Begi Intake Manifold
#162
![Default](https://www.miataturbo.net/images/icons/icon1.gif)
Ill post up some pics here shortly but we are working on some 1.6L short runner manifolds and will be done with the first batch shortly. Orders can be placed via Welcome to Flipside
Gary
P.S. didn't mean to steal thread, started one https://www.miataturbo.net/forum/t27887/#post328318
Gary
P.S. didn't mean to steal thread, started one https://www.miataturbo.net/forum/t27887/#post328318
you haven't poste a price for people to order it.
#169
![Default](https://www.miataturbo.net/images/icons/icon1.gif)
Scott, hope you're still playing around with this stuff now that you have your own manifold. Have some ideas:
The improvement I see in this one is the lack of turbulence around the velocity stacks between their entrance level and the "floor" of the manifold.
What I see here is the stack gets the #1 inlet up to where it can begin to "reach" into the flow column coming from the throttle body.
The angle is different, so it's hard to tell, but it looks like when you moved the throttle body plate farther forward, you also moved it up away from #1 inlet, so the velocity stack is prevented from reaching into the flow again.
I think as a practical matter, the throttle body should be oriented parallel to wherever the throttle cable is going to come from so that the cable will have as straight a run as possible from its bracket to the t/b "wheel".
This looks the best for #1, but I think the throttle body angle is a little large. Note that the sharp transition between the inlet and the "roof" of the plenum causes separation at that point, building to a large swirl of turbulence over #3 and #4 runners.
What my comments have been building toward are the following suggestions. I'd love to see a sim if you care to continue geeking out with me. Start with something like your 3rd or 4th sim above, and make the following changes:
1. Keep the velocity stacks on the runners, but put them on the outside of the manifold, so the entrances are on the level of the "floor" and there's no turbulent swirl around them.
2. Put a velocity stack on the throttle body flange. I think it would be best to have the flange outside the manifold, as with the runners, to avoid the stagnant, turbulet areas behind the mouth. In general, this should more evenly distribute the intake charge into the volume of the plenum and reduce turbulence. Put some care into how the short-side radius meets the #1 inlet.
3. Play with the angle of the tapered "roof" and throttle body (which as above I think should be parallel) and throttle body fore-aft position, along with short-side radius to #1, until you become a god of Miata intake manifold design.
Then we will all
Thanks, I'm learning a bunch. This is way better than copying pictures on HondaTech.
What my comments have been building toward are the following suggestions. I'd love to see a sim if you care to continue geeking out with me. Start with something like your 3rd or 4th sim above, and make the following changes:
1. Keep the velocity stacks on the runners, but put them on the outside of the manifold, so the entrances are on the level of the "floor" and there's no turbulent swirl around them.
2. Put a velocity stack on the throttle body flange. I think it would be best to have the flange outside the manifold, as with the runners, to avoid the stagnant, turbulet areas behind the mouth. In general, this should more evenly distribute the intake charge into the volume of the plenum and reduce turbulence. Put some care into how the short-side radius meets the #1 inlet.
3. Play with the angle of the tapered "roof" and throttle body (which as above I think should be parallel) and throttle body fore-aft position, along with short-side radius to #1, until you become a god of Miata intake manifold design.
Then we will all
![bow down](https://www.miataturbo.net/images/smilies/bowdown.gif)
Thanks, I'm learning a bunch. This is way better than copying pictures on HondaTech.
#172
![Default](https://www.miataturbo.net/images/icons/icon1.gif)
Not on a personal computer. That kind of processing power won't be available for another 10-15 years in a personal computer. If done at high resolution these runs can take you well past 20 hours on a high performance processing center. A transient like that, could take weeks, if the ram is available to handle that information.
#173
![Default](https://www.miataturbo.net/images/icons/icon1.gif)
Not on a personal computer. That kind of processing power won't be available for another 10-15 years in a personal computer. If done at high resolution these runs can take you well past 20 hours on a high performance processing center. A transient like that, could take weeks, if the ram is available to handle that information.
#174
![Default](https://www.miataturbo.net/images/icons/icon1.gif)
A lot of it comes down to what you want to pickup. The general flow patterns are probably about right on. Where the difference shows up comes from the harder questions like how much flow, what temperature, what pressure, and small geometry changes like velocity stacks. This simulation probably should detect something like changing the throttle body orientation, but more than likely couldn't tell the difference between a 1 inch radius and a .125inch radius on a runner stack.
Another weird thing about these simulations is that you can run the same exact simulation twice and get 2 answers that are pretty different. It's like solving the solution for a x^6+x^5+x^4+x^3+x^2+x-5. A polynomial with more the 4 roots is not solvable, and that’s been proven. So the machine has to just basically guess to find the roots, and depending on where it guesses you get different closeness to the true answers.
