Adventures in Airplane Trading: Part 2

When I traded my Hatz CB-1 biplane for a homebuilt Rose Parrakeet I was aware that there were issues with the fuel system. The seller and previous owners had told me that the engine would starve for fuel in a steep climb with the tank less than half full. All of the previous owners had just lived with the issue, as lowering the nose and/or reducing power would fix it. Despite the issue, the airplane had amassed over 300 hours and made a number of long cross-countries, but it still wasn’t right. As somebody who’s well off-center on the “engineering spectrum” (if that isn’t a recognized mental disorder it probably should be), I knew I would have to try to fix it.

Furthermore, on my third flight in the plane after trucking it home and reassembling it, the engine sagged and partially lost power even in level flight at cruise power. I didn’t know if this was due to the new engine or carburetor—the previous owner had recently replaced the engine with a used C-85 out of a Cub, while the original engine was an A-75—but it made fixing it more important.

There was also some fuel leaking from the carburetor when not running. This is not unusual with Stromberg carburetors, but I wanted to fix it if I could. In many cases, the leak is caused by a poorly fitting float needle and seat. Early Stromberg carburetors had a steel needle, which had to be lapped for a good fit. A neoprene rubber-tipped needle was introduced in 1943, which sealed well and required no lapping, but there were issues with the rubber tip swelling and getting stuck, either due to poor quality control during manufacture or different fuel formulations. In 1963 a Delrin (plastic) needle was introduced, which (depending on who you talk to) either worked well or didn’t. Many owners just shut down the engine with the main fuel valve so there would be no fuel in the carburetor bowl to leak out.

I decided the first order of business was to disassemble and check the carburetor, just to rule out any problems there. This turned out to be worthwhile, as not only was the fuel level in the float bowl too high (which could explain the leakage, if the fuel level was above the idle bleed holes in the venturi), but the float was also dented, making it float too low. This would not only make the fuel level in the bowl too high but also limit the amount the needle could open, restricting fuel flow and further exacerbating the fuel flow problem. By temporarily soldering a piece of brass rod to the float, I was able to pull out the dent and get the level set within specs.

After reassembling the carb and putting it back on the plane, ground running showed that with 3 gallons in the tank (one-quarter full), the engine would start and idle just fine (once I readjusted the idle mixture to compensate for the lower float level). Partial power was also OK, but it would still starve for fuel after about 30 seconds at full throttle. On the plus side, correcting the float level seemed to have fixed the dribbling problem when not running.

At this point an unrelated issue arose. While running the engine, I noticed oil dripping on my knee from behind the panel. At first I thought it was a leak in the oil pressure gauge line, but it turned out to be coming from the back of the tachometer. Further investigation revealed that the seal where the tach cable attaches to the engine was bad, allowing oil to be pushed down through the cable housing to where it dripped out. This was a straightforward fix but caused more delay.

At the same time, I was looking at the rest of the fuel system. The big problem was the geometry. The tank in my plane is a copy of a Taylorcraft tank instead of the flatter and wider tank of the original Parrakeet design. Why it was built this way I don’t know, but such things are hardly uncommon in homebuilts, especially older ones; it may simply have been what was available. Added to this is the relatively long engine mount, which moves the engine up higher relative to the tank in a climb attitude.

Previously while draining the tank to replace the fuel lines, I had roughly timed the fuel flow. Generally, the available fuel flow should be at least 150% of the engine’s maximum consumption, with the plane in a steep climb attitude. The C-85 burns 7.65 gph at full throttle; 150% of that is 11.5 gph. A proper test would be with the plane pitched up higher than the three-point landing attitude (sometimes called the “tail in a hole” test) and the fuel draining from the carburetor drain plug, but at the time I was just wanting to empty the tank, so I disconnected the fuel line at the carburetor. There was plenty of flow at first, but the last gallon took over 7 minutes, which works out to only about 8.5 gph with the plane in a three-point attitude. Clearly this would be (and was) inadequate with the added restriction of the carburetor inlet screen and needle, and/or in a higher pitch attitude.

I also did some measurements of the head heights to get some numbers. I imported a side photo of the plane into my CAD system, scaled it, and then drew the lines and angles over it for the level flight, three-point attitude (9.6° nose up, measured between a bubble level and the cockpit longerons) as photographed, and 15° nose up representing a steep climb. The Stromberg carburetor is designed for, and the float level set at, a fuel pressure of 0.5 psi, which is 19 inches of head in a gravity-feed system. This works fine in planes with wing tanks or tanks mounted high in the fuselage. On this plane, however, that’s about where the top of the tank is in level flight, with only 6 inches of head at minimum fuel. It’s even worse in a climb, with head ranging from nearly 15 inches when full but less than 2 inches in the three-point attitude, and just over 12 inches with full fuel to half an inch below the carburetor inlet when 15° nose up. No wonder there were problems!

