Last updated 07/29/05

Purpose

This article documents the considerations and changes made to the cooling system and ducting on our RV6A in order to acheive reliable function. This is intended as a practical, real world guide to others contemplating a liquid cooled engine installation in their aircraft. We changed one thing at a time and had very comprehensive temperature and pressure probes instrumenting the engine and cooling system. The conclusions and solutions contained here are the result of actual scientific flight testing, not armchair conjecture. The turbocharged engine presented additional cooling hurdles vs. the naturally aspirated EJ22/EJ25s normally fitted. This was an experiment to provide adequate cooling with the lowest possible drag, not just shoving in a big car radiator and blowing a lot of air through it with drag as a secondary concern.

The unblemished stock type cowling lines when we started. The aircraft did not cool in flight properly in this configuration.

History

A turbocharged Subaru EJ22 engine was chosen to power the Racetech RV6A as an alternative to the usual Lycoming O-360 engines. I've never liked the decades old technology and high cost of certified engines and I wanted something different under the cowling. A turbocharged, fuel injected, liquid cooled engine seemed to offer possibilities. Our initial aim was to reduce the cooling drag penalty suffered by air cooled engines by designing a properly ducted radiator setup. This proved difficult from placement/ time to build and ducting standpoints, in part due to the nose gear support structure on the 6A. A compromise solution was developed where the rads would be mounted behind the factory cheek air inlets with short diffuser ducts. At the time and in retrospect, this proved to be an inferior choice however, it was made to work with many hours spent changing various parts of the cooling system. If you have the space, a longer duct placed in the outer half of the prop arc, incorporating a diffuser and a converging exit into a low pressure zone is likely to give better ground cooling, flight cooling and lower drag than our setup.

Basic Considerations

Heat rejection rates of a given radiator or heat exchanger (HE) are dependent on the mass flow (weight) of air through the core and temperature difference (Delta T) of this air vs. temperature of the coolant. We cannot control the ambient air temperature nor to any great degree the coolant temperature so most of the focus is on air mass flow. Mass flow through that core at a given airspeed or velocity is dependent on air density and pressure differential across the core (inlet to outlet or Delta P). We cannot control the air density and this varies with altitude so we have less mass flow available at higher altitudes at a given airspeed. We can really only make changes to the Delta P. Inlet pressure at a given altitude is determined by velocity and the pressure recovery of the duct. As air slows down within a diffuser duct, its pressure increases which is exactly what we want. If the duct is too short, we get poor pressure recovery and lower inlet pressure. The other side of the coin is the outlet pressure. The more air you let into the cowling and the less air you get out, the more back pressure you have acting against the rad. This means lower Delta P and lower mass flow. Its is very important to note that with Delta P at 0, there is no net airflow through your rad. In other words if the air cannot get out the back side of the rad, no amount of air or size of inlet on the front will do anything. The higher the Delta P, the better the rad will work. Air should therefore exit into the lowest pressure area possible.

With liquid cooled engines we have three basic considerations:
1. Adequate ground cooling under idle and taxi conditions
2. Adequate cooling at full power in climb at low airspeeds
3. Minimal drag in cruise configuration

These 3 conditions are often at odds with one another. On the ground, while we are only producing 5-20 hp, we still must have adequate airflow through the rads to keep the coolant temperatures stabilized. Most applications rely on prop blast for this airflow. If the propeller is geared down as in our case at 2.2 to 1, the propeller is turning only 454 rpm at 1000 engine rpm. This reduces airflow through the rads. If rad air is taken beside the spinner as on the RV6, the propeller airfoil section may be less than ideal here as well. All of these things lead to relatively low airflow and pressures available at the rads during low rpm operation. Because we need minimal rad frontal area for low drag in cruise flight we are forced to use a small area, deep core to obtain sufficient rad volume to cool at high hp levels. These require substantial pressure differential between the front and back of the rads for the air to penetrate the core and remove the heat.

In the climb, we are developing 150-170 hp so the heat rejected into the cooling system is the highest under these conditions. Fortunately we have an airspeed of around 85 knots to help us out here. There needs to be enough mass flow through the rads to stabilize coolant temperatures again.

