General 16v Inlet Port Splitter Modification Strategies

[PersonaGrata]

This is my first posting on GCRE, so hello there everyone and thanks in advance for any help you can give me.

The Background:
I’m modifying a 4G93 Mitsubishi designed cylinder head, as manufactured by Proton in this instance. Its a four valve per head DOHC aluminium one and belongs to my budget trackday car which is a five door Proton Persona Xli. My aim is to gas flow the head, intake system, raise compression ratio and swap onto E85 ethanol, but remain on the standard cams. I don’t know the duration of the cam currently but I do know the lift is 9.0mm.

The Problem:
I’ve started the porting work by smoothing out the Short Side Radius (SSR) and widening each port before and adjacent to the SSR to reduce gas velocities, and allow the gases more chance to make the turn, or at least increase the range of velocities that they can do this at.

In doing this I’ve machined the port walls both on the outside but also on the inside on the ‘splitter’ side.

This is the standard port with untouched splitter

P1050624.JPG (583.42 KiB) Viewed 12607 times

This is my inwork, modified port with narrowed splitter

P1050631.JPG (520.72 KiB) Viewed 12607 times

The splitter is around 43mm long, measured to the furthest par of the valve seat.

So far, I’ve not been able to test the results because my flow bench is so very Heath Robinson and not returning repeatable results so far (this is in hand). However, I’ve just got thinking about the form of the splitter and am worrying now that this was wrong thing to do.

My theory is that because of the variability introduced by hand machining there will be differences in flow velocities at the point the gases are separated. This will cause the stagnation point to move to one side or the other of the splitter, flow to occur across the front of the splitter and vortices to be formed on the side that had the higher velocities. This then reduces the cross sectional area available to gas flow in that port. In this scenario, the form of the splitter becomes all important whereby a nice rounded profile reduces the harm cause by the cross flow. A sharp profile (like mine) guarantees turbulent inefficient flow.

Hand drawing showing issue (apologies for the quality)

P1050637.JPG (522.7 KiB) Viewed 12607 times

I really don’t think that I can physically form a very sympathetic profile so I’m into damage limitation strategies. The best I can come up with so far is to open one valve side at a time and measure dynamic pressure in the region adjacent to the splitter. When there are differences I plan to widen the port on the faster flowing side until the velocities are matched. I’ll do this with no valve in to maximise the absolute value of velocity so that small differences are more likely to be spotted.

Matched velocities would minimize the absolute quantity of cross flow and turbulence, and make the ‘best of a bad job’. I’ll then profile the splitter as best as I can to reduce the effect of those flows.

Any thoughts on the issue will be gratefully received Guy (and anyone else for that matter). Apologies if this issue is old hat and answered comprehensively elsewhere on this site.

Many Thanks
Iain McLean

[Guy Croft]

First point I'd make is the long splitter was used for decades to separate the two barrels and the short splitter - a relatively recent introduction) with its big radius nose (like the one in your shots) was introduced for two reasons:

1. To reduce the viscous loss along the splitter wall
2. To get rid of the 'horseshoe' vortex that exists on the long splitter (those I have tested - Vauxhall 16v HS and XE, Fiat 16v all have it and I now know it's a well known phenomenon on aerofoil sections like that)

Both of these lead to flow loss. Before you do anything els I'd like you to 'blow thru' from the chamber and out thru the valve seats with a cotton trace and see if you get the horseshoe vortex forming in your head - standard and modified splitter. The rotation will always be opposite - if both traces go the same direction of rotation they are not vortices. Picture below describes.

GC

[PersonaGrata]

Many thanks for the reply Guy

I will do just what you suggest though my equipment is set up to 'suck' only so far. It might take a little bit of time to perform the test.

Iain McLean

[PersonaGrata]

Hello Guy, I've managed to find enough time to do the blow through tests as you asked. The setup was slightly ropey in that I ended up using my daughters bouncy castle blower to provide the puff - so the pressure behind the port is pretty low as a result. Not being an expert in interpreting the movement of the threads, its best that you have a look at the results and decide whether vortices are present. As far as I can tell, the movement doesn't fit your description....

Standard
http://www.youtube.com/watch?v=JTzOsVN3cAk
Modified
http://www.youtube.com/watch?v=7oKnfH2Bkjg

I also took the liberty of investigating the port velocities whilst in ‘blow through mode’.

