S54 Header Design talk

[3staxontheradio]

The last chart with exhaust pressure, doesn't that just indicate the system is slightly out of time? You can see the low pressure is continuing after evc and everything is just happening later. What rpm is that chart at?

[digger]

The last chart with exhaust pressure, doesn't that just indicate the system is slightly out of time? You can see the low pressure is continuing after evc and everything is just happening later. What rpm is that chart at?

It is at 7,500 rpm (By default shown in the legend).

Yes, that's the effect i am talking about, with the collector the 540 mm goes out of tune w.r.t the zoomie due to the addition of the collector, and the length that is more intune once the collector is added at that higher rpm is the shorter primary.

Ignoring the zoomie for the moment the 540mm is more "intune" 6500-7500 vs the 290mm which is more intune 7500-8500. There also a cross over point at lower rpm where one works better and then worse. This is the issue with length tuning it doesn't work across the board (unless you are way way out to lunch)

[Chrisyphus]

Man, this thread was a great read. I have a lot I want to respond to, and I think I will be quite busy with this tomorrow, picking some of your minds.

But first, I want to respond to this:

Great idea — I'll just leave this here as well.

There’s a company called Celeritech that produces 100% equal-length headers based on precise 3D Chassis Scans.
They currently offer what are widely considered the best S85 headers on the market right now.






Links:
Celeritech Website

https://youtu.be/4XHE-eXWz50

Price List (PDF)

Their prices are quite reasonable, and once you’ve finalized your simulated sizes, they could be a solid option for production.

It would also be great to include the E36 chassis in a scan for header design, as the differences are very minimal.
With this, the headers could potentially be made to fit both chassis.

We spent over 50K with Tjiabring (celeretech) on this project. Unfortunately, once the current inventory on these headers is exhausted (lol), they will become unobtainium. We will never work with him again. He is... a challenge. After about a month of emailing back and forth with him on the topic, he decided he would never talk to me again because I am... impertinent. Typical Dutchman- always happy to tell you in no uncertain terms exactly how dumb, lazy, and incompetent you are, but will be mortally offended at any and all criticism you give back.


His whole claim to fame is that he is able to design headers that are "true equal length to the millimeter." Forgive me for not being impressed by making a set of tubes the same length in CAD after I spent countless hours 3D scanning the chassis, subframes, existing exhaust, and motor, assembled and disassembled, and compositing everything together to define the packaging constraints. Once you have all of that, getting everything to be the same length is CSWA-level design IMO.

To his credit, He does do EXCELLENT fab work.

[0-60motorsports]

Man, this thread was a great read. I have a lot I want to respond to, and I think I will be quite busy with this tomorrow, picking some of your minds.

But first, I want to respond to this:


We spent over 50K with Tjiabring (celeretech) on this project. Unfortunately, once the current inventory on these headers is exhausted (lol), they will become unobtainium. We will never work with him again. He is... a challenge. After about a month of emailing back and forth with him on the topic, he decided he would never talk to me again because I am... impertinent. Typical Dutchman- always happy to tell you in no uncertain terms exactly how dumb, lazy, and incompetent you are, but will be mortally offended at any and all criticism you give back.


His whole claim to fame is that he is able to design headers that are "true equal length to the millimeter." Forgive me for not being impressed by making a set of tubes the same length in CAD after I spent countless hours 3D scanning the chassis, subframes, existing exhaust, and motor, assembled and disassembled, and compositing everything together to define the packaging constraints. Once you have all of that, getting everything to be the same length is CSWA-level design IMO.

To his credit, He does do EXCELLENT fab work.

I bet you people on here, in this amazing community will replicate this easily as they come together. And im sure if an S54 header design will come out from the hard work being done here at least it'll be tested and the honest results will be present here.

[digger]

making the collector even bigger helps power and starts to get closer to open primaries



improves the trace a bit but still time shifted (it mostly just help drop the pressure lower)

7500


8000

[Obioban]

we're going to need cam versions before this is through.

Happily that's really only two variants on this chassis-- 280/272 and 288/280.

