My Emissions Story Options

[pbanders]

Well, I always dread December here in AZ, because that means I have to get the 914 smogged. If I'd only stayed in CA, I would be home free by now!

In the past, I have never trusted the AZ MVD test equipment, which as you'll see, was probably a mistake. Always seemed high on CO and HC to me. As some of you may know, I have a garage full of 914 D-Jetronic test equipment, and I also have a CO meter of my own.

I didn't play it smart by checking my CO first, I just went and warmed the car up by driving for about 30 minutes, backed off the ECU idle mixture knob a few clicks leaner, set the idle speed and took the test. Here in AZ, they test CO and HC at load (about 2000 rpm on the dyno in 3rd) and CO and HC at idle after the load test. The limits for my year at load are HC>500 ppm / CO>4.20%, and at idle HC>500 and CO>5.50%.

The car has less than 2K on a full stock 2.0L rebuild. Passing those standards should be no problem. Well, it failed big time. Loaded HC=115 ppm / CO=4.85%, and idle was HC=890 ppm (!) / CO=5.85% (!). Waaaaay rich, wow. I groused at the guy that it couldn't be that rich and drove it home.

First thing I did was check my timing, it was spot-on. Next, the CO. Yep, I got just below 6%! I could fix that by adjusting the ECU, and I figured the high CO at load was because I'd swapped in an new MPS this past year, must have been richer.

I took a few minutes to compare the MPS's calibrations with my tester before I swapped them. Both were identically calibrated, with 8.9 ms injection pulse width at 0" Hg vacuum (wide-open throttle conditions), 6.3 ms at 5.0" Hg vacuum (moderate part-load), and 4.2 ms at 10" Hg vacuum (light load, near idle levels). Since they were the same, it made me concerned that I'd fail again for loaded CO. But, hey, I didn't trust their numbers, so I figured I'd just hit a different test lane this time and it would pass.

I did the swap, set the idle CO down to 3%, and went back for the retest. And as you might suspect, while it did better, it did fail as I feared for loaded CO. Results were loaded HC=127 ppm (nearly the same as last time) / CO=4.57% (fail, and nearly the same as last time), and idle HC=335 ppm (MUCH better) / CO=2.71% (MUCH better, and the same as I read on my meter). OK, now I know what to do.

Went home, yanked the MPS back out. The one I'd put in was one that I'd removed the epoxy seal so that I could do my own calibrations. I removed the full-load stop screw (the big screw slot in the end of the MPS) to expose the main load adjustment and part-load transition screws. While connected to the tester, I held the part-load transition screw fixed with one screwdriver, and screwed in the main load adjustment screw while watching the injection time on the tester. I dropped the full-load duration from 8.9 ms down to 8.5 ms, which resulted in the 5" Hg reading dropping to 5.9 ms, and the 10" Hg reading dropping to 3.9 ms. I figured this would make it so lean that it would drive like crap, but would pass the test easily.

Stuck it back in, set the idle CO to around 2.5%, set the idle speed. Surprisingly, the car ran fine. Drove it hard back to the test facility. You know what they say, "third time's the charm".

Guy who tested it this time owns an '80 911SC, so he was really interested in seeing if it would pass. And it sure did! Loaded HC=57 ppm (Wow!), / CO=1.58% (Wow!), idle HC=113 ppm / CO=0.99% (!). Got my cert and left.

Obviously, I went a bit too far with my adjustment, as I'd like to see the loaded CO at about 2%, and I'm going to bring my idle CO up with the ECU knob a few clicks. But I learned a few important things. First, something I've always known, which is to set idle and loaded mixtures correctly, you've got to do it on the dyno. Guessing with road results doesn't cut it. Second, the MVD's numbers were right and I should have paid more attention to them in the past, as they've always shown me to have quite a rich loaded CO. Third, the base calibration on the 2.0L MPS's is quite rich. I've got two NOS MPS's and both are closer to where I started from.

Long story, hope it had some useful info in it and wasn't too boring. Now that my car's registered for 2010, I'll be having a happy New Year, and I wish all of you out there in 914-land a Happy 2010!

[pbanders]

I did some more characterization yesterday, perhaps I don't need to do a dyno tune on it after all. I have one tester that uses an interposer to tap off the signal at an injector, so that you can get the injection pulse width. I also have a Heathkit CO meter, and the readings I get with it match what I got at the MVD for my idle levels. Both can be used with the car on the road, with a bit of futzing around, so I went for a drive yesterday and noted some data. All the CO levels I report below are in 0.5% increments, it's an analog meter and that's as good a resolution as I feel comfortable with reading it.