An example would be:
If he runs the same problem twice maybe the machine guesses some good guesses on the first run, and comes out pretty close. If the machine on the second run picks some bad guesses then he would get quite a different answer. If he cranks up the resolution there are more guesses all total, and the computer can tell that where there used to be one cell pointing in a reasonable direction there are now 18 cells, and that previous vector makes no sense.
Also these simulations very much build like a snowball effect. Its what you would call an N+1 problem. If it takes one of the elements near the throttle body, and the machine just guesses it wrong, everything else behind that cell is very likely to be screwed up. A lot of times a truly good solution involves some advanced techniques; variable meshing density, turbulence setting manipulation, complex boundary conditions.
The simplicity of the interface of these programs is a little evil. When you first start using these you think you can solve world hunger. After a few years of using them, you realize if you can’t do an analysis like this by hand, then you probably shouldn’t touch it. I can’t do them by hand, but I’ve had some skilled hands give me a good enlightening on the subjects, and I’ve had around 5 years of experience in using this. Its funny because I started out with, gasp, modeling intake manifolds.
My actual parts flowed reasonably close to what the machine predicted in 2003 computer technology. I wasn’t doing anything fancy though, symmetrical designs, and 90* direction change from throttle body to runner. All I was trying to get out of it was if the flow was going to be evenly distributed and it did its job.
Another weird thing about these simulations is that you can run the same exact simulation twice and get 2 answers that are pretty different. It's like solving the solution for a x^6+x^5+x^4+x^3+x^2+x-5. A polynomial with more the 4 roots is not solvable, and that’s been proven. So the machine has to just basically guess to find the roots, and depending on where it guesses you get different closeness to the true answers.
An example would be:
If he runs the same problem twice maybe the machine guesses some good guesses on the first run, and comes out pretty close. If the machine on the second run picks some bad guesses then he would get quite a different answer. If he cranks up the resolution there are more guesses all total, and the computer can tell that where there used to be one cell pointing in a reasonable direction there are now 18 cells, and that previous vector makes no sense.
Also these simulations very much build like a snowball effect. Its what you would call an N+1 problem. If it takes one of the elements near the throttle body, and the machine just guesses it wrong, everything else behind that cell is very likely to be screwed up. A lot of times a truly good solution involves some advanced techniques; variable meshing density, turbulence setting manipulation, complex boundary conditions.
The simplicity of the interface of these programs is a little evil. When you first start using these you think you can solve world hunger. After a few years of using them, you realize if you can’t do an analysis like this by hand, then you probably shouldn’t touch it. I can’t do them by hand, but I’ve had some skilled hands give me a good enlightening on the subjects, and I’ve had around 5 years of experience in using this. Its funny because I started out with, gasp, modeling intake manifolds.
My actual parts flowed reasonably close to what the machine predicted in 2003 computer technology. I wasn’t doing anything fancy though, symmetrical designs, and 90* direction change from throttle body to runner. All I was trying to get out of it was if the flow was going to be evenly distributed and it did its job.
#175
![Default](https://www.miataturbo.net/images/icons/icon1.gif)
I second TravisR's comments. I don't do this stuff at work, but I'm involved in discussions of the results, and it's been said more than once that if you don't have a pretty good idea what answer you're going to get before hand, the FEA isn't going to be much help.
I remember a project in school where we modeled a simple 2D divergent equation. The point was the limitation of a numerical, rather than analytical, solution. More, smaller steps means more precision, but at the same time, more calculations. Each calculation involves rounding, which affects the input to the next calculation. At some point, the accumulated error of rounding starts diverting the results. So in something as basic as setting up your mesh size, there is actually a peak in quality of results, and a maximum accuracy that can be achieved.
Until I can get to a library....
I can speculate about ways the pulsed flow changes the requirement. My main thought is to reinterpret the results with the idea that at some times, most of the flow is going to go into #1, and at other times, most of the flow is going to go into #4.
There are probably some good lessons, in that we want to make sure we aren't hosing one of the cylinders by poor flowpath design, as I commented above.
I suspect the continuous flow into cylinders #1-3 in the sims is protecting us from the problem of stagnation against the back of the manifold increasing flow into #4.
In general, I think that as much as possible, we want the plenum to "look" the same to each cylinder at the time of intake valve opening. That suggests extending the plenum a bit past the #4 runner to provide volume on the trailing side for it to draw from. I'm not sure about the reduced plenum cross section moving toward the back. Edelbrock doesn't seem to do it. It may be that as long as you are above a certain size, being larger or smaller doesn't make much difference. It could be that the only reason to have a taper is to be able to provide good contours on the short side radius between the t/b and the #1 runner.
My instinct is that #1 and #4 are where the work needs to go, and if any applicable "best practices" are transferred, #2 and #3 will kind of take care of themselves.
The only real way to know is data, of course. Individual cylinder EGTs are one way. I knew an old-school drag racer who used those on his V8. Not a sophisticated guy, but he data logged individual EGTs and jetted each barrel of his dual four-barrel high-rise setup to equalize them. Light years beyond a lot of import drag racers I've seen in operation.