[3staxontheradio]

Those variable sized collector tests are interesting but also adding to my feeling like something is missing. My understanding/biases/assumptions don’t fit well with the 3rd vs 5th order harmonics comparison. I'm happy to be wrong about that but I have a few ideas about what is happening. Also, I don’t think most of us would assume running 3” exhaust on these cars is going to go well on a stockish naturally aspirated car making like 365 crank horsepower so something seems off.


First off, adding a merge and collector just seems to do more harm than good in this simulation and I don’t think that's replicated in reality? Maybe the zoomies fall off like a stone below the RPM range shown in the chart but my general understanding is that zoomies are not what people turn to for peak engine output and area under the curve even if allowed to by class rules, and that correctly designed merges are a net benefit for scavenging. Pipemax is not an engine simulation and it seems like it considers primary length tuning in the simplest terms of just being a rarefaction wave on the same cylinder/primary it was generated. Which yes, that happens in the same cylinder but also when you add a merge you get an additional wave that travels up the other primaries as the high velocity gasses pass the other primary exits into the collector. Do you know if that gets modeled in engmod4t?


On collector sizing: it makes sense that a bigger collector diameter = a bigger pressure differential at primary exit = bigger wave amplitude. As you said, it's operating closer to a zoomie. It sounds like right now collectors are dumping into the atmosphere when they terminate? My thought is the 1d sim is overweighting the strength of collector effects in the output and that's skewing the result or how much the collector pressures would interfere with primaries. Can you swap a Y pipe to single 3.5” or an H/X pipe to continue dual 2.5” at the end of the collector and then continue the exhaust? I'll get you an approximate length in a few minutes.


If you frame the job of exhaust scavenging as efficiently translating exhaust gas energy into useful negative pressure at the exhaust ports its hard to come up with a good reason why dividing that energy into 5 waves and only having 1 of those waves be in time would produce better scavenging than dividing that energy 3 ways and having 2 be wasted. That's obviously a simplification but I have a hard time thinking the design that involves way more unusable/incorrectly timed waves is the efficient one, at least for peak scavenging power. I know pumping efficiency has to fit in there too and impacts longer runners more than shorter runners. Maybe this perspective is just prioritizing scavenging effects over pumping efficiency too much but I would at least think the sims would show a more aggressive scavenging peak with a 3rd order harmonic and that doesn't seem to be happening so far.

[siamkazi]

we're going to need cam versions before this is through.

Happily that's really only two variants on this chassis-- 280/272 and 288/280.

so does that mean different headers for different cam profiles? i.e. if one were to go 280/272, then another set of headers from this analysis would be optimal? 😮

[Obioban]

so does that mean different headers for different cam profiles? i.e. if one were to go 280/272, then another set of headers from this analysis would be optimal? 😮

I believe so, but this is WELL outside of anything I know. Happy to be told I'm wrong.

But, with 3D printed headers-- get exactly what's optimal for your cars setup, no?

[3staxontheradio]

I believe so, but this is WELL outside of anything I know. Happy to be told I'm wrong.

But, with 3D printed headers-- get exactly what's optimal for your cars setup, no?

Exactly that. There isn't a production cost or penalty for one off designs, its just design effort. Headers can be designed to be application specific and that's a combination of someone's engine build, how they want to use the car, and any other constraints unique to them.

For a changing cams -> you move the timing and size of the window for scavenging around a little but you also increase total gas flow. For the same intended application you would want to up size the runner diameter.

[siamkazi]

I believe so, but this is WELL outside of anything I know. Happy to be told I'm wrong.

But, with 3D printed headers-- get exactly what's optimal for your cars setup, no?

holy moly that is cool. I was planning on cams next and having headers that are setup for specifically for certain cam profiles is such an awesome concept! Appreciate the insight!

[digger]

Those variable sized collector tests are interesting but also adding to my feeling like something is missing. My understanding/biases/assumptions don’t fit well with the 3rd vs 5th order harmonics comparison. I'm happy to be wrong about that but I have a few ideas about what is happening. Also, I don’t think most of us would assume running 3” exhaust on these cars is going to go well on a stockish naturally aspirated car making like 365 crank horsepower so something seems off.


