# PC high voltage RF capacitors

Double sided printed circuit board (PCB) is often recommended in ham circles as an inexpensive and effective way to fabricate a high voltage RF capacitor.

For example:

• (G3NGD 2011) If High Voltage Capacitors cannot be obtained, it is possible to make HV Capacitors using 'Double Sided Printed Circuit Board'. One could make the coil using Coaxial Cable, the Cable providing the Capacitance; and
• (Rule nd) recommends PC capacitors for his traps.

One of the problems in selecting dielectric materials is the dearth of reliable information.

# WX7G's example

(eHam 2012) contains comments by several posters, and includes a recommendation by WX7G who works up a quantitative example. His example is a piece of 1/16" FR-4 PCB of 8 sq in at 14MHz for a capacitance of 114pF and a capacitive reactance of 100Ω. He throws some numbers around and comes to a loss of 16W at 1000V impressed on the capacitor.

Using WX7G's assumed D factor of 0.02, the equivalent series resistance (ESR) of the capacitor is D*Xl=2Ω. If we assume that his 1000V is RMS (it is not qualified, and most readers seem to have taken it as RMS in context), then the current in the capacitor (Z=2-j100Ω) will be approximately 10A, and the power dissipated in the capacitor will be I^2R=10^2*2=200W... a long way from the 16W he calculated.

In WX7G's sketchy posting, it appears that he is calculating using a more complicated method depending on the definition that Q=1/D=2*π*EnergyStored/EnergyDissipatedInOneCycle. In that case, the EnergyStored in a capacitor is C*V^2/2, and so 114e-12*(1000*2^0.5)^2/2=114e-6J. The EnergyDissipatedInOneCycle=2*π*114e-6*0.02=14.33e-6J, and Power will be the EnergyDissipated in 1s, or 14.33e-6*14E6=200W.

So, this 8 sq in of board (eg 2"x4") is dissipating 200W with 1000V RMS impressed upon it at 14MHz, and in still air, it is likely to quickly reach excessive temperature and sustain damage.

Some other posters with apparently better experience cautioned that WX7G's enthusiasm might have been wrongly placed.

# End fed matching network

The article Matching a 5/8λ vertical - example 1 explores a matching network solution for a 5/8λ vertical. A solution not offered in the article is a simple L network, with a shunt capacitance on the load side and series inductance on the source side. Using the D factor for FR4 and assuming Ql=100, a solution to match the feed point Z=120-j445Ω to 50Ω is C=15.8pF and L=3.17µH. At 1000W of RF input, the dissipation in the capacitor (24.4mm square, less than an inch by an inch) is calculated to be 44W. Such a capacitor will not be capable of dissipating that power in free air, or worse, in a protected enclosure. Sure the capacitor will work, but the system is limited in power, probably less than 100W continuous power rating, and that is why the solution was not offered in the article.

Even at low power, more than 4% of the transmitter power is lost in the capacitor alone whereas a quality capacitor would have a loss of less than one tenth of that.

A 160mm length of RG213 as an o/c stub would be a better capacitance solution in this instance, having a quarter of the loss and much larger surface area to allow dissipation of the heat.

# Traps

(G3NGD 2011) and (Rule nd) recommend PCB capacitors for traps based on their ability to withstand high voltage. Whilst they might withstand high voltage without flashover, the preceding examples show that dissipation from these capacitors may be excessive at even modest RF voltages and that will often serve as the practical limit to application.

Note that excessive temperature increase might not be observed on 'low duty cycle modes', eg SSB telephony... but an extended tuning session might destroy the capacitor.

# Other dielectrics

One of the problems in selecting dielectric materials is the dearth of reliable information. Nevertheless, if a sample is available, a test capacitor can be fabricated and measured.

FR-4 is not alone in being a commonly used electrical insulator having a poor D factor, PVC is another.

The author's experience is that PVC is variable, D may be frequency dependent, but performance seems to depend on the sample and it is suspected that pigments, fillers, plasticisers and alloys with other polymers may affect performance. Commonly quoted D factors for rigid PVC vary from 0.02- to 0.04+, so it would appear to be worse than FR-4. Nevertheless, it features in RF capacitors such as in (Salas nd).

Contrasting FR-4's D factor of 0.02, PVC is worse,  Delrin four times better at 0.005, and PTFE is seventy times better at 0.00028,.

# Conclusions

• PCB capacitors using FR-4 may have high voltage withstand, but the lossy dielectric and attendant dissipation may severely limit the applied RF voltage.
• The fundamental problem is that the dissipation factor D for FR4 material is fairly poor when compared to the dielectric used for quality RF capacitors.
• FR4 is not the only material available for PCBs, and much better materials are used at microwave frequencies, but better materials are expensive.
• The RF performance of capacitor dielectrics should be evaluated effectively, some traditional insulating materials are lossy dielectrics.