It would be easy to keep any engine cool on an aircraft if drag were no consideration. A standard automotive rad mounted in a duct would certainly do the job but the plane would be very slow indeed. Low frontal area and space limitations dictate small face area and deeper cores than would be used on an automotive application. Cooling drag is influenced by many factors: The most important would be inlet area and geometry, the actual restriction of air passing through the fins and over the core tubes and the turbulence created by the exiting airflow on the a free stream airflow. The smallest inlet combined with the most efficient radiator design and duct shape, discharging the air at about the same velocity as the free stream air would give the lowest drag. These are very difficult to achieve over a wide speed range in reality with the two factors above also considered, hence the problems that many encounter with liquid cooled engines. Drag was a primary consideration on our RV6A so we tried getting away with the smallest of everything. This would prove to haunt us for some time.

Radiators

We first settled on 2 rads. One is a Chev Cavalier evaporator core mounted in the left cowling cheek at a 30 degree angle. It measured at 10 X 7 X 2.85 inches giving 70 square inches of face area and 200 cubic inches of core volume. It is important to avoid obtuse rad angles unless inlet guide vanes are used to re-align the airflow. A second Ford heater core was mounted underneath the oil pan. It measured around 7.75 X 7.75 X 2 inches gave us 60 more inches of face area and 120 cubic inches more core volume. Total core area was 130 square inches, total volume 320 cubic inches. Both cores are aluminum. Both cores were fed by diffuser ducts, tightly fitted to the cores. The lower rad had a 440 CFM electric fan fitted behind it for ground cooling. Water flows through 3/4 inch silicone heater hose from the top engine discharge pipe to the left rad, then through the lower rad, then to a header tank, then back to the inlet of the water pump. The deeper the rad with any given type of construction, the less efficient it is at dissipating heat. I feel that rads deeper than 3 inches should not be used on aircraft in the 200 knot and under range unless they have a very open tube and fin construction.

Left: main rad is a Chev Cavalier evaporator core. Right: lower rad added later is a Ford van aluminum heater core with Spal electric fan for ground cooling. The fan was later deleted after hot weather testing showed it had no effect on dropping coolants temps in ground running.

While these two rads eventually proved to be adequate in winter flying conditions, flight in higher ambient temperatures proved problematical and a third heat exchanger was eventually added. This was in the form of a barrel shaped head exchanger with a matrix about 3.5 inches in diameter and 6 inches long. Total core volume about 58 cubic inches. This is fed by a 2.125 inch duct from the right cowl cheek inlet. Water source for this heat exchanger is a 1/2 inch hose from the right head, going throgh the turbocharger center section, through the HE and back to the thermostat housing. In the end, we had total core volume at 378 cubic inches and total area at about 139 square inches. Undoubtledly, a single radiator with roughly these dimensions or even less mounted in a properly shaped duct on the belly would do the job and create less drag.

Barrel radiator added later to solve high coolant temperature problems

Header Tank and Rad Cap

We use an aluminum header tank fabricated from tubing measuring about 2.5 inches in diameter by about 8 inches tall. The tank contains a pressure transducer and temperature probe. A 7 lb. pressure cap is currently fitted. The water pump inlet fitting is located in the bottom, the water inlet fitting from the rad exit is located in the top at a tangent to the tank to allow entrained air swirl out. We found it VERY important to leave an air space of about 2-3 inches in the tank to allow the coolant to de-areate and expand properly, otherwise large amounts of coolant were discharged overboard. The pressure in the system is determined by the cap relief pressure. The header tank is mounted at the highest point of the system.

Coolant

Initial testing used 25/75 Ethylene glycol/water due to the cold ambient temperatures. I dispute the recomendations of many to use a 50/50 mixture unless it is required that you fly at -50C. The higher the proportion of glycol, the lower the thermal conductivity (heat transfer capability). We used the minimum amount of glycol to prevent freezing for the seasonal conditions. This permits smaller rads and less airflow (drag) to do the same job. Straight water is about 50% more effective than a 50/50 mix at conducting heat. Straight water also takes 15-20% more energy input to raise its temperature the same amount as a 50/50 mix. This makes a measurable difference in reducing coolant temperatures. The downside of more water in the mixture is slightly lower boiling points. This can be compensated for with a slightly higher rated pressure cap however. We also found that Redline Water Wetter decreases coolant temperatures by reducing the coolant viscosity and increasing heat transfer rates. For summer conditions run 90% water, 10% glycol and a few ounces of Water Wetter.

After many problems encountered, we switched to Evans NPG+ Propylene glycol (non aqueous) in an attempt to eliminate perceived problems with vapor bubbles entrained in the coolant. The NPG+ has a boiling point of 375F. Although we now use this coolant, I believe that standard EGW type coolants would also work fine.

