Capacitors... AC Applications

The following list contains examples of common generalizations of DC and AC ratings for capacitors:

1. As long as the peak value of the RMS voltage does not exceed the DC voltage rating, the capacitor is acceptable for the AC circuit. This is only partially true. There are other factors that must be considered and will be covered later.
2. If a capacitor is failing in an AC circuit the failure can be eliminated by using a higher DC voltage rating. This is not necessarily true. A higher DC voltage unit may result in a decrease of failures, but will not eliminate them if the cause of the failures remains uncorrected.
3. Any AC rated unit can be used in an equivalent DC circuit. This is generally true but the reverse is not!

AC Versus DC

The application of a DC voltage to a capacitor results in an electrical force field (voltage gradient) stress on the dielectrics. This stress imposes a plus (+) and minus (-) electrical relationship on the dielectric molecules which in turn must then assume a dipole characteristic, and line up directionally with the force field. (See figure 1).

During the initial transient (short time) period of charge, electrons move out into the dielectric and initiate this dielectric polarization. As long as the force field remains on the capacitor, these polarizing electrons are bound to the dielectric molecules.

Figure 1 (a) illustrates a dielectric with the molecules in their natural non-polar orientation, i.e. with no force field applied. (It is interesting to note that various dielectrics will have varying degrees of natural polar orientation).

Figure 1 (b) illustrates the dielectric polarization effect that takes place when a voltage is applied. Removal of the force field frees the bound electrons and allows the dielectric molecules to reassume their natural orientation state.

Once in a fully charged condition, and unless subjected to an internal breakdown of the dielectric or an external change or variation, the capacitor sees no other stresses except the DC voltage gradient across the dielectric and the normal environmental conditions.

The application of an AC voltage to a capacitor results in not only the voltage gradient stress but additional stresses that must be considered. Two of these added stresses are extremely critical as early failure mechanisms. They are corona internal heating.

Corona

Corona is the term used to describe an electrical discharge mechanism occurring as the result of the process of continuous ionization of a gas. Ionization is the phenomenon wherein a normally non-conducting gas, consisting of essentially neutral atoms or molecules, is changed into a conducting medium consisting of positive and negative ions. This change results from some external source supplying sufficient energy to strip some electrons from atoms or molecules of the gas. The atom or molecule that has lost an electron becomes a positive ion.

These free electrons, being in constant motion, can then collide with other atoms (molecules) or positive ions. Depending on the energy of the collision, the electron will then create other free electrons. There is a possibility it will recombine to create negative ions by attachment to a neutral molecule, or rebound and remain a free electron.

The external sources that can supply the energy required to initiate this ionization process can be quite varied and in many cases is a combination of several of them.

1. The application of a voltage force field. This supplies external electrons for bombardment of the molecules. The intensity of the force field controls the point at which ionization begins by controlling the energy level of the externally introduced electrons.
2. Ultra-violet light, x-rays, alpha rays from radium, etc. can be used to ionize a gas. It must be noted that the energy levels of these sources must be high enough to initiate the process.
3. Externally applied temperature (heat energy) acts to increase the velocity of the molecules of the gas. When sufficiently high, the energy of impact between the molecules will strip some electrons thus initiating the ionization process.

As the ionization process continues, positive and negative ions form into separated space clouds. At some critical point in time these space clouds will recombine locally, thus releasing their excess energy in the form of a minute local arc discharge. When the formulation and recombination of these space clouds becomes practically instantaneous and continuous, still on a local basis, the phenomenon of corona discharge results.

The energy released during corona discharge manifests itself as heat, light, and electro-magnetic waves. The heat is a result of a very high-density concentration of current flow; the light appears as a purplish-blue haze; and the electro-magnetic waves can cause interference (noise) over a wide frequency band in nearby electronic equipment.

In a capacitor, the presence of corona discharges, which take place in the tiny air film or pockets adjacent to the dielectric surface, cause rapid deterioration of the dielectric due to the hot spot temperatures resulting from the heat concentrations during the discharges. Ultimate failure of the capacitor then follows in a relatively short time.

Corona is not just an AC phenomenon. It is a factor that must be considered in DC applications, but the relative voltage levels producing DC and AC corona are considerably different. Other factors in DC applications generally control design parameters such that DC corona does not become of critical concern except in a few special cases. In AC applications, because of frequency and the relatively low voltage levels at which AC corona initiates, designs must always factor in corona considerations.

Internal Heating

The application of an AC voltage to a capacitor, unlike a DC voltage, results in continuous heat generation within the capacitor. The total heat generated is from two distinctly different sources.