Standard
http://www.youtube.com/watch?v=N0agwrnNMKM
Modified
http://www.youtube.com/watch?v=XHEsOxRVsJY


The findings are:
1. There is very little flow to speak of making it around the SSR in either the modified port or the standard. Pushing the probe slightly furher into the port shows sudden and huge increases in velocities
2. There is marked a ‘velocity shadow’ cast by the valve guide.

Could you cast your expert eye over these videos Guy, and give your opinions???

Many Thanks
Iain McLean

[Guy Croft]

Very good work!

The (excellent) films with traces and vel probe tend to show:

1. There isn't a vortex tendency at all (and that is what I'd expect from the short splitter)
2. The air doesn't 'see' the region between the inlet contact face and region of the SSR close it - thus one can assume it 'skips' over that area and just heads straight for the valve stem.

Of course one can argue all day about test pressure and velocities but with some 5 years extreme flowbench dev and vast dyno feedback on those tests - I'm confident enough to say:

1. The splitter - as an intrusion into the airstream - might benefit form being knife-edged - your own tests will determine whether this is better than a radius on its entry section - you can't really go wrong with this as making it sharp certainly won't make it any worse.
2. The velocity distribution isn't brilliant - more downdraft would encourage the air flow to further round the SSR - but it is typical of a port layout of that kind. You could be wasting your time trying to improve the distribution by radically reshaping the SSR, the best you can hope for is that maybe increasing the chord radius will make the air cling better to the SSR profile - but it could be a 'near-run thing' - you might make the flow worse. I would make that alteration by grinding say 0.5mm off the middle of the curve and blending it up-and-downstream. If that doesn't make the vel distr better I would just smooth the existing curve at 80 and then 120 grade and leave it at that.

Interestingly the valve shape can tell you a lot about airflow distribution - and of course it's interesting to consider how they are actually designed and which phase in the valve lift regime is considered most important (eg: low lift 0-4mm, medium lift 4-8mm and over that), because I know from exp that you'll never get a valve that does everything perfectly. In a sidedrafted port of this kind - in my exp- you tend to want a valve with a fairly modest amount of material on the non-firing side, ie: verging towards 'penny_on_a_stick'. Though I emphasise 'tend' because it is not always true, a lot seems to depend on the 'drop' between the end of the short-side radius and the valve contact face - and the seat angles too.

The more work I do on heads, the fewer 'Golden Rules' I uncover. What works on one type of head may not work at all on another or markedly similar style! The more I bless Superflow. A flowbench is a demanding mistress - but she certainly takes the 'guesswork' out of head dev! Gain on flow doesn't always mean more power - but loss DEFINITELY means power loss!

GC

[PersonaGrata]

Thanks a lot for your that Guy, I really appreciate the effort you have put in there and will digest it over the next couple of days - I don't want to post up some knee jerk nonsense in reply.

Regards
Iain McLean

[PersonaGrata]

Hello Guy, I was hoping to have a period of time thinking in depth about what to do next, but everyday life happened instead. So this is a bit 'off the cuff' I'm afraid.

Your first comment was about valve shapes - so here is a photograph on one of them.

P1050779.JPG (602.72 KiB) Viewed 12259 times

In the fullness of time I will get one machined to remove some of the 'back' side and do a head to head against the standard valves. Ideas about just how to do that will be gratefully received of course!

Something I alluded to in the title of the thread, but didn't explain properly was strategic thinking regarding the port modifications. What I mean by that is that this head will be run with a single plenum manifold and throttle body and its quite probable that a transverse component in air direction will be produced by the contortions of the manifold (from the blow through tests there is a definite transverse component at least for cylinder 1). Given the imperfections of the manifold it might be a risky strategy to go for knife edge splitter as cross flow will be badly affected by the sharpness of the profile. In these situations it might be best to assume that cross flow will occur and machine a profile that reduces the negative effects of this. As a compromise, I might knife edge the two inner cylinders but leave the outers with a rounded profile. Some imbalance may occur, but judging by some earlier postings from yourself this is not an especially big problem. Does this seem sensible?

Talking of imbalance, it occurs to me that it might be a good idea to have one side of a port flow more than the other by design in order to promote swirl at the expense of outright flow figures? I'm not sure if that is a stupid idea or not - do 8v heads give the same power as 16v heads given the same CFM flow capabilities? If they did, then that might give some credence to the idea.