Thermostat

After encountering several thermostat problems initially due to the cold side placement on the EJ22 engine, we are currently operating with no thermostat installed for summer flight. The OE setup relies entirely on the bypass hose water flow to open the thermostat. If you plug this hose, the engine will almost certainly overheat!!! The factory thermostat starts to open at 78C and is fully open at about 93C.

Getting the Air Out

Getting all the air out of the system and filling the system were big problems. These were helped by fitting two air bleed petcocks, one on the left upper rad tank and the other on the engine coolant crossover pipe. Air in the system will expand into large bubbles and cause local boiling in some cases due to no heat transfer. Explosive release of coolant occurs which quickly leads to serious overheating. All high points should have an air bleed fitted. One of the final problems discovered was that air trapped elsewhere would find its way into the upper part of the left rad leading to wild temperature and pressure fluctuations. Some people have placed small diameter bleed lines from system high points to the header tank to automatically purge the system. These may be a very good idea. Air trapped in the system will lead to major heachaches!

Rad/ Inlet/ Exit Changes

We first started ground testing with the GM core and a 10 X 10 X 2 inch Earls oil cooler as the other radiator, both fed by their respective stock Van's sized inlets. A water to oil heat exchanger was fitted. In ground running, the oil temp went up over 120C in the pan so the water/oil cooler was tossed and the right Earls cooler was changed over to an oil cooler using a sandwich plate and thermostat from Mocal. This solved the high oil temp problem but now the engine would not stay cool. Where to mount another rad and how to get the air to it? We finally came up with mounting the Ford heater core, diffuser and fan in the lower cowling fed by an elliptical hole cut in the lower right front cowling and a 3 inch hose. This proved adequate finally for ground running and taxi testing at temperatures up to about 23-25C. A revised round boundary layer type scoop replaced the flush scoop after tuft testing revealed high turbulence here.

Flight testing in 0 to -10C ambient temperatures showed that the engine was not being cooled well. Temperature of the water exiting the rads was not a lot cooler than the engine coolant temperature. Lots of water was being puked out on the belly of the aircraft after filling the overflow bottle completely. Not good. The cowlings were off after every short hop. Finally we did not fill the header tank all the way and things improved a bit. The coolant stayed in the engine. Next we fitted some sensitive pressure gauges to measure the pressure differential between the radiator inlet and internal cowling area where rad air was discharged into. We immediately saw that there was not much differential here. No differential, no flow through the cores. Air was reaching the rad but there seemed to be insufficient exit area to allow it out of the cowling. I built a temporary aft facing exit scoop to take the place of the oil filler door on top of the cowling in a known low pressure zone. Test flight confirmed reduced pressures and temps.

I took the step of cutting a 4.25 by 6.625 inch rectangular hole in the side of each cowling and bonding inward facing exit ramps into the cowling interior to direct the airflow aft. Test flights saw internal cowling pressure drop way down along with the coolant temps. I was able to climb at high power without the temps getting too high for the first time. Eventually I added sheet metal ducts over the exit holes to try to help suck the air out of the cowling.These showed more gains. I did not want to fill the cowling with holes but did not have a choice. The factory RV6A exit area was clearly inadequate to discharge enough air to cool a 200 hp turbocharged liquid cooled engine. This can possibly be explained by the fact that air cooled engines operate at much higher temperatures than liquid cooled engines so the temperature differential between the cooling air and heads is 300-400F instead of 100-200F on the liquid cooled engine. Basic thermodynamics says that the higher the temperature differential, the more heat will be dissipated. The liquid cooled engine therefore is likely to need more air mass flow to stay cool. I would say that on our setup with less than ideal inlet and outlet ducting, cooling drag will likley be higher than with a Lycoming engine. That being said, I believe that even with a properly designed diverging, converging duct containing the heat exchangers, that the liquid cooled engine probably cannot achieve cooling drag numbers similar or superior to an air cooled installation. Others may not agree, but I don't see any of the liquid cooled RVs exceeding the Lycoming powered RV speeds at lower altitudes with similar installed hp.

First change in air exits involved inward facing exit ramps and flush hole in cowling sides.

Left:Temporary sheet metal wedge shaped exit ducts racer taped to cowling sides Right: Final configuration with composite exit ducts and intercooler inlet duct

We currently have 28.5 square inches of radiator inlet area and 15.4 square inches of oil cooler inlet area for a total of 43.9 to cool the engine. Exit area totals 66.7 square inches which comes to about a 1 to 1.5 ratio inlet to exit area. This is what works on this installation but I believe there are no accurate rules of thumb to apply here. I know that with my rad setup, the 1 to 1 stock ratio did not allow the engine to cool properly.

