Resolving Amplifier Arcing

A vacuum tube high power RF amplifier is full of high voltages. The large physical size of the tubes requires high voltage between cathode and anode to function. The combination of high output impedance and high power results in a high RF voltage (by Ohm's Law). Prevention of high voltage arcs requires careful design, component selection and operation. Those arcs are potentially destructive: to the tube, power supply and your ease of mind.

Arcing is more common in older amplifiers due to chemical degradation of insulators, insulator cracks and component warping from thermal and electronic stress, and the oxidation and burning of the contacts in physical switches. Although arcing is rare in solid state amplifiers since they operate at relatively low voltages that is no panacea since they suffer from other ills. But this article is about tube amplifiers.

Not long after putting my recently acquired vintage Drake L7 to work it suffered from intermittent arcing at the high end of its power range. Because the high DC plate voltage is independent of power level the arcing was almost certainly at RF. Although the arcing has been eliminated I expect it to return because the faulty part requires replacement. It's on my very long to-do list and it'll take a while until I get around to a implementing a long term solution.

If you've never worked on a high power tube amplifier my relating of how I investigated and resolved the arcing may be of interest. Despite the large size and lethal operating voltages and currents tube amplifiers are actually quite uncomplicated. With attention to safety and the peculiarities of dealing with high voltage, current and power they are not too difficult to work on.

Locating the arcs

The case of the L7 has no opening except on the bottom and rear for forced air cooling. These are useless as windows into the amplifier's interior. The case must be removed. This is the first problem in any properly designed amplifier. There are interlocks to prevent accidental electrocution and burns to budding technicians who are out of their depth.

The Drake L7 has two interlocks: one shorts the 3000 VDC plate supply and other (well hidden) prevents the amplifier from being turned on. Both interlocks must be disabled to operate the amplifier without the case. Before you try this, or even consider doing so, you must educate yourself about what you're attempting to do. Better still, have a knowledgable friend help you out. If you get yourself killed don't say I didn't warn you.

Yes, that's really a brick! I needed something flat, heavy and non-conducting to safely disable the high voltage interlock. The mains interlock (not visible) on the bottom is disabled with a chunk of plastic wedged under it. In the picture the amp is on and idling with a plastic (non-conducting) LED desk lamp for added illumination. Arcs are so bright that no ordinary lighting will wash them out.

I mentioned my suspicion about the loading capacitor in an earlier article. I dutifully straightened the multitude of plates until they maintained a decent gap for their full rotation. Unfortunately that repair resulted in no improvement. Hence the deep dive into the amp's innards.

As I increased drive the amp arced as expected but not where I expected. The loading capacitor sat quietly when the fireworks began.

It was the plate capacitor that was arcing. Although the spacing was generally good (and much wider than the loading capacitor) a number of rotor plates didn't track well. I straightened them without needing to remove the capacitor from the amp. There is an unrelated problem that will require its eventual removal for repair but that can wait.

Straightening the plates did not fix the problem. Worse, the arc location was seemingly random. Each one occurred in a different location. The arcs were evidently due to a fault elsewhere that caused an excess voltage condition beyond the rating of the capacitor. The capacitor is working just fine.

This is interesting so let's take a detour to review the design of a tube amplifier's output network.

Amplifier pi-network

The simplified schematic below is that of a typical pi-network found in many tube amplifiers. It transforms the high impedance of the tube output to the low impedance of the antenna system. The operator adjusts the plate capacitor to resonate the plate circuit and the loading capacitor for a high efficiency match to the impedance presented by the antenna system.

The resonance condition is a typical feature of an impedance transformation network, as previously covered in this blog. The circuit also attenuates harmonics since it is a low pass filter.

The blocking capacitor keeps DC out of the antenna circuit and the choke keeps the RF out of the power supply. Band switching (not shown) alters the range of the variable capacitors and the inductor value. The T/R switch (input side not show) bypasses the amplifier during receive.

RF voltage is determined by the power and impedance in accord with Ohm's Law: E = SQRT(PZ). From the data sheet for a pair of 3-500Z tubes a little arithmetic suggests there is approximately 3000 volts across the plate capacitor at 1000 watts RF output. The recommended capacitor rating is 4.5 kV because the impedance, and therefore voltage, can be higher depending on operating parameters.

On the high bands the L7 places a fixed capacitor in series with the variable capacitor to reduce the capacitance. This also lowers the voltage across each capacitor since capacitors in series act as a voltage divider. Because of this arcing incidence is greater on the low bands. However it is not eliminated. Something more dire is going on to cause arcing when the RF voltage is low.

