EAA - Experimental Aircraft Association  

Infinite Menus, Copyright 2006, OpenCube Inc. All Rights Reserved.


Tools:   Bookmark and Share Font Size: default Font Size: medium Font Size: large

EAA Experimenter

[ Home | Subscribe | Issues | Articles | Q&A | How To | Forum Review ]
[ Hints for Homebuilders | Glossary | Polls | Around the Web | Submit an Article]

Exhausting Stuff

by Richard Mole, Costock, England

The original exhaust for the JPX 85-hp engine in my Jodel D18 had the header from the front cylinder flowing into the header from the rear cylinder, giving an outlet pipe on each side of the engine. It had served me well over the years. It was simple, was quite light, and had never cracked. However, I came to suspect that it was causing uneven exhaust gas temperatures (EGTs) from one cylinder to another. EGTs can also be skewed by poor induction design - but that’s an entirely different story.

Original exhaust system
Photo 1. Richard’s original
exhaust system (shown here mocked up)
worked fine for 1,000 hours.

Cheap mild steel exhaust pipes are available for automotive VW engines, and it was easy to modify them to fit my uncowled engine. Would four independent exhaust pipes have any effect on EGTs or the maximum static rpm?

A test at maximum static rpm, with the cowl halves removed, should provide a useful indication to the climb EGTs that would have occurred had the aircraft been flown. Tests with the original exhaust demonstrated that, in comparison to climb EGTs, the rear cylinder EGT would run about 35ºF cooler and the front EGT would run about 65ºF cooler.

It was necessary to allow sufficient time to thoroughly warm the oil before opening the throttle. And I had to be aware of overheating the cylinder heads, while allowing sufficient time at wide open throttle (WOT) for the EGTs to approach their steady state.

The four 30-inch-long exhaust pipes (Photo 2) produced an average EGT that was 23ºF cooler than before, but this hid a surprising variation. On the port side of the engine, the EGT No. 2 was hotter by 43°F, and yet EGT No. 4, which was previously the hottest, was markedly reduced by 103°F. The impact on the starboard side EGT was relatively minor. The EGTs of both of the rear cylinders, No. 3 and No. 4, were now around 1,290°F and much closer together than hitherto. Overall, the spread of the EGT (from the hottest cylinder to the coldest) had reduced from 227°F to 135°F.

Exhaust pipes
Photo 2. Not a pretty sight. Ground testing independent exhaust pipes.

Put another way, this test had demonstrated that my original exhaust system was indeed increasing the disparity between the EGTs at WOT.

A further 10 tests were made, shortening the four pipes by some 2 inches each time. Talk about a no-brainer! The EGTs were unchanged, to my surprise. Furthermore, there was scant evidence to support the idea that the maximum static rpm depended upon the length of the pipes. However, the data did suggest that four independent exhaust pipes, of whatever length between 10 and 30 inches, produced another 20 static rpm, or thereabouts, compared to my original exhaust system.

Fired up by these discoveries, I threw the remnants of the mild steel pipes in the trash and reinstalled the original system! But I resolved to learn more about exhaust systems when my enthusiasm returned.

Power Flow exhaust system
Photo 3. Cessna 172 G-BJXZ sports the first
Power Flow exhaust system to be installed in the United Kingdom.

A local maintenance organization run by my friend and mentor had just secured CAA approval for fitting the Power Flow tuned exhaust to the Lycoming O-320 powered Cessna 172. (Photo 3) This was a good opportunity to take a long look at the hardware and to get stuck into evaluation test flights. Power Flow development engineers spent several years before they were completely satisfied. Now, with thousands sold for various aircraft, they have a well-engineered product.

Power Flow
Photo 4. The Power Flow (upper left) and stock Cessna
(lower right) exhaust systems side-by-side.