Cooling at High Altitudes

The aircraft was originally fitted with an adjustable cowl flap on the stock lower exit. In flight testing, it was shown that closing this increased speed by 3-4 knots however the temperatures started to rise on a winter day at high altitude. It was thought that inadequate cooling at higher ambient temperatures would result so it was removed. After calculating mass flow at various altitudes I saw that theory confirmed the experimentation. The turbocharged engine can produce 200 hp at 15,000 feet if desired where the atmo 180 hp Lycoming will only be producing about 100 hp here. The cooling requirements would be double in this case but air density is only about half. You can see the problem here. Let's take some examples:
1. Sea level 200 mph, inlet area 44 square inches, theoretical mass flow 412 lbs./ min.
2. Sea level 90 mph climb, mass flow 185 lbs./ min.
3. 15,000 feet, 220 mph TAS, mass flow 256 lbs./ min.

From the examples above we see the worst scenerio for either engine is in climb and this becomes even worse for the turbo engine at high altitude as it pulls the same hp up to its critical altitude. For the Lycoming, power drops off in almost direct proportion to altitude so assuming it cools in the climb at SL, it should cool everywhere. Consider that the turbo engine climbing at 15,000 feet at 90 mph IAS has less than 85 lbs./min. mass flow to cool the engine! This negates the need for a cowl flap on the turbo engine operated at higher altitudes. A cowl flap could probably add speed to both engine combinations at low to medium altitudes in cruise as this much airflow would not be required to cool the engine. Fortunately ambient temperatures at higher altitudes are almost always lower and true airspeed is also higher. This helps to increase mass flow and Delta T slightly helping out a bit.

Inflight testing of revised intercooler inlet and cowling air exit scoops

Oil Cooling

A turbocharged Subaru engine will almost certainly require a dedicated oil cooler to dissipate the extra heat rejected into the pistons, chambers and from the turbo bearings. The piston oil squirters add to the oil heat load. We use a 10 X 10 X 2 inch Earls oil cooler plumbed with a Mocal oil thermostat which opens at 70C. This provides rapid warmup and ensures no pumping problems in cold weather. The oil cooler is fed by the right cheek inlet. The cooler drops the temperature 20-30 depending on ambient conditions. Oil temperature after the cooler generally runs 100-115C in the climb and 85-105C in cruise. As test flying progressed into the summer months, we found that the oil temperatures started to get too high in extended climbs above about 18C ambient with in-the-pan temperatures reaching 135C. We relocated the compressor inlet tube from the ram duct directly underneath the spinner to a NACA duct on the lower left cowling and use the original duct to blow cool air over the oil pan. So far, flight testing at +16C has shown this to drop sump temperatures 20-30C and oil temperatures after the cooler 10-15C.

Oil cooler. Note sheet metal scoop which ducts air to the intercooler core through one of three, 2 inch Scat hoses.

Scoop configuration July 2004. Left to right 3 inch duct feeds lower rad, 2.5 inch duct below spinner feeds cooling air to oil pan, NACA duct feeds turbocharger, 2.5 inch boundary layer scoop feeds intercooler core. Twin lower intercooler air exits visible on belly.

Intercooling

All turbocharged engines in aircraft should be fitted with an intercooler. Our RV-6A is fitted with a 9.25 X 6 X 3.5 thick Spearco core. We typically see compressor discharge temperatures in the range from 60-90C. These are the air temperatures that the engine would see if no intercooler was fitted. Not only is the air density severely reduced at these temperatures but heat load into the engine is increased, taxing the oil and water cooling systems. Furthermore, power and pre-ignition limits are decreased. The final setup drops the compressor air about 30-35C. Calculated effectiveness is about 50% in climb and 55-62% in cruise. Not great compared to racing cars which are in the 90-98% range, but better than the 10% we started with. We went through many changes in intercooler ducting to arrive at a satisfactory solution. The air volume required here is immense to get really good performance. We started out with a single NACA duct and one 2 inch SCAT hose feeding the intercooler plenum but found that the air volume and pressures supplied were sadly lacking. We ended up with three, 2 inch hoses fed by forward facing ram ducts for maximum pressure recovery. Air originally exited the intercooler into the cowling but we found that the pressure differential between inlet and cowling was not that great. Much work was then done to isolate the the intercooler air exit from the cowling. In the revised setup, the core is sealed to the firewall and air exits through the firewall into 3 SCAT hoses with 50% more area than the inlet hoses, then out 3 custom designed, aft facing exit ducts under the belly. This decreased the exit pressure, improving the Delta P and increased the core effectiveness substantially.

