Excellent report on bellmouth design for throttle bodies

myhui

myhui

Experienced Member
Joined
9 September 2006
Messages
928
Location
Orange County, CA, USA
Thank you very much for posting that.

It makes sense that a large entry diameter to a bellmouth (the flaring portion of an intake pipe) is good, but I wouldn’t have expected that the efficiency decreases when you make the bellmouth portion of the intake too long. So an intake bellmouth should have an elliptical shape (not a simple radius), the length of the bellmouth should equal its exit diameter, the entry diameter should be 2.13 times the exit diameter, and once the entry to the bellmouth flares out beyond 90°, it should start turning backward on itself a bit with a corner radius 0.08 times the entry diameter. Good to know!
 
Very good info - thanks for sharing.

I wish every design aspect had this level of data to back it up. I looked at some of the references and saw that there exists a UK publication called Race Engine Technology and that some of the referenced articles were about "Airbox Design." I may have to pick up some of those back issues...

Thanks again for sharing.
 
Here are two related questions and some answers to them.

1) Optimal places to inject fuel depends on RPM. An entry in Jenvey induction system's FAQ deals with that. You want secondary injectors at the entry to the ITB for high rpm operation.

2) Volumetric efficiency is best when ITB length is proportional to engine rpm. Lots of references are available on this, as you want the same number of pressure waves inside the ITB across a wide range of rpm. The waves are closer together at higher rpm. Fabricating a variable length trumpet (trombone?) is not too hard. But quickly varying its length upon computer control is tricky to do reliably.
 
The bellmouth article is freely available for download from the author’s website, but just in case he takes it down, it’s on NSXPrime now as well: http://www.nsxprime.com/w/images/9/..._Design_of_an_intake_bellmouth_Sept._2006.pdf

One picture in the article shows the shape of an optimal bellmouth, on which I’ve noted his design specifications:

Race_Engine_Technology_Optimal_bellmouth_shape_2006.jpg
 
Regarding ITB length and fuel injector placement, here is what it looks like inside the airbox of Ferrari’s 2008 Formula 1 engine. It operates at a vastly different rpm than an NSX engine, but it’s interesting to look at nonetheless.

 
The Mazda 787B had a really good variable intake system.

26b_3.jpg



Here's a good look at a F1 Fezza variable intake w/ exposed telescoping airhorns/bells that lengthen the runners overall:

51409549.jpg
 
Last edited:
greenberet said:
Regarding ITB length and fuel injector placement, here is what it looks like inside the airbox of Ferrari’s 2008 Formula 1 engine. It operates at a vastly different rpm than an NSX engine, but it’s interesting to look at nonetheless.

Click to expand...

This is interesting. The article mentions that rectangular shaped trumpets are the least efficient in their tests and the photo of the intake above is either rectangular or square.

Fezza must know something or were limited by design formula...
 
Yeah, the article states that, ““As it is rather difficult to design a rectangular profiled bellmouth, it will inevitably have a rectangular entry, the general conclusion must be that rectangular intake ducts and rectangular intake bellmouths should be avoided by design if at all possible.”

Looking at the space limitations inside the Ferrari airbox, I guess they had the choice between larger rectangular or smaller circular bellmouths. Given an equal cross section, round bellmouths surely let the air flow more efficiently. But if you need a larger cross section and the only way to get that is by moving away from a circular shape and going more towards a rectangle, maybe the tradeoff is worth it. Since Ferrari won the 2008 Formula 1 World Championship with that engine, it couldn’t have been a very bad choice. Also, since the exits of those F1 bellmouths are oval, maybe circular wouldn’t be the best shape for the entrances even theoretically.
 
When designing a high-performance engine, it's not just about optimizing each component separately and then slapping them all together.

They all have to work together harmoniously, and this involves countless time, iterations, and money to optimize the entire engine given your goals and design constraints. Often, this leads to different configurations than you would have believed with simple textbook theories.

For example, ITB bellmouth shapes should be optimized based on airbox geometry, head geometry, intake valve number and layout, etc. I'm sure Ferrari optimized that bellmouth based off of CFD modelling of their entire cylinder, head, and intake valve designs, then did a bunch of dyno tests with various ITB geometries to validate the modelling.

Also, on variable-length runners, is that really worth a small gain in a narrow RPM range to justify the added cost and weight (let alone if the system even keeps up with a quickly-accelerating engine) of it?

