This is the long version of my adventures with a surplus 4815A. Hopefully it will be useful to anyone contemplating similar repairs, or provide entertainment to those familiar with the restoration of classic test equipment.
The instrument was acquired from a local surplus dealer who purchased it as excess inventory from a large company. Since its condition was unknown and the deal strictly "as is", the price was only $100. Cosmetically it was fair to good, with some small scratches and scrapes, but it had the probe, and the possibility of adding RF impedance measurement to my home lab was irresistible. One of my major interests is the characterization of passive components, such as bypass capacitors, filter networks, and piezoelectric ceramics.
The instrument worked reasonably well the first time it was powered up. Calibration wasn't terribly far off, but the oscillator stopped oscillating when the frequency was reduced to the lower portion of the top few frequency bands. Removing the covers gave the first clue that something was amiss, as a thick sticky oil was seeping out from the castings surrounding the oscillator.
This first power-up also confirmed what a fantastic instrument this would be if its troubles could be cured. In only a few minutes it taught me how much more I needed to learn about the imperfections of passive components at high frequencies. I had never realized how quickly current and voltage begin to shift phase, even in small value capacitors.
One look at the castings, dial drive, and general construction of this unit was enough to convince me I needed a manual. HP doesn't stock it anymore, but I was lucky enough to copy one owned by a retired HP service person. Disassembly is somewhat difficult, and not for the inexperienced. The right side panel must be removed and the front panel components have to be warped out of position to get at critical screws. Without the manual you would probably break something in the process of disassembly. Note- don't let the hinged casting at the rear of the instrument fall down; it's unsupported once the side is removed. There are a bunch of screws of slightly different lengths to keep track of. Most difficult is unsoldering the connections from the oscillator that connect to the band switch. One is hidden underneath the other, and it's almost impossible to get a soldering iron tip in there! There are also several other easier connections to be unsoldered. Finally, alignment and calibration involve many steps and quite a bit of specialized test equipment, though not beyond what most techs can get access to.
The main tuning capacitor is a conventional air variable that runs in an oil bath. The gasket on one end of the casting was in fine shape and appeared to be shiny neoprene. The other gasket was a softer, more porous rubber material, and had completely disintegrated in the central exposed area. I have no idea why it was different, but the leak never would have occurred if it had been the same material as the other end (see later notes on this under reassembly). The leak allowed oil to reach the end cap of the casting, and run out a vent hole that allows for slight expansion and contraction of the gasket. The oil and capacitor plates were loaded with black material from the gasket. The material may be conductive, causing the oscillator to fail, however HP engineering suggested another cause described later.
This instrument was no longer serviced by HP as of 1987, so the "not field serviceable" capacitor module had to be serviced. Once removed, the capacitor assembly looks somewhat like a carburetor, so it was cleaned by spraying with Gumout carburetor cleaner. This was very effective at removing the black goo. Other areas of the instrument were cleaned with alcohol and Q-Tips, or whatever would do the job. Not knowing what the oil was, I took extra precautions to avoid contact, and avoid contaminating tools and workspace. The remaining oil (100 ml.) was saved for further tests and possible filtering and re-use, though the volume was insufficient due to loss. 150 ml would probably be more than a full refill. A new neoprene gasket was also cut to replace the defective one.
Even the HP engineer who worked on these instruments ten years ago said the factory wouldn't tell him what the oil was. Apparently it was a closely guarded secret, since it is not listed on the internal HP parts list for the capacitor assembly either. Since I appeared to be on my own here, I would have to test the oil for dielectric constant, and do some research on what was currently available to replace it.
The oil has the consistency of honey, though it eventually flows into every possible space. It increases the value of the tuning capacitor, but its main function is to increase stability. It prevents pressure and humidity changes from affecting the tuning capacitor. Air changes density since it's compressible; oil doesn't! Very clever design work on someone's part back in 1966 or so when this thing was introduced. Of course, if they had just added a couple plates to the capacitor, and run it in air, you wouldn't be reading this now. Maximum performance is often at odds with simplicity and reliability.
For testing, a multi-plate trimmer capacitor was adjusted to 99.6 pF in air (as close as I could get to 100 pF). Dipped into the oil, its value slowly rose to a value of 242 pF, so the stuff has a dielectric constant of 2.43. Since the oscillator frequency was nearly correct at each end of the low band, I'll assume the oil has maintained its properties, and that 2.43 is a reasonable value.
I proceeded to measure as many oils as I could lay my hands on, and rapidly reached the conclusion that ordinary oils rarely get above 2.3. Silicones all seem to be higher than desired, some very much so. Nor are the two miscible, so there is little prospect of blending a suitable material.
Further research uncovered some disturbing information. The 4815A oil is miscible with ordinary mineral oil, thus it is not a silicone. The only oil I found reference to, described as thick and honey-like, with a dielectric constant of 2.4, is polychlorinated bi-phenol (PCB). Pure PCBs are said to have a slight mothball odor, and the 4815A oil does have a similar odor. PCBs were the transformer and capacitor oil of choice until they were banned in the '70s as a carcinogen. It is illegal to manufacture them, you can no longer buy them, and you don't really want them in your test equipment. They also have to be handled and disposed of in accordance with strict federal laws.
I might have scrapped the 4815A right then and there. With all the screw holes and rough surfaced castings, it would be impossible to remove all traces of oil from the unit. Given the media reports, the EPA regulations, and the millions of dollars spent on specialized clean-up facilities, PCBs must be incredibly deadly. Or are they?
