An adjustable, DC, current-limited, breakdown voltage tester [BVT] -- also known as a "hi-pot" -- can be used for nondestructively measuring the voltage withstanding ability of relays, capacitors, transistors and silicon rectifiers. A BVT can be used to test amplifier-tubes for the presence of gas, and oxide-cathode type [8877, et cetera] amplifier-tubes for the presence of sputtered gold. A BVT can help you identify a tube that could seriously damage your amplifier if you inadvertently plugged it in.
For most amateur radio applications, a maximum voltage capability of 15kV is adequate. For testing commercial components, 40kV or more is usually needed.
Troubleshooting an RF power amplifier without a BVT is difficult and problematic. Building an RF power amplifier without a BVT is risky.

Circuit

An adjustable transformer, T1, delivers approximately 0-132vac to the primary of the HV transformer, T2. A 120v, incandescent lamp is used to limit the primary current in T2. A full-wave voltage-doubler rectifier with capacitor filters is used to produce the DC. R8 protects the device under test [DUT] from excessive current during testing. The typical current during a test is 2 to 5 uA.
A voltage-quadrupler could be used instead of the voltage-doubler. Quadruplers have one advantage and two trade-offs: Only half as much secondary voltage is needed from T2. Four filter capacitors are required instead of two. If a voltage-quadrupler is shorted, two of the four filter capacitors discharge through two of the rectifiers and the transformer secondary winding. This can damage rectifiers and/or the transformer. If a quadrupler is used, it is appropriate to use long leakage path current limiting resistors in series with all of the capacitors and a robust MOV across the transformer primary.
T2: A 50 - 60 Hz HV transformer from an old junked copy-machine or oscilloscope might work acceptably. January, 1998 Parts Sources

Construction Notes - Due to the high voltage levels, the ideal material for constructing a BVT is not metal. Polyurethane varnished plywood, Lexan® [polycarbonate], acrylic, or fiberglass are satisfactory choices. A metal cabinet with a non-conductive front-panel and chassis can also be used.
It is not good engineering practice to pack the components in a BVT too tightly. In dry air, at sea-level, 1kV will jump about 1mm. In moist air and at elevations greater than sea level, less voltage is needed to jump the same distance.
To minimize corona current-leakage into the air, solder connections that are at high-potential should be rounded "blobs." Avoid using sharp points on conductors. Corona-effect is easy to check for: In a darkened room, turn the voltage up to maximum and watch for blue lights accompanied by a hissing sound and the odor of ozone. [O3]. If needed, apply more solder and/or smooth trouble spots with a file. Mounting components too close together can also cause corona problems.
Current Limiter/Detector--R8/M2 is at high potential--so these components should not be placed close to M1, which is at ground/common potential. M2 and R8 should be mounted on an insulated panel with no low potential conductors nearby. The current limiter/detector is wired to banana-jacks so that it can be connected to the positive or to the negative output terminal as needed.
D1 and D2 can be made from a series of individual 1kV diodes or a series of multi-diode HV low-current rectifier units [a.k.a. focus rectifiers]. So-called "equalizing-resistors" and "equalizing-capacitors" should NOT be connected across modern rectifier diodes. Use diodes with similar ratings and they will equalize themselves.

Testing Components

The basic technique behind non-destructive voltage testing is: slowly increase the applied voltage across the DUT until a small current flows through the device. Record the V and stop the test. If the voltage were allowed to increase beyond this critical point, the component could be damaged. The applied voltage usually does not need to remain on the DUT for any longer than it takes to read the voltmeter, M1, and the current-meter, M2. For testing devices with lower breakdown voltages, only half of the doubler should be used. One end of the DUT is connected to Common instead of to one of the HV OUT terminals. Thus, the voltage readings on M2 must be divided in half. If the breakdown voltage of the DUT is expected to be less than 1000v, you may want to connect a DMM from the output side of R8 to Common so that the voltage can be read more accurately. Be careful. Most DMMs can not safely tolerate more than 1000v without using a voltage multiplier probe.