I mean, you're not going to really notice this on normal street driving, and then when you want to drive spiritedly, you're usually above 5 or 6k RPM's anyways. Just shoot for an optimum runner length at 7000RPM's and call it a day :smile: Sure, on an F1 engine with a greater RPM operating band, but ours?

My $0.02.

Dave
 
Last edited:
I plan to use an engine dyno to develop this, running at constant rpm, while I tune the itb length, injector timing and duration (2 injectors for each cylinder), with throttle plate angle being the independent variable. Exhaust mixture for each cylinder will be measured. The goal is to maximize engine torque.

Do this over a range of rpms and throttle plate angles, and use the ECU to interpolate over the unmapped points, plus add a few magic parameters to anticipate things before they occur.

I'll have a machine shop make a hand-turned ITB trombone in aluminum. The weight will be less than 24 lbs for all six of them. A stepper motor will vary its length, and a linear potentiometer will sense its actual length. Total weight is still under 30 lbs, and this includes feedback to the ECU to continuously check for proper operation.

At WOT I think it'll take about 5 seconds to rev from 3k to 8k rpm, when under max load, so moving the trombone over maybe 3 cm in 5 seconds is not too hard.

But is all that worth it? I'll get a more powerful engine compared to not doing that, so yes, I think this is worthwhile to do.
 
Cool!

It looks like you've already done some research, but it appears as if the electrical stepper motor ideas haven't worked too well in the real world (heat, slow speed, vibration). I thought the more robust was a vacuum-operated type, which would only give you basically two runner lengths when you have some good vacuum at low engine speeds and then none at WOT.

But then you're back to doing a lot of work for little gain low in the RPM band.

Anyways, good luck!

Dave
 
Mac Attack said:
Also, on variable-length runners, is that really worth a small gain in a narrow RPM range to justify the added cost and weight (let alone if the system even keeps up with a quickly-accelerating engine) of it?
Click to expand...

The whole idea of variable-length runners is that the intake runner change length to optimize hp torque at any given RPM that the engine can handle. So there is no "narrow RPM range." A longer intake increases airflow speed creating better turbulance which aids combustion which then promotes low RPM torque. A shorter intake at high RPM will let the engine ingest a greater amount of air which helps with high RPM power. And this isn't even bringing up tuning the variable intake for Helmholtz resonances. Even fixed runner length ITBs, which your fits description in operating best in certain RPM range (ie 5-8k in our example), can be tuned to acheive 120% volumetric efficiency. Cosworth did that in the 90's.

Mac Attack said:
I mean, you're not going to really notice this on normal street driving...
Click to expand...

I'm not so sure. The NSX already has a type of variable-length runner system, though nothing like the sophistication on the 787B or F1 engines, is effective. And we notice low RPM tractability b/c of another variable engine system (VTEC) - and definitely notice it at high RPM. The benefits of a variable-length runner system could produce similar results in driveability.
 
myhui said:
I'll have a machine shop make a hand-turned ITB trombone in aluminum. The weight will be less than 24 lbs for all six of them. A stepper motor will vary its length, and a linear potentiometer will sense its actual length. Total weight is still under 30 lbs, and this includes feedback to the ECU to continuously check for proper operation.
Click to expand...

The airhorns, even variable length versions should be much lighter than 6lbs for each. The fixed airhorns that I've handled were very light (less than a pound) and you'd probably use thinner walled airhorns in your variable system. The most significant weight would likely be in your controlling mechanism.

To lighten it even further, you can also go carbon as well. Though costs would be exponentially more, I'm sure.
 
Last edited:
Ponyboy said:
The whole idea of variable-length runners is that the intake runner change length to optimize hp torque at any given RPM that the engine can handle. So there is no "narrow RPM range."
Click to expand...

You misunderstood my point.

Our usable NSX powerband extends from around 3000-8000RPM (or a 5k RPM spread) on a 3.0 or 3.2L N/A engine. Compared to an F1 engine with usable RPM powerband ranges of double that, then yes, our RPM band is pretty narrow and will see less benefit of a variable-length design than they would.

I'm all for the idea of doing something new to our cars. But, this is certainly not a cheap proposition to implement effectively and reliably, making it a low "bang-for-the-buck" mod.