Intercooler left, revised triple feed hoses right

Exit holes in firewall left. Later another hole added. Right: Exit holes in belly. Later one more added to improve core effectiveness.

Temporary sheet metal exit duct left. Final configuration of intercooler exit ducts right. Note guide vane in right duct to improve flow and reduce exit turbulence.

Instrumentation and Quantifying Data

It was clear that many prevoius liquid cooled aircraft engine projects failed due to insufficent information about temperatures and pressures in various parts of the system. Wild conjecture seems more often applied than anything remotely scientific. We had envisioned certain developmental difficulties with our cooling system before building the aircraft so had incorporated many channels of temperature sening is various areas of the engine. Had we known how many problems we would encounter, many more channels would have been added beforehand. I'd recommend that each heat exchanger have inlet and outlet temperatures monitored so that each component can be studied with precision. Additionally, we found that sensitive pressure gauges were a must to determine Delta P through the ductwork and cowling inlets and exits. Without these items, your chances of complete success first time out are very limited. We made great use of temporary sheet metal scoops and ducts which could be fabbed up in a few minutes and taped on to the aircraft to test a concept. If the results proved out, proper ducts were fabbed and painted. This saved much time when something did not work as expected.

Sensitive pressure gauges are much more stable to read than fluid manometers. Available from Davis Inotek. We looked for a minimum 4 inch H2O differential between inlet and exit pressures at 125 knots. These gauges can measure both pressure and vacuum.

Wool tufts revealed turbulence in certain areas which allowed revision for cleaner airflow and less drag. Videotaping in flight from chase plane allowed these results to be reviewed more easily. ME 109 type scoop on left cowling ducts air to intercooler.

Ongoing Development (will it ever be done?)

We have now installed a new rad in the rear fuselage fed by a NACA duct in the belly. Air exits out new rear facing ejector ducts fitted underneath the stab on both sides of the fuselage. Pressure distributions were checked in CMARC and plotted in POSTMARC to find the best duct locations for this experiment. We verified findings with actual pressure probes in flight with the NACA dact and ram ducts fitted over top for comparison. This rad replaces the cowling mounted barrel rad added some months ago. This also provides a welcome shift in the C of G another 1/2 inch aft.

Aero Classics oil cooler gives 37 square inches of face area and 135 cubic inches of cooling volume

Aluminum diffuser attached to rad, fed by 3 inch SCAT hose

Assembly mounted in rear fuselage

Nowhere else to run the 1/2 inch aluminum coolant lines. Nice hand warmers for winter flying!

Air exits out twin ejectors mounted under horizontal stab

Complete assembly

Belly mounted NACA duct feeds rad via 3 inch SCAT hose

It was a bear to get the air purged from the system with the new longer lines.

Flight testing of the new rad setup went well. Temperatures do not exceed 80C in the climb at ambient temps up to 22C. With the NACA duct only at 100 knots IAS and 5000 feet, inlet pressure was 2.5 inches of water, exit pressure was .5 inches in the rear fuselage. Exit temp from the rear ejectors was 34-36C so it is doing something back there. Ambient temp was 8C at altitude. I landed and taped on a sheet metal ram duct over the NACA duct to check pressure differences. This brought pressure up only .5 inches to 3.0 on the inlet. The NACA duct works quite well in this location on the belly.

Still was not happy with the coolant temps in the climb on warmer days, I've now removed the Ford aluminum heater core/ rad from the nose scoop and replaced it with a 7 X 10.6 X 1.25, 2 core motorcycle rad. The Ford heater core I felt had a tube and fine spacing too close for proper airflow, causing high pressure drop. The SPAL electric cooling fan was mounted behind this and in hot idle ground running did nothing to reduce coolant temps so this has also been removed. Air to the new rad setup is routed via SCAT hose from the right cheek and the 3 inch scoop below the spinner. Some air from the 2.5 inch scoop under the spinner for the oil pan can also flow into this rad around the oil pan contours as demonstrated by the shop vac test. Later, this rad was completely boxed in and fed by the twin SCAT hoses. This showed a significant drop in coolant temperatures and again reinforced that you MUST not allow any space between the air feed and HE face as the air will choose the path of least resistance and flow around it. Seal the rads tightly to the ducts!