For dyno work, you could build a configurable manifold. Play with volume behind the #4 runner, play with some sort of filler material in various places to change contours and flow paths, etc.
It might be possible to attempt single-cylinder flow analysis by configuring the runners to reflect the flow conditions at various point in the operating cycle. Say, one cylinder at a time, or have #4 flowing wide open with #1 and #2 closed, and #3 at some reduced flow rate. Maybe tapering the runner to a venturi to control flow rate. Obviously, it won't simulate the possible interference of other runners' pulse events.
I don't actually know any of this, except the V-8 guy story. I love making stuff up on the internet.
I need a better library.
I remember a project in school where we modeled a simple 2D divergent equation. The point was the limitation of a numerical, rather than analytical, solution. More, smaller steps means more precision, but at the same time, more calculations. Each calculation involves rounding, which affects the input to the next calculation. At some point, the accumulated error of rounding starts diverting the results. So in something as basic as setting up your mesh size, there is actually a peak in quality of results, and a maximum accuracy that can be achieved.
Until I can get to a library....
I can speculate about ways the pulsed flow changes the requirement. My main thought is to reinterpret the results with the idea that at some times, most of the flow is going to go into #1, and at other times, most of the flow is going to go into #4.
There are probably some good lessons, in that we want to make sure we aren't hosing one of the cylinders by poor flowpath design, as I commented above.
I suspect the continuous flow into cylinders #1-3 in the sims is protecting us from the problem of stagnation against the back of the manifold increasing flow into #4.
In general, I think that as much as possible, we want the plenum to "look" the same to each cylinder at the time of intake valve opening. That suggests extending the plenum a bit past the #4 runner to provide volume on the trailing side for it to draw from. I'm not sure about the reduced plenum cross section moving toward the back. Edelbrock doesn't seem to do it. It may be that as long as you are above a certain size, being larger or smaller doesn't make much difference. It could be that the only reason to have a taper is to be able to provide good contours on the short side radius between the t/b and the #1 runner.
My instinct is that #1 and #4 are where the work needs to go, and if any applicable "best practices" are transferred, #2 and #3 will kind of take care of themselves.
The only real way to know is data, of course. Individual cylinder EGTs are one way. I knew an old-school drag racer who used those on his V8. Not a sophisticated guy, but he data logged individual EGTs and jetted each barrel of his dual four-barrel high-rise setup to equalize them. Light years beyond a lot of import drag racers I've seen in operation.
For dyno work, you could build a configurable manifold. Play with volume behind the #4 runner, play with some sort of filler material in various places to change contours and flow paths, etc.
It might be possible to attempt single-cylinder flow analysis by configuring the runners to reflect the flow conditions at various point in the operating cycle. Say, one cylinder at a time, or have #4 flowing wide open with #1 and #2 closed, and #3 at some reduced flow rate. Maybe tapering the runner to a venturi to control flow rate. Obviously, it won't simulate the possible interference of other runners' pulse events.
I don't actually know any of this, except the V-8 guy story. I love making stuff up on the internet.
![Jerkit](https://www.miataturbo.net/images/smilies/jerkit.gif)
#176
![Default](https://www.miataturbo.net/images/icons/icon1.gif)
Well if anything the intake manifold is much worse in the simulation that brainiac is doing then in real life. That is the good part. Velocity is not your friend, and more then likely he put pressure openings on the runners (28inches of water or something) and atmosphere at the T/B. That means the air is probably flowing through the throttle 4 times faster then it oughta, because only one cylinder fills at a time, and its pulsed flow. With the velocity much much lower, the gas "hugs" the walls much better, and disrupts flow in the cylinders much less, thats because less force=less differential field flow shear going into the runner.
#177
![Default](https://www.miataturbo.net/images/icons/icon1.gif)
And I just remembered that Cosworth's 350Z manifold has trumpets in the plenums.
http://www.nyxracing.com/cosworth-ca...2.html?image=0
As does the Duratec manifold:
http://www.cosworth.com/popup.php?ur...U2NDUwLmpwZw==
Looks like they are working on some different theories. Air moves past the trumpets to fill the plenum, then reverses into them. Avoiding the velocity across the runner openings and resultant potential for uneven flow.
Also, countering my suggestion above, it looks like the trumpets may provide additional volume of air accessible to each runner to prevent adjacent cylinders from robbing each other.
http://www.nyxracing.com/cosworth-ca...2.html?image=0
As does the Duratec manifold:
http://www.cosworth.com/popup.php?ur...U2NDUwLmpwZw==
Looks like they are working on some different theories. Air moves past the trumpets to fill the plenum, then reverses into them. Avoiding the velocity across the runner openings and resultant potential for uneven flow.
Also, countering my suggestion above, it looks like the trumpets may provide additional volume of air accessible to each runner to prevent adjacent cylinders from robbing each other.
Last edited by SolarYellow510; 11-23-2008 at 03:32 PM.