Again, just my $0.02. :smile:

Dave



Ponyboy said:
The NSX already has a type of variable-length runner system, though nothing like the sophistication on the 787B or F1 engines, is effective.
Click to expand...

Really? The effectiveness of our variable volume/length system is debatable, as shown here over the years. It produces negligible gains in torque/power before 4800RPM, and is then a flow hinderence from that RPM on.
 
Last edited:
Mac Attack said:
Our usable NSX powerband extends from around 3000-8000RPM (or a 5k RPM spread) on a 3.0 or 3.2L N/A engine. Compared to an F1 engine with usable RPM powerband ranges of double that, then yes, our RPM band is pretty narrow and will see less benefit of a variable-length design than they would.
Click to expand...

One of the best parts about the NSX is it's flat torque curve with about 88% of it's torque being available in about 73% of the engine's RPM range. F1 engines, OTOH, are tremendously peaky. F1 engines actually have a fairly narrow powerband b/c of design restrictions. They make up for it with high RPMs and gearing to keep the engine at the ideal RPM when the driver shifts.

Mac Attack said:
I'm all for the idea of doing something new to our cars. But, this is certainly not a cheap proposition to implement effectively and reliably, making it a low "bang-for-the-buck" mod.
Click to expand...

Of this, we can agree! It is not inexpensive and specific results aren't guaranteed. But it can be a fun project and fortune favors the pioneers.

Mac Attack said:
Really? The effectiveness of our variable volume/length system is debatable, as shown here over the years. It produces negligible gains in torque/power before 4800RPM, and is then a flow hinderence from that RPM on.
Click to expand...

Do you know of a dyno with a stock car w/o the VVIS system in place to compare it with? I know the 2 stage variable intake system doesn't work well w/ FI but I know of no stock NSX dyno w/o the VVIS to make a claim that it doesn't work. It made appreciable power down low (8hp max below 4500rpm) w/ my NA I/H/E though the dynos were done at different times when I baselined and then removed the VVIS.
 
Last edited:
You can talk about our torque curve, but real-life shows this to be a pretty gutless motor until you get some decent power around 5k RPM's. It's still a small-displacement N/A engine, and instead of N/A rights bragging, you can do some forced induction for much better return on your investment. But, some people like .... so to each their own.


The point is, if you "punch it" in second gear around 4000RPM's, there is just a little bit of time before you hit redline, shift to third, and then the RPM's drop by 2000. As you would really only need a variable-length induction system for ultimate performance and not everyday cruising, you have a limited RPM band that this would be useful for.

The same resonant and induction length tuning theories pretty much apply for N/A or F/I, with changes to account for air density, etc. F/I before/after VVIS deletion dynos are applicable to N/A.

It's always amazing to me that Honda spent a lot of time and money to design and manufacture our VVIS system for a peak gain of 5-8HP below 4800RPM (and sacrifice a little on the top end in exchange), yet the C30 engines came with cast-iron manifolds. In engineering, there is no perfect design. Everything is a tradeoff.

Dave
 
Remember the dyno sheet FastraxTurbo posted up in this thread comparing an n/a NSX with the stock VVIS plate, a gutted VVIS plate, and a removed VVIS plate?

The car showed a:
- 3.5 horsepower gain at 7800 rpm with a gutted VVIS plate or without a VVIS plate
- 11.5 horsepower loss at 4300 rpm without the VVIS
 
Yeah, over an extremely narrow 1500RPM range, the VVIS added a PEAK of ~10HP.

My opinion on post 31 of that thread still stands: It's amazing the time and money spent on that part for such a small gain in performance (in the grand scheme of things).

Dave
 
IMHO, when it comes to nsx, in most cases you cannot 'out-engineer' the factory. too much care and development went into it- they seem to have found that elusive synergy in performance.
 
If you have the right tools and the right engineering knowledge to use those tools, then maybe you can improve the NSX motor's performance.

The FLUENT CFD software for automotive use can be purchased and set up with the ITB's geometries, and the NSX engine's geometries can be obtained eitiher via measurement or by pleading with Honda to release their CAD files.

The best outcome might be a "sculpted" ITB that requires a 3D laser printer to realize, somewhat like how "sculpted" the Red Bull 2010 F1 car is on the outside.

Here is what the software can do, according to http://www.fluent.com/solutions/automotive/auto9.htm

FLUENT 6 provides leading-edge functionality with a robust and intuitive, yet powerful, unstructured framework for modeling all of the critical components in internal combustion engine design including fixed-lift port flow analysis, fuel injection and detailed transient moving valve and piston simulations.

