Ecoboost Speed Density tuning

[murfie]

Yes the filtering is a MP and multiple graphs will give you the fueling error in each MP. You will see they will populate different RPM ranges. you would apply large error to the slope and small error to the quadratic. These table represent an amount of air. so if its adding fuel you would add(x by 1.01+) and if its removing fuel you would remove(x by .99-).

If you want you can add a column of loads and then take the average percent of the different loads and apply it. It should come out the same, but you will be able to see what loads need more fuel and what loads correlate with what MP the most.

[metroplex]

So if the filtered table shows a value of say, -20% then that means multiply the slope value by 0.8? The load columns sound like a good idea.

[murfie]

well 20% is a lot of error. I would first check how many cell counts are in that cell and see if it is an outlying piece of data that can be ignored. things like this are possibly caused by DFCO or tip in /tip out areas. you can also increase the MP % to make sure the probability that the data applies to the appropriate MP increases. Its just like tuning a MAF curve just spread over the MPs and different sections. apply as much as you can to the slope then the quad to correct fuel, then the remainder calculated MAP and measured MAP difference goes to the BT or offset.

[metroplex]

Ok, say it was -5%. I chose 20 as an example, but I've seen +13% on my factory tune. What would I do if I saw -5%

Do I multiply the slope by 0.95 or 1.05?

[murfie]

-5% would mean that the ECU thinks there should be 5% more air than the O2 sensor is saying there is. it is compensating by removing 5% of the fuel. to correct this you would remove 5% of the air the ECU thinks should be there. so multiply by .95.
compare the difference of commanded lambda and WB lambda to the fuel trim. The fuel trims correct the commanded to get lambda 1. You correct the commanded lambda with the values of MAF/MAP/SD in the ECU.
fuel trim.PNG

[Blown383]

On my Focus ST with increased boost, I was running into points where the STFT would be +25% and the car would be at 0.92 lambda when 0.85 lambda is commanded. Way too lean for 22psi of boost. I adjusted the slope in the SD tab by the inverse of the correction. So say mapped points 5 and 6 were 65% and 45% at 2750rpm. I applied a -25%, or multiplied the values at 2750 rpm by 0.75 in mapped points 5 and 6. After getting enough data and blending the values in, my WOT fuel trims are now within +/- 3%. I no longer have any lean spikes under boost. I'd recommend erring to the side of being a little rich so the system has a negative fuel trim value. The ECUs respond extremely fast but not quite fast enough on a car like the Focus ST. Boost was coming on a little too quickly for the fuel trims to compensate. Murfie, the problem with filtering the histograms is that it doesn't blend the values as well as the ECU does. If the histograms could follow the exact blending that happens with the mapped point weighting, it would be extremely beneficial to adjusting both fueling and timing. Also, the Ecoboost has no MAF and only a MAP sensor. There is no MAF curve or VE table. The mapped points are a bunch of VE formulas that tell the ECU how to fuel. On MAF based Ford systems, SD tuning isn't totally required although it does affect some transient fueling.

This image is an example of how the fuel trims don't quite respond fast enough to rapid changes. Eventually, the trims catch up but not until after a lean condition has occured.

Capture.PNG

[murfie]

Have you right clicked the parameter and adjusted the polling speed to be faster of mapped point weights? you can make the weights update up to 20 times faster than they do with no modification.

This is a good example of when you need to adjust the SD. When the fuel trims are not the same as the difference of the commanded and actual lambda.

MAP after intake event make sense why they include BT and offset. This would explain why appling the inverse is the correct way to modify them. Very good info.

[metroplex]

I had an old log with the MP weights logged, so I was able to create my own Math Parameter based on the predefined EQ Error (the PIDs in that predefined parameter weren't available for the SHO, but we have similar ones). It looks like the bulk of the error is below 3500 RPM. Anything above 3500 RPM seems to be within +/- 3%-6%. Below 3500 RPM, some of the MP's will have up to 40%-50% error at certain RPM ranges.

In theory, it seems like this would achieve the same as monitoring STFT but in a cleaner way... unless I am seeing something wrong.

So just to ask again to make sure I'm on the same page, if I see a positive EQ Error value (say, +5% or +39%), I'd go to the Slope Table for that particular MP, and multiply the value associated with that RPM by 0.95 or 0.61 respectively (-5% and -39%)?

Also, how would I filter out the actual EQ Ratio readings when DFCO is enabled? My actual lambda can go as high as 2.00 when I am off the throttle and cruising down. This obviously throws off EQ Ratio.

[murfie]

Blown383 says apply the inverse and has had it correct the fuel error. Go with that. A way to eliminate the DFCO is to view the zoom data only and zoom in a a part of the log that doesn't have DFCO in it. The other option is to type AND (caps matter) and create another PID actual lambda<1.2lambda next to the MP function.

This is just a representation of what to do not actually what to do. you can also use OR >.78 to get rid of the COT and other random sensor drop outs.
Multiple functions.PNG

[metroplex]

Thanks!! Should we use STFT, STFT+LTFT, or Wideband Delta?