Shop vac was used as poor man's wind tunnel here to test airflow through various ducts and the new rad

With the present (final ?) configuration which is really kind of a mess, we have 181 square inches of rad area and 428 cubic engines of rad volume. If we figure 180 hp for takeoff, we have 1 square inch of rad area per hp and 2.37 cubic inches per hp. Cooling seems quite good now although I finally came to the conclusion that the new VDO water temp gauge reads about 20F high! Not impressed. This was leading me to think that things were getting too hot. I'll be checking the calibration of all new gauges I install from now on with a lab thermometer.

After more flight testing, I rearranged the cabin heater air control to be able to duct air through the heater core and out the belly or into the cabin. This provided a major reduction in coolant temps in the climb and added only about1 pound of weight. I also added a boundary layer duct to the rear mounted rad to increase Delta P. This provided a further gain over the original NACA duct, again flight verified with pressure and temperature measurements. Finally, the front bike rad was fully boxed in so that all ducted air would have to pass through it. All these changes have now resulted in temps not exceeding 90C in an 85 knot climb at 35 inches/ 4600 rpm at 25C ambient through 7500 feet MSL. I'm finally satisfied.

Trials and Tribulations

Our first major discovery was that the stock RV6A cowling had insufficient exit area to cool the engine with our less than efficient rad placement. We simply did not have enough Delta P across the rads. This was solved by sawing some holes in the cowling sides and glassing exit ramps over the holes on the inside to allow more air out of the cowling. This greatly improved the Delta P. Much time was spent in the latter part of development trying to understand why the engine would cool fine only with the heater valve opened and would rapidly overheat with it closed. With only 2 small NACA ducts feeding the heater core and the valve only half open or less, there did not seem to be an explanation of why the heater core was so effective in cooling the engine when the rads were not. In the end, we had to add the third HE and also found by accident that the main rad was 1/4 full of air in the top despite bleeding the system when we filled it last. This was discovered after we could not explain temperature readings that we were getting across the rad which showed zero Delta T which was deemed impossible. On removing the air bleed to install a new temperature sensor, we saw the coolant way down in the rad. Once the system was completely purged of air and the third HE fitted, everything stayed cool without the heater on. For once, we weren't sweating! Erroneous temperature readings and the vexing problem of air in the system probably wasted a couple of months and led to much hair pulling. Problems with the thermostat closing due to insufficient coolant flow through the bypass hose also caused some anxious moments.

Conclusions

1. 3/4 inch hose is adequate to plumb the radiators for a 200 hp EJ22T

2. Full temperature and pressure instrumentation will save you time, money and stress during development

3. A liquid cooled engine will likely require 1.5 times the rad exit area compared to inlet area due to the lower Delta T compared to air cooled engines. This rule of thumb applies to intercoolers as well.

4. Some means to remove ALL trapped air from the system is imperative. You MUST have an adequate air space in the high mounted header tank to allow for coolant expansion. You don't really require an overflow tank if you have this.

5. You should aim for a Delta P of at least 3 inches of water at 125 knots on your HEs. The higher the Delta P, the smaller your rads can be. Exits placed in low pressure zones increase Delta P. We measured no decrease in aircraft speed by doing this. Forward facing (ram) ducts recover far more pressure than flush type NACA ducts.

6. HE ducts which are too short do not have good ram recovery and may also cause turbulent airflow leading to reduced cooling efficiencies. A rule of thumb is that the duct length should be a MINIMUM of half the length of the longest HE dimension. Longer is better.

7. The 1/2 inch bypass hose MUST be used on the EJ22 if a thermostat is fitted.

8. Each HE should drop coolant temperature a minimum of 15-20C in the case of series rigged rads and 30C in the case of parallel rigged rads.

9. Because of the all aluminum construction and design of the EJ22, coolant and oil temperatures are closely linked. If one goes up, the other will closely follow.

10. If I could do it all again, I'd use a belly mounted scoop with a single radiator. This would have an 18 inch long diffuser and a 12 inch long converging duct with controllable exit flap. It might have exhaust augmentation to scavenge air and eliminate the electric cooling fan for ground running. Oil cooler and intercooler would be fed from the stock cheek inlets and exit out the sides of the cowling, similar in fashion to our present exits. The stock bottom exit would be faired flush to allow the belly scoop to pick up air properly.

Would I do it again? Now that I've gone through it and understand things better, I'd certainly use another Subaru engine. It performs well, sounds fantastic and is something different from all the Lycosaurus powered RVs out there. I love the thrust, whistle, easy management and climb performance of the injected, turbocharged engine.

For more info on the test flying and development of Racetech's RV6A: RV6A Flight Testing/ Development