First off, adding a merge and collector just seems to do more harm than good in this simulation and I don’t think that's replicated in reality? Maybe the zoomies fall off like a stone below the RPM range shown in the chart but my general understanding is that zoomies are not what people turn to for peak engine output and area under the curve even if allowed to by class rules, and that correctly designed merges are a net benefit for scavenging. Pipemax is not an engine simulation and it seems like it considers primary length tuning in the simplest terms of just being a rarefaction wave on the same cylinder/primary it was generated. Which yes, that happens in the same cylinder but also when you add a merge you get an additional wave that travels up the other primaries as the high velocity gasses pass the other primary exits into the collector. Do you know if that gets modeled in engmod4t?


On collector sizing: it makes sense that a bigger collector diameter = a bigger pressure differential at primary exit = bigger wave amplitude. As you said, it's operating closer to a zoomie. It sounds like right now collectors are dumping into the atmosphere when they terminate? My thought is the 1d sim is overweighting the strength of collector effects in the output and that's skewing the result or how much the collector pressures would interfere with primaries. Can you swap a Y pipe to single 3.5” or an H/X pipe to continue dual 2.5” at the end of the collector and then continue the exhaust? I'll get you an approximate length in a few minutes.


If you frame the job of exhaust scavenging as efficiently translating exhaust gas energy into useful negative pressure at the exhaust ports its hard to come up with a good reason why dividing that energy into 5 waves and only having 1 of those waves be in time would produce better scavenging than dividing that energy 3 ways and having 2 be wasted. That's obviously a simplification but I have a hard time thinking the design that involves way more unusable/incorrectly timed waves is the efficient one, at least for peak scavenging power. I know pumping efficiency has to fit in there too and impacts longer runners more than shorter runners. Maybe this perspective is just prioritizing scavenging effects over pumping efficiency too much but I would at least think the sims would show a more aggressive scavenging peak with a 3rd order harmonic and that doesn't seem to be happening so far.

the collector is more about torque and area under the curve and it shows up below the peak tq rpm in the plots above


6-2 vs 6-2-1

6-2 is 540mm primary, 1800mm collector
6-2-1 is 540mm primary, 1800mm collector and 1900mm single tail pipe 3.5" pipe (1800+1900= 1st harmonic collector length)

[digger]

Man, this thread was a great read. I have a lot I want to respond to, and I think I will be quite busy with this tomorrow, picking some of your minds.

But first, I want to respond to this:


We spent over 50K with Tjiabring (celeretech) on this project. Unfortunately, once the current inventory on these headers is exhausted (lol), they will become unobtainium. We will never work with him again. He is... a challenge. After about a month of emailing back and forth with him on the topic, he decided he would never talk to me again because I am... impertinent. Typical Dutchman- always happy to tell you in no uncertain terms exactly how dumb, lazy, and incompetent you are, but will be mortally offended at any and all criticism you give back.


His whole claim to fame is that he is able to design headers that are "true equal length to the millimeter." Forgive me for not being impressed by making a set of tubes the same length in CAD after I spent countless hours 3D scanning the chassis, subframes, existing exhaust, and motor, assembled and disassembled, and compositing everything together to define the packaging constraints. Once you have all of that, getting everything to be the same length is CSWA-level design IMO.

To his credit, He does do EXCELLENT fab work.

True equal length is obviously a pure marketing term to try and differentiate themselves which is fine as they are a business trying to make money. For awareness they use the centreline distance for measurement and i hate to be the one that points out the obvious but the exhaust gas particle doesn't follow the centreline (and pressure waves diminish when going around corners) so you would need to adjust the length based on the number of bends, bend radius, and the angle of bend etc as each runner is different. It doesn't mean you shouldn't try and get it fairly close but mm accuracy is an utter waste of time and hence money for the consumer.

[Chrisyphus]

True equal length is obviously a pure marketing term to try and differentiate themselves which is fine as they are a business trying to make money. For awareness they use the centreline distance for measurement and i hate to be the one that points out the obvious but the exhaust gas particle doesn't follow the centreline (and pressure waves diminish when going around corners) so you would need to adjust the length based on the number of bends, bend radius, and the angle of bend etc as each runner is different. It doesn't mean you shouldn't try and get it fairly close but mm accuracy is an utter waste of time and hence money for the consumer.