The voltage across the loading capacitor is easier to calculate since the antenna system impedance is nominally 50 Ω. At 1000 watts the potential is a less than 300 volts. The voltage will often be higher when the SWR is greater than 1, which is very common for most hams. If the SWR is too high the loading capacitor with its smaller plate spacing can arc.

Sequencing

Since the capacitors appear to be in good shape and the output impedance is well within the acceptable range the problem must originate elsewhere. Most likely is a fault that affects the impedance at the output port. The antennas, external switches and transmission lines were ruled out by additional testing. High power can aggravate weak components and loose connections in an antenna system to create intermittent and permanent impedance changes.

By this process of elimination I focussed my attention on the T/R relay. It is common for slow or faulty amplifier relays to cause plate capacitor arcing when transmitter power appears at the amplifier input before the output relay contacts have settled. Unsettled contacts cause a momentary high impedance (open condition) at the output port. Once an arc starts it can continue after the relay stabilizes since the path to ground is always lower impedance than the antenna system.

The solution is sequencing to ensure the amplifier relays settle before power is applied to the input port. This can be coordinated with the transmitter, signal source (e.g. PTT in advance of transmit) and even with the amplifier itself by having the output port relay close faster than the input port relay.

The open frame relay in the L7 is typical of many vintage amplifiers. It iss slow at best, and with age the contacts are suspect. Replacements can be found but better solutions are available.

While arcing at turn on was occurring its incidence was less than that of arcing during a transmission. That is sufficient evidence to rule out sequencing as the cause of my problem. It does not mean sequencing isn't a concern, just that it isn't responsible for the observed behaviour.

Relay woes

I had good reason to suspect the relay. I earlier had to clean the contacts on the input and output of the bypass side due to intermittent signal attenuation on receive. With an ohmmeter connected to the centre pins of the input and output SO239 jacks you clean the contacts until you reliably read 0 Ω. Since the relay arms can shift laterally it is important to test for this by manipulating the relay arms.

The bad contact can be isolated by connecting one ohmmeter probe to the bypass bridge seen on the right. The output port is at the bottom of the picture and the input port is at the top. The centre arm applies tube cutoff bias during receive. Just my luck that both sides of the relay were corroded.

I had to resort to aggressive cleaning with an abrasive when a deoxidizing contact cleaner and non-abrasive buffing were insufficient. Do this only when absolutely necessary since abrasives can easily damage the thin contact coating, assuming there is any left (usually silver) after several decades of use. This is discussed in more detail by W8JI. I used a thin strip of 3000x sandpaper to be as gentle as possible. It worked.

Checking the through amplifier relay contacts is more difficult since an ohmmeter cannot be easily employed to measure resistance. Deoxidizing cleaner and buffing didn't resolve the arcing problem, but it did seem to reduce its frequency. Having gone that far I resorted once more to the sandpaper. After thoroughly clearing the contacts of debris I did another test. This time the arcing vanished entirely.

Permanent solution

I don't know how long the repair will last. The relay needs to be replaced. In any case it is slow and loud. Not only is that very annoying it is cause for ongoing worry. I don't want to take the risk of it failing during a contest.

Designs and even kits to replace T/R relays in vintage amplifiers are available. Some are fast enough to enable QSK operation. I don't need QSK just a solution fast, reliable and quiet.

Another problem with the existing T/R switching is the long lead lengths along the bypass path. On 6 meters it is enough (almost 0.03λ) to significantly raise the SWR. With 6 meter season in full swing I manually bypass the amplifier when I am not using it. Any T/R relay replacement will need to address this issue.

Vintage amplifiers are cost effective assuming you have the time and motivation to repair aging equipment and add modern features. My next amplifier will likely be a new purchase, one with a stiff power supply, silent operation and 6 meters. That way I can tolerate the quirks and faults of its older cousin at the secondary operating position.

80 Meter Stinger Version 2.0 (and Pipe Fitting)

You might think that because I write a blog about antennas and station building that I always do things right. I only wish! Perhaps I make fewer mistakes than some but I have my share of them. The original stinger for the driven element of my 80 meter array is one.

I was economical with the aluminum tubes and pipes I had on hand, saving the longer pieces for other projects. The stinger was 21' long, with 19' (6 m) projecting above the tower top. That was topped by over 3' (1 m) of 1" PVC pipe to get more height for the parasitic wire element support ropes. The total amount of aluminum and PVC above the tower was 7 meters long. That's a lot.

