Why do we truly need to advance the injection ???
Turbo Mini Forums

Below is an email I sent to Marcel previously... I was bombarding him with emails at the time, so this one might have slipped by - anyways, it is definately best discussed here - amonst us lot.

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
I want to throw this thought your way, tell me what you think... (Yes - it's A-series injection again).

Do you agree with this statement:
The maximum injector duration we're going to need is basically going to be at max power - which for the sake of argument, we can say is near-enough max RPM.





Now imagine our (virtual A-series) engine at WOT. Inlet port velocity is high. Let's assume it's infinite - ie as quickly as fuel leaves the injector, it's at the valve. This is the worse-possible case for sure. Hence, if our inner valve is fully closed at 50 degrees ATDC, we need to start injecting at 50 degrees ATDC (or later).

Next phase: At the end of our injection 'window' we have a valve that is to all intents, closed at 40 degrees after BDC. Again - assume infinite port velocity so when we turn off that injector at 40 degrees after TDC, the last drop of fuel leaves the injector and goes down the valve.

Now, I know all the above is great at infinite port velocity, but we have to consider a much reduced port velocity.

Well - here is the crunch; at lower port velocities, the injector is going to be open for a shorter length of time. We have already sized an injector large enough for getting all the fuel into the (theoretical) engine at WOT, so the END of injection point at 1000rpm, is going to be happening only a few degrees after the START of injection. Sure, this start of injection point is now happening much later than we first planned, but there is plenty of time to get the 'injection pulse' into the injection window at these much lower port velocities.


Now - I know you went over all this on your engine testing, but I'm (now) not following why you got the results you did. (!!!)
From what I've just described above, if you're able to get all the fuel into your engine to run at WOT; Switching ON the injector when the inner valve has shut, then Switching OFF the injector several (?) degrees before the outer valve shuts - you should be able to run the engine inside of this window as the injector length needed is so much less.

I'm interested to hear what you might think on all the above... Do you think some of the effect you experienced could be that the inner cylinders were basically taking in a homogenus (sp?) mix of fuel and air - whereas the outer cylinders were taking in fuel in drips and flow ??? Would this make that large a difference? I suspect i'm clutching at straws here!!!
Any other thoughts?


^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Could this also explain a need for a VE map aswell?
Let's not forget the Rover MPi mini however - Basically a sequential injection system... Do 'OE' sequential setups have VE maps - anyone?

As to why we need a VE table and do OE systems have them....
Yes, OE systems have VE tables, IF they run based on a MAP sensor rather than an MAF. An MAF sensor measures the actual volume of air entering the engine, so a VE table is unnecessary. The amount of air moving past the MAF is equal to the amount of air in the engine, regardless of engine size, but a MAP sensor can only read the pressure level. Based on the VE table and knowing the engine size, the pressure directly correlates w/ the amount of air in the engine. You need to know the amount of air going into the engine to properly meter fuel. You could rely solely on the O2 feedback, but that's usually too slow and isn't even used in open loop mode.

-Tyler
DO NOT TOUCH MY HORNS OF DOOM!

But isn't the mass of the air just calculated from throttle position, MAP and RPM - a simple 3D table - (with air temp correction) exactly as the DTA that Marcel initially used?

That would only be if the engine were ideal. The VE table tells the computer how well the engine can fill it's cyls. From MAP and RPM the ECU can only tell how much air SHOULD be going thru the intake, but the VE table is needed to modify this so it's closer to how much IS going thru it. Our engines are relatively poor at this(probably a max less than 85%) compared to a modern engine such as a honda(which can actually get over 100% in some conditions)

-Tyler
DO NOT TOUCH MY HORNS OF DOOM!

Also, as to marcel's results, I believe that was due to treating it like a 2 stroke. There are actually 2 cylinder events that occur per port when the ecu was expecting 1. The ecu did not have any concept of how to divide the fueling between the cyls, only how much the two cyls should use together and most of this was going into a single cyl.
I've thought of two approaches to solve this. One, i believe will be more accurate, and that is to treat it like a 4 stroke, firing each injector twice, once for each window for each cyl. You need to advance this to compensate for the time delay of injecting, versus injesting. Figuring out the velocity is the hard part because for several degress, both ports are sucking, then one closes. I think this is what makes this hard.
The other solution I thought of was indeed to treat it like a 2stroke motor. You will still need to advance, but instead of calculating it per cyl, you can calculate it per port, and "balance" the pulsewidth on the overlap. Basically, it's a single injection event, half being prior to the middle of the overlap, and half being after, but the entire thing advanced to account for port velocity.
This last idea, which simpler, I feel will have problems on the lower end because the PW will be so short, that I'm afraid I can't say which cyl will actually get the fuel. basically PW will be smaller than overlap.