1. Dielectric heating is the result of the energy required to first polarize the dielectric in one direction and then repolarize the dielectric in the other direction for each succeeding half-cycle of the AC voltage applied.
2. Resistance heating is due to the continuous current flow in the series resistance elements of the capacitor. This current flow is the result of charging in one direction and then discharging, and charging in the other direction during each cycle of AC voltage applied.

Dielectric Heating

This is a natural phenomenon wherein the amount of heat generated varies with the inherent polarization orientation of the dielectric material, the magnitude and frequency of the applied voltage, and the geometrical character of the voltage wave-shape.

Resistance Heating

There are three (3) major elements comprising the series resistance and thus contributing to the total resistance heat generated:

1. The resistance of the metal in the leads, electrodes, solder, and end-metal spray.
2. The resistance of the oxides in the interface connections between the leads, solder, electrodes, and metal spray (for metallized units).
3. The effective resistance of the dielectric material.

Figure 2 is a block and schematic representation of a metallized dielectric capacitor, depicting the series resistance elements. (Distorted for illustration purposes.)

The basic formula for calculating the heat generated due to the series resistance is:

It should be noted that any action or circumstance that tends to increase Rs will tend to increase the heat generated.

The factors controlling Rs and the circumstances controlling or affecting these factors are:

DF (dissipation factor) for these capacitors is the ratio of the equivalent series resistance to the capacitive reactance. Therefore it follows that:

Equation 1

For a given current, frequency, and capacitance value, the DF figure can be used as a direct criteria for measuring the comparative heat generating capabilities of different capacitors under an AC application condition.

The critical nature of the current amplitude is apparent from the formula above. An increase in the current becomes a squared increase in heat generated.

The major portion of the total heat generated at low frequencies comes from the dielectric material itself. As the frequency increases, the heat generated by resistance heating becomes greater. Variables such as the purity and condition of the dielectric, type of electrodes, end connections, and other external factors determine the specific frequency at which the variables contribute a major portion of the total heat generated.

Fundamentally, prior to the generation of heat energy, the capacitor is in a state of temperature equilibrium with the surrounding environment. The appearance of heat energy in the capacitor starts an immediate increase in the internal temperature of the capacitor at the point of generation.

Since heat flows from any hot point to a cooler point, a temperature gradient is rapidly established in all directions from this point of origin. During the process, some heat energy will be dissipated by conduction through the lead connections, some energy is used in the process of establishing the temperature gradient, while the remainder of the energy is conducted to the outside surfaces of the capacitor. Here, the energy continues its process of establishing a temperature gradient with the surrounding environment by radiation. If there are any objects in contact with the outside surfaces, the conduction process will also continue at these points of contact.

With time, and pre-supposing that a thermal runaway condition (i.e.: the case where equilibrium is not achieved and failure results from excessive heat generation) does not prevail, a condition of equilibrium is reached wherein the heat dissipation is equal to the heat generated and a stable temperature gradient exists. The extent or level of this steady-state temperature gradient will depend on the speed and ability of the capacitor and its environment to dissipate the heat energy as it is generated.

The interior of the capacitor will be operating at some temperature higher than the surrounding ambient temperature with a commensurate loss of service life. There will be a change in parametric behavior characteristics associated with this higher temperature. If the change in parametric characteristics (i.e. capacitance, variation, dissipation factor value, and lower insulation resistance) associated with this higher temperature are such that circuit malfunction results, then either a redesign of the circuit application must take place, or some means provided to increase the speed of heat dissipation to sufficiently lower the temperature gradient.

In a given application, an optimum capacitor may have been selected in order to minimize the amount of heat energy generated. If the circuit design cannot be further altered, and an excessive temperature rise still exists, reduction of the temperature must then be accomplished by some external means. This can be done by increasing the speed of heat dissipation in three basic ways.

1. Air cooling by using air blower fans to increase the rate of convection cooling.
2. Liquid cooling by providing for a flow of liquid in or adjacent to the capacitor and conducting the subsequent absorbed heat away.
3. Solid cooling by providing an auxiliary conduction channel to conduct the heat away from the part. This is commonly called a heat-sink.

The use of evaporation, refrigeration, etc., are in reality special variations of the three basic methods.

Many capacitor manufacturers are blamed for making a poor capacitor when an uninformed or careless end-user installs a 1 Mfd. 200 VDC capacitor in a circuit that applies 115 VAC 400 Hz across the capacitor, and experiences a failure. The unit may have been perfectly suitable for a 200 VDC application, but was not designed for an AC application because of its inherent dissipation factor characteristic and method of end terminations.