Although the stinger is guyed by the catenaries for the wire elements the tension produces a downward force (compression). It was obvious during setup that a butt joint between 1-½" x 0.095" aluminum tubes was not up to the stress. My hope was that it would last long enough that I could focus on other projects until I would have to replace it. All seemed well for a year as it survived one wind storm after another. Then the joint suffered a fatigue failure and the top half of the stinger fell down.

Stinger version 2.0

The stinger needs to be robust but not necessarily lightweight. The stinger must resist modest horizontal tension of the parasitic element catenaries. It must also be up to the compression force due to those same catenaries, especially with regard to high stress points where the yield point could be exceeded in a high wind and icing, or from tension imbalance among the 4 catenaries. It is laterally stabilized by those same catenaries which act as guys.

Being lightweight is beneficial when installing the stinger since it is long and must be lifted overhead to be dropped into the tower top. It can be assembled in pieces and raised from below at the price of more time and effort. As you will see I did a bit of both.

Rather than a butt joint between two lengths of 1-½" aluminum tubes the main improvement is a butt joint between two lengths of 1-½" aluminum pipe (1.9" OD). This allows me to use the existing 2" saddle clamps that secure the stinger to the tower plates. I also have a supply of these surplus pipes on hand and I know where I can get a few more at a good price.

Pipe fitting

My introduction to fitting pipes together for antenna construction was simply out of expediency: I had sources of cheap surplus aluminum pipe and tubes are expensive. I later came across the same idea in W6NL's book Physical Antenna Design (now out of print), which increased my confidence.

Although aluminum pipe is almost always 6061-T6 -- excellent tensile strength -- they have a seam, hard as it can be to find one on these aluminum pipes. They are theoretically weaker than seamless tubes although I have yet to see a seam failure. Indeed, some of my stock comes from commercial antennas that have survived harsh Canadian winters at great heights.

Aluminum pipe follows the same size schedules as steel pipe in the US and Canada. That's very convenient. Plastic pipe that is now in common use for water pipe and conduit -- ABS and PVC -- are similarly sized. I'll provide several examples of how pipes and tubes can be mated with respect to my 80 meter stinger, and additional ideas covered in earlier articles. There are other combinations of pipes and tubes that can work well together.

On the left are two schedule 40 pipes: a 1-½" pipe inside a 2" pipe. The inner pipe OD is 1.9" and the outer pipe ID is 2.067". The gap is 0.167", or 0.083" all around. Depending on the application the pipe can be simply bolted together. For improved rigidity a shim made of aluminum flashing can be used. An alternative is to increase the amount of overlap to reduce wobble, at the expense of greater weight.

The middle example is almost the same except that the outer pipe is schedule 80 with an ID of 1.939". That is a much better fit. Both are options to butt join two lengths of 1-½" pipe. I considered using the schedule 80 pipe until I found what is, to me, a more favourable solution. I prefer to save the schedule 80 pipes for my various yagi projects.

On the right is the 1" schedule 40 PVC pipe that is fit to a 1-½" aluminum tube at the top of the stinger as a non-conductive extension for the catenary attachments. The OD is 1.315" and the ID is 1.049". The 1" pine dowel purchased at a hardware store provides structural strength and fits well enough for the intended use.

This size PVC pipe will also fit well over 1" tubes. I may use PVC pipe as an insulator on the driven element 1" centre segments if I decide to use a beta match. Insulation is not needed for a gamma or T match.

Here are a few more pipe fitting ideas. When I used a 2-½" aluminum pipe to mate with 3" tubes for my first set of long yagi booms I had a machine shop turn down the pipe a few hundredths of an inch. The mast on my 150' tower is a 2-½" (2.875" OD, 0.25" wall) steel pipe slipped inside a 3" (3.068" ID, schedule 40) steel pipe used as the drive shaft for chain driven prop pitch motor. There the fit is so poor that I needed shims to prevent slippage caused by the large mechanical load. For one long yagi boom I fit 2" OD heavy wall tubes into both ends of a 2" schedule 40 pipe, which is a good fit. I bought several lengths of heavy wall 2" tubes to mate with these pipes to make booms for the 20 meter and 15 meter long boom yagis I am building.

There are other applications of pipes that I will discuss in future articles about those projects. Consult charts of pipe sizes and trawl through surplus yards for cost effective solutions in your antenna farm.

Stinger butt splice

In my junk box are Hy-Gain yagi parts that have been collected over the years. Hy-Gain booms are mostly 2" OD and spliced at the centre for the longer yagis. I have two of these surplus brackets. I tested a bracket on 1.9" OD pipes and was successful in achieving a secure fit despite the smaller size.