-Tyler
DO NOT TOUCH MY HORNS OF DOOM!

It was used in '2 stroke mode' but all that meant was it was firing the injectors every 360 degrees instead of once every 720 degrees.
I proved with some very nice traces from a digital trace recorder that the DTA injection event happened not at TDC, but at the same time as the Ignition event (ie 25-40degrees BTDC).
Marcel managed to get a 'one off' update from DTA which then allowed him to 'swing' this injection event relative to TDC - ie he could get the injection event to start 20, 30, 40, whatever degrees after TDC.

I now realise that this isn't so obvious from his webpage... I of course knew this, as Marcel and me were coersing at the time! Everyone else is clearly not psycic!


This again leads to what i (now) don't fully understand. At WOT VE is at its max - hence, if you advance the injection point to (say) 50degrees ATDC, and you have a (magic) injector large enough to get the required fuel in before the injector closes at (say) 20 degrees ABDC, why shouldn't it work with MUCH smaller injection durations at lower RPM's ???

Edited by TurboDave16V on 4th Jan, 2006.

Those 2 methods are the ones described in Rover's patent for siamese-port injection.

As you say, I don't think that the second method of a single pulse can be successfully used at low rpms especially with bigger cams. Also, the first method becomes a problem with higher rpms and higher load since you have to open and close the injector in rapid succession and you might not be able to. So the 2 methods are actually complementry and necessary if you want to cover a big enough range.

Unfortunately, with the current MS hardware with only 2 injection drivers, you will not be able to do those 2 methods and have injection staging which might be an issue for turbo engines.

However, that was something I have suggested previously and might implement in a later version of the code because it might allow the use of a smaller single injector per port than the single injection per revolution of the current implementation.

Regards,
Jean

http://www.jbperf.com/

TurboDave,

The issue is that at low rpms, the TIME is not the same for the same advance. So if you take into consideration the injector opening time and the fact that the fuel has a certain speed due injection and is not only carried by the velocity of the air, then you will be injecting in the inside cylinder instead of the outside cylinder. At least, that how I see it.

http://www.jbperf.com/

I realise that - but assume an infinte port speed (as would be needed for WOT max RPM) then if you only had a 'fixed' injection point (say 50degrees ATDC), then at 1000rpm when the port velocity is so much reduced, the fuel would be getting to the valve later (eg 65degrees ATDC) not earlier...

Edited by TurboDave16V on 4th Jan, 2006.

Thanks for getting this forum together, Dave.

The concept for a 4D map came from the final experiments I did with the DTA ECU setup in 2 stroke mode. Although DTA were belligerent and unable to belive that just sticking it in 2 stroke mode would not solve all the problems, they condescended to give me an EPROM image that allowed me to use the ignition advance knob of the dyno control to move the injection event around WRT TDC. The experiment involved holding the engine at fixed load and RPM, trimming the centre branch O2 sensor then advancing or retarding the injection initiation point until the outside O2 sensor started to go weak. Only a small amount more advance shut the ouside cylinder off. From this the injection opportunity window was established and injection advance table built.

It occurred to me that these parameters (window of opportunity and advance table) were specific to a particular engine design, so that's where I got the idea to not anticipate the 4th dimension (ie, pre written table), but let the ECU build it by using the O2 pair to provide feedback thus:

*inside and outside O2 sensors equal and at stoich, do nothing, write advance value to the 4th D table
*inside and outside O2 sensors equal and but lean increase pulse width
*inside O2 sensor at stoich, outside weak, advance injection event until both are equal at stoich

Unless you can somehow anticipate what that 4th D will be in order to establish the table, I can't see how this can be done without a second O2 sensor. And since MS can't do that, I don't know how you can use an existing MS board? Anyone got any ideas?

www.starchak.ca and www.TDCperformance.ca

I can't see how anyone would be able to get a system to work 100% without using two o2 sensors, and generating a 'strategy' - unless you have massive injectors that gives you a 'degree of freedom' to simply 'hope' that it's working right...

Edited by TurboDave16V on 4th Jan, 2006.

