Compound Turbocharging SI Engines
Compound/series turbocharging is often done on diesels, but when it comes to spark ignition engines it's very uncommon. This leads to a lot of internet myth and needless debate. I've been doing it myself since 2009 on a Mitsubishi 4 cylinder, have built several successful setups, and helped others build dozens more on all kinds of platforms. Normally I'll point people to a thread on Yellow Bullet that I started in 2010, but it's a very long thread and hard to follow (still a good read if you have the time). Here I will attempt to briefly cover the fundamentals, answer some of the most commonly asked questions, and give some examples. As always, I don't intend to portray myself as the world's premier expert on any of this. I speak from my own experience, and my views on the subject are based on that experience. Others may view/do things differently. I've made every effort to verify my findings in every way I can (the car is pretty well instrumented), but in the end I'm a racer, not a lab technician.
Compound Turbo Basics
Compound turbocharging has a few meanings, but what we are talking about is running two or more turbochargers in series. Like a two stage air compressor, there is a larger primary/atmospheric turbo that compresses the air once, then passes it on to a smaller secondary/high pressure turbo to compress it again. The work is shared between the two turbos. In the typical arrangement, the order of airflow is large compressor, small compressor, intercooler, engine, small turbine, large turbine, exhaust. There are some slight variations of this that can work, but this is the basic layout. Diesel setups will be similar, but we do things a little differently. Turbo sizing, wastegating, and boost control are the major differences.
While diesels use compound (or staged) turbochargers to run very high boost levels, SI engines typically use them to improve spool up. Particularly on small displacement engines. A turbo 4 cylinder might still run the same 45 psig boost, but each turbo will now only be running a Pressure Ratio of 2:1, and “feel” like it's only making 15 psig. By choosing a small turbo that will spool easily, and a large turbo that will make enough airflow for the power goal, you attempt to achieve the best of both worlds. This is especially helpful with automatic transmission drag cars, which with compound turbos, can then get up on the converter without nitrous. They can also be used to simply make high boost of course. While there are modern turbochargers that will make the pressure ratios required, once over 60-70 psi, compounding two compressors starts to make some sense. The rules with most sanctioning bodies for drag racing don't really account for compound turbos, there is some gray area there. They will often be considered twins, or two power adders, when in reality they do NOT add together. You are still limited to the flow of the atmospheric compressor.
Common questions and answers:
Q. It'll never work!
A. OK, that's a statement and not a question, but I've heard it often enough to give it the number one spot here. The answer of course is that it does work. But not always. I've seen more setups that did not work well, than setups that did work well.
Q. Why is the big compressor feeding the small compressor? That's backwards!
A. Nope, that is the correct and only way to do it. This goes for all multi-stage compressors, not just turbos. The key point to keep in mind is that the first, atmospheric compressor has only the pressure of the atmosphere to feed it. You are therefore always limited to the capability of the atmospheric compressor. The smaller secondary compressor is able to keep up because it is being fed compressed air, which lets us “cheat” a bit there. More on that later.
Q. Won't the small compressor be over spun to death?
A. No. Shaft speed sensors will confirm that speeds are perfectly normal, and in fact the small compressor will spin much more slowly than it would as a single being forced to work much harder. The key point to remember here is that turbos move a given volume of air. When they are rated in mass flow rather than volume flow, the engineers assume a standard set of inlet conditions to generate those mass flow numbers for the compressor maps. Each turbo manufacturer uses a different set of standard conditions, and those conditions can be found somewhere in their documentation.
When we put a larger turbo in front of the smaller one, we are simply messing with those inlet conditions. Instead of feeding the compressor air at atmospheric pressure, we've boosted it to, say, 15 psig. This roughly doubles the density (or weight for a given volume) of the air. A turbo that moves 50 lbs/min at atmospheric pressure will now move 100 lbs/min. In most cases it will be somewhat less, since temperature has likely been increased, which reduces density, but it illustrates the point for now. The small turbo can move it's rated mass flow times the Pressure Ratio of the large turbo, to get in the ball park without doing too much math. Use Density Ratio (which accounts for temperature as well as pressure) to get a closer figure. Similarly, to find your location on the small compressors map, divide actual mass flow by large turbo PR to get in the ball park (or DR for a more accurate figure).