Vacuum-capacitors and vacuum-relays should be tested for gas before installation. I have seen new, unused vacuum-devices that were defective due to very slow air leaks. Such leaks typically show up several years after manufacture. In use, a gassy glass-envelope vacuum-relay can often be identified because the ionized air inside emits a blue light around the open contacts. However, the ionization in a defective glass-envelope vacuum-capacitor is usually deep inside the concentric meshed plates and can not be seen. Variable vacuum-capacitors should be tested with the plates fully meshed. When a vacuum-capacitor goes bad in an amplifier, reduced PEP in one result. However, other things can cause the same problem. When troubleshooting an amplifier, it is a good idea to routinely test all of the vacuum capacitors and vacuum relays with a BVT.
Vacuum-capacitors that have been in storage for a long time may develop "whiskers"--microscopic filaments of copper. Sure, it sounds weird. This anomaly causes the breakdown-voltage to initially be lower than normal. It is possible to burn-off these whiskers. During the process, the capacitor may self discharge--as indicated by a "tink" sound. Repeatedly forcing the capacitor to self discharge will result in a decrease in breakdown voltage. Five tinks is usually a good point to stop.
Semiconductor testing is important when building HV power supplies that contain groups of series-connected rectifier diodes. If one or more of the rectifiers fails during use, it could start a "domino-effect" and short-out the other rectifiers in that group. Shorted rectifiers deliver AC to the filter capacitors. This is not a serious problem for non-polarized capacitors. However, with polarized electrolytic capacitors, even a small episode of reverse-current can damage the capacitor--or even cause it to explode. Thus, one defective 10¢ diode can trigger the destruction of many dollars worth of good parts. It is better to cull-out bad parts prior to construction.

Silicon-rectifiers: Each diode should be tested individually where possible. Increasing reverse voltage is applied to the DUT until a leakage-current of approximately 2uA is detected. [high-current diodes can withstand more reverse current] At this point it is important to observe the current meter. If the leakage-current randomly fluctuates without adjusting the voltage, the diode has a manufacturing defect and it should be discarded. Since it is difficult to mark each diode, I usually sort diodes into labeled drawers according to PIV. Individual diodes which test greater than about 1.3kV should be viewed with suspicion because this usually indicates a doping problem. A forward voltage-drop test at the rated current can be used to discover whether such a diode has a problem. At 1A, the forward voltage drop in a silicon PN rectifier junction should be less than 0.9V.

Transistors are now made with voltage capabilities that are similar to silicon-rectifiers. Some transistors are rated at 1500V. Testing these devices is similar to testing silicon-rectifiers--except that a resistor of roughly 100 Ohm should be connected from the base to the emitter, or from the gate to the source.

Air-variable capacitors: Identifying too-closely spaced points that need realignment is easy with a voltage breakdown-tester.

Gridded power tubes need a good vacuum in order to function properly. A vacuum test is made with no filament voltage applied. HV is applied between the anode and a grid. A healthy 3 - 500Z will typically exhibit less than 10uA of current-leakage at double the rated anode-voltage.
The BVT can also be used to check the alignment of the filament in a 3-500Z. When its filament is cold, a healthy 3-500Z can withstand 7 - 8 kV between its cathode and filament. If the filament is not concentric with the grid, the breakdown voltage will be lower. This problem is usually brought about by intermittent VHF parasitic oscillations--a condition that generates a large pulse of cathode and grid currents. The principle is simple: a flow of electrons is always accompanied by a magnetic force. The larger the current, the stronger the force. During an intermittent VHF parasitic oscillation, the magnetic force is sometimes strong enough to bow the hot, tungsten filament wire helices toward the grid. If the [cold] filament to grid withstanding voltage of a 3 - 500Z is less than 6kV, a grid to filament short may occur when the tube is hot.