Case History - The "Borderline" Unit

Ten apparently "identical" (same style, rating, design, built at the same time from the same lot, and all passing the required tests) DC rated capacitors were applied in the same AC circuit under identical operating conditions.

At the end of 8,800 hours of operation, the status of the units was:

1 unit failed (open) at 17 hours.

1 unit failed (short) at 628 hours.

8 units still operating.

Analysis of the failed components was performed and the following conclusions were drawn:

1. The "open" unit failure mode was the destruction of the interfacial juncture between the metal spray and electrode material on one end of the capacitor. The area around the failure exhibited a discoloration due to excessive temperatures at the end termination site.
2. The "shorted" unit failure mode was a dielectric failure. Upon further investigation it was determined the failure was due to multiple punctures at one end of the section roll. This same end exhibited massive charring and discoloration due to excessive heat.
3. Both units showed many clearings around the failure area indicating that the self-healing characteristic of the metallized dielectric was operative prior to the time of failure.

It was apparent that both failures were a result of the same cause, a high resistance point on the end of the capacitor section creating excessive heat generation. The generated heat gradually raised the temperature internally which lowered the voltage breakdown strength of the dielectric until the clearings began. These localized "clearings" contributed additional heat and finally, with time, a thermal runaway situation developed leading to the eventual failure.

In the case of the "open" unit, the connection between the metal spray and the metal electrode material became totally disconnected before massive dielectric failure could occur. For the "shorted" unit, the dielectric failure occurred first.

This was not a case of using known non-conforming units in the application - both of the failed units tested the same as the eight units that are still in operation. Standard tests of capacitance, dissipation factor, dielectric voltages, and insulation resistance failed to isolate the two units from the rest of the lot. Referring to equation 1, the frequency (f), capacitance (C), and current (I) will have the same effect for all ten capacitors if any variations occur. The single factor in the formula that is controlled by manufacturing and workmanship is the total series resistance (RS). The excessive heat generation in the two failed units is directly traceable to an increase in the RS value with time.

Two units out of ten have RS parameters that have normal values at the time of factory tests but increase with time under AC operation. These units are designated as "borderline" units for AC operation.

The problem then is to find an effective tool, or method of identifying these "borderline" units prior to installation and field operation.

Considering the of ten "identical" units, it should be noted that the probability is extremely low that RM or RD (see figure 2) would react differently in two of the units under AC operation. There is practically nothing in the manufacturing procedures to cause variations between units for these factors that wouldn't be apparent during final factory tests. RO, however, is almost totally dependent on the manufacturing procedures and workmanship standards. This, then, is the variable that is primarily responsible for variations in RS between "identical" units from the same lot.

The fact that the oxide films comprising the interfacial junction points (RO) are highly responsive to frequency gives us a tool or method for locating these "borderline" units. Consider the following illustration covering 12 units manufactured to meet a DC specification where the DF requirement is a 1.0% max. at 25°C and 1000 Hz.

Notes:

1. All units passed the DC requirement for DF.
2. Units 1 and 2 are certain failures on AC. These are not even borderline units since they show abnormally high DF values (for AC operation) even at the standard 1 KHz factory measurement frequency.
3. Units 3 through 12 all show normal DF values at 1 KHz, however:
   1. Units 3, 4, 6, 7, 9, and 11 also show the normal increase in DF at 10 KHz for this type of unit.
   2. Units 5 and 8, because of their highly abnormal increase in DF at 10 KHz are, "borderline" units and probable failures in AC operation.
   3. Units 10 and 12 show a slightly abnormal increase in DF at 10 KHz and are listed as "borderline" units and possible failures in AC operation.
4. Regardless of the numerical values involved, the final criteria for failure in any AC operation is the AC circuit itself. Obviously, the stresses imposed by the AC application - relative to frequency, voltage magnitude, current amplitude, and environmental conditions - determine the final DF figure that makes a unit "borderline" or good.

The method for identifying "borderline" units for AC operation is really a screening operation using the higher frequency DF measurement as a tool. This does not "eliminate" AC failures - but by discarding those units from a lot that show an abnormal increase in DF value at 10 KHz, you can sharply reduce the number and probability of failures in the AC application.

A final word of caution - it is important that all measurements for the DF values be accomplished on the same instrument. This reduces the problem to one of variation in the DF parameter rather than absolute values of DF. Each type of DF bridge or comparator, indeed many individual instruments within a type, will exhibit wide variations in absolute DF values.