Another way to look at this is to consider what happens to a turbo at high altitude. The air is less dense, so for a given pressure ratio mass airflow is down, compared to sea level. This is like going 10,000 feet below sea level, for the secondary compressor.
Q. Why doesn't everyone run compound turbos then?
A. It's definitely not for everyone. There are as many cons as there are pros, mainly in cost, weight, and complexity. If you like things cheap, light, and simple, this isn't for you. The people that tend to do well with compounds are those that like to learn, tinker, and build things with no guarantee of a successful outcome.
Q. What is the hardest part about compound turbos?
A. Boost control. It seems as though any good fabricator can successfully stick two turbos onto an engine and get it to make noise. But when it comes to boost control, balancing pressure ratios, achieving good spool and good back pressure, and tuning for it all within the limitations of the particular ECU in use, the majority of people begin to struggle. Expect to use wastegate springs, wastegate sizes, manual boost controllers, electronic boost controllers, and pressure regulators in combination to get the right effect. There is also a tendency for most or all of big turbine back pressure to be added to the small (recirculated) gate's springs. The thread on Yellow Bullet goes into many different control schemes, but I'll cover three in my examples below.
Q. Why not bypass the small turbo once the big turbo spools?
A. The short answer is that this can be done, or at least attempted, in many ways. You start getting into sequential operation though, which can get exceedingly complicated. I've yet to see a really successful setup that bypasses the small compressor. Bypassing the small turbine (at least most of it) is much more feasible. But despite the challenges, the majority of new compound setups being built seem to go for some kind of sequential strategy. In my experience, a traditional compound setup is much more likely to work, with less effort, and with most of the benefit.
Q. What about intercooling between turbos?
A. This is certainly of benefit, though not necessary. It mainly comes down to packaging. If nothing else, it spreads the cooling work over two intercoolers (assuming another is installed in the usual place). But there are gains to be found all around with intercooling. I've run one air-air IC after both turbos, air-water ICs between and after the turbos, and non-intercooled. I'll cover each briefly in the examples below. This is a vast topic, and can get pretty far into thermodynamics if you want it to.
Q. Won't the small turbine choke the exhaust flow and kill horsepower?
A. No, not if you have working wastegates. Any exhaust the small turbine does not need to make the boost you've asked of that turbo will be wastegated. Typically the small turbine's gates are rerouted back into the exhaust, so that exhaust energy that is gates goes on to drive the big turbine. If the wastegate(s) can't flow enough to bypass the small turbine, you will get boost creep, by definition. In other words, you'll know it. In these cases you simply need more wastegate on the small turbine. This is one thing that needs special attention, small turbine wastegate capacity. I prefer two 44mm gates to one 60mm gate, since the 60mm gates can be hard to keep shut when you start pushing the boost and back pressure. The small turbine is obviously a factor in total back pressure however, more on that below.
Q. What will back pressure be?
A. For me, this has proven very difficult (and very educational) to predict/calculate in advance with any real accuracy. But the short answer is, it will likely be higher than the big turbo would be on its own, but much less than any turbo that would spool anywhere near as quickly, even if it could make the power of the compound setup. It's a good compromise to make in many cases. Someday I'll get all the math right on this.
In short, big turbine drive pressure will be multiplied by small turbine expansion ratio (or pressure ratio, in reverse). So any increase in big turbine back pressure will be multiplied by the small turbine. And any increase in small turbine back pressure will multiply the big turbine back pressure by a larger multiplier. So it still pays to keep back pressure low at each turbo. Run the biggest turbines you can spool to your liking, same as with any single turbo setup. It's hard to go too big on big turbine size. If you get into trying to bypass the small turbo you can attempt to get the best of both worlds, at your own risk.