If you have or want a 4815A, you should research the PCB topic and reach a decision you can be comfortable with. Current data suggests PCBs aren't nearly as bad as we've been led to believe. Basically, the mice didn't develop cancerous tumors, and long term studies of people who worked in the stuff daily for years showed no ill effects. The ban came about because the EPA doesn't differentiate between benign and cancerous tumors, and mice fed huge quantities of the oil developed benign liver tumors. PCBs do concentrate as they move up the food chain, so you don't want them in the natural environment. Check the GE web site for detailed information on all this; my forte is electronics, not politics, law, or chemistry.
Remember, I have not proven that the oil in the 4815A is or is not a PCB, but in light of my strong suspicions, I did additional clean up work, disassembling more panels and hardware to be sure there were no reservoirs of oil trapped in hard to reach places. I intended to have the oil analyzed, but it wasn't necessary as I found additional information described at the end of this write-up.
Finally, if you have a 4815A, it is bound to leak sooner or later. If you deal with the capacitor before it leaks, the instrument will stay uncontaminated.
GE suggested one silicone that might be a suitable replacement. Their SF1147 silicone fluid has a dielectric constant of 2.3. At first I considered raising the constant slightly with another silicone fluid, but then realized that we're only talking a few percent here. My other concern was that the fluid might have high losses at 100 MHz, or cause some other misbehavior of the oscillator. It isn't nearly as viscous as the original oil, but I couldn't see where that would cause a problem. Lacking good alternatives, the 4815A was powered up on its side, and the silicone poured in. Success! A quick check of the oscillator showed the frequency reasonably close at the end points. In addition, it worked well over the entire range, including the HF areas where it had failed to oscillate previously. A thorough check would come later. (If you can't get the silicone, the highest DC oil I found was Permatex Hydraulic Jack Oil at about 2.3- haven't tried it, but it's a place to start.)
Since the capacitor had been filled with the end cap off, replacing that was the next order of business. I noticed that the cap rocked on the gasket, so it was unlikely to seal on the fairly stiff Neoprene. The surface of the end cap was probably belt sanded smooth, but never machined truly flat.
Speculation- HP probably had a sealing problem as a result of the unmachined end cap and the fact that it uses fewer screws than the opposite end. Using a softer gasket was an easy and inexpensive solution. There was probably no clue to indicate the material would dissolve in the oil over a twenty year period.
I set the end cap up in a mill and skimmed off a few thousandths until the surface was completely flat. It now seated on the gasket perfectly and was screwed into place. (You could accomplish the same thing with sandpaper, a flat surface, and a lot of patience.)
With the right side of the instrument raised just slightly, the pipe plug in the capacitor casting was removed to top off the new silicone oil. Just before replacing the pipe plug, the right hand gasket was pushed in very slightly via the vent hole. Don't use anything sharp! This raises the oil level even with the top of the casting. The pipe plug is then installed. The result should be no air bubble and no internal pressure on the gasket as the pipe plug forces it back to the neutral position.
The rest of the reassembly was just the reverse of disassembly. It was tricky to get apart, and tricky to get together! Despite my best efforts, I still mixed up some of the screw lengths on all the little Philips screws holding the castings and cover plates together.
The good news is that the unit works at all frequencies and on all bands. In hindsight, I should have doctored the silicone oil a bit to raise the dielectric constant from 2.32 to 2.43. A few percent of GE Dielektrol VI from a modern oil filled cap might be a good place to start. Though the calibration adjustments get the scales very close, they are not perfect. Probably just a couple percent outside specification. It may be possible to do better by rotating the dial drum slightly, but I haven't tried that. If I ever have the unit apart again, I'll do an oil change with modified oil. That's a much easier task than the complete capacitor tear down. It isn't much of an issue, as I always keep a frequency counter tied to the RF output. One more note- do not unscrew the trimmers all the way or they will fall down into the casting. You will have to take everything apart again! I didn't do this, but came very close.
Russ Caroli of HP engineering, (973-586-5342) says he saw many units that failed to oscillate just as I described. His field fix from his schematics follows:
Locate the base of Q1 on the tiny board attached to the tuning capacitor casting. It ties to C1, a 0.1 µF capacitor to ground. Lift one end of the cap and place a 30 Ω resistor in series with it. Change the ground point of the cap to the upper right hand PCB mounting screw with a small lug. This fix never made it into a service note.
I still believe my unit failed to oscillate due to all the gasket gunk coating the capacitor plates. I didn't make this modification, and the oscillator seems to work fine for now. Possibly failure to oscillate is an early indication of gasket breakdown or contamination of some sort.
Russ is in engineering now, but it sounds like he started out doing service or support. He remembered Manny Pires, the local HP service engineer, now retired, who supplied me with the manual for the unit.
This unit was probably built in the mid '70s judging from the serial number prefix. It may have been one of the last, as the series was also discontinued in the mid to late '70s. It is very clean, uses excellent components, and the probe appears good. The probe is the weak point of the 4815A, as they are almost impossible to field repair, and are no longer serviced by HP. Note that the current (when this was written) HP 4194A Impedance Gain Phase Analyzer has an impedance probe option (HP 41941A or B) that appears very similar, but the probe alone is $2245!
I'm extremely grateful to Hewlett Packard Corp. for going beyond the call of duty to help with this restoration. My personal notes were organized into this write-up at the request of Theo Klewansky of HP (theodore_klewansky@hp.com). I've rebuilt quite a lot of classic test equipment and am always happy to discuss it with fellow enthusiasts.
August 1998
Conrad R. Hoffman
George Sandford, the former HP4815A Technical/Production Supervisor at Hewlett-Packard's NJ Manufacturing facility, had a business servicing the 4815a. I talked to him on the phone and he told me what the oil was. It is not a PCB, but is Polybutene. He has since ceased servicing the units due to a general lack of customers and a lack of surplus units to work with. He's no longer supplying the oil.