Custom Exhaust Design for the Turbo EP912STi – Part 1

Design Considerations

My custom-built SR-1 race plane has a narrow firewall, and the stock turbo sat completely outside the cowl (the SR-1 is powered by an EP912STi, which keeps the stock turbo Rotax exhaust layout). It was therefore necessary to move the turbo to a more centerline location. This was possible since we are not using the large Rotax muffler, just a simple downpipe. The move also allowed the turbo to be rotated about the yaw axis in order to find a better compromise orientation for packaging the exhaust collector, turbo downpipe, and compressor inlet.

Although I’m always chasing after weight savings, the cost of a titanium exhaust was beyond the project budget. Not the tubing so much as the welding, which is relatively expensive due to the level of expertise required and more extensive shielding gas and fixturing required. My research yielded mixed messages on the advisability of ti for exhaust systems. I was told by an aircraft exhaust welder that a ti exhaust might present issues with cracking, especially being a turbo (with the extra weight and heat). That said, I may come back and re-do the downpipe in ti, as it sees less stress than the exhaust runners, would represent a significant weight savings, and is an easier weld-up compared to the long runners. (Note on terminology: I use pipes and runners interchangeably in this article.)

On the other hand, there was no point in putting the effort into a custom exhaust and using mild steel—stainless steel (SS) was the clear choice. You’ll find COTS (commercial off-the-shelf) SS exhaust components in 304, 316, and 321. All are fine—they are functionally interchangeable and can be welded to each other; just be sure to let your welder know what you are using so that they use the appropriate filler rod. 321 has better high-temp fatigue properties than 304. On the other hand, 304 is usually less expensive and more COTS items are offered in 304 than 321. You’ll also need to decide on the tubing wall thickness. 0.049 inch might be considered typical for homebuilt aircraft, but 0.035 inch and 0.065 inch are also available.

The Rotax uses 1.25-inch OD pipes (actually 32 mm). These can be a little difficult to source; I found a couple of sources, including SPD (Specialty Products Design in Sacramento, which specializes in custom exhaust systems), which sells straight tubing and custom mandrel bends. The large online retailers like Summit typically sell stock mandrel bends (45, 90, 120, 150, and 180°). If your design has, say, a section with a 137° bend angle, your options are either a custom 137° bend or to buy a 150° mandrel bend, trim it to the desired angle, and butt-weld on a section of straight pipe. The upside to the latter approach is price and availability; the downside is an extra butt-weld op and somewhat less smooth internal flow due to the extra weld bead. You can also tweak pre-bent sections, but don’t expect to get more than a few degrees of angle change.

The Rotax exhaust tubes have socket ends that slip into the cylinder head exhaust ports, which seat against a beveled land and are held in place with a hold-down flange. The socket ID is about 0.050 inch smaller than the 0.049×1.25-inch exhaust tubing ID. I purchased my first set of sockets from a vendor that maintained a constant ID on the socket, which resulted in a hard step at the transition from the smaller socket ID to the 0.050-inch larger tube ID. I was told “a good welder” could handle the diameter mismatch. That may be true (it sounded more like an excuse for lazy machining), but it also didn’t seem ideal to have a hard step in the exhaust tubing right at the exhaust port (where velocity is highest).

I mentioned this to Thomas Huaklien, CEO of Edge Performance, who built my 912STi, and he informed me that Edge sells these sockets, which are machined with a nice taper from the smaller ID at the port to the slightly larger ID where the socket butts up to the exhaust tubing. I highly recommend the EP sockets and flanges. Rotax also sells sockets and flanges if you prefer to source OEM parts.

The SR-1 exhaust is similar to the stock Rotax in that the forward cylinders’ exhaust makes a near 180° turn toward the rear of the engine straight out of the sockets. Rotax sells circular donuts that can be cut to achieve this desired angle—theoretically, a single donut cut in half (think two “C” shapes) would suffice to fab both forward exhaust bends, but I was advised against using the donuts on account of how they are made: two formed halves (think of bagel halves) welded together to form the donut, which leaves two rough weld lines on the interior surface. Again, this is right at the exit port and we’d like this to be as smooth as possible. In addition, the centerline radius (CLR) of the donut was larger than desired. Fortunately, I was able to source some tight CLR mandrel bends that gave me a bit more clearance from the cowl and would be smooth internally.