Marcel,

The MS has an input for a second O2 sensor so if the software can be done the hardware will support it.

However, I'm not sure it's as simple as you say. In your third point, how can you tell that you need to advance more and not retard more? Because you're going to get the same result if you're too much advance or retarded. Also, you might simply be running out of injector capacity in which case no amount of advance or retarding will help you but the ECU still has to do something (completely cut fuel to save the engine?).

In any case, why do you think the ECU has to determine the advance? Or course, it would be nice. But it's another tuning parameter that the A-series engine requires and as with the other tuning parameters its specific to each engine setup.

Regards,
Jean

http://www.jbperf.com/

I'm a little confused. What are the 4 axis?

-Tyler
DO NOT TOUCH MY HORNS OF DOOM!

Marcel will correct me if I'm wrong but I think it's the standard 3D VE map with the addition of the advance so you have RPM, MAP, VE, advance as the 4 "axis".

http://www.jbperf.com/

oh, the 4 (Dimensional) Axis....


Time to go to bed.

Laterz!

Oh, that make sense. I guess it's a conceptual thing at that point because I wouldn't consider that a 4D table, but 2 3D tables(tho they share the same base axis). 2 tables that represent different data, tho to be useful they work together. The reason I separate them is VE is used in many other calculations that advance would not...
Anyway, i think I'm on the same page....

-Tyler
DO NOT TOUCH MY HORNS OF DOOM!

Ok, so I reread the patent again this morning...
I read it a while back, but I guess I didn't quite understand the problems at that point. Having the patent as a reference, they don't appear to use any "injection advance" and in fact there is a line that kind of indicates they don't worry about it at all(tho I can't find it right now).
That makes me question the whole approach we've taken to the problem.

-Tyler
DO NOT TOUCH MY HORNS OF DOOM!

Edited by curta_crankn_daddy on 4th Jan, 2006.

www.starchak.ca and www.TDCperformance.ca

Edited by TurboDave16V on 4th Jan, 2006.

Marcel - Email me your full address and i'll get this MPi camshaft over to you...

Marcel,

I understand all this. What I meant by retarding is reducing the current advance. At some point (high rpm and load), the starting point for the injection may have to be moved before A on your graph due to injector opening time and travel time. If the value you start with (taken from a table or computed by some algorithm) is actually too great then you might start having fuel going to the inside cylinder instead of the outside one. The only thing you'll see is this discrepency in AFR and no amount of additional advance will correct it (unless you go to 300+ degrees).

This could be caused by a bad table or an over correcting algorithm or it could just mean that you're running out of injection window. If you want to automate the injection advance you need to be able to cope with these things otherwise you may blow your engine by having a cylinder go lean at high rpm and load. That is my main concern with what you propose in terms of automated advance with O2 sensors. In addition, I'd be very concerned about those sensors reacting fast enough and accurately enough at high rpm and load to rely on them.

That's why I think that the advance table has to be built manually with the help of O2 sensors in a very careful manner. You probably would also want to make sure that you have some margin for the injection time compared to the injection window at your max load and rpm. And some monitoring of the O2 sensor should be done by the ECU to turn things off in case something goes lean.

Regards,
Jean

http://www.jbperf.com/

So, Marcel suggested a max IDC of 58%(I think), but looking at the Rover MPi information I've seen, they're hitting a max duty cycle of closer to 40%. Here's how I came up w/ this.
Rover MPi injector size : 440cc(so I've heard)
Brake Specific Fuel Consumption : 0.4 - 0.5 is generally accepted for N/A motors(we'll use .45 as a midpoint)
according to this: http://en.wikipedia.org/wiki/Austin_A-Series_engine
we get
Rover TPi Mini HP : 63hp - 77hp(we'll use max)
using our formula of :
10.5 * (hp * BSFC)/(no. inj * IDC) = flow rate
we can reorg to get :
10.5 * (hp * BSFC)/(no. inj * flow rate) = IDC
plugging in :
10.5 * (77hp * 0.45)/(2 * 440cc/min) = 0.41

Obviously the cam would have something to do w/ this, but the window seems smaller than previously thought?

-Tyler
DO NOT TOUCH MY HORNS OF DOOM!

Jean; consider this - the max injection advance you'd need is at full RPM's and WOT.

Now set this as the start of advance for every RPM site and VE. There is no reason you should have a problem as you've already advanced the start of injection to the 'worse case' point...