One of the previous owners had altered the fuel tank cap with a forward-facing vent to try to slightly pressurize the tank. The ram air pressure at 100 mph is 0.18 psi, which is the equivalent of raising the tank 6.6 inches higher (since ram air pressure is proportional to the square of the airspeed, at 80 mph climb speed, it’s more like 4 1/4 inches). Since the tank was still empty, I went into mad scientist mode to test this using a leaf blower and a homemade water manometer.

First, I used the manometer to measure the ram air pressure right out of the blower. I got 4.5 inches of water pressure, which works out to 95.4 mph … not the “120 mph” claimed by the sticker on the blower (shocking!), but good enough for my purposes. Note that 4.5 inches of water is equivalent to 6 inches of fuel head since the density of avgas is 75% that of water.

Next, I hooked up the manometer to the (empty) fuel line where I had disconnected it from the carburetor. With the blower pointed at the vent inlet, I measured 3 inches of water pressure, which is the equivalent of raising the fuel level in the tank by 4 inches (2 1/2 inches at 80 mph). This would help, but not enough. And clearly there were losses somewhere, since I wasn’t getting the same pressure that I measured right out of the blower. One place where there would be leakage is the hole where the wire from the fuel gauge cork passes through the cap. I taped around the wire to seal it, and then observed the same 4.5 inches of pressure I saw straight from the blower.

I then removed that tape and taped over the forward-facing vent to seal it and pointed the blower across the cap where the wire exited. This resulted in a suction of 1.5 inches in the tank! This would explain why fuel sprayed out over the windshield during a brief period of negative G on my first flight. I also experimented with a piece of tubing with the front side cut open around the wire exit to change the suction to a positive pressure area, which changed that 1.5 inches of suction to 2 inches of pressure. It’s not clear how all the pluses and minuses would add up, and my leaf blower wasn’t large enough to simultaneously blow on both the vent and the wire exit, but it was clear that while the forward-facing vent might be helping a bit, it wasn’t enough. Ultimately, I may replace the cork-and-wire gauge with a sight tube in the cockpit (retaining the forward-facing vent), if only to keep fuel from spraying out the gauge hole during negative-G flight, but that’s a project for another day.

At this point it was clear that I had two main options: I could design, build, and install a new fuel tank more like the original Parrakeet design, which would require major surgery on the plane including removing the wings again as well as the cabane struts and the top of the fuselage. Even that wouldn’t meet the Stromberg specs, though a lot of Parrakeets are flying that way. Or I could install a fuel pump, which would be a lot less work.

The obvious solution would seem to be an engine-driven fuel pump. The engine does have provision for the usual mechanical fuel pump, operated by a lever riding on the camshaft. Until I saw the prices … $1,350 for an overhauled unit, or $1,650 new! Not only that, I would also have to replace the carburetor, as a different carburetor is used for pumped applications. The only difference seems to be the float needle seat size to handle the increased pressure, but that part seems to be unavailable separately, and a new carburetor is around $1,200, adding to the expense. Furthermore, that would introduce a new single point of failure; if the pump failed there would be no backup. Most aircraft that depend on an engine-driven fuel pump have an electric backup pump, but my plane has no electrical system. Some low-wing planes, like the Ercoupe, use the engine-driven pump to fill a small header tank from the main wing tanks, gravity feeding the engine from the header which provides a reserve, but I had no room for a header tank.

I did later discover that some C-85 engines were made with two engine-driven pumps for redundancy, but the parts to install the secondary pump are even rarer.

There were some other theoretical options. I heard a lot of things like, “Oh yeah, the Continental pump is a copy of the pump used on a 1937 flathead Ford” (or 1934 Buick, or whatever … ), but nobody had specific information or personal experience, and I wasn’t about to stick the lever of some random pump into my engine and hope it didn’t crash against the camshaft. And I’d still have the pressure issue requiring the different carburetor, unless I added a fuel pressure regulator, adding another potential point of failure. My experience with the poor reliability of industrial pressure regulators made me prefer to avoid this option. I briefly daydreamed about stranger options, like hacking the front end and gear of an original generator to drive a lower-cost mechanical automotive pump from a newer vehicle, but I didn’t consider that seriously.

The alternate was an electric pump, which meant adding a battery and wiring. I could, of course, install a generator or alternator and other electrical system components, but aside from the cost and weight, that would also mean I’d have to add a transponder and ADS-B Out (even more cost and weight) to fly through a nearby Mode C veil I occasionally pass through, as I’d no longer be under the exemption for aircraft not having an “engine-driven electrical system.” That would be getting even further from the simplicity that’s one of the appeals of this plane’s 1930s vibe. I wasn’t crazy about going electric at all, but it really seemed to be the best solution.