After a cold start, my idle goes to about 1700-1800 RPM while the AAR is open. Pulse width is about 4.5 ms. I didn't get a CO level because I didn't put the meter in the car until it was warmed up. Once the car is fully warmed up, I got a consistant idle CO of 3.0% (spec level for my engine), with an injection pulse width of 3.6 ms at 1050 RPM. Part-load cruising with constant throttle angle on a flat road at 2500 RPM in 4th gear resulted in a CO of 2.0% (which was my part-load target level), and an injection pulse width of 3.8 to 4.0 ms. Full-load (wide-open-throttle) conditions starting from 40 mph gives CO levels over 4.0%, and injection pulses of greater than 8.0 ms (varies widely as a function of engine speed). Overrun conditions (coasting downhill, throttle closed) gives CO levels between 2.0 and 3.0%, with injection pulse widths as narrow as 2.0 ms (indicating manifold vacuum levels of more than 15 inHg - note that my overrun valve is set to 18 inHg).

The car ran quite well, once fully warmed-up. Good power at all throttle angles and load levels, good idle stability, good transistion characteristics. As some of you who know me and my car may remember, I've always had issues with idle stability with electrical loading (e.g. lights on, heater blower, etc.) due to alternator loading (verified this as the cause through some experiments with my battery and manually controlling DF). With my current tuning, this problem is nearly eliminated. Engine idle with lights on drops from 1050 RPM down to about 800 RPM, more of a drop than I'd like, but a level I can live with. CO stays the same, but the injection pulse width increases to 3.8 ms. I can now drive my car at night without gritting my teeth and massaging the throttle at every stop light!

Where the car can still use some improvement is with the cold-start and warm-up phase. One thing that's totally different now that I have a much leaner idle setting is that under a cold-start (engine off for 3+ hours, ambient temp above 32 deg. F), after cranking, with no throttle input, my engine immediately starts and jumps to 1700-1800 RPM (due to the AAR being open). Previously, I would have to open the throttle (not to give more gas, but to give more air) to get it to start, and it would stumble under 1K for a few moments until it crept up to about 1500 RPM. While this is a nice improvement, the problem is that the AAR seems to close too quickly, in something less than 5 minutes. It would be much better if the AAR closed more slowly, over a 10 minute period, as it takes quite a while for the engine to reach full operating temperature (yes, I have a fully working thermostat and air flaps system (IMG:style_emoticons/default/smile.gif)). What I think I'm going to do here is to experiment with adding external ballast resistance to the AAR heater circuit. The normal resistance of the AAR heater is around 13 ohms, I'm going to see if I can dig up a 5 ohm / 5W power resistor to ballast in. I may also experiment with letting the AAR close with the heater open.

The other warm-up issue is that IMO, it's too lean. I'm judging this by a combination of some light exhaust popping during overrun, rough idle, and throttle transition behavior during warm-up. What I did to help that was I added 50 ohms of ballast resistance to the head temp sensor. The CHT circuit in the ECU is designed to richen the mixture when the NTC resistor in the sensor is at a high value (cold). Adding more resistance increases the richness, but you have to be careful. Depending on the ECU, the CHT circuit is designed to stop leaning out the mixture once the car is fully warmed-up, which is indicated by when the sensor's resistance drops below 300 ohms. Once you've driven the car for 30 minutes or so, the sensor resistance drops to around 50 ohms. What this indicates is that you've got about 200 ohms or so of ballast resistance you can play with to vary the warm-up phase richness and duration. I suggest adding resistance in 50 ohm increments, up to 200 ohms, as you zero in on the best performance.

One last thing I noted was that while I was doing all this tuning, I had an issue with the CHT sensor's connection to the wiring harness. I have one of those really super harnesses that Jeff Bowlsby makes, and the CHT sensor connection has a plastic protector, just like the factory one. The CHT sensor connector is exposed, so this protector prevents shorting should it touch a metal part in the engine compartment. Unfortunately, the down side is that it's a real PITA to get that sensor connector to engage with the spade part that's deep in the plastic protector. Often, actually doesn't slide onto the spade, but ends up wedged between the spade and the side of the plastic shell. First time you hit a bump, it either arcs or pops out, and your car burbles or stops running. Installing the 50 ohm ballast resistor helped fix that, as I'd made up a resistor with male and female spade connectors on each end. The extra length of the resistor and connector assembly makes it easy to engage the spade deep in the plastic protector, and exposes a spade connector on the other end that's easy to attach the sensor's connector to. I use a tie wrap to hold the exposed contact away from any metal parts to avoid shorting.