I butt spliced 10' and 7' pipes in the bracket. Holes were drilled through the pipes to make use of the bracket holes for that purpose in lieu of using the inner perimeter holes intended for a mast clamp.

I briefly experimented with plastic pipe as insulators to electrically isolate the pipes. I would need do this to attach a switchable coil to add 160 meters to the 80 meter array. I slit a scrap length of 2" white PVC pipe to slide over the pipe. It had to be slit since the ID is slightly less than 1.9".

At the end of the dressed pipe is a round insulator made of pressure treated lumber, cut with a hole drill, to provide mechanically robust isolation between the upper and lower pipes. The lower pipe does not need an insulating sleeve except perhaps to achieve a consistent diameter within the bracket.

I put the idea aside as not quite ready for implementation. It can be retrofit later. I first need to ensure high voltages between the pipes when the coil is active (not shorted) cannot jump the gap through the slit while maintaining high mechanical strength. It may be as simple as a wrap of thick polyethylene sheet and a seal to keep water out.

Upper stinger

Spliced to the upper pipe is a 7' (2 m) length of 1-½" x 0.095" aluminum tube. This is the only piece of the original stinger used in version 2. The 0.11" gap (the pipe ID is 1.61") is filled with a wrap of aluminum flashing coated in conductive grease. Stainless screws with nylocs hold it together. The 1" PVC pipe (1.315" OD) with its inner wood dowel are attached to the 1-½" tube (1.41" ID) in a similar fashion.

The original holes for attaching the rope catenaries are reused, and holes drilled through the wood dowel inside. The raw pine is protected by a cap of pressure treated wood, the top of which is sealed with caulk. The hose clamp adds tensile strength to the PVC and wood to better withstand the tension on the catenaries. The tension isn't high but I want to ensure years of trouble-free service.

The final stinger is ~2' (60 cm) longer than the original. This is intentional. I found that with my parasitic T-element design there is some slack on the vertical wire. The longer stinger removes the slack, and is easier than rebuilding the wire elements.

Installation

The new stinger was raised in two steps. The lower 10' pipe with Hy-Gain bracket went up first and dropped into the tower clamps until the bracket rested on the top clamp. The nice thing about aluminum pipe is high strength-to-weight ratio. The 10' of pipe with bracket attached is only ~9 lb (4 kg). It is easy to hold it vertical over my head as I fuss with inserting it through the tower clamps.

The catenary ropes are tied to the top of the stinger and detached from the wire elements in preparation for the next step. I briefly considered leaving the elements attached until I realized that the lateral tugs of those small weights would prevent safe lifting of the 17' long upper stinger.

With everything in place I lifted the upper stinger and dropped it into the Hy-Gain bracket. Once that was secured the elements were reattached. The complete stinger was then pushed up through the tower top and clamped in place.

Back on the ground I tensioned all the catenaries. The test for tension equalization was to have the stinger straight and in line with the tower. That was after the adjacent picture was taken.

Although a simple procedure it is tedious. I took an overnight break at one stage when I felt that I was too tired to do the lift with complete safety. A brief delay is preferable to an unnecessary risk.

Matching network changes

The new stinger's mechanical length is 2' longer and its electrical length is 3' longer. The reason is explained below. But having done so the new electrical properties of the driven element must be dealt with. For a λ/4 monopole on 80 meters the approximate rate-of-change (dF/dL) is10 kHz/6 cm; that's ~150 kHz lower due to the 3' extension.

There is no reason to make the driven element resonant at any particular frequency. The low impedance still requires a matching network, both as an omni-directional vertical and in yagi mode. I am using a switchable L-network.

My next task is to remeasure the antenna's impedance across the band, in both omni-directional and yagi modes. I will then use TLW to determine what changes are required. They should be small. I'll describe the details in my final article about this antenna, which will be written once it is complete and fully operational.

Catenary rope length changes

Changing the stinger height presents an interesting geometry problem: to lengthen the catenary ropes in a manner that keeps the parasitic elements vertical and preserves yagi performance. My first inclination was to ignore the problem since the change is quite minor. However it isn't difficult to check this on paper so I took that precaution.

Since the two sides of the right angle triangle are approximately equal each increment in height lengthens the full catenary (including the T-top of the wire element) by 0.7 increments; that is, 1.4' for a 2' height increase. However we only need to lengthen A to B, the distance from the top to the vertical wire, since we can freely add rope at the bottom. With A and B only 40% of the total length -- 10.5/25.5 -- we need only 40% of 1.4' or 6" (15 cm).