FLUENT’s usability raises the industry standard for modeling complex problems that involve moving geometry, such as IC engines or fuel injectors. Our innovative “dynamic-mesh” model requires the user only to set-up the initial mesh and specify the motion of the boundaries and FLUENT automatically changes the mesh to follow the motion as the transient simulation proceeds. FLUENT’s dynamic-mesh model is compatible with our expansive suite of combustion and spray models.

FLUENT’s robust cavitation model can resolve highly cavitating flows even for very high pressure conditions such as those that occur in high-pressure diesel injectors.​

I am just an electronics engineer by training and profession, not a mechanical engineer or aerospace engineer. But I know of some NSX owners who are aerospace engineers, so maybe they can lend a hand. The hardest part is to write the software, and I have no idea how to do that, so I buy it from ANSYS, Inc. instead. The second hardest part is knowing how to optimise the design, and I might not know how to do that either, at least not to the level a trained professional would be able to do.
 
myhui said:
If you have the right tools and the right engineering knowledge to use those tools, then maybe you can improve the NSX motor's performance.
Click to expand...

I found some general formulas to use in determining runner lengths that may help you:

SQRT [ (target rpm for peak torque x Displacement x VE)/ 3330 ]

SQRT = square root

VE = Volumetric Efficiency in %

Displacement in Liters

eg.

So if we want peak torque at 5800 rpm at 95% VE, VE = 0.95

SQRT [ (5800x 1.8 L x 0.95)/3330]

= 1.73 in. or 43.8 mm (1,73 x 25.4 mm/in.) is the ideal runner diameter.

And/or tune by Helmholtz resonances:

where f = the rpm at which you get peak torque ( the natural frequency of pressure oscillations in the acoustic chamber ) , c = the speed of sound (= 340 m/sec.) , S = runner area, L = runner length, V = displacement per cylinder

RPM for peak torque =

642 x c x [ SQRT (S/[L x V] ) ] x [ SQRT { (CR-1)/ (CR+1) } ]


= 218,280 x [ SQRT (S/[L x V] ) ] x [ SQRT { (CR-1)/ (CR+1) } ]

Or just screw the formula and use this website. :)

In any event, I wish you much success in your project and look forward to hearing about how it goes!
 
myhui said:
I am just an electronics engineer by training and profession, not a mechanical engineer or aerospace engineer. But I know of some NSX owners who are aerospace engineers, so maybe they can lend a hand. The hardest part is to write the software, and I have no idea how to do that, so I buy it from ANSYS, Inc. instead. The second hardest part is knowing how to optimise the design, and I might not know how to do that either, at least not to the level a trained professional would be able to do.
Click to expand...

Have you done a patent search on this for some more ideas? There might be some good stuff out there. I know there are for a similar idea - variable-length tuned exhaust systems....

Or, you may try to contact Jenvey to see if they've experimented with this idea. If so and were unsuccessful due to the electronics portion, then there you go :wink: You could work together to make a cool system and probably a few bucks too.

Not to sound even more discouraging on the CFD side, but my first job out of school was working with a group of PhD's at a defense contractor on anything CFD-related the government needed help with. My group wrote their own RANS solver :cool: and the government provided multple supercomputer parallel processors around the country to solve our work. The most important thing to know about CFD solvers is: Garbage in = Garbage out. It takes a lot of experience to know what to model, how fine of detail to model it (the mesh), and then if your results seem reasonable.

FLUENT, STARCD, etc are OK for most commercial apps, but even then, require a lot of training to use.

Dave
 
Mac Attack said:
FLUENT, STARCD, etc are OK for most commercial apps, but even then, require a lot of training to use.
Click to expand...
I can relate to that totally. It's like having a beautifully designed analog amplifier circuit simulating quite well in SPICE. Then you're asked to optimize its performance in a specific way. Only the very skilled will know what to adjust (change component values, change connection, add/remove components) to realize that goal. It's a bit like how F1 is right now. Every team can afford the very best CFD software. But look at how few of them know what to adjust on the car to make it go faster.
 

Similar threads

D
Grade 9 student murder, WITNESS ACCOUNT, MEDIA REPORT translated.
Replies
0
Views
2K
downwiz2
D
Back
Top