Also, just as a quick update, the port I modified that's shown in the videos luckily turned out to give more flow than standard. I say luckily because it wasn't done incrementally and checked at each step because the flow bench wasn't working very well. Here is its results relative to standard

Cyl1CFM_1.jpg (37.77 KiB) Viewed 12259 times

Don't place too much importance on the absolute figures in this chart because my bench isn't calibrated. I've tried to calibrate it using sharp edge orifice plates of different diameters but they give very, very different results so the figures are only really of use in a comparitive sense.

Regards
I McL

[Guy Croft]

1. Valve shape - I don't think you'll better it.
2. Splitter affecting flow one barrel to the other? Test it by blocking off one barrel with a lump of modelling clay (or put a valve in) each should flow the same and their sum - if they flow equally - will be the same as the whole port.
3. Splitter knife-edged? Test it and see if it helps.
4. You need to state your test depression on cfm results, sorry if you have - I did not see it.

G

[PersonaGrata]

1. Heck. I'll sacrifice a couple of valves for the sake of getting a definitive result anyways
2. OK, will try it
3. I've got a scrap head just for testing on its way - so I'll have a go on that
4.The CFM results are all normalised to 10" static. In reality the low lift static depression is up to ~70" and the high lift is as low as ~3" static but all the results are normalised to 10". I've tested this method on a constant restriction across many static pressures and found it to reliably give the same results (give or take 1%)

Thanks again
Iain McL

[PersonaGrata]

Hello again Guy.
I was testing out the differences between the different sides of ports by closing a valve at a time and measuring the flow through the remaining valve - and have found quite a strange effect.

The first thing I found was probably to be expected in that the flow summed across both valves does not equal the flow measured when both valves are open. It is always lower, see below:

IndividualValveFlowComparison.jpg (53.72 KiB) Viewed 12111 times

This can probably be explained by the flow having to cross the front edge of the splitter and causing a layer of low velocity air in the controlling section next to the splitter. This is then an impediment to higher velocity flow in this area.

What is perhaps less easily explained is the occasional result where my flow bench shows a higher static depression but higher pressure differential through the venturi implying greater flow. If the bench is to be believed I have ports that offer more resistance to flow but flow more at the same time!

The only explanation I can possibly come up with is that a signifcant vortex is being generated when measuring this side, and that vortex is lowering the static pressure measurements in an unexpected way especially in the small diameter part of the venturi. This in turn leads to the greater pressure differential. I'm reasoning that if the gas is rotating as well as travelling along the axis of the pipework then the static pressure measurement will have to be lower.

The other feature of measuring one side at a time is the lack of stability in the readings taken from the venturi. This does not happen when both valves are open. The static depression is taken from the exhaust side of the cylinder about an inch from the head face and is stable.

Have you ever experienced anything like this before?
Regards
Iain McL

[Guy Croft]

"..my flow bench shows a higher static depression"

Sorry but I don't fully comprehend the technicalities of this yet!

I may have misunderstood. My flowbench is adjusted to run at constant depression, is yours not? Sure the depression manometer (vertical one) goes down and if the flow goes up (and vice-versa) - but at each new state you adjust the valve to stabilise at 10".

G

[PersonaGrata]

Sorry, i didn't explain my flowbench properly.

Its pretty Heath Robinson, being made up from timber, polycarb, old plumbing parts, home made O rings and a domestic vacuum cleaner. Instead of throwing up a frankly embarrasing photograph of it I've produced a schematic:

FlowBench.JPG (27.36 KiB) Viewed 12117 times

NOTE: The vortex shown on the diagram isn't a feature of the rig, but a possible explanation of the odd strange result I get from it (see previous postings in this thread)

Because the vacuum source is of a constant power (more or less), the static depression versus the flow differential relationship is linear - high resistance to flow results in high static depression across the cylinder head and low flow differential through the venturi, lower resistance to flow results in the lower static depression and higher flow.

Because static depression is an output of the system and can't be adjusted, I have to do the following
1.Adjust the valve drop.
2.Take my flow differential measurement
3.Note the static depression that they were produced at
4.Calculate the flow based on the differential
5.Extrapolate (or interpolate) to a flow value that would have been produced if the static depression had been 10inch.

