ITBs + Turbo with Speed Density ECU 2

[wgknestrick]

I am in the process of designing/fitting a set of roller barrel ITBs (Individual Throttle Bodies) to my Subaru STI and was looking for some insight on some things. This is something that is underway and going to happen so please don't try to talk me out of it, or say it won't make any HP, etc. That is not the point here.

My question is reguarding tuning of ITB systems. Most NA guys switch to an Alpha N based fuel map (meaning load is in terms of throttle position, TPS) instead of MAP/speed density/MAF. This is because of the near ATM MAP (very non linear) values read from the MAP sensor when the throttle is opened. MAP is fine (but lower) on idle, then pegs to around 0psi (ATM) for the upper 75% of throttle. This is because you now have 4x (or # of cyl)of throttle opening leaking the vacuum away.

Alpha N works great for NA apps to offset this, but I am running a turbo, so I NEED speed density to account for boost. The one thing helping me is that I would be in the upper PSI levels at WOT, unlike a NA app where they are in a flat part (0psi) of the fuel map. I have good resolution in the high load sections unlike NA guys, but I still have this window (where most driving occurs sadly) that will be difficult to map fuel around 0psi

I've read of Miata/Honda guys making a vacuum manifold or accumulator from each port, then tapping the MAP into that. The theory is that the accumulator dampens out some of the noise from each port. This is trying to replicate the a normal plenum that has one throttle on it. The problem is that I don't want to make this too big because it affects my throttled volume (main reason for the ITBs in the first place).

ITBs are used with turbo systems in 2 OEM applications, RB26DETT and the SR20 GTiR engine, both Nissan motors. They both however run MAF based loading systems. There are tons of guys that modifiy these engines and I would guess someone has made speed density work with the OEM ITBs. Please note my stand-alone ECU cannot run MAF loading.

So about what volume do I need in this vacuum accumulator for a 2.5l engine, and what ID ports should I connect to this? I was leaning towards a port ID of .7mm, but if anyone here had suggestions it would help a great deal.

Here are some Solidworks pics of my design:

http://users.rcn.com/wgknestrick/itb transparent.jpg

http://users.rcn.com/wgknestrick/itb section.jpg

http://users.rcn.com/wgknestrick/TURBO ass'y.jpg

http://users.rcn.com/wgknestrick/TURBO ass'y2.jpg

Vac accumulator would be connected to the hose barbs from each cylinder port.

 

[oculi2]

just a quick note to add, nearly everyone who modifies the RB26DETT or the GTiR SR20DET gets rid of the ITBs and uses a large single throttlebody for simplicity and ease of tuning.

why not make a single big barrel and use it in roughly the stock location?

AFAIK multi throttles on a turbo engine are not a good value mod.

good luck!

 

[wgknestrick]

I appreciate the info, but that really isn't the case here. I designed them myself, and am getting the machining in a barter trade. This is something that I want to do to challenge myself and because I really like these types of things.

I'd imagine that people generally take the path of least resistance, but that path is not for me.

There is a reason that Nissan put ITBs on their halo cars to begin with. There is also a reason that the WRC banned them on their cars. Neither of these reasons were because they are too easy to tune / etc.

All I am looking for is greatly increased throttle response along with more mid-range/top end HP from my car.

BTW the engine management I am using is the Hydra Nemesis

Does anyone have experience with vacuum accumulators for Hondas or Miatas?

 

[wgknestrick]

http://users.rcn.com/wgknestrick/TURBO ass'y2.jpg

http://users.rcn.com/wgknestrick/TURBO ass'y.jpg

More pics of the setup because 2 of the top links were not working.

 

[wgknestrick]

It is chopping off the last part of the link. You just have to copy this into your browser. Good thing we can edit our posts (rolls eyes).

 

[BrianPetersen]

Can you use alpha-N but use the intake manifold pressure as a multiplying factor?