The Power Flow system (Photo 4) consists of four stainless steel “headers,” 43 inches long on average. The header pipes bolt to the individual exhaust flanges and come together at a four-into-one collector from where a larger diameter, 35-inch-long stainless tailpipe exits the cowl. An absorption type silencer within the tailpipe attenuates some of the higher frequency components of the exhaust note.

The headers’ outside diameter is only 1.5 inches and the internal diameter is expanded out to 1.65 inches at the flange, to match the exhaust port in the cylinder head. The 43-inch-long headers are folded up in tight 180-degree U-bends within the heater shroud, which is not evident unless the shroud is removed. (Photo 5)

Power Flow
Photo 5. The Power Flow system with the shroud removed.

The standard Cessna exhaust is quite different. Much shorter headers empty into a cylindrical expansion box, with an integral heater shroud. The tailpipe is a relatively large diameter and stumpy affair. According to an older version of the Power Flow website, a dynamometer test suggested that the system might create more back pressure than allowed under FAA regulation; this needlessly robs some of the potential engine power!

The ground runs showed that the maximum static rpm with the Power Flow system had increased by about 50 rpm, as established by a handheld optical tachometer and the aircraft tach. The immediate impression in flight was quite unexpected. The ride quality was much improved by the smoother-running engine, and the engine note seemed lower. Climb-out was certainly faster, and standard leaning techniques in cruise had more clear-cut consequences than before. This is probably because a tuned exhaust allows all four cylinders to operate much closer to the optimum conditions for combustion. This is good for fuel consumption, to say nothing of prolonging engine life. By contrast, a standard exhaust tends to cause one cylinder to run comparatively lean while another runs comparatively rich, and leaning-out is less precise.

There is some scatter evident in the points plotted on the graph. This is quite understandable on the basis of data from just two “before” flights with the Cessna exhaust and two flights after fitting the Power Flow exhaust. The results suggest a reduction of about 2.5 liters per hour (0.66 gph) or some 7 percent, amounting to useful financial savings. Further savings are available from a K&N air filter.


A baulked approach and go-around is a tough test in a C172 at gross. I let down almost to runway level before applying full power at fully rich mixture and turning off carb heat. Leaving in the 40 degrees of flaps, we re-trimmed the aircraft to 55 knots indicated airspeed. When it settled, typically at about 200 feet above ground level (AGL), we timed the climb for a further 500 feet. A height gain of 500 feet required 90 seconds with the Cessna exhaust giving a rate of climb of 333 fpm. This was reduced to 79 seconds with the Power Flow exhaust (averaged from four overshoots) giving an average rate of climb of 381 feet/minute. Assuming a ballpark 66 percent propeller efficiency in this regime, the improved climb suggests a useful increase of some 5 shaft horsepower or approximately 5 percent extra power output at 2,350 rpm.

How does a tuned exhaust system work? Several good references to exhaust system theory are given at the end of this article. The theory is complex, and I do not pretend to understand the subtleties. Basically, there are two main ways in which an exhaust system can affect the scavenging, or removal, of spent exhaust gases from the combustion chamber: wave scavenging and inertial scavenging.

Inertial scavenging is due to the momentum of the exhaust gas traveling toward the exit of the pipe, at somewhere between 150 and 300 fps. A slug of fast-moving exhaust gas creates lower pressure behind it. This assists the process of induction during the brief period of valve overlap, when the inlet valve opens before the exhaust valve closes. Inertial scavenging depends upon the velocity of the exhaust gases.

Stainless steel exhaust pipes have about half the thermal losses of mild steel, so the exhaust remains hotter, and faster, for a longer period.

Then again, smaller diameter headers produce a higher initial flow velocity, and the exhaust gas decelerates less fast, because the smaller surface area reduces the rate of heat loss. Small diameter header pipes tend to yield good torque at below 3,000 rpm, whereas progressively larger diameters are required for high power at higher rpm. Perhaps this explains why the Power Flow header pipes are only 1.5-inch outside diameter stainless steel, albeit flared out to match the 1.65-inch inside diameter of the exhaust port in the cylinder head. Other reasons probably include the lower weight and the smaller bend radius. In a well-designed four-into-one system, the inertial scavenging is further improved by merging the pulsating flows of the individual pipes to obtain a more sustained and regular flow in the tailpipe leading to the exit.

