do cause small problems). Ceramic discs are temperature rated and often as little as 5% over -20 C to +80C. But there are temperature stabilized caps that are good for 1/2 % over the same range. There used to be some made by a Company called Stabilex. Again tubular polypropylene caps have excellent attributes, are extemely low in temp coeff drift and sensitivity to sound or vibration, and amazingly enough, electroman you can find high voltage versions that will hold a charge for days or weeks and knok off your socks!!!! but in capacitors that is a GOOD thing.

Other than Tantalums, and electret microphones, show me any non-polarized cap that will charge itself, "out of circuit" once the two leads have been connected together for a tenth of a second!!!!!!!

And in defense of disc caps, is there any electronic engineer out there with twenty active years designing circuits with caps that hasn't seen thousands of ceramic caps used to filter noise, for instance RFI in power supplies. They may be cheap and imperfect but they work pretty well. If they generated (much) noise all by themselves, then how come virtually every computer monitor power supply and computer switching power supply, indeed most consumer devices use ceramic caps to filter noise?

I humbly suggest that anyone interested in the characteristics of various caps, get a bunch of electro-junk which is ubiquitous in the US, perhaps not India and take apart a few things and actually "measure" some of the properties of different capacitors. If they are so great as noise generators try using a ceramic disc as a noise generator, connect it to the aux input of your stereo and see how much noise it generates, and since it is also equivalent to a condensor mic why not use it also as a microphone the next time someone comes over to sing Karaoke, or if it is a great temperature sensor use one in a circuit to acurately measure the outside temp, ( that is the only example that I can think has a posiibility of working!!

Forgive my defense of the humble Capacitor. I didn't think that a capacitor's ability to store a charge was something worthy of complaining about because that is what they are for.

Geoffrey Reed

Dear Geoffrey Reed

John is proposing to write a book and I expect that he will place some valuable information about capacitor world that is not so obvious to the electronics professionals unless they run into problem when they actually work with the device.

Q = C * V ................. (1)

E = 1/2 C * V * V ...... (2)

These are very fundamental equations for the ideal capacitor.

When we talk about memory, we also talk about the condition under which capacitor has been or stressed to. Capacitor dielectric is basically some solid or liquid or gas material. This material undergoes changes in time and due to electric field, frequency of the electric field, the current transient, and temperature. Perhaps there may be other things also such as external magnetic field of high strength, humidity, pressure and change in acceleration due to gravity. We will not talk about quantum capacitors here and let us remain in realistic world of large size board level capacitors only.

It has been noted by scientists that capacitors retain memory of stress under which they were subjected to. This is what we are talking about.

Geoffrey, I think we are getting our wires crossed here...

You are talking about a capacitor holding a charge, that is beyond dispute, of course they can and do.

What I and I suspect the others are referring to is the memory effect otherwise known as dielectric absorption.

This is the effect where a capacitor, of whatever dielectric, when its charge is changed the dielectric takes time to release the expected charge, so if the charge held on a capacitor is changed the dielectric will change the remaining charge held in the capacitor over a period of time...

A simple experiment with a good aluminium electrolytic shows this effect well, just charge a 100 microfarad capacitor up to 10 volts ,say, and then quickly discharge it and then watch the terminal voltage increase afterwards.

This is dielectric absorbtion or memory effect and is exhibited by any dielectric capacitor...

Its most noticable when using a capacitor as the integrating capacitor in a A/D convertor for a digital display, the roll over error is easily seen when connecting equal positive and negative voltages to the input and getting an error in the reading on the output due to the memory effect or dielectric absorbtion of the integrating capacitor....

If you don't believe me or you can't do a simple experiment just look at the data sheets for the 7109 series or 7135 series of A/D convertors, the manuafacturers go into detail about selection of the integrating capacitor to minimise this effect.

I think what you are confusing is the capacity of a capacitor or perhaps the ability of a capacitor to hold a charge which is nothing at all to do with memory / dielectric absorption effects.

John.

Geoffrey, i seriously think you have totally misread or understood what this memory effect is that we are talking about?!

There are so many different types of capacitors that its not very sensible to argue with you about what you term a 'perfect' capacitor to be...

But all capacitors will drift with temperature.

All capacitors exhibit memory or dielectric absorption effects.

All capacitors store charge.

All capacitors cause some noise. etc... etc...

As for your comments about the use of ceramic capacitors in filtering out noise, they do work extremely well, in low impedance applications such as supply decoupling, as you stated... However, you then state that if they generated much noise why are they used to filter noise?

Its simple! In low impedance applications the noise characteristics of a ceramic capacitor are minimised... remember the very high impedance of a piezo ceramic device?