A low-pressure pump would allow keeping within the original gravity-fed carburetor’s 4 psi pressure limit. If the pump failed the engine should continue to run, at least at a moderate power setting and pitch angle, just as it always had. Many experimentals, and some standard certified aircraft, use Facet-Purolator (formerly Bendix) electronic “cube” pumps, which are small and lightweight, have low power consumption, and are reasonably priced. Facet does warn that they’re “not for aircraft use,” but they’re widely used in airplanes anyway. I have even heard reports of the “not for aircraft use” marking still being found underneath an FAA-PMA sticker! There are a wide variety of pressure and flows available, with their 40163 model rated at up to 17 gph and 1.5–2.5 psi max pressure, delivering 7 1/2 gph right at the carburetor’s 0.5 psi design pressure … perfect! But how to power the pump? Without a generator or alternator, it would have to be a total-loss system. Many people run electronic ignitions or other systems this way, charging the battery on the ground when necessary. The pump draws 1.6 A, so a 5 Ah battery would run it for over 3 hours, which is more than the plane’s fuel capacity. Modern lithium iron phosphate (LiFePO4) batteries of that size are small, don’t have the fire danger of lithium-ion or lithium-polymer batteries, last for thousands of charge cycles, and weigh under 2 pounds. Furthermore, the pump should only be needed for a short part of each flight (takeoff and climb), so there should be plenty of reserve capacity. I could even use two batteries with a switch to select them, for redundancy or additional time.

According to the manufacturer, it takes 0.25 psi to push fuel past the internal spring-loaded check valves of a non-operating pump. This is no problem when using the pump as a boost pump in line with another pump but would be too much restriction for my already inadequate gravity-feed system. This could be handled by installing a bypass with a springless check valve (to prevent backflow) in parallel with the pump. When operating, the pump would pump fuel around the closed check valve and the downstream pressure would hold it closed; with the pump not operating, the check valve would open and allow gravity flow around the pump. It would take a simultaneous failure of both the pump and the check valve to completely starve the engine of fuel, and I considered that unlikely enough that I’m comfortable with it. After choosing this approach, more poking around online revealed that other people had done something similar on different aircraft.

As my day job was designing machinery for the automated production of check valves for automotive fuel systems, among other things, my first thought was to grab an “engineering sample” valve from the lab and machine a custom manifold to adapt it to the Parrakeet’s fuel system. I sketched up a couple of versions of a manifold, but I soon discovered that Aircraft Spruce carried a check valve already in a housing with #6AN connections, which would be a lot less work. The plan would be to tee off the fuel line at the gascolator discharge, with the gravity-feed line to the check valve on the straight run of the tee (for minimal restriction) and the side connection going to the pump, then back into another tee between the check valve and the carburetor. Simple in theory, though choosing all the fittings and getting all the hose lengths right would be a tedious process.

Before diving too deep and spending money on more parts, I ordered the pump and only the required fittings to plumb it directly inline without the check valve, just for ground testing. With the pump zip-tied to the engine mount and temporary wiring to a leftover starter battery, I did my “sanity check” test run. As expected, the head pressure from 3 gallons of fuel in the tank wasn’t enough to push fuel past the pump’s internal check valves. With the pump running and the engine not running, there was no leakage at the carburetor, and it started easily, idled smoothly, and ran at full throttle for several minutes, until I shut it down. That night, I ordered more fittings, battery, pressure gauge, etc., for the permanent installation. I also ordered a 74-micron filter screen (the size Facet recommends) for the gascolator to replace its original 120-micron screen. The hoses would be ordered as components were mounted and measurements taken; the local speed shop usually had a one-day turnaround on custom-length orders from the manufacturer, who was also local.

There was some question as to how to mount the fuel pump. The manufacturer is vague about it; although they recommend that the pump be mounted near the tank with the outlet oriented somewhere between 45° upward to straight vertical “to prevent vapor lock,” they also say that straight horizontal is OK. On the Parrakeet, the most convenient place to mount the pump would be in the engine compartment above the gascolator, on the firewall with the outlet horizontal. There is existing structure behind the firewall at this point where something had once been mounted and removed, and the plumbing would be simple, keeping within the hose’s 3-inch bend radius limitation. Elsewhere on the firewall there would be nothing but the 0.016-inch-thick firewall itself to mount it to, which I didn’t consider adequate. I decided to try it that way, as unlike a setup where the pump is pulling fuel from a tank below the level of the pump, there would always be some head pressure. Furthermore, with the Parrakeet’s Cub-style open cowl with engine well forward of the firewall and the cylinders and exhaust completely outside, it shouldn’t get too hot, lessening the probability of vapor lock. I ordered some temperature witness stickers to put on the pump, however, to record the actual temperatures during testing. If necessary, I could add a cover with a fresh-air duct, as is done on some other aircraft.

