RF Capacitors: NP0, Silver Mica and X7R for Radio Use

A capacitor looks like the simplest part in the junk box: two plates, a dielectric between them, a value printed on the case. At audio and DC that picture is nearly true. At radio frequencies it is dangerously incomplete. The same 100 nanofarad capacitor that smooths a power line perfectly may behave as an inductor at 20 MHz, and a ceramic disc that tunes a filter on the bench may drift by ten percent when the room warms up. The right choice among RF capacitors decides whether a low-pass filter attenuates its harmonics, whether a VFO stays on frequency and whether the capacitor in a tuner survives a long FT8 transmission. This guide explains the types you will meet, the few parameters that matter, and the choices that work in a ham station.

What a Real Capacitor Is Made Of

An ideal capacitor has only capacitance. A real one is a small circuit: the capacitance in series with a tiny inductance, called the equivalent series inductance or ESL, which comes from the leads and the internal plates, and with a small resistance, the equivalent series resistance or ESR, which represents the losses in the dielectric and the conductors. Together they set the behaviour at high frequency. A parallel leakage resistance also exists, but at RF it is usually negligible.

The series inductance creates a series resonance at which the capacitor has its lowest impedance. Below it the part behaves as a capacitor, above it as an inductor. The self-resonant frequency is f = 1 / (2π √(L C)). For a 100 nanofarad capacitor with 10 nanohenries of total inductance, including the leads, this is about 5 MHz. A 1 nanofarad part with 5 nanohenries resonates near 71 MHz, and a 10 picofarad part with the same inductance near 710 MHz. Shorter leads and surface-mount packages push the resonance higher, which is why the best bypassing on a VHF circuit is done with small chip capacitors soldered right at the pin.

Dielectrics: The Heart of the Matter

The material between the plates determines stability, loss and size. Class 1 ceramics, labelled NP0 or C0G, have a low dielectric constant but extremely stable behaviour: the capacitance changes by only about 30 parts per million per degree, hardly changes with voltage and does not age. Their loss is very low, so the Q is high. They are the correct choice in tuned circuits, filters, oscillators and matching networks. Their weakness is that high values are large and costly, and above about 10 nanofarad they are rare.

Class 2 ceramics, such as X7R, X5R and Y5V, have a high dielectric constant, which allows a large capacitance in a small package. The price is that the capacitance changes with temperature by 15 percent or more for X7R and by much more for Y5V, drops with the applied DC voltage, and ages with time. They can also be microphonic, which means that vibration produces a voltage. These parts are excellent for decoupling and DC blocking, where the exact value is irrelevant, and unsuitable for anything that is tuned. The overview of ceramic capacitors lists the codes in detail.

Silver mica capacitors have a layer of silver deposited on mica sheets. They are very stable, low in loss, accurate to one or two percent and tolerant of high RF current, which is why they are the traditional choice for transmitter tank circuits, filters and the frequency-determining parts of oscillators. They are expensive and appear in values from a few picofarads to a few nanofarads. Film capacitors made of polystyrene have low loss up to a few tens of megahertz but cannot tolerate temperatures over 70 degrees, and polypropylene is somewhat better. Polyester and similar films are intended for audio and have too much loss and inductance for RF work. Electrolytic and tantalum capacitors are used for power supply filtering only and are useless at RF because of their high ESR and inductance.

Decoupling and Bypass: Parallel Combinations

A very common task is to bypass a supply line or a control wire so that RF does not travel along it. A capacitor with a low impedance at the frequency of interest short-circuits the RF to ground. Because every capacitor has its own self-resonance, a single value covers only a part of the spectrum. The usual practice is to place two or three capacitors in parallel, for example 100 nanofarad and 1 nanofarad, or 10 nanofarad and 100 picofarad, so that each covers a different part of the range. A small ceramic part near the load deals with the highest frequencies and a larger one with the lower ones.

There is a pitfall: between the self-resonance of the large capacitor and that of the small one, the large part looks inductive and the small part capacitive, and the two form a parallel resonance with a high impedance, called an anti-resonance. A small resistor or a ferrite bead in series damps this peak. The beads and chokes you might use are explained in the article on the ferrite mix. For cables and power wiring, the bypass capacitor works together with the common mode choke, as described in common mode chokes.

RF Voltage and Current: The Rating Nobody Reads

The voltage rating printed on a capacitor is a DC rating, but an RF circuit also applies an AC voltage and passes an AC current. In a transmitter filter these can be large. Take 100 watts into 50 ohms: the RMS voltage is the square root of 100 × 50, which is 70.7 volts, with a peak value of 100 volts. A capacitor in a low-pass filter whose reactance is 50 ohms passes an RMS current of 70.7 / 50 = 1.41 amperes. At 3.5 MHz that capacitor is about 909 picofarads. The power lost in its ESR is the current squared times the ESR, so even a small ESR of 0.05 ohm gives 0.1 watt, which is not a problem. A cheap capacitor with an ESR of 1 ohm would dissipate 2 watts, enough to make it hot and drift.