The static measurements range from over 70inches at 1mm lift to about 3 inches at high lifts - but most flows are taken at less than 10inches and have to be extrapolated

An extra complication is the 'Coronation Street Effect' which is where people brew up at soap opera breaktimes, reducing the voltage available to the Vacuum cleaner and reducing the flows measured. However since both static and flow pressures are reduced i find the system still produces the same extrapolated result (+/- a small percentage)

If you look at the red dotted area on the diagram this is my supposed vortex - which only get sensed by the low diameter part of the venturi leading to higher, but unstable, measured flow - even though the static depression tells a different story by returning a relatively high result. This is the theory anyways.

Thanks
Iain McL

[Guy Croft]

I might be tempted to suggest that the vortex is being created by the flow thru the two valve throats right into your exit duct. Unlike the Superflow rig yours has no damping so the vortex is passing right thru your system. It's easy to imagine but not so easy to prove unless you can induce smoke and film it.

Regarding extrapolation I (of course) have frequently converted readings to and from 20" (mine operates at 10") to compare other people's data and I don't think it's very accurate although Superflow data says the conversion is straightforward, so how much confidence you can have in yours I wouldn't care to say..

I would perhaps prefer to see 1) more power and 2) a damper chamber, see att dwg. With two motors you might be able to test at constant depression, 5" or water (or more?) and there will be fewer confusing variables to worry about.

Not sure if you have temperature correction - the air is being 'worked' and is heating up, this has to be allowed for.

I hope this is marginally helpful,

GC

1210 hrs 23 Nov - dwg edited by GC

[PersonaGrata]

Thanks again Guy,

You are probably correct in what you say - vortices might be present even when both valves are open.

I plan to do the following

1. fit a long length of straight pipe in the run up to the venturi, and a bend before he straight pipe (reasoning that a vortex is unlikely to survive a 90 degree bend, and any flow bias introduced by the bend will be taken care of by extended straight pipe flow)

2. Increase the size of the vacuum source and fit a variable aperture vent near the vacuum source. I will then pick static depressions to test at for all parts of the lift regime and measure flows at these predetermined depressions.

Additionally, it may be wise to sample pressure across both sides of the pipe so that an average can be determined, and increase the area that the static pressure is sampled across.

The temperature correction is an interesting point. Does it affect the absolute figures only (CFM)? or could it lead to significant differences from day to day measuring the same item. NB. I've recorded the air temperature for every test.

Regards
Iain McL

[PersonaGrata]

I've rebuilt the bench and incorporated the features detailed in the previous post.

It looks like this now:
Schematic

NewBenchConfig.jpg (40.08 KiB) Viewed 11824 times

Photos
Vac Box

P1050972.JPG (575.11 KiB) Viewed 11824 times

Elbow

P1050978.JPG (605.68 KiB) Viewed 11824 times

Variable Air Admittance Valve

P1050975.JPG (658.51 KiB) Viewed 11824 times

OK, so it has obviously been nailed together from odds and sods and some old plumbing equipment but so far the resulst are good. - This rig does seem a lot more stable in that results so far have been repeatable and expected.

The biggest improvement by miles is the air admittance valve (AAV) which allows me to set the static depression across the head to a single value (per type of test). For example if I am measuring BPF I start the vacs then adjust the AAV till the static depression is at 120mm allowing me to directly compare flow readings with previous ones. If I am measuring a single port then I have a different static pressure to use. The vacs simply aren't strong enough to use 10inch for every type of test.

This rig is also calibrated (after a fashion) as I fitted a sharp edge orifice plate where the head would usually go and measured venturi FLOW differential pressures across the range of flows that I would expect from the head. Using a calculator on the efunda website I calculated what the flow would be expected across the orifice plate for each STATIC pressure differential used. This gave a correct flow (efunda/orifice plate) versus the rigs estimation (own calcs/venturi). A linear interpolation could then be generated that will produce errors for each estimated flowrate.

I've also looked into the effects of temperature/humidity and barometric pressure on density so mass flow rates can be calculated.

And from the hundreds of hours of work on both the flow bench and the cylinder head I have (fairly) reliably determined that the head is flowing a measely 7% better than standard - at approx 129CFM versus 121CFM standard. This is at 10inch with standard 33mm valves lifted 10mm

If you don't mind Guy, I'll start a new thread soon on tackling this issue of poor performance. This current thread has turned into a flow bench discussion.

Thanks
Iain McLean