While waiting for parts to arrive, I made a mounting plate for the pump and got it mounted, using rubber washers and bushings to somewhat damp the vibration. I also made a “snubber” for the fuel pressure gauge, which is an 0.006-inch orifice intended to damp the pump’s pressure pulses so the gauge doesn’t vibrate. The orifice also slows fuel loss if the line to the gauge should fail for any reason. I wanted the gauge to verify pump operation, since in most flight conditions there would be no other indication that the pump is in fact operating.

When the fittings arrived, I was finally able to work out the hose routing and lengths. There were several routing options, but I finally came up with an arrangement I was satisfied with and ordered the remaining hoses. Then it was just a matter of finishing the plumbing and finding a place for and mounting the pressure gauge in the cockpit.

I spent some time considering the electrical side of the setup. The battery I ordered should be sufficient to run the pump continuously for 3 hours or more, though I didn’t anticipate it would need to run continuously. At the same time, it would be nice to have power available so I wouldn’t have to rely on the handheld radio and phone’s internal batteries on longer flights. The battery is light and compact enough that using two wouldn’t be unreasonable, with one to power the pump and the second battery for the electronics, with a selector switch so that the secondary battery could power the pump if necessary.

Finding a place to put two batteries (or even one) was a problem, though, even as small as they are. Finally, I decided the only logical place was on the floorboard between my legs, just forward of the stick. The floor is wood, so a wooden box wouldn’t look out of place and would keep everything unobtrusive and neatly protected. I decided on a box big enough to hold two batteries, switches (battery selector and pump on/off), a power outlet for portable electronics, and a voltmeter. I even had most of a sheet of mahogany plywood left over from a friend’s Minimax project. But that could wait; it was getting close to winter so I was running out of good flying weather. The box and second battery could be a winter project, but in the meanwhile, a simple cradle holding a single exposed battery and wiring would suffice for testing and short flights during the colder months.

I was almost done when I encountered another small setback. One of the hoses I ordered came in a half-inch shorter than I ordered and couldn’t be used. The required length happened to be the exact same length as the original tank-to-valve hose I had already replaced, so I thought I could use that instead while waiting for the replacement … until I saw the puddle of fuel on the ground after turning on the pump. Apparently, the old hose (which hadn’t been leaking when removed) was sufficiently old and brittle that it must have cracked when bent to a new shape. Although the cracks were invisible under the hose’s fabric cover, fuel was oozing out all along its length, so further testing would have to wait until the replacement hose arrived.

However, I had already verified that I was getting gravity flow through the check valve with the pump off, and up to 2 psi with it on. Once the new correct-length hose came in and a permanent version of the fuel gauge baffle was installed, I was truly ready.

Ground testing showed that with 3 gallons in the tank and the pump not running, the engine would run continuously at up to 2,000 rpm, while at full throttle (around 2,200) it would starve after 30 seconds, just as before. Switching on the pump brought it back immediately, and with the pump running it would run at full throttle indefinitely while indicating about 1/2 psi, or as near as I could tell on the 15-psi gauge. Satisfied, after another thorough check for leaks or anything out of place, I buttoned up the cowling, waited a day for lighter winds, and made a test flight.

This time, there was no hesitation from the engine. With the pump running, I was able to climb steadily at full throttle all the way to 4,000 feet. Level flight or moderate climb with the pump off was no problem. In a very steep climb with the pump off, the engine would falter, but switching on the pump immediately brought it back. Back on the ground, the temperature sticker showed that it hadn’t reached even its lowest 130° F mark, though I will continue to keep an eye on it, especially in warmer weather.

After the first few flights, I exchanged the pressure gauge for one with a 3-psi range. At 2 1/2-inch diameter (the old one was 1 1/2-inch) it’s larger than I’d like, but I couldn’t find anything smaller. That was interesting; at full throttle with the pump running it’s a steady 1 1/2 psi, while in level flight with a full tank and the pump off, the gauge shows 1/2 psi, just as predicted. The indication drifts a bit as fuel sloshes around in turbulence, but that’s unsurprising. On the ground afterward with the pump off it was reading somewhat lower due to the three-point attitude and the fuel burned, which is also as expected. It responds quickly enough without showing any vibration of the needle, which means the 0.006-inch orifice is appropriate.

There was still more to do. If I can’t find a smaller pressure gauge, I plan to move the existing one to the panel, so I won’t have to worry about banging it with my knee while getting into the cockpit. The switch location on the floor next to the battery wasn’t hard to reach, but it’s inconvenient. Having it near the throttle will be a better location, though the other switches and electrical components will still be in the future battery box on the floor. All of that is for a winter project, though, as the airplane was flyable as it was. Life is good.

(to be continued … )