Wave scavenging utilizes the energy of the sound wave or pressure wave produced after the exhaust valve opens. This travels at around 1,750 fps, much faster than the exhaust gases themselves. A pressure wave moves from the originating exhaust valve to the relatively low atmospheric pressure at the exit of the tailpipe. It is immediately reflected as a rarefaction (i.e., depression) wave, which travels back up the exhaust pipe. Ideally, the rarefaction wave arrives at the original cylinder while the exhaust valve is still open, and the depression will increase the pressure gradient leading from the combustion chamber. This encourages more of the remaining burnt gases to exit the combustion chamber into the exhaust header. Hence the term “wave scavenging.” The depression will also induce a greater mass of fresh charge for the next power stroke during the brief valve overlap period. This is the main feature of a tuned exhaust.

The skewing of the EGTs by my original exhaust system also has a theoretical explanation. Most flat-fours fire in the order 1-3-2-4 with a power stroke every 180 degrees of crankshaft rotation. Merging the exhausts on each side of the engine produces an unbalanced situation whereby the rear cylinder (No. 3 or No. 4) fires only 180 degrees after the front cylinder (No. 1 or No. 2, respectively), there being 540 degrees before the front cylinder fires again. Since an exhaust valve remains open for about 240 degrees, one might expect some interference from the rear cylinder in the final stages of the exhaust cycle at the front cylinder.

Peak pressure normally develops around bottom dead center (BDC), near the start of the exhaust header, adjacent to the exhaust port. Smith and Morrison give empirical evidence (in the book The Scientific Design of Exhaust & Intake Systems, page 111) that the pressure rise, at the end of any closed branch leading off the header, follows almost simultaneously with the rise in pressure next to the port. Furthermore, the peak pressure at the closed end may greatly exceed the peak pressure in the header. These effects pertain even when the closed branch merges with the header at an acute angle, at a narrow Y junction, designed to direct the gas flow to the exit and away from the closed branch.

The pipe work that links the front and rear cylinders in my original exhaust system has more in common with a closed pipe than an open pipe leading to an exit at atmospheric pressure. So a strong pressure pulse from the rear cylinder might very well affect the front valve around the time that the front cylinder is passing top dead center (TDC). It would then almost certainly impede the scavenging action at the front combustion chamber.

In summary, my original exhaust system was very possibly counter-productive. Hard evidence in support of pathology includes the uneven EGTs during a WOT climb. The trial of the four independent exhaust pipes suggested a slightly increased static rpm with more uniform EGTs. On the other hand, I have flown behind the system for almost a thousand hours, without complaint. Quite the reverse, knowledgeable passengers have commented on my smooth-running engine!

I wanted to reduce the spread in EGTs during the climb, and I hoped for a small increase in maximum static rpm and rate of climb. But how? There was not enough space around my tightly cowled engine for a copycat Power Flow tuned exhaust, and I shied away from all that extra weight and complexity, despite the promise of extra power. I considered using four independent pipes, but they would also be quite heavy and the sound quality is unattractive.

A crossover system would be lighter. This merges the exhaust headers from the front cylinders at a Y junction; the upper branches of the Y represent the headers, and the stem represents an outlet pipe. A similar but separate arrangement is used for the rear cylinders. The cylinders with merged headers will fire 360 degrees apart. Thus the rise in exhaust pipe pressure at BDC of the power stroke of the one cylinder will occur while the exhaust valve of the paired cylinder is fully closed. Sometimes, the two outlets pipes are themselves merged at a third Y junction, or two-into-one collector, to give a single tailpipe.

