Another 998 (now turbo) MPi trial
Turbo Mini Forums

Hi,

I'd like to share my experience with an MPi setup on a NA 998 A+.
To add information for those who follow Paul's experience.

I'm going to leave for few years, so this is a kind of report for the community.

Some pictures here:
http://www.turbominis.co.uk/forums/index.php?p=vt&tid=222687

The engine:
Rebuit and balanced 1988 998 A+, CR 9.3:1, -1kg lightened Minispares flywheel
998 ported head, unshrouded valves, head machined 1.5mm (ended in the oil passage, made weld the head and modified the passage to go through a stud hole).
SW5, maniflow, MPI intake manifold with a 1991's Polo GT dual-throttle body.
front mounted Metro rad, motobike fan

Management
Bosch Motronic ML4.1 (Alfa Romeo 75) hardware, my own firmware and PC tools
static wasted spark ignition, Lodge HL plugs, I've been using it since last summer with a carb
Injection: one EV5 Bosch injector (4 jet spray) per port (450cc), both injectors driven by the same signal, once per revolution (Edit: changed, see my next post).

The injection management code is very fresh, I'm still improving it

The engine idles very well, I can make it idle smoothly at 600RPM with a 13.5:1 richness
I mapped it up to 5000 RPM, I made a 2 hours drive/map test today. It runs well, even if I still have to improve transients.
Everything was mapped to run at 14.7:1 in steady state, except in idle where it's around 13:1

Looking at the spark plugs it's obvious that outer cylinders run leaner. They are chocolat, while inner cyls are black.
So it seems it's possible to use this setup to have a decently running engine.
Unlike Paul I don't have 2 WB lambda, only one for all cylinders, so I don't know the real richness of every cylinder.

Now I'm going to put a phase sensor (modify an ignition distributor) and drive every injectors separately.
I'm still convinced the main richness problem is with wall wetting, nothing to do with charge robbing or anything else.
So the code will still inject once per revolution, the phasing will only allow individual cylinder trim.

http://apavlov.club.fr/mini/management/dscf2622.jpg

http://apavlov.club.fr/mini/management/dscf2623.jpg

Added:
I believe MS uses a 0.1ms injection resolution. Like me.
I had to change that into a 0.05ms resolution, which gives a 12ms max pulse. The engine idles with 1ms+0.9ms pulses and refuses to start with 5ms crank pulses when dead cold (0C outside temperature), I tried to scale down to 0.025ms.
At 90kpa intake pressure the inj pulse is around 4ms after air temp correction (30-40C air temp), this gives 5ms for -20C air temp.

Edited by alpa on 30th Jun, 2008.

std 998 A+, g295, MD266, RHF4, 109hp @0.8bar/5400rpm

Well done. The more ideas we try, the quicker we will crack this problem.

Edited by Paul S on 27th Jan, 2008.

Saul Bellow - "A great deal of intelligence can be invested in ignorance when the need for illusion is deep."
Stephen Hawking - "The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge."

Hi,

That's it, it works. I'm using again my car to go to office.

NOTE that I have only one common wide-band lambda. However transients are very smooth, plugs have good colors and the injection map is very consistent.
I still have to work on some details like setting CO level at idle, and to measure the torque/power curve to compare with what I had with a carb.

The engine does not ping with advance that was pinging with the carb (autoignition when shutting down).
With the carb I had roughly the same AFR numbers than now in the same RPM/load conditions.

I had to make a phase sensor. I used an old ignition distributor which opens the switch only once per distributor's revolution (every second engine revolution), I removed 3 of 4 bosses of the cam. It's opened by the cam and rests closed during most of time, this lets the switch the time to close to avoid glitches when sampled.

The system opens both injectors once per revolution, but they have different timing. Inner squirt pulses start 10 degrees ATDC. Outer pulses start 10 degrees ATDC up to 3000 RPM and then linealy advance up to 53 BTDC for 6500 RPM.

The injection pulse width of inner cylinders is 80% of the outer pulses. This relation does not depend neither on RPM nor MAP, unlike I was expecting.

The engine is difficult to tune in transients when dead cold, the first 1-2 minutes after the start. The accel enrichment (AE) is very different when cold and warmed, the AE duration depends on the engine temperature (from 1.5 to 0.5 seconds), the gain depends on engine temperature and MAP (hot+low MAP -> low gain). When cold (0 degrees) the wall wetting is extremely high. At around 50 C it becomes more or less stable.