An Expensive Paperweight

The four exhaust tubes come together at a 4-into-1 collector that bolts to the intake flange of the turbo and directs the exhaust into the turbine. This collector uses double slip joints identical to those on the OEM turbo exhaust. The double slip is two short sections of tubing—one that slips outside the mating exhaust pipe, the other inside—think circular tab-in-slot. The depth of the slip (i.e., overlap) is about 1 inch, and the clearance is only a few thousandths of an inch. Properly aligned, the pipes slip into the collector relatively easily and have a tiny amount of wiggle to accommodate bolt-up. Once the exhaust heats up, the tubing expands and the gap shrinks to seal against exhaust gas leakage. The sockets at the cylinder head are also a slip fit (held in place as mentioned above with hold-down flanges) with even less clearance. The upside to this approach is that individual runners can be relatively easily installed or removed when the engine is cold.

The stock Rotax collector is a stamped unit that resembles a palm with four fingers. I’m sure there are manufacturing and layout reasons Rotax uses this design, and it obviously works fine since they use the same collector for all of their turbo models. But it wasn’t something that I could integrate into my layout, so I clearly needed a custom collector. Although companies like Burns Stainless sell pre-made collectors, I was unable to find anyone carrying collectors for 1.25-inch exhaust.

I contacted a vendor that makes experimental aircraft exhausts (the same who supplied the sockets), but the collector I received was poorly made—tapped with the wrong size threads, asymmetrical, and no smoothing of the bullet (the interior cone shaped by the four converging pipes). This is right in front of the face of the turbine, and again, we’d like nice smooth flow here. Long story short, I ended up with a $600 paperweight. With any custom parts, you need to be very, very clear about your requirements and expectations. Vendors cannot read your mind.

In frustration, I started a new search for a shop that I could trust to build the collector. I found SPD and was assured that they could make a collector and that what I would receive would look as good as the showpiece pictures on their website. They answered all of my questions and took the time to understand the very specific requirements/desirements of the project. This was in stark contrast to the previous vendor that showed little interest in resolving the issues with the faulty collector they’d fabricated—buyer beware. SPD delivered, and the collector was beautiful. Having seen their work on the collector, I decided I would have SPD weld the exhaust as well.

A Plastic Precursor

At this point, I should roll the story back a bit. Before buying any stainless, I bought a lot of PLA—that’s the filament I use in my Prusa 3D printers. PLA is cheap and easy to print with. The sockets, collector, and mandrel bends were all 3D-printed before ever ordering any metal. 1-inch PVC pipe served as straight tubing, and all of the parts plug into each other. This approach worked great. If you don’t have a 3D printer, you can buy mockup kits from vendors like iceengineworks.com, although I’m unaware of anyone selling kits for 1.25-inch exhaust. These kits look great but are not cheap. You could buy a nice 3D printer for the same price or less.

The first step in creating the proxy exhaust was establishing the location and orientation of the turbo in my SolidWorks model. I then needed to replicate this on the real engine. This was accomplished by mounting the turbo to the engine mount using adel clamps and scrap 3/8-inch O-temper aluminum tubing, which is easy to bend as well as flatten into a mounting spade at the ends. Thus the starting point (engine exhaust ports) and finish point (turbo inlet flange) were rigidly fixed relative to each other vis-à-vis the engine mount. Additionally, the engine was clamped to a welded assembly/support frame that allowed relatively easy access to the exhaust while sitting on my workbench and the ability to later easily fork the engine in and out of my Forester.

Although some features of the exhaust were modeled in CAD ahead of time, this was more of a rough fit check than a final model. The firewall forward is only partially modeled, as a complete detailed model would require a significant amount of time and not be of any particular benefit for my purposes. For the most part, only items that impact the cowl outline are included in the FWF CAD assembly (such as the engine, radiator, and intercooler). To be clear, all the exhaust bends ARE modeled for the purposes of 3D printing, but only a few are actually included in the CAD assembly.

At this point in the project I had three 3D printers all printing furiously as I tweaked the angles and clocking of all the bends, iterating until I had a layout free of interference issues and sufficient standoff from other components. Once the proxy model was finished, it was time to order steel. We’ll pick up from there in Part 2 of this article.