I wanted a compact exhaust system that would resist premature cracking even without additional physical support. This implies quite short headers of similar length, and merging them smoothly together at an acute-angled Y junction, as close to the engine centerline as possible. It seemed inevitable, however, that the exhaust pipes from the front cylinders to the outlet would be longer than those from the rear cylinders.

New crossover
Photo 6. New crossover exhaust tack welded on a jig,
made from an old dresser drawer and a pair of
discarded VW cylinder heads.

My cross-over design is shown above. The crossover distance from valve to valve averages 43.6 inches. The pulsating gas flows in the two intermediate outlet pipes are brought together at a two-into-one collector, in the hope that a more sustained mass flow through the tailpipe would improve the inertial scavenging, despite it being very short. This arrangement should also soften the tenor of the exhaust note and help to provide a robust and self-supporting assembly.

Ignoring the two-into-one collector just before the exit, the active exhaust valve sees a pipe leading to the exit, and sees the paired header as a blanked-off leg branching off at an acute angled Y junction. The distances from the valves to the exit are much shorter than the tuned length, being between 42.8 inches for No. 1, and 29.5 inches for No. 3. So a significant contribution from wave scavenging was not to be expected.
In The Scientific Design of Exhaust & Intake Systems, Smith and Morrison, (pages 108-111) describe the complex working of this configuration:

  • An initial pressure peak develops at the top of the header adjacent to the opening exhaust valve. This has a lower amplitude than would occur for an independent exhaust pipe.
  • An initial pressure peak occurs almost simultaneously at the end of the blanked-off leg, i.e., at the closed exhaust valve at the top of the paired header. The lag is typically only 10 degrees to 15 degrees of crankshaft revolution, because the pressure starts to rise in the crossover connection almost immediately when an exhaust valve opens.
  • The initial pressure peak at the end of the blanked-off leg may well have greater amplitude than the initial peak adjacent to the opening exhaust valve. Both peaks have short duration.
  • A second pressure peak develops near the opening exhaust valve. It lags the first by an interval that corresponds with wave development in a pipe slightly shorter than the headers. The headers are each 532 mm long, and so at 2650 rpm the lag is less than 32 degrees, say 30 degrees.
  • A second pressure peak also occurs at the end of the blanked-off leg, lagging the first peak developed there by the same 30 degrees at 2650 rpm.
  • The pipe from the exhaust valve to the exit also sustains wave action, quite independently, over its entire length. Thus two rarefaction waves will return to the exhaust valve that started the process, still some 30 degrees apart.
  • Together, these provide a relatively prolonged period of depression. The depression can be timed to coincide with valve overlap to improve the scavenging, but this requires a tuned length.

Taking these points into account, along with possible harmonics, resonance, and so on, there could be no guarantee that the new crossover exhaust would work any better than the original. The degree of inertial scavenging was questionable, and there could be no deliberate wave scavenging. Furthermore, there could easily be some undesirable interference effects between the cylinders.

I was not expecting any dramatic changes, and indeed the ground run produced more or less the same maximum static rpm as before. The highest EGT had reduced, but only by 23ºF. Observers told me that the engine had lost its Vee-Dub character and now sounded like the other aircraft engines on the block. Oh, and it idled well.

During the climb, the highest EGT was about 45ºF lower than before, and the spread in EGTs had reduced by the same amount. The rate of climb and the rpm were both improved to some small extent. The cruise EGT had risen by 10ºF to 20ºF, so the aircraft should be a tad more economical to operate. The engine still feels smooth and purposeful - no change there!

Whether a new exhaust system is likely to improve matters will obviously depend upon the quality of the old one. I had become concerned about my original exhaust system only because a newly installed digital instrument provided accurate EGTs for the first time. It did not seem right that the highest EGT in the climb should exceed the highest EGT when properly leaned-out in a low-level cruise. My new exhaust system has corrected this with a 55ºF differential. 