So it's feasible !

std 998 A+, g295, MD266, RHF4, 109hp @0.8bar/5400rpm

Great info!

I spent 2 hours today trying to adjust the CO level.

I need a 14:1 AFR to get 1.5% level, around 13.5:1 to get 2.5% (max allowed level for a carb car).
The ignition advance has no impact on CO.
With the carb I was getting 1.5% easily with a 13:1 AFR.

The engine can idle "smoothly" at 500 RPM, like an old US V8, very funny. "smoothly" is quoted because 500RPM makes a very low firing rate. It's also very quite, like an industrial diesel.

The engine starts going lean at about 90C engine temperature, especially because the air temperature rases above 50C. It's VERY difficult to adjust because a 0.05ms change makes it go from 13 to 15 (it idles with 1.05ms + openning time pulses). While the engine temperature is below 80C everything is OK. The rad fan switching ON/OFF disturbs a lot.
All that with 5C ambient air temperature.
I can not imagine what this will give in summer with 35C ambient temperature.

I see two ways to solve the problem:
- make a 1/4 AFR adjustement step by adding 0.05ms to one, two or three injectors with a round-robin policy.
- make the pulse duration toggle +/- 0.05 ms with a 1-2 second rate to sythesize the required average AFR.

Or may be should I mimic the carb with a inner/outer AFR difference ?

std 998 A+, g295, MD266, RHF4, 109hp @0.8bar/5400rpm

I added a special low load injection map used only for idle. It has 0.05/8 = 0.00625ms resolution and allows adjusting idle AFR in a precise way. Now ambient temperature is 15-20C.

std 998 A+, g295, MD266, RHF4, 109hp @0.8bar/5400rpm

That's it, I drove it 30minutes with a low boost VNT15 setup. Intake abs pressure limited to 1.2bar. The NA maps looked to work fine.

Still have to adjust the nozzle control, at 2000 RPM it's already spinning like hell.
That's funny because this VNT15 was designed for a 1.9 TD 110HP.

std 998 A+, g295, MD266, RHF4, 109hp @0.8bar/5400rpm

I've been using the car with VNT15 for last 2 months.
115kpa max boost.
The fuel consumption is definitely lower than NA, while the car has more torque. I know that because I live near mountains, there are lots of hill over there. For some technical reasons I was not able to measure power curve, and I don't have time to make mods for that.

The NA consumption was 7-7.5L/100km with a carb and 7L/100 with MPI.
Now it's 6L/100. It's far from 5L/100 I'd like to get, but still a 15% improvement.

In idle the CO level is between 0.6 and 1.1%. I have to stay above 13.5:1 richness for that.

The turbo control is:
- under some defined throttle position the nozzles are wide open. So the turbo "isn't used", but I can hear it so it spins.
- above that position there are 2 steps: 105 and 115kpa.
During transients the boost is limited (injection is cut on every second cylinder) at 135kpa. And it's easily reached in 2nd gear, the turbo is very hard to control.
The boost starts at 2000RPM.

I did not measure MPG when turbo was "used" all the time, even in partial load. Such mode is dangerous as the turbo is quickly in surge.

Another conclusion (even if it's not a surprise): the std 998 head is a piece of shit. No way to make it work well above 4000RPM. I've used a Metro 998 head which has 1mm bigger intake seats.

Edited by alpa on 30th Jun, 2008.

std 998 A+, g295, MD266, RHF4, 109hp @0.8bar/5400rpm

or 47mpg in english.

Very impressive. I've managed 45 with thw 998 turbo, but that's not driving slowly and is a mix of town and a road driving.

Fastest 998 mini in the world? 13.05 1/4 mile 106mph

www.twin-turbo.co.uk

Great result. I'm sure there is more to come.

Have you been over the Vercors or the Chartreuse in the Mini?

Great drives in any car.

Saul Bellow - "A great deal of intelligence can be invested in ignorance when the need for illusion is deep."
Stephen Hawking - "The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge."

I forgot to say that my driving style isn't really gentle. Front tires look worn, they have about 5000km :) .
I drive in town every 2-3 day, but when the traffic is light.

Chartreuse and Vercors are what I see from my house window :) . No I did not go to mountains in Mini, it's usually a family trip so we go in biturbo or Kappa.

Edited by alpa on 30th Jun, 2008.

std 998 A+, g295, MD266, RHF4, 109hp @0.8bar/5400rpm