When the SWR is high, voltages rise. A capacitor in an antenna tuner may see several times the nominal voltage, and the current in a coil-capacitor circuit with a high Q can be much greater than the line current. Select capacitors with at least double the expected peak voltage, and prefer types with a stated RF current rating. Silver mica parts for transmitting are specified in this way, as are the high-voltage NP0 chip capacitors used in power amplifier filters. For the maths of power, voltage and current, see the article on Ohm’s law for ham radio.

Capacitors in Filters and Matching Networks

In the low-pass filters of the transmit path, capacitors set the cut-off together with the coils, and their tolerance and stability are directly visible in the attenuation of the harmonics. The guide to the low-pass filter shows the values. Use NP0 or silver mica parts and measure them. Parts of 5 percent tolerance are usually acceptable, and combining two parts in parallel, for example 680 picofarads and 220 picofarads to get 900 picofarads, is a standard way to reach an unusual value. A NanoVNA can verify the result, as described in the guide to L network matching.

In oscillators and VFOs, temperature compensation is a classic technique. A capacitor with a negative temperature coefficient is combined with the main NP0 parts to cancel the drift of the coil, so that the frequency remains constant as the box warms up. Class 2 ceramics must never be used in the frequency-determining circuit, because their drift is large and they also change value with the applied voltage.

Variable Capacitors and Trimmers

Air variable capacitors have the lowest loss and the highest voltage capability, but they are large and increasingly hard to find. Their plate spacing sets the breakdown voltage, with 1 millimetre good for about 3,000 volts in dry air. Ceramic or plastic film trimmers are small and convenient for alignment, with a Q of 100 to 500 in the good ones. Piston trimmers, with a glass or quartz dielectric, offer the best stability and are used in precision circuits. All trimmers are sensitive to the way they are adjusted: use a plastic or ceramic tool, because a metal screwdriver changes the value when it is near the trimmer.

DC Blocking and Coupling

Another everyday job is to pass RF while blocking DC, as in the bias circuit of a PIN diode switch or the output of an amplifier stage. The value is chosen so that its reactance at the lowest frequency of use is small compared with the circuit impedance, usually a tenth of it or less. In a 50 ohm circuit on 160 metres, at 1.8 MHz, a capacitor of 10 nanofarad has a reactance of about 9 ohms, which is acceptable, while 100 nanofarad gives 0.9 ohm, and the part also has to carry the current. A larger value is not always better, since the self-resonance falls with the value, and a 100 nanofarad part will be inductive on 10 metres. Many designers choose 10 or 100 nanofarad for HF coupling and add a small parallel capacitor of 1 nanofarad to keep the higher bands in order.

Coupling capacitors that sit in an audio path are a different case. There the noise, the leakage and the microphonic effect of class 2 ceramics can be audible, and a film capacitor is the quieter choice. The same part can therefore be right in the RF section and wrong in the audio section of the same radio, which is why the choice depends on the job and not on the capacitance alone.

Selecting and Testing in Practice

For a practical rule: use NP0 or silver mica anywhere that a value matters or that the part carries real RF power, use X7R for decoupling and blocking of DC, use film or electrolytics for the power supply and avoid mixing the roles. Check the markings: the capacitance code, the tolerance letter and the voltage. Buy from reliable sources, because a counterfeit part is a real risk with cheap bulk lots. Keep leads short, keep chip components flat against the board and ground them with several vias. Measure suspect capacitors with an LC meter and, for the Q, with an instrument that tests at the working frequency.

Finally, remember that a capacitor can fail in a way that is hard to see. Ceramic parts crack under mechanical stress and then fail short or intermittently. A part that has overheated may show a brown discolouration, and a mica capacitor may lose its silver. If a filter that used to work begins to leak harmonics, or a tuner begins to arc, the capacitors are the first parts to check. Replace them with the same type, never with a cheaper class.

The Bottom Line

RF capacitors have a series inductance and resistance that set a self-resonant frequency, above which the part acts as an inductor. Use NP0/C0G ceramic or silver mica where stability, low loss or RF current matter, as in filters, tank circuits, tuners and oscillators. Use X7R for decoupling and DC blocking, where the value is not critical, and keep electrolytics for power supplies. Combine values in parallel to cover a wide range and watch for anti-resonance, check the RF voltage and current against the ratings, keep the leads short, and replace failed parts with the same type.

Frequently Asked Questions

What is the best capacitor type for RF?

NP0/C0G ceramic and silver mica are the best for tuned circuits, filters and power use because of their stability and low loss. X7R is acceptable for decoupling.

Why can a capacitor act like an inductor?

Every capacitor has a small series inductance. Above its self-resonant frequency the inductance dominates and the part no longer works as a capacitor.

Can I use X7R capacitors in a low-pass filter?

It is not advisable. Their value changes with temperature and voltage, so the cut-off and the harmonic attenuation will not be reliable. Use NP0 or silver mica.

Why put two capacitors in parallel for bypassing?

Each value has its own self-resonance. A larger and a smaller capacitor together cover a wider range of frequencies, but a small damping resistor or a bead helps avoid a parallel resonance between them.

What does ESR mean?

Equivalent series resistance represents the losses in the capacitor. It turns part of the RF current into heat, so low ESR matters in power circuits.

Are electrolytic capacitors usable at RF?

No. They have high ESR and inductance and are meant for power supply filtering at low frequencies. Add small ceramic capacitors in parallel for RF bypassing.