The time interval in seconds for a wave to travel D inches at V fps is D/(12V) seconds or D/(12*60*V) minutes. Equivalently, we can express the interval as the fraction D/(12*60V)*rpm of crankshaft rotation, or as D/(2V)*rpm degrees of crankshaft rotation.

We would like the pressure wave leaving the valve to be reflected as a rarefaction wave at the open end of the pipe and to arrive back at TDC during valve overlap. If the valve opens at N degrees before BDC, then (180+N) degrees of rotation are required before TDC.

However, Smith and Morrison (see page 83) state that about 120 degrees of rotation is required before the valve has opened sufficiently to start an effective wave on its way to the exit. The rarefaction wave is then required to return to the exhaust after a duration equivalent to (180+N-120) degrees or (60+N) degrees. Note that N≈60 degrees (57 for the IO-360, 59 degrees for the JPX, 50 degrees for the A65, for example), so we need a delay equivalent to about 120 degrees of crankshaft rotation or thereabouts.

Thus D/(2V)*rpm = (60+N) so that D = (60+N)*2V/rpm. If a wave is assumed to travel on average at V=1,700 fps, then D=2*1,700*(60+N)/ rpm.

The tuned length L will be the length of the exhaust pipe plus the distance between the valve and the exhaust pipe flange. The wave travels down the pipe to the exit and then returns to the valve so that L=D/2. Thus L=1,700*(60+N)/rpm (see Smith and Morrison page 83).

This formula assumes an average wave velocity of 1,700 fps and that the wave origin and destination are the exhaust valve, rather than the moving piston crown. Note that L is inversely proportional to rpm. With these assumptions, one finds that almost 7 feet is required at 2500 rpm. More precisely, the formula yields L= 80.9 inches when N=59 degrees. In practice, good exhaust system performance can also be expected within the range 2350 rpm to 2650 rpm.

Allowing 3.3 inches for the distance between the valve and the flange gives a pipe length of 77.6 inches at 2500 rpm. It is interesting that the overall length of the Power Flow tuned system averages about 77.75 inches according to my measurements, plus or minus about an inch variation between cylinders.

It can be shown that a wave velocity of 1,750 fps and an exhaust gas velocity of 300 fps yields the given average of 1,700 fps. Yet the formula is no more than an educated guess for the tuned length. For example, the wave velocity is proportional to the square root of the absolute temperature of the medium, but who knows how the temperature of the exhaust pipe contents cool down in any given installation, en route to the tail pipe. And the target delay for the return of the rarefaction wave of (60+N) degrees is based upon an assumed timing of the peak pressure in the exhaust port, which really requires laboratory determination in any particular case.

The Scientific Design of Exhaust & Intake Systems, P.H. Smith & J.C. Morrison, Bentley Publishers
“Aircraft Exhaust Systems,” Sport Aviation January 1997 (also January 1996, March 1996, May 1996)
“Tuned Exhaust Systems for Aircraft Engines,” Sport Aviation November 1980

About the author…Richard has published about 20 articles in Popular Flying, the magazine of the Popular Flying Association (now known as the Light Aircraft Association- or LAA), that implements delegated airworthiness responsibility from the governmental Civil Aviation Authority (CAA) in the United Kingdom. An LAA inspector must sign off on various stages of an experimental aircraft build and thereafter for the annual permit renewal. Richard’s inspector, Martin Jones, is a licensed engineer and runs an airfield and flying club. Under Martin’s careful supervision, and with approval from the LAA engineering staff, Richard has enjoyed modifying his aircraft in several ways.
Richard writes the sort of articles he wishes he could have read prior to embarking on some engineering change or another to his Jodel. He has a degree in aeronautical engineering and completed an apprenticeship with Vickers Armstrong’s Aircraft. Recently he retired from the OpenUniversityBusinessSchool.

Copyright © 2014 EAA Advertise With EAA :: About EAA :: History :: Job Openings :: Annual Report :: Contact Us :: Disclaimer/Privacy :: Site Map