Now try to use a ceramic capacitor in a sensitive, high impedance section of an audio amplifier etc... They will generate sufficient noise to make you see what we are talking about.

So whilst I'm not de-rating your comments, i am asking that you re-read the previous posts and perhaps read a few data sheets on these devices.

John.

Dear John

Now a days, we see that many ADCs use capacitive method rather than current, resistive array and voltage comparator method. How come they are made to 16-bit precision and at such a low charge and such a high speed?

Diode capacitance used for radio frequency sources is all known and have become very much standard. However, I am still not very much aware of the charge integrating capacitors that are used on the chips for very low leakage applications.

There were few Burr Brown chips that were meant for charge integration and also you can see them in CCD devices using MOS capacitors similar to memories. If those are good one then why we do not use them in discrete circuits?

I often see X7R preferred for power supply filter along with high and small value caps. One advantage I see is that they are polarity free.

I have tried using aluminum capacitors for AC signal isolation from DC and they work fine due to low ESR and very low impedance to AC. Only problem was when signal was near zero and had dual polarity. I am now biasing the amplifiers to one side DC and then separating the AC. This works fine and I do not have to lose fortune on high value NPO or X7R.

Another concern is high temperature capacitors. What kind of capacitors are good for 150C to 300C? Perhaps only air capacitor and Mica may work well. What else? Alumina ceramic capacitors? This area I want to look into such that my electronics then can be rated for MIL grade +.

Dear John

Problem with air dielectric is that most of them are open air dielectric and sensitive to moisture very badly. This makes capacitors to drift easily. Capacitors show different value if instrument is hot or cold, it is raining our side or there is summer hot wind. I use them as moisture sensors.

Other concern is the inductance of a capacitor. Generally large film capacitors are very slow for fast signals. Recent development of multiple parallel plate capacitors is good one as they reduce inductance and offer short multiple plates for greater capacitance. I often use many in parallel to reduce inductance effect.

ESR at high frequencies is a good parameter to know the capacitor's likely performance in HF circuits.

I am still not fully aware of method for least noise capacitors. I often use signals in the range of 10kHz to 10Mhz and noise in this zone need to be the lowest. I also will like to remove the harmonics generated in the amplifiers. Hence, capacitors for applications differ.

Another concern is high voltage isolation capacitors in the range of 10nF to 1uF and rated for 1kV to 10kV range and also small charge storage capacitors 0.5pF to 10pF range rated for 1kV to 10kV range. High value capacitors I use X7R and low value NPO. I think metal film and Mica may also be tried but not yet used as they are bulky and my circuits are miniature. I avoid Aluminum or any polarized capacitor as they are likely to be damaged due to polarity change by the user and they are also bulky. Only in power supplies I am using low ESR aluminum capacitors. I think Tantalum is Aluminum compound and has funny damage characteristics and makes the board dirty if gets damaged. They are small so often used in nuclear instruments. I am not using them as they are sure noisy and never used in audio amplifiers as they cause funny noise.

Shyam, I would suggest 100 volt polypropylene, tubular, 1% caps, with corrected temp coeffient if you need less than 1% change fom 0 - 100 degrees C operating temp. If you need a higher operating temp, then 400V precision glass, tubular ceramic, or presicion silver mica caps may be the way to go. Silver mica caps tend to be quite large ( esp. for a .1 at 100-200V). All of these can be found with temperature compensation. I'm not sure who makes these nowadays. Corning used to make precision glass caps. Sprague, Mallory, Cornell Doublier, Elemenco, Illinois Caps, Nippon? and Matsushita?(not sure on spelling or specific products). make polyprop, and used to make higher quality tubular ceramics. Tubular ceramics typically have more robust foils for high current, quick discharge, and are much less microphonic, as well as operating probably around 500 degrees F, which is (above?) the melting point of solder. .1 uf at 100V 10% polyprop should be around $.50 cents to $1 in single quantities. The ceramics may be similar. Glass and mica are more expensive. you will pay a lot more for 1% than 10%. I have no idea what the price would be for thousands or millions. Geoffrey

You are right and some of these capacitors are capable to perform better in terms of stability with voltage and temperature. I am procuring lots of them with different ratings and different types in different values. I may end up in having a big store house, but will like to use the right type at right place. I will also conduct research on their actual behaviour in the designs.

I think people also cheat and give different material than what is ordered. Some companies charge inhibiting cost but may be OK for few parts as their parts can be better trusted.

I am right now making a list and I am in touch with manufacturers and will purchase good number of them. Problem with manufacturers is that they do not want to take up the job for less than 10K/20K numbers for any type. I may end up buying few million numbers. I am worried about storage and usefulness over time. Any way, I will buy these now that have entered the list.