One quick point I would make on back pressure, is that while lowering back pressure is always a worthy cause, it just doesn't have the huge effect people usually expect it to. Even large reductions in back pressure seem to produce only modest improvements in power. I will gladly trade a little more back pressure for a setup that spools quick and is easier to race.
Q. What is some of the math involved in working with compound turbos?
A. There can be an awful lot of it, if you want to get into it. For the basics, most people will do well to start with figuring out how Pressure Ratios are calculated. PR is simply a compressor's absolute outlet pressure divided by its absolute inlet pressure. To use a common example, two turbos at 2:1 PR. I use 15 psi for sea level for easy math, adjust accordingly for your area.
- The 2:1 PR multiply together (not add) to 4:1, which is 4 times atmospheric pressure, which is ~45 psig at sea level, or 60 psia.
- The primary compressor at 2:1 will put out 15 psig, or 30 psia, or two times atmo pressure.
- The secondary compressor at 2:1 will multiply that 30 psia times 2, for an outlet pressure of 60 psia, or 45 psig (60 psia minus 15 psi atmo pressure).
- Each turbo “feels” like it is making 15 psig, with a PR of 2:1. Neither one is struggling to make that 45 psig boost.
- If you want to split a given PR evenly between two turbos, just find its square root.
- The same PR math works with ER on the turbine side.
- Approximate position on the small compressor's map is total airflow divided by big turbo PR (or DR for greater accuracy).
- Big turbo can be plotted on its compressor map as usual.
This setup I built in the '08-'09 off season. 2008 was my first year racing this car. It went 8.96 at 156 on a single GT4294 at ~45 psi. Intrigued by some other DSMers talking about compound turbos, I decided to build something and see what would happen. What could possibly go wrong!
It used an old school T3 50 trim with a "stage 3" turbine in a .82 housing (~49 lb/min capable), and an inexpensive off the shelf 75mm Borg Warner S475 with the 96mm turbine wheel in a T6 1.32 housing (~105 lbs/min). It was still slow to spool on gasoline, but even the 4294 was completely impossible to spool without nitrous, so this was an improvement. The weight added by the second turbo and extra plumbing was completely offset by removing the nitrous. It used a single cheap air-air IC after the second turbo.
Initial testing with a 44mm WG for each turbo resulted in extreme overboost. Adding a second V44 to the manifold to further bypass the small turbine took care of that. At the same boost as the single turbo, it ran the same ET and MPH. With no more nitrous to deal with, it was a win. Back pressure was 1:1 with boost up to 8k rpm, and then it just passed boost slightly at 9k. As it turns out, at the low PR I ran on the S475 (about 2:1), it was only good for about 95 lbs/min. This is one thing to keep in mind when sizing compressors. It was made very inexpensively with mild steel materials, and welded with a 110v flux core mig welder and just about zero skill. But, while it was just meant to test the concept, it lasted three seasons of racing. The best ET was 8.80.
I wasted a lot of time before realizing that the car was effectively launching on the 50 trim, since the 475 came in about 1 second down track. The ~1.35 sixty foots were all that could be expected. The spread between small compressor and large turbine is important. It also took some time to realize that the primary compressor won't reach its max airflow at such a low PR and does limit power somewhat. The 96mm turbine in the 1.32 T6 housing was at the limit of what will work on these motors. Dropping that down, and/or increasing the size of the small compressor would improve spool significantly. But the fact that it could spool that turbine at all with 2.0 liters (122 inches) shows what's possible here.
Boost control was simply 18 lb springs in the small gates (resulting in 28 psi boost contribution since big turbine back pressure is added to them, more or less). With big turbo boost routed to the top ports to bring small turbo boost up 1:1 with big turbo boost. The big turbo just ran an MBC on it's own outlet, which can be varied to easily adjust total boost up and down. It was set to around 17 psi, for 45 psi total.