[metroplex]

I used STFT with the filters, and EQ Error with the filters. The STFT data shows I am within 4%, but EQ Error shows up to -15%
What's weird is that for Mapped Point 3, the STFT data is inverse of EQ Error. EQ Error is all negative values up to -15%, while STFT shows positive maxing around 3-4%

This makes me wonder which values to use

[metroplex]

I just did a quick WOT run. The EcoBoost loves this cooler denser air.
Here we see the Wideband error ratio showing the greatest delta percentage.
STFT doesn't look too bad. LTFT+STFT averaged is mainly biased with the LTFT data. The LTFT data looks to be positive in the 6-7% range which is skewing the results that way. On my SHO, the mapped points 0-3 are the ones that have the most useful data. The trend is about the same or worse for MP1-3 using the EQ Ratio Error data (we're talking -10% to -17%) at high RPMs.

I'm not so sure I need to adjust anything at this point. Even with 18 psi of boost and the extra air flow (52+ lb/min of air mass vs stock 38 lb/min), the actual wideband data shows slight lean during initial WOT, and then it follows the commanded lambda and dips down to 0.78 for some reason right near 6000 RPM (COT shouldn't be kicking in as the cat temps are still good, and commanded lambda is still 0.85). Once it upshifts, it will remain at the commanded 0.85 maybe dipping richer briefly.

[Bugasu]

Correcting just the slope is incorrect. There is something hopefully in the works for a more thorough explanation including blowthrough for Speed Density, but it's not completed yet.

The basics are when you're not in blowthrough, you're correcting a quadratic fit, but its a special kind of quadratic, not just Quad*X^2 + Slope*X + Offset. Quad actually has a weight in both the linear and offset terms as well. You have to separate the slope-offset fit from the quad fit. The quadratic is "orthogonally" fit in this sense. That is, the value it takes is independent of slope and offset.

When you're in blowthrough the ECU is actually tracking two airflows. An in-cylinder airflow that is limited, and a blowthrough airflow that represents fresh-air that has blownthrough to the exhaust. Fresh air skews oxygen sensor readings and thus control of in-cylinder lambda, so its tracked to correct in-cylinder lambda as well as for emissions purposes (fresh oxygen affects the balance in the catalyst).

I've posted the patents in prior posts. The system is not simple, and there's a good deal of math involved in getting it all calibrated correctly. As well, the equations are for MAP = f(aircharge). The ECU inverts this (with the quadratic equation), and there are special cases for no valid-real solution, one-solution, or two solutions.

The problem is not solved by simply changing the slope.

[metroplex]

Bugasu: What are we supposed to datalog and what are we supposed to change? The patents and theory aren't very helpful until I see there is an actual way to adjust fuel trims.

[murfie]

Bugasu: What are we supposed to datalog and what are we supposed to change? The patents and theory aren't very helpful until I see there is an actual way to adjust fuel trims.

https://onlinecourses.science.psu.edu/stat502/node/203

This is a "simple" plant density to grain yield orthogonal analysis. To correct fuel trims replace it with engine RPMs and MAP(as air mass). You would need to collect data to populate a table then run it through the IML procedure in SAS. The current scanner is capable of collecting the data but you would need to use this extensive and complicated third party software(this is where the math and programming is needed) to repopulate the tables correctly.

If you find correcting the just the slope values helps your fuel trims I say go for it. Just don't expect the SD to be reliable for sensor failure.

[cyruss]

How do you know how and why the ECU will choose which distance table to use ? drivability or Fuel economy ? (see screenshot).

hpt2.jpg

I know we don't care as soon as the mapped point map are updated but it is to understand.

About the weighted mapped map, do you think it is a good solution to disable them and keep only one enable so we can tune one map at the time. I mean we put mapped 1 enable, all off, so that will lock the camshaft in only one position, then we could adjust the VE for that map. and so on with one by one mapped 1 to 14 .... (or 1 to 5).

[Trent_Petersen]

How do you know how and why the ECU will choose which distance table to use ? drivability or Fuel economy ? (see screenshot).

I'm pretty sure if you look at the "Enable Max Load" and "Enable Min Load" tables on the right most column, that is how it decides what distance tables to use.

[cyruss]

I'm pretty sure if you look at the "Enable Max Load" and "Enable Min Load" tables on the right most column, that is how it decides what distance tables to use.

Ok thanks I got it.

[ghostofnyc]

Thank you for sharing

[cyruss]

I just did a quick WOT run. The EcoBoost loves this cooler denser air.
Here we see the Wideband error ratio showing the greatest delta percentage.
STFT doesn't look too bad. LTFT+STFT averaged is mainly biased with the LTFT data. The LTFT data looks to be positive in the 6-7% range which is skewing the results that way. On my SHO, the mapped points 0-3 are the ones that have the most useful data. The trend is about the same or worse for MP1-3 using the EQ Ratio Error data (we're talking -10% to -17%) at high RPMs.

I'm not so sure I need to adjust anything at this point. Even with 18 psi of boost and the extra air flow (52+ lb/min of air mass vs stock 38 lb/min), the actual wideband data shows slight lean during initial WOT, and then it follows the commanded lambda and dips down to 0.78 for some reason right near 6000 RPM (COT shouldn't be kicking in as the cat temps are still good, and commanded lambda is still 0.85). Once it upshifts, it will remain at the commanded 0.85 maybe dipping richer briefly.

Hello Metroplex,

I'm a bit confuse So which data we need to take to do the correction ? LTFT+STFT or the EQ Error ?