I know people have put turbochargers on fuel-injected motorcycle engines, and those all have individual throttle bodies and no MAF sensors, and folks who turbo them aren't always using stand-alone controllers. The OEM system on Suzuki and Kawasaki motorcycle engines has a throttle position sensor, an intake manifold pressure sensor, a barometric pressure sensor, and an intake temperature sensor. The intake manifold pressure is a single sensor for all four cylinders and measures the average to minimize the fluctuation. There is a hose barb on each throttle body and rather ordinary looking vacuum hoses join through T-fittings to the single intake manifold pressure sensor. Nothing special, and it doesn't take up much space.

I don't know the contents of the ECU programming, but the shop manual says that it uses pressure/RPM at light engine load and throttle-position/RPM at heavy engine load, corrected by barometric pressure and temperature. There is no discernible change-over, and if one of the sensors fails then it switches over to the other algorithm as a fail-safe, so I suspect that both strategies are fully mapped throughout the speed/load range. It probably applies some kind of "weighting" factor i.e. it takes, say, 90% of the outcome of the pressure-based map and 10% of the outcome of the TPS-based map at zero throttle, and the other way round at full throttle, and blends them at part throttle, this would eliminate any "steps" due to change-over between map strategies while using each strategy in its most accurate range.

 

[wgknestrick]

From what info I was able to gather, It seems as most (meaning the 1-2) people that have run a turbo with ITBs use what is termed Alpha-N - MAP blended maps on Motec ECUs. This is kind of what you were aluding to in your post.

This is great and all, but the Hydra ECU can only use Alpha N or Speed Density, not a function of both. I also must keep this ECU as I don't have a local tuner or $$ for much else(Motec). I do have a compensation fuel table for TPS/RPM. There are user defined tables allowed, but I don't know if I can have them interact with other parts of the table. (I don't claim to be a EFI/Hydra/tuning expert) You can get them to run external signals and ect, but getting them to interact with base fuel table may not be possible.

The motorcycle area is a great place to look for info as there are some very crazy people putting turbos on bikes now and most of those have ITBs. Thanks for that lead.

I was hoping that this could just have some sort of "workaround" using a MAP sensor (always best for turbo) for the base loading and then TPS fuel trims versus RPM. I think there are still driving situations that aren't accounted for this way though.

 

[mattsooty]

wgknestrick

What you are attempting is pretty ambitious and I admire you for attempting something so difficult......However what you are suggesting sounds like a recipe for disaster; if you do not have near perfect knowledge of the amount of air that is in each cylinder at any one time then you risk very severe engine damage. My worry would be that the noise in the 'MAP' signal would be sufficent to make your base spark table nonsense and impossible to calibrate. Also what will the difference be between transient and steady state running with the set up you have???

Why dont you try and find an alternative challenge which is actually do-able and wont risk a hugely expensive engine bill at the end of it.

Honestly not wanting to pi$$ on your chips - but just dont want you to wreck your motor!

MS

 

[wgknestrick]

I appreciate your concern for myself and my engine, but this is not as much of an issue (with respec to the engine) as it is with NA ITBs. Maybe I am missing something here, but my engine will still have good load resolution at high torque areas of the maps.

To try to describe the behavior of the MAP sensor in my app (ITBs + turbo):

MAP sensor is fine at idle and very small throttle openings.

MAP sensor is fine once turbo is making +boost at any throttle opening. Vac leaks from ITBs are gone at this point. This means I still have great resolution at max torque/HP load cells which means minimum danger to the engine.

MAP sensor has plateau or lack of resolution at large (>25%) throttle openings and no boost, MAP readings near 0psi. This is because for every cylinder pulling vac, there are 3 that are leaking it due to the nature of ITBs and 4 throttles.

Once the turbo is building boost, this problem is gone. Problem is most normal driving would venture into the problem window with my setup.

I can rescale my MAP load scale on the fuel table to account for this (giving me more MAP resolution around near ATM vacuum levels) but I really need a TPS overlay with respect to fuel and RPM.

I was thinking about using some sort of flow regulated check vavle system to eliminate the 1vac:3leak throttle problems. I could have 2 check valves (facing opposite directions) per cylinder where 1 has a reducing orifice inside. This would act like a "leaky" check valve on each cylinder to allow the vacuum accumulator to keep the engine vac, but still respond to boost.

There are 2 ways to solve this problem, mechanical (fixing the vac leaks from the ITBs) or electrical (finding another load signal to map to under this condition).

