OK, so the bits and pieces of a custom intake for my 05 RS are winging or rolling their way across the country to my door. While us NA guys generally have less of a problem with intakes confusing the MAF sensors than the turbo guys, I still see the potential for a problem due to the construction of this intake. So my questions are:

1) Using only the stock telemetry from the car (no WBO2), what would it look like if the new intake was fuzzing the MAF?

2) Using only the stock telemetry from the car (no WBO2), how do you adjust the tables in the ROM to make the MAF read correctly again?

Log A/F correction, A/F learning, throttle position, MAFv, engine load. Once the car is fully warmed up, log smooth consistent driving at as many points on the MAF v scale as you can. You want to avoid tip-in fueling from quick throttle angle changes. When complete, open your datalog and create a custom column in the spreadsheet that adds both correction percentages together. This value should be consistent for a certain MAF voltage, and indicates the percentage that your g/s at that particular MAFv needs to change. I have found that it helps to make a gigantic scatter plot graph.... with enough data, your MAF correction curve will become apparent!

The top end of the MAF curve I have no idea. I can't get the ECU to see those values without going into open loop. I suppose I just need to adjust it until my AFR from the tailpipe matches that of my map. But I'm not sure.

You need a WBO2 to correct the portion of the MAF tabel you hit in OL fueling. That is the only way. You can tune the lower portion by checking the fuel trims.

The top end of the MAF curve I have no idea. I can't get the ECU to see those values without going into open loop. I suppose I just need to adjust it until my AFR from the tailpipe matches that of my map. But I'm not sure.

Yes, use a WBO2 sensor with Airboy's spreadsheet. There is an AFR interpolation tab in it.

However, watch out for d(MAF)/d(time). If MAF is changing rapidly then you don't want to use that data. For turbocharged cars, you can set WGDC = 0 (disconnect BCS or reflash or plumb the hose directly from nipple to actuator) and then do a 5th or 6th gear log, you should get fairly smooth MAF values. For the very top end, do a lower gear log.

OK, some questions since the front half of the intake showed up today and definitely is causing a ruckus with the MAF...

1) I drive a manual. How do I minimize dMAF/dt? I have to move the throttle pretty quickly each time I shift. Is there some way to make Excel filter out lines with too great a delta?

2) I'm never going to hit the top end of the airflow range in closed loop as I have the tables set up. If I temporarily set the entire OL Fueling table to 14.7 (i.e. prevent the car from exiting closed loop at all, obviously with the appropriate compensation to timing to prevent det), can I actually trust the data coming from the stock O2 sensor at 1.24g/rev and 5800RPM necessary to hit max MAFv?

3) Once I have a good clean data set for a given MAFv, and have summed AFL and AFC together to get a total error, how do I adjust the g/sec in the table at that MAFv? For instance, if the total error were +20% fuel (about the typical error I'm seeing), do I increase the g/sec or decrease the g/sec by 20%? I presume I increase the g/sec by 20% to tell the engine there's 20% more air, and therefore it will put in the 20% more fuel, but I want confirm.

1. You won't have this problem because you don't have a turbo. Turbocharged cars go from 1 g/rev to 3 g/rev in about 1000 rpm. That causes MAF to go up very quickly.
1a. Do a 5th gear log. Let the rpms slowly drop to 2000 rpm and then go WOT to as high an rpm as you safely can.
1b. Do a 3rd gear log. Let the rpms slowly drop to 2000 rpm and then go WOT but this time go to redline.
2. I don't know since you don't have a turbo (high pressure at stock front O2 location) and am not familiar with the O2 sensor variant that is used in the RS, but I would recommend you get a LC1 WBO2. You could sell it after you're done with it.
3. In closed loop, you use total correction to adjust it. In open loop, you compare target AFR to actual AFR using Airboy's spreadsheet.

1. You won't have this problem because you don't have a turbo. Turbocharged cars go from 1 g/rev to 3 g/rev in about 1000 rpm. That causes MAF to go up very quickly.
1a. Do a 5th gear log. Let the rpms slowly drop to 2000 rpm and then go WOT to as high an rpm as you safely can.
1b. Do a 3rd gear log. Let the rpms slowly drop to 2000 rpm and then go WOT but this time go to redline.

Ah! I was thinking I'd need like 30 minutes or more of driving around town to get enough data for this, not just a couple of pulls. This car will go from ~0.2 to ~1.2g/rev in 300-500ms. Is that enough to mess up the math?

Ah! I was thinking I'd need like 30 minutes or more of driving around town to get enough data for this, not just a couple of pulls. This car will go from ~0.2 to ~1.2g/rev in 300-500ms. Is that enough to mess up the math?

Yes. So don't use that data.

OK, so I've been looking at the bottom half of the airflow range since I can do that in closed loop. I've followed the above suggestions above and have noticed some very worrysome things.

1) Correction at a given MAFv isn't a constant. In fact, it's not even close. A given MAFv, in steady-state (meaning no changes to load or throttle position), can have total corrections that span a 3-7% (I mean that AFL+AFC might range from 3% to 10%, or 0% 5o 5% without MAFv changing at all) range. WTF?