Testing for parasitic damage in 8874s, 8877s, 3CX800A7s and other oxide-cathode type tubes:
Such tubes have the following things in common: indirectly heated strontium-oxide/barium-oxide cathode, high gain, ultra high frequency capability, and gold-plated grid. The oxide-coating is an efficient electron-emitter. The gold plating helps to reduce primary electron-emission from the grid. This improves performance. There is a tradeoff. If the gold evaporates [sputters], the loose gold particles can cause serious problems.
In a vacuum, gold does not begin to evaporate unless it is heated to more than 1000ºC [1832ºF]. Heating the entire mass of the grid to >1000ºC requires more energy than is available. However, if there were a way to heat the gold plating, without heating the entire grid, gold evaporation would be possible. VHF/UHF energy has a substantial "leg up" when it comes to heating metal plating. VHF/UHF current travels exclusively on the surface. During an intermittent VHF parasitic-oscillation, the VHF grid-current can become so large that the surface of the gold plating briefly becomes hot enough to evaporate gold. The resulting gold vapour cloud can then move about freely inside the envelope. As the gold cools, it solidifies into tiny balls. In a low power microscope, they look like dew drops of water on the petal of a flower. Some of the evaporated gold lands on the emissive coating. Gold poisons the cathode's electron-emitting ability--causing a reduction in anode-current and power output. [A copy of a letter from Eimac describing this phenomenon is available on request to this author.] Evaporated gold can also land on the inside of the ceramic anode-insulator. This can cause arcs between the anode and the adjacent (grounded) grid-ring. An arc is most likely during the crest in the anode-voltage swing when the tube is not conducting--as the HF tank-circuit/flywheel swings to its positive voltage peak. At this instant, the peak anode-voltage is normally about double the positive supply voltage. If the insulating ability of the ceramic has been compromised by the presence of gold, trouble is probable. When the anode arcs to the grounded grid, the HV-positive circuit is virtually grounded. Thus, the HV-negative circuit tries to rise above ground to the voltage in the HV filter capacitors. This typically causes damage to components in the HV negative circuit. A common problem is an arc between the cathode and the filament or between the cathode and the grid. Test for compromised voltage withstanding ability between the filament and the cathode--and a burned out filament.
The Loose Gold Test: There is a simple test that can confirm the existence of loose gold particles without sawing open the suspect amplifier-tube. The only piece of equipment needed is a BVT. The principle behind the test: Like charges repel and unlike charges attract.
Procedure: Remove the amplifier-tube from the amplifier. The positive and negative DC-voltages that are applied between the anode and the grid should be two to three times the operational anode-voltage.
Loose gold particles can be moved around by changing the polarity of the anode-voltage. If the anode is positive, the gold particles are attracted toward the anode-insulator. This causes the indicated leakage current to increase. If the anode is made negative, the gold particles are repelled and the leakage current will decrease. If the leakage current is equal with either polarity, the presence of gas is indicated.
Another method of confirming the presence of loose gold particles is: apply positive anode-voltage and record the leakage current. Shut down the BVT and, with the tube vertical and the anode-cooler up, repeatedly bang the anode-cooler, both vertically and horizontally with a approximately 2oz. soft face hammer. This will cause some of the loose gold particles to fall to the bottom where they will be less attracted by the positive voltage at the anode. Keep the tube vertical. Re-apply positive anode-voltage. If the leakage current decreases, you have loose gold and you are making progress. If the leakage current does not decrease, either you need to bang harder or no more improvement is possible. This procedure has been used by some amplifier owners to get more operating hours out of gold sputtered tubes. Banging also causes the gold to fall off the cathode coating. This increases electron-emission. However, if the tube is turned upside down, the loose gold becomes redistributed and the banging process must be repeated to move the errant gold to a safer place.

CIRCUIT ALTERATIONS:
There are many things that can be altered to accommodate available parts--or to change the voltage capability of the tester. Ohm's Law is the most important guide.
The wattage of the current-limiting lamp should be approximately twice the wattage or VA rating of the HV transformer.
It is important to note that resistors have a dissipation-rating and a voltage-rating. The voltage-rating always takes precedent over the sometimes-phony dissipation-rating. For example: A 5.1M Ohm, 2W, ±5% resistor has a dissipation-rating that could lead a designer to assume that the resistor could safely dissipate 2W. Using Ohm's Law, E = [PR] 0.5 = [2W x 5.1M] 0.5 = 3193V at 2W. However, there's a catch. The resistor is limited to 500V maximum by the technical specifications. This means that the "2W" max. resistor is really {P = E2 / R, = 500v x 500v / 5.1M Ohm =} a 0.049W max. resistor! It pays to read the specifications. High voltage resistors have a long, spiral, resistive, deposited film that is designed to handle voltage.
- R8: The actual value of R8 is not critical. If you can not locate a single resistor for R8, there is a commonly available spiral-film resistor, that is marketed by ECG-Phillips, which has the right properties. These resistors have a tolerance of ±2% and a dissipation rating of 2W. They can be series wired on a perfboard to build a satisfactory R8. One can be used for R9--whose function is to isolate the charged capacitance [to ground] of M2 from the DUT.
- C1 AND C2: It might seem necessary to use capacitors with a Working Voltage [WV] rating that is equal or greater than the actual V that will be encountered. However, voltage testing components is an intermittent application that produces virtually zero ripple-current. WV is a continuous duty rating--usually at an elevated temperature and for a substantial amount of ripple current.
For BVT intermittent service at normal room temperatures, a WV-rated capacitor can usually be used at 1.5 times its rating with no difficulty. Sprague Vitamin-Q capacitors are very conservatively rated. They can usually be used at double their WV rating in BVT service.
If you are going to use tubular glass-case capacitors with metal screw end-connections, it is important to put a stop-nut between the capacitor and the mounting hole. The stop-nut prevents the screw [which is soft-soldered to the end-cap] from being pulled out of the capacitor when the outside fastening nut is tightened. Four nuts are required to properly mount each capacitor. It is not appropriate to mount this type of capacitor by only one end. Metal-cased capacitors should be mounted without metal clamps on an insulated surface. Use silicone rubber adhesive for mounting.

OPERATING NOTES:
Reasonable care should be used with a BVT. Although the steady-current capability of this BVT is quite limited--even without R8--a substantial [as in rudely awakening] current pulse can be delivered by a charged capacitor.
When testing with voltages above about 8kV, it is possible to have a false current indication caused by corona from sharp points. To check for this problem, disconnect one end of the DUT, increase the voltage and observe the current reading on M2. . . Rich, AG6K (805)-386-3734.