 

[TurboXS]

Just so I have this clear. You are running 1 MAP sensor connected to all 4 throttle bodies. At low loads all 4 throttle bodies show vacuum. At high loads basically all throttle bodies show boost.

What does your Hydra log show for MAP? Is it fluctuating, are you moving around load sites at constant light load?

 

[wgknestrick]

Just to clear things up.

This is theoretical at this point. The ITBs have about half of the parts done in the CNC. My car has not even started (or could've at this point) with this setup on it. I am trying to solve problems before I get to them, like an engineer should.

This problem has been documented by others who did the same (ITBs + Turbo) on multiple engine platforms.

 

[TurboXS]

I haven't had experience with multiple throttle bodies but I have designed engine management hardware and software.

I would expect the main problem is that a MAP sensor located at the throttle body is so close to the cylinder that it gives very high flucuations in its readings. This is because it is seeing the pumping action of the cylinder. Normally this would be damped by the manifold and the location of the MAP sensor. Hence the use of an accumulator by other people. Even so you can normally see pressure flucuations on the MAP sensor and I normally software filter these so that the engine management doesn't jumped from load point to load point.

In you situation I guess I would try to run a MAP sensor per cylinder. I would sample the MAP sensor at a particularly part of the cycle, say at BDC of the intake cycle and use that as a load indicator for that cylinder. Additional filtering would probably be required as well. This would allow you to do individual fuelling on each cylinder and if the engine management was up to it, individual ignition timing.

Obviously the easiest way of doing all of this is a special software load for your Hydra. Try talking to Element Tuning as I believe they distribute the Hydra in the USA.

Failing that an electronic solution could be made to sample and filter the MAP signals and send them to the Hydra.

 

[BrianPetersen]

The motorcycle systems solve the 1 vacuum + 3 leaking situation simply by using a single common MAP sensor connected to all four cylinders by ordinary vacuum hoses and T-fittings. No check valves, no accumulator (other than the hoses themselves), no multiple sensors. It is quite possible that it is taking a sample at a given crankshaft position in order to cut down on uncertainty. BUT, they only rely (predominantly) on that sensor at light load, they blend over to N-alpha at higher load.

The vacuum signal given by this type of arrangement is much lower (closer to atmospheric pressure) than you will get with one common throttle for all cylinders. For example, typical bike engine at idle pulls around 20 cm (8 inches) Hg on the common sensor.

One other thing to remember ... you have to synchronize the throttle movement very well between cylinders. If the linkage is designed right, you don't have to touch it much after the initial setting.

 

[SPOrotary]

Ok if you have the Hydra nemisis ECU you should be able to create a map that based from information from one cylinder can control 4 injectors (hydra should be able to control up to 8)

You can then run a simple "cheat" to get the fueling right when you are running >25% Throttle and 0 psi.

Basically in those conditions you only need to run one injector per cylinder meaning you have 4 to control in total. You can then take the vacuum line from only one throttle body (bypassing you vacuum leak problem) and make a map where 4 injection times are programmed from that one input. Basically you are telling the ECU that the info it gets is from the whole engine.

This is what my colleague did when he turboed his kawasaki.

I know you don't really want any comments on your system but I have to say:
-the intercooler has too much internal volume (unless there are internal baffles) and not enough surface area for sufficient cooling.
-Roller type throttles have shown to produce better WOT power and torque levels but how also shown to hamper performance at less than WOT conditions...

I would also like to say that I like your idea of having the single turbo feeding into the middle of the IC and feeding two air lines out of the ends.

Current project car: '98 BMW 316i
Main targets: 950kg, 250rwhp, min. 1G lateral grip. Special: Twincharger system, hydraulic ebrake, New suspension.