I decided to retire that poor first setup after three seasons, but wasn't sure exactly what I wanted to do with the next compound setup, so I went back to single turbos for a couple years. While trying out a stock EVO3 16g turbo for a bit, I learned a lot more about turbos. Being compressor limited really can make you think differently about where to find more power. After running a 10.3 on that turbo, and 8.89 on a cast wheel 3582, 8.23 on a precision 6766 (with nitrous down track), and finally 7.90s at up to 180 on both a Forced Performance Super 99 HTA and a Forced Performance 4505HTA, I decided it was time to revisit compound turbos and try to get them into the 7s (this was finally achieved in October of 2017).
This setup has seen two iterations. Non-intercooled methanol with lots of upstream injection, and air-water intercoolers for each turbo.
The non-intercooled setup was a huge science project. I read some text books and learned enough thermodynamics to get me by. With a massive spreadsheet (affectionately named "the spreadsheet of everything") I was able to test hundreds of possible scenarios, for different turbos and every possible combination of upstream injection options and intercooling. I narrowed it down to a couple worth testing. The final version was a Forced Performance 3582HTA coupled with the existing Forced Performance 4505HTA. There was 1 Fuel Injector Clinic high Z 1650 per runner in the normal port location, with an additional 4 1650s upstream. Two in the large turbo compressor cover, and two in the pipe between compressors. At the fuel pressure in operation, that was 8-10,000 cc/min of straight methanol flow through the compressors. Well, it all went through the small compressor, but the large turbo had it added in the housing for packaging reasons (compressor inlet is above the hood). The air is compressed (and therefore heated) in the cover anyway, so I expect it still provides most of the benefit. Some observations:
- The upstream injection worked very well. It turns out there are some limits to how much methanol you can vaporize under pressure, but at 55-60 psi boost charge temps were in the 225-250 range as predicted based on the properties of methanol. Compressor outlet temps are calculated to be over 500 degrees, so that's a nice drop. The goal was to stay within the ~300F limit of the air temp sensor, since it is used for speed density airflow calculations.
- The 4th injector was dropped in favor of a 2200cc fixed nozzle fed from the same mechanical pump and fuel system as the other 7 injectors, due to packaging constraints. Tuning for this on the stock ECU (with software from ECMtuning) was an extensive math lesson for me. There is no easy way to do it, but it can be made to work.
- If charge temps remain above the cooling capability of methanol at a given pressure, additional cooling will occur at the port injectors, which is not seen by the IAT sensor. This will show up as an increase in VE when tuning.
- If charge temps reach the lowest they can for the current boost pressure, any additional injection will not vaporize and will not drop temps any further. It will proceed on to the cylinder and vaporize/provide cooling on compression. In my case a little but of the upstream injection and all of the port injection falls into this category.
- This setup was super light, and very compact.
- It also spooled insanely quick. 0-60 psi on pedaling a drag run would occur in as little as 1.2 seconds.
- The car easily went 8.3 at 160ish mph this way, but struggled to go any faster despite making more HP with more boost. In hind sight, this was a torque converter problem. The potential of the setup was never reached.
- On this setup I tried a new WG arrangement. Rather than reroute two V44s to the large turbine, and run a third V44 for the big turbine, I ran only the two V44s on the manifold collector. One is rerouted to the big turbine, and controls the small turbo. The other is dumped to atmosphere and controls the big turbo. The idea is that if the big turbine doesn't need that exhaust energy, the small one doesn't either. It works very well and saved me another gate, more plumbing, and more fab work. At some point however (over 60 psi boost in this case) back pressure starts to become a problem for the gate that dumps to atmosphere, since it sees the full pressure differential across it. With that large differential the gate has poor resolution at low lift and causes large shifts in back pressure as it cycles near closed.
- There is an inherent risk with this much upstream injection. As one example, at .50 lambda, and 50% of your fueling is upstream, AFR in the charge pipes is stoich. With all of the compressor housing and piping volume, this is a pipe bomb. Unfortunately, an engine failure did let the fire out on a dyno pull and sent the whole setup into near earth orbit, ending this experiment. I don't have one single regret. I learned more from this process than just about anything else I have ever tried.