By not lengthening the top rope the wire element will lean towards the driven element ~4" (10 cm). As confirmed by modelling (and as you'd likely guess) this is negligible. Since I had no slack to lengthen the upper rope section I built and installed long insulators between the rope and upper end of the T shaped wire element. These are made from PVC pipe. The height of the vertical component of each wire element was increased by 0.8' (25 cm) -- 40% of 2' -- which took up all of the slack. Mission accomplished.

Returning to work

With the stinger rebuilt, better and stronger than before, work can resume on the yagi. All the parasitic switch boxes are installed and working. Tuning of the elements is partially done. Then comes the final step: the main switching system at the base of the driven element (tower).

Progress on the antenna has slowed due to more urgent projects, especially the 20 and 15 meter stacked yagis. Not to mention 6 meter DXing and otherwise simply enjoying the warm weather. With my 80 meter interest being DXing and contesting the 80 meter yagi is not urgently needed. It can wait until late summer, but may be completed earlier depending on circumstances.

In consideration of the weather and my busy schedule don't be surprised by a slowed rate of articles through the summer.

Potential for FT4 on 6 Meters

The E_s (sporadic E) season is well underway. It peaks at the solstice, which is less than 3 weeks from now. Although the season is reasonably long it is less so for DXing, which requires multiple clouds and long path lengths. When it does occur the openings are usually fleeting.

Openings for any one signal are often too short for most stations to complete an FT8 QSO. Bigger stations -- power and antennas -- or those in excellent locations do better. There are in principle two factors to be overcome in achieving DX success on 6 meters E_s:

• Weak signals: The combination of multiple bounces and forward scatter keep DX signals in almost all cases quite weak, even for those with big antennas.
• Brief opportunity: For the same reasons each signal can be in and out in less than one minute, the minimum duration for an FT8 QSO.

In the first case the challenge can be overcome with more power and more antenna. Unfortunately that is not practical for most hams, especially when it comes to towers and and antennas. But if you can do it you will see a tremendous improvement in your results.

In the second case you can increase the window of opportunity with more power, a bigger antenna or both. However, for the majority the greater opportunity is with modes that allow for quicker QSOs. That way you can exploit the propagation peaks and not just their long heads and tails.

The diagram is copied from the article linked above since I believe it makes clear the QSO duration challenge. See the article for a description of the diagram.

Traditional modes are fast, but...

Although CW and SSB are faster than FT8 (and even the twice-as-fast FT4) you are rarely in the right place at the right time. I discussed this "discovery" problem in an article last summer.

It is a major reason why FT8 has been so successful on 6 meters that it reduced CW and SSB activity by at least 80% in just one E_s season. You can't argue with success. Well you can but arguing won't put DX QSOs in the log. It's the only reason I made the move to FT8. Trying CW in a recent opening saw me through only 3 QSOs, including one Caribbean station I have not heard on FT8. SSB activity was a little better.

TEP, tropospheric ducting and (we hope) F₂ benefit less from the speed of FT4. FT8 is fine but then so are CW and SSB for these propagation modes. When signals persist the comparison between digital and traditional modes is little different than for HF. To give you my take on this I will merely state that I do not use FT8 on HF.

Expectations

I am intrigued by FT4. As of its latest incarnation the time slots are half that of FT8 (7.5 seconds), promising QSOs in half the time. Indeed it may be even faster since there is the possibility of fewer message repeats due to QSB during a lengthy FT8 QSO.

Despite not having used FT4 and the few reports I've heard from those using it I am comfortable predicting that FT4 most likely will be a good fit for 6 meters. In particular to E_s DX openings.

• Loss of a few db of sensitivity is more than compensated for by speed
• Ability to squeeze more QSOs out of brief DX openings
• Favours the small station better than FT8; while this may seem surprising consider the ability to better exploit brief propagation peaks

I have not yet used FT4. I am content to wait for it to stabilize. This is scheduled for July. That leaves me with enough time to play with it at the tail end of this E_s season. Others are using the beta software on 6 meters at its proposed slot of 50.318 MHz.

But more than a choice of mode my main concern right now is propagation. DX opportunities have been slim lately on 6 meters despite a promising beginning in mid-May. DX has been worked though nothing new or of especial note.

I listen most days with WSJT-X monitoring 50.313 MHz while I go about other activities. Between 6 meters and chasing 3D2CR there is enough DXing to fill the gaps in my busy warm weather schedule.