 

[wgknestrick]

SPOrotary:
Yes, there looks to be infinite room in the designs, but it is the largest IC that can fit in that location. It uses (2) 10" x 3.5" x 10"" cores. You always have to compromise when modding an engine. I chose to minimize tubing between the turbo and the heads over raw intercooler size. This isn't a drag car, but a hill-climber and needs good response. The intercooler is much bigger in person than what it appears on the designs 29"x10"x3.5" however

This whole piping setup is modeled after JUN's Hyper Lemon WRX shown here:

http://users.rcn.com/wgknestrick/wrxjunmotor.jpg

http://www.youtube.com/watch?v=KqD5L-tY4rk

For reference, that car has around 550-600hp out of a smaller IC (only 3" deep) so it "seems" to work

I have no idea what you are talking about with the Hydra ECU. Have you ever used it? There is only 1 injector per cylinder, of course I have to use it to fuel that cylinder. You cannot just connect 1 cylinder to a MAP sensor and use it to fuel the car unless there is software filtering to take out the noise in the signal. This is why people run the "T-ed" accumulator.

There are no "cheats" either: up, up, down, down, left, right, left, right, a, b, start will not make my car run.

 

[Majik]

god I know why it won't start. That's because your doing it wrong. it's up up down down left right left right b a b a start It's also select start if your bringing a passenger along.

I've never known hill climbers needing much more than just sheer power. You'll also find out that the hyper lemon overheats with that v mount setup. Well it's more a 20 minute demo car than a 1 hour race event winner. Don't forget the air needs a place to go once it's in the engine bay. I also don't think if you're planning a hill climb car that their IC dimensions will work with what your planning. Especially if it's a drift dirt hill climb instead of a tarmac one. Good luck on your project though. Again Mounting your radiator as low as the lemon you might need to bring some spares. One rock is gonna take that cooling system with it.

 

[wgknestrick]

I think the reason for the Hyper Lemon overheating (first I've ever heard of it) would be because they have such a small radiator in it and then mounted it at a 45 angle (the wrong way). It really makes no sense to me with all the time and development that went into the car.

I'm using a 3 row/copper radiator made for rock climbing Jeep Cherokees (only radiator core sized for lower bumper location at 36"x12"). This is mounted perpendicular to air flow (but would be prone to rocks) -thats why it's copper. A little solder and there you go.

Subarus have about the best cooling of any engine, ever (just split one in half to see why), as long as they have a decent radiator with any airflow. These aren't Miatas, or LSX engines here.

A small intercooler would limit power/boost and tend to knock at higher ambient temps-not your typical "overheating". I'm also using Meth inj which helps cool the intake charge, but I'm sure there will be problems I haven't thought of once I get it all together.

 

[SPOrotary]

wgknestrick:
I was referring to the solid works pictures of your IC. It is too simple not enough surface area and too much internal volume. As to it's size it looks to be big enough. The picture of JUN's Subaru shows how your intercooler should look. lots of relatively thin air passages (because you want to have the minimum of internal volume for good response)and lots of surface area due to all the fins added in between the air passages.

No I haven't used the Hydra ECU but I have some experience with ECU tuning, this is a relatively standard thing when the vacuum signal from one cylinder is corrupting the value. The ECU should have a signal filter doesn't it? At least the ECUs I have worked with have had a filter. By "cheat" I meant you could get the ECU to think it was getting signals from all cylinders when it is only getting a signal from one.

Current project car: '98 BMW 316i
Main targets: 950kg, 250rwhp, min. 1G lateral grip. Special: Twincharger system, hydraulic ebrake, New suspension.

 

[pmrobert]

Personal anecdote; I have a well running NA Mazda rotary w/ ITBs running solely speed-density. I have compensated for the erratic MAP signal with 2 approaches. First, teeing both ITBs together at the ITBs and utilizing the 10 feet of vac line necessary to get to the MAP sensor as an accumulator of sorts. This adds up to ~75cc of volume (in the vac line). Secondly I use a software averaging approach in the ECU code to further smooth the signal. Also, there is no TPS used; as previously noted the TPs signal is extremely nonlinear epecially at tipin and low load settings. I use MAPdot to trigger my accel enrichment. This approach did take some time and effort and a bunch of logging and analysis of logs to get to a point where a smooth, rapid processed MAP signal is obtained.

 

[wgknestrick]

pmrobert
Excellent info. That is my situation (ECU under passenger footwell), so that saves me a ton. If I need more volume, I can just add it.

I would like to thank everyone for the help.

Hydra has ways to enrichen sudden TPS changes too.

Bill