At this point it was time to rebuild, and it was only a few weeks before the 2016 DSM/EVO/GTR Shootout. RAD Auto Machine, FFWD connection, Race Ready Fab, and many of my other supporters and friends came through for me by getting me another engine built in time, and building a whole new setup for the car. This time around I wanted to try an idea I had been trying to figure out exactly how to accomplish for years: two air-water intercoolers. Separate air paths for each compressor, but a common ice water tank built onto the cores. Packaging was a huge nightmare, but the end result worked incredibly well. Same two turbos (after Forced Performance repaired the poor 3582). I moved the 4 upstream injectors to a secondary Magnus Motorsports fuel rail on the manifold.
- The intercoolers worked reasonably well. Charge temps were in the 130F range at 60 psi. A number of deficiencies were discovered and will be addressed for this 2017 season (see last bullet). Still, that was better than the air-air I previously had on it. There is one bosch lightning pump pulling water through one core and into the other core, just to circulate the water from the tank through the cores. In hindsight, like any centrifugal pump, it hates the inlet restriction. Water flow (and current draw) are very low.
- The car is significantly heavier, nearly 100 lbs. At least it's on the nose.
- It still spools incredibly quick, same as the previous setup, 0-60 psi in 1.2 seconds on respool. Idle to 60 psi on the launch 2 step takes 3-4 seconds.
- HP per boost is higher naturally, due to the lower charge temps and proportionally higher charge density.
- Interestingly, ET remained stuck at 8.3, despite making up to 400 hp more on the dyno. This was still a converter problem.
- Back on pump E85 with the intercooler. The fuel system is now very much overkill, which means it works perfectly.
- Back pressure is approximately 1.5:1 at 60 psi. The small turbine housing is still a .63 T31 housing. This setup spools quickly enough, I can easily go to a larger one at some point.
- Small turbo shaft speed reaches 75k rpm early on, then starts to drop as the large turbo spools. At full boost the large turbo is running 90,000 rpm and the small turbo is cruising comfortably at a very low 25,000 rpm.
- For 2017 I intend to improve the IC setup, the converter problem, and try a new boost control scheme using the same WG arrangement to drop back pressure a little more.
- Update: The IC system was indeed improved. There are two Bosch lightning pumps. Each pulls from the bottom of the ice water tank and pumps water into each of the two end tanks. One pump per core with no inlet restrictions. This dropped charge temps a further 50 degrees to approximately 80F at 60 psi boost. The new boost control scheme was not successful at higher boost, neither was the new WG arrangement. I'll be going back to the old setup from 2009 to run higher boost. The new converter helped but still needs adjustment. The car did run 7.90 at 173 with sixty foots as low as 1.20 in marginal conditions. There's still a lot more improvement to be found, in many areas.
This setup was sized to spool up on a very tight converter and make about 1000 hp. It's a Forced Performance 71HTA (68HTA compressor, ~51 lbs/min) paired with a Forced Performance 4205HTA (76mm compressor, 115 lbs/min). There is a single Treadstone air-air intercooler after both compressors. It uses one recirculated Tial 60mm gate for the small turbo with excellent priority, and a single 44mm PTE gate dumped to atmo for the large turbo, in the usual arrangement. The main issue here is keeping that large 60mm gate under control. It blows open incredibly easy, and limited us to 40 psi boost for a couple years.
- It makes that 40 psi at 3500 rpm on the 2 step, and launched that way (AWD Auto) it produces 1.35 second sixty foot times.
- It's best time slip to date is a leisurely 9.3 at 148.
- We've run this setup on E85 primarily, but at altitude (DSM Shootout) we've run it on methanol to get it to spool.
- With a very tight Bradco converter (2300 rpm at 0 psi), it's slow to spool, but it does. Once it reaches about 10 psi, it shoots straight to 40 in 1-2 seconds.
- I've since changed the boost control scheme. Rather than fight the 60mm gate to keep it shut using compressed air to the top port, and so on, I decided to just use big turbo boost or back pressure to blow it open and be done with it. It was not this simple, but it did eventually work. On the dyno I saw 60 psi (5 bar map sensor limit) at ~4000 rpm, by accident of course. We've yet to test this on the track.
- Update: Boost control continues to be a struggle.