View Single Post
Old 11-19-2014, 12:08 PM   #1 (permalink)
EcoModding Apprentice
Join Date: Aug 2014
Location: Adelaide, Australia
Posts: 120

Emerald - '97 Honda Civic CXi
90 day: 40.13 mpg (US)
Thanks: 53
Thanked 53 Times in 32 Posts
Talking MPG optimised header ideas

So, I've been thinking about building an MPG-optimised header for my civic CXi... (Why? To see if it works, of course ) probably will be a while before I make a start on it even if I do decide to do it of course, given I have 1000 other things to and haven't even gotten around to installing my MPGuino yet but I thought I'd throw a couple of ideas out there. Posting in Unicorn Corral as it's a potentially controversial suggestion and, of course, it may remain mythical due to time / money constraints.

Why it might work:
  • I'm not talking about fat "high flow" headers off ebay or from a performance shop, this would be specifically designed and built to optimise low rpm torque (without sacrificing too much top-end of course - having the same peak power as stock would be fine by me).
  • I've heard the VX has a specific header designed for efficiency, but looking at it, it's no great prize - short cast iron pipes straight into a cat, the only thing it seems to have going for it is small primary runner diameter. Oh, and a small amount of exhaust pulse overlap from the 1-2, 3-4 pairing *may* help the scavenging of cylinders 1 and 4, but it's dubious.
  • "tri-y" or 4-2-1 headers can enhance low-end torque as well as high-end. Of course, people tend to measure at WOT, but if the torque is higher at WOT then presumably the original torque will be available at some value of partial throttle.
  • We’re basically unrestricted with exhaust header design here, as long as you don’t eliminate the cat, oxygen sensors or other emissions equipment - and the cat on this car is the ‘under the body’ kind, not the sort that sits right on the exhaust ports, which makes longer tuned pipes a real possibility.

Background, based on a bunch of internet reading (feel free to jump in!):

Conventional wisdom is that short fat primaries are good for top-end rpm, while long thin ones are good for low rpm. However, it seems that within the 'butter zone' of primary diameter where you're neither choking the engine nor slowing the exhaust excessively, going a little smaller hurts the top end less than going larger hurts low-end torque.

There are two traditional designs, the 4-1 (all cylinders collect at one point) and the 4-2-1 (pairs of cylinders are connected using Y merge junctions, then the resulting two runners are connected using a third Y merge). The 4-1 can be tuned for a high and narrow peak power band but can sacrifice low-end torque even compared with a stock manifold, while the 4-2-1 apparently can give smaller gains across the board, including the low rpm range, at the expense of some peak power.

Equal length tubes, again, can be used to optimise for specific rpms, while slightly unequal lengths can spread out the resonances and produce a smoother torque curve.

At each Y merge in a 4-2-1, it’s recommended to step the size of the tube up slightly to cope with the combined flow - 15-30% in area seems to be recommended Y merges for ’street’ headers, though it’s more important with the last Y, as the first ones should have exhaust pulses coming in at 180 degree phase to each other (360 degrees of crank rotation). See

Stainless steel is a better insulator than mild steel, and ceramic coating helps even more - this is important because apparently keeping the exhaust speed up is vital for scavenging especially at low rpm, and gas that cools gets smaller and therefore slows down.

A newish development appears to be a 'hybrid' / 'long tri-y' design, utilising primary runners similar in length to a traditional 4-1 design (generally much longer than the primaries in a 4-2-1) - or even longer, with a step in the primary diameter, and longish secondary runners. Putting a step from smaller diameter to larger apparently helps stop reversion (high pressure reflected exhaust pulses going the wrong way) as well as providing an extra reflection point in a longer runner. There are also ‘reversion preventers’ out there that have a step followed by a gentle taper back to the original size, but the jury is still out on them.

Another 'new' idea is a slight 'restriction' in diameter at junctions to produce a greater venturi effect in sucking out exhaust gasses from the other branch.

There is also a set of simple formulae I've found in various places on the webs which are 'guaranteed to spit out something useful' as a header design, which are implemented in this calculator here.


Using the above calculator, with values of
Exhaust Open BBDC. = 60
Exh close ATDC. = 20
CC of one "Cylinder" = 400

at 5800 rpm (roughly peak torque for a honda D16)
5800rpm (peak torque)
P = 32”
primary ID = 1.42”
P1 = 15”
P2 = 17”
secondary ID = 1.86”
CL = 5.6”
TP ID = 1.91”
TL = 29.8”

at 2400 rpm (5th gear at 100kph, a nice highway cruising speed)
P = 82”
primary ID = 0.91”
P1 = 15”
P2 = 67”
secondary ID = 1.198”
TP ID = 1.23”
TL = 81.8”

… and now for the hybrid ‘long-runner tri-y’ mashup. Note that the formula & calculator always use 15” for primary length, apparently “this is the best length” or something.

So, the trouble with primary diameter… the actual exhaust ports on the engine are oval 1.75” x 1.3125”, which is similar in area to a 1.5” ID tube - the above equations suggest anything over this is a waste of time and ideally smaller would be good, but we don’t want a ‘step’ in the wring direction, so 1.5” ID it is.

The primary length to the first set of Ys (our ‘P1’) is then the ‘P’ from the 5800 rpm calculation - 32”. Of course, many sources are quite insistent on the ’15 inch’ thing, and it would be nice to take advantage of anti-reversion effects, so we may as well put a step there, out to 1.625” ID, which is the next standard tube size.

At the first set of Ys, we don’t want too much increase in size (see above), so we only go up to the next standard size, 1.75” ID (area increase of 16%). This is about 10% less area than the formula spits out for 5800rpm, but hopefully should be OK.

Now, the total header length for the 2400rpm calculation is pretty darn long, with P=82” - I don’t like the idea of trying to put the last Y all the way back there, because I’d probably have to move the cat, which I’d rather not. So rather than that, I may as well use the cat as the last reflector for the 82” resonator, and bring the last Y forward a bit… so we can put the last Y at, say, 75” and run a 2” ID collector, with a restriction of around 1.875” (15% increase at the venturi, total step up of 31%). References say that you can almost ignore everything after the cat due to its damping effect on resonance, so I will

Sanity checks:

Ebay headers for D16s appear to have primaries of about 1.75” - this probably makes fabrication easier (you don’t have to dolly the tubes out to match the ports, you can just weld round tube straight on), but it has to be hurting the low RPM power, based on the header formulas etc. (‘serious’ performance headers are larger, but intended for racing with heavily modified engines) … even 1.5” is more appropriate to 6500rpm according to the calculator, but given the port size I think it’s the smallest that can reasonably be fitted.

On the other end of the scale, some info on stock manifold primary tube sizes would be interesting for comparison - however, I strongly suspect that a nicely fabricated mandrel-bent ceramic-coated header with ballpark sensible tube sizes is going to flow nicer at any speed the low cost cast iron jobbie even if the length is a bit longer and the tuning isn’t exactly spot-on. Well, I hope, anyhow.

I’m pretty sure this will all fit, but getting that extra length before the cat will need some interesting design - I’m thinking a ‘ramhorn’ style manifold, in the style of the “k-tuned” headers might do it… luckily I quite enjoy TIG welding

Wrap up:

Crazy idea I know, and I have no idea if or when I’ll get time to try it, but it’s fun to play around with. What would be quite nice is to get access to some 1D exhaust simulation code to validate the above design, but the software seems to be on the “very expensive” side. On the other hand I think the design above looks pretty sensible (to my inexperienced eye), so the old “try and see” approach may just work.

  Reply With Quote