100% correct. This is one of the myriad reasons I am working through the process of defining timing-based pressure inlet profiles for ANSYS Fluent. The existing aftermarket tools are great, but limited in their capacity to accurately simulate exhaust pressure wave and mass flow in such an incredibly dynamic environment. It is really easy to do steady-state CFD studies, and not terribly complex to do transient studies, but creating pulsatile behaviours on the appropriate timescale AND coming from multiple inlet boundaries, each phased 120 degrees from each other AND in different stages of their cycles is proving... challenging. I am wishing now more than ever that I had taken more software programming courses because I am getting lost in the sauce.

I am working through it, but I think there might be literal steam coming off my head at this point.


Having said all of that- There does seem to be some really valuable data in your ENGMOD sheets- I would really like to replicate the pressure wave ( or at least use it as a starting point) it generates as a function in my inlet boundary profile, but there's some detail that I think we just can't see on these graphs:

Exh Pressure

On the pressure graph, it is pressure ratio across crank angle. What is the pressure a ratio of? is it just relative to some arbitrary normal? Or is it a ratio of some known variable like 100kPa?

Also, since I am asking a total stranger for all this help, could you run one of these simulations with a primary that is effectively infinite on this timescale? 100 Meters or something? With a diameter that matches the OEM port diameter as closely as possible? In a perfect world, I would like a pressure wave graph that is as close to uninfluenced by the reflected waves in the system as possible, so that I can let the actual physics in Fluent simulate what would happen with a pressure wave that matches the actual pressure wave generated by the exhaust stroke as close as we can estimate. I need to find a solution to measure exhaust gas pressure in real time (in the range of 10,000 Hz, since the entire exhaust valve open-close duration is 14 milliseconds at 7000 RPM, even with an insanely long-duration cam like the Schrick 292 exhaust cam) on our engine dyno if I really want the sauce.. This is something we have in the works, but I do not want to wait around for that to be completed to complete this project.

Anyways, any information you can provide would be excellent!

[digger]

100% correct. This is one of the myriad reasons I am working through the process of defining timing-based pressure inlet profiles for ANSYS Fluent. The existing aftermarket tools are great, but limited in their capacity to accurately simulate exhaust pressure wave and mass flow in such an incredibly dynamic environment. It is really easy to do steady-state CFD studies, and not terribly complex to do transient studies, but creating pulsatile behaviours on the appropriate timescale AND coming from multiple inlet boundaries, each phased 120 degrees from each other AND in different stages of their cycles is proving... challenging. I am wishing now more than ever that I had taken more software programming courses because I am getting lost in the sauce.

I am working through it, but I think there might be literal steam coming off my head at this point.


Having said all of that- There does seem to be some really valuable data in your ENGMOD sheets- I would really like to replicate the pressure wave ( or at least use it as a starting point) it generates as a function in my inlet boundary profile, but there's some detail that I think we just can't see on these graphs:

On the pressure graph, it is pressure ratio across crank angle. What is the pressure a ratio of? is it just relative to some arbitrary normal? Or is it a ratio of some known variable like 100kPa?

Also, since I am asking a total stranger for all this help, could you run one of these simulations with a primary that is effectively infinite on this timescale? 100 Meters or something? With a diameter that matches the OEM port diameter as closely as possible? In a perfect world, I would like a pressure wave graph that is as close to uninfluenced by the reflected waves in the system as possible, so that I can let the actual physics in Fluent simulate what would happen with a pressure wave that matches the actual pressure wave generated by the exhaust stroke as close as we can estimate. I need to find a solution to measure exhaust gas pressure in real time (in the range of 10,000 Hz, since the entire exhaust valve open-close duration is 14 milliseconds at 7000 RPM, even with an insanely long-duration cam like the Schrick 292 exhaust cam) on our engine dyno if I really want the sauce.. This is something we have in the works, but I do not want to wait around for that to be completed to complete this project.

Anyways, any information you can provide would be excellent!

It is a pressure ratio relative to atmospheric pressure of 101.325 kPa.

Couldn't run more than 9m primary, There appears to be a maximum number of grids/mesh and the size is not user definable so it reaches a limit.



coupling 1D with 3D is ideal but finding the optimal solution might not be apparent, but would work for optimizing/refining a known design

TFX sell some relatively affordable pressure measurement equipment