2) At the VERY bottom of the range (idle and just off idle), MAFv fluctuates randomly. As an example, if you hold a TINY bit of throttle into it, sufficient to raise it to 1800RPM in neutral lets say, the MAFv will wobble up and down randomly. While MAFv is fluctuating, the engine will actually hunt up and down in RPM. I logged both pedal angle and throttle angle to confirm that neither was changing when the MAFv and RPM were hunting. Again, WTF?

OK, so I've been looking at the bottom half of the airflow range since I can do that in closed loop. I've followed the above suggestions above and have noticed some very worrysome things.

1) Correction at a given MAFv isn't a constant. In fact, it's not even close. A given MAFv, in steady-state (meaning no changes to load or throttle position), can have total corrections that span a 3-7% (I mean that AFL+AFC might range from 3% to 10%, or 0% 5o 5% without MAFv changing at all) range. WTF?

2) At the VERY bottom of the range (idle and just off idle), MAFv fluctuates randomly. As an example, if you hold a TINY bit of throttle into it, sufficient to raise it to 1800RPM in neutral lets say, the MAFv will wobble up and down randomly. While MAFv is fluctuating, the engine will actually hunt up and down in RPM. I logged both pedal angle and throttle angle to confirm that neither was changing when the MAFv and RPM were hunting. Again, WTF?


1. Did anyone ever say it was? The lower the MAFv, the more erratic the MAF is. That's why I wrote a MAF scaling spreadsheet for the low MAFv range to do this statistically. It gets steadier in open loop at 70+ g/s. The spreadsheet is more modified than when I posted it on enginuity since I added injector scaling/latency calcs to it. I'll have to write instructions for it again. Kascade incorporated a more slick feature into his alpha version of the enginuity logger. He posted pics on enginuity.org before it went down.

2. It's going to fluctuate. There's NOTHING you can do except get it close. My AFR Learning A varies between -6 to +2 depending upon if I am idling when the engine is cold or engine is hot. Frankly, it doesn't matter. AFR Learning A is not that big a deal except when it gets way off. In that case, the car will be hard to start but once the front O2 sensor is warmed up, it will rely upon AFR correction to steady itself.

2a. Actually, my AFR Learning A varies over time but it doesn't hunt like you're describing. Sounds like you might have a small leak somewhere.

1. Did anyone ever say it was? The lower the MAFv, the more erratic the MAF is. That's why I wrote a MAF scaling spreadsheet for the low MAFv range to do this statistically. It gets steadier in open loop at 70+ g/s. The spreadsheet is more modified than when I posted it on enginuity since I added injector scaling/latency calcs to it. I'll have to write instructions for it again. Kascade incorporated a more slick feature into his alpha version of the enginuity logger. He posted pics on enginuity.org before it went down.
Actually, the first reply to my initial post did state that it would be consistent for a given MAFv. Since it will apparrently NOT be so, how close am I shooting for?

Can you link me to the most recent version of your MAF scaling spreadsheet and give at least a vague outline of how to use it?

2a. Actually, my AFR Learning A varies over time but it doesn't hunt like you're describing. Sounds like you might have a small leak somewhere.

AFL A doesn't hunt up and down, the engine speed does. Holding a constant throttle position near idle results in both MAFv and engine speed varying up and down.


Thanks for the help

OK, another day of beating my head against the wall and spending a long time processing 7400 log lines into a graph by hand. Here's what I've got after the first revision to the MAF scaling. Using the stock scaling with this intake had 3 of the AF learnings pegged at +20%, so this is an improvement. The graph is clipped vertically, there are some single data point outliers up at +-20%, but not many anymore. What do you think? Am I headed in the right direction?

http://www.siriusphoto.com/forum_posts/20080106_maf_plot.jpg

Did you exclude data where d(MAF)/d(time) was significant?

To the best of my ability, yes, I was eyeballing it and I'm sure I either missed changes or just wasn't paying enough attention. There were too damned many lines of log!

I'm hoping that for whatever my ADD let through, the fact that it was a lot of steady state cruise control work with little human intervention will have caused the ones I missed to fade away as noise.

BTW, how big of d(maf)/dt are we talking 1V/sec? 10V/sec? 100V/sec?

Something I haven't seen mentioned...

Make sure you exclude the first 10 minutes (at least) of data. The coolant may be warmed up in a few minutes, but the engine isn't fully up to temp. I noticed a lot of inconsistancies when I included that data. When I dropped it, I had my intake tuned in 1 day of commuting to work and back.

Also, if you notice a trend over the whole MAF scale (IE 15%+ off over the whole range, then you'll be better off rescaling the injectors before tuning the intake.

OK, so I'm asking this again now that I have a semi-automated way to filter this out of the logs:

How big a change, either in terms of dMAFv/dt or dPedalAngle%/dt can be allowed before data needs to be thrown out?

You will have to look at your logs.

I'm not familiar with normally aspirated cars.

If the AFR corrections when d(MAF)/d(time) are different than when d(MAF)/d(time) is constant then you know that needs to be thrown out. Try driving at 80 mph on the freeway at cruise and 15 mph on a long street. That will give you two datapoints that you are positive.

I think you have a leak if your rpm is fluctuating at idle. You should double check that first before proceeding.