Ford Speed Density tuning on EcoBoost...

[GapRider]

I'm still glad you're working on it though. I've been comparing the original cobb sheet with your quad rev1 sheet, trying to understand more.
The linear calculation is a simple way to calculate aircharge, right, using slope and offset (MAP, air density for reference)?
I don't understand quadratic math being used to calculate aircharge, from what I've read, all quadratic formulas result in a parabola, is Ford using a portion of the calculated points in the parabola so you end up with a "line" whose slope changes from aggressive to less aggressive?

[murfie]

The goal of speed density is prediction of Air Mass based on pressure and temperature. The more accurate the better.
Y=Mx+b with a given M and B would be a diagonal line. The slope M would then be (y2-y1)/(x2-x1) describing the rate of change. In this form you can see the closer the values are together the more accurate the slope will be to the true rate of change. 0 slope is no change in air mass= pressure change of offset (b). This means all predicted air mass values have a inherited degree of error. This is not an accurate way to describe the pressure= weight relationship. The variables in PV=nRT are rapidly changing. Pressure and temperature effect density directly and can change not only in value, but also in uniformity in a given volume. Density has a direct effect on n. n is what we are really solving for. n=PV/RT

Yes Y=Ax^2+Bx+c and at given A,B,C would be a parabola. Knowing that rate of change was critical to the speed density calculations accuracy what this equation does is describe the rate of change of the slope(which describes the rate of change). The main advantage to the quadratic equation is the rate of change is built in to the equation. You no longer have a limit of accuracy at every airmass, but rather just one point of 0 airmass.

[GapRider]

"Whatever the pnenonic representation in the digital computer computation of volumetric efficiency or its equivalents, the significant factor is the use of the PIOPE or PEOPI ratio of absolute pressures. These pressures in ratio and when combined with a second factor provide direct and accurate indications of current or real-time engine volumetric efficiency, i.e., volumetric efficiency as of the time the absolute pressures are determined. (This, of course, assumes the intake and exhaust conduit pressures are measured or determined at the same or insignificantly different times). The second factor mentioned above is representative of the dyanamic forces of friction and inertia that act upon, and tend to retard the flow of, the gaseous mixture in the engine's intake conduit; these forces are proportional to engine speed and other engine operating parameters of lesser significance. The second factor, and also the constants K and K₂ above, can be determined by multiple regression analysis of data obtained by testing a particular engine design on an engine dynamometer. This method for determining the second factor typically results in the second factor being defined by a quadratic equation, having known constants K₃₉ K₄ and K₅, as follows:-
second factor = K₃ ⁺ (K₄) (engine RFM) + (K₅)(engine RPM²)."
http://www.google.com/patents/EP0026642A2?cl=en

[murfie]

A patent from 1980 describing the linear and quadratic approach to sped density. WOW. the same basic principal applies. Instead of PIoPE or PEoPI ratios and RPMS they are now using PIoAM. This is why air mass is not just air mass, but air mass per event. just like RPM and RPM^2 describe the speed the engine is rotating and the rate of change the engines rotation speed is changing, airmass/event and airmass/event^2 describes the airmass in the cylinder and the rate of change of air mass in the cylinder. PIoAM is just another method to improve accuracy over PIoPE or PEoPI. Interesting history stuff.

[GapRider]

[the volumetric efficiency, is related to the ratio of the intake manifold absolute pressure and the engine exhaust system absolute pressure. The relationship is almost hyperbolic. If the ratio is inverted, it is almost linear.]

Intake pressure/Exhaust pressure, the higher that ratio = higher flow. I can see that being non linear but a hyperbola and a parabola are both "U" shaped. I can't picture how that PIoPE relationship could go up, then down like the hyperbola and parabola curves do?

Also:
[The exhaust conduit pressure is a simple quadratic function of engine air mass flow rate, that is, exhaust pressure is equal to a constant times the square of the air mass flow rate.]

Exhaust pressure = K(ExhMAF^2). That is a basic principle they taught us in the Navy. In the tables we have available - OFFSET, SLOPE, QUADRATIC TERM, where do the individual parameters fit in?

[murfie]

We have offset(pressure), slope(pressure/mass), blow through slope(pressure/mass), and quadratic(pressure/mass^2). For quadratic units my hptuners doesn't show that but copying the table into excel does.
All of these describe the final conditions of cylinder pressure when the cylinder volume is at its greatest.
Offset- describes the pressure the cylinder will be at when change in actual air mass(weight) has no change from one event to the next. This pressure came from part of the manifold pressure.
Slope- describes the pressure of new air mass in cylinder from one event to the next. This pressure came from part of the manifold pressure.
Blow through slope- describes pressure of air mass that is lost from the cylinder during valve overlap. This pressure was part of the MAP but air mass would be subtracted for fuel calculations.
Quadratic- this describes the change in pressure of new air mass. The pressure of this would also have been part of the MAP.
Add them all up and you get MAP just before event.

Think m/s describes a speed and m/s^2 describes acceleration. Acceleration is a change of speed. A constant speed is linear a change in speed would be a parabola in practical application and hyperbola in mathematical application. A parabola will have a maximum or minimum a hyperbola will not. Rate= distance/time. Change of rate = (distance/time)/time.

PIoPE in a na engine using absolute pressure
PI ranges from 7-8inhg to 31-32inhg
PE ranges from 26-28inhg to cylinder pressure at exhaust valve opening so pretty high.
The ratio would constantly be going back and forth from very small number to a little over 1. Think about the pressures right at the valve events.

[GapRider]

Original post:

I know a few people do this for a living, but have any of the DIY'ers developed some kind of guide? Essentially, all I'm interested in doing is being able to correct/compensate for an aftermarket air intake or at least fix the stock tune so it is more dialed in. With the older Ford MAF systems, I could do this fairly quickly thanks to a variety of guides/resources available. Which PIDs do we log, and which tables do we modify? Where do the mapped points come into play?

Still the elephant in the room : )

[metroplex]

Original post:


Still the elephant in the room : )

I know, right!

The theoretical discussion on how the ECU handles the data is nice, but there's nothing that I can see being easily used to basically do this: "Detect +13% lean fuel trim, correct XYZ to fix that lean condition". It can't be that hard. The other half of the challenge is figuring out the EcoBoost tuning strategy.

[Witt]

I know, right!

The theoretical discussion on how the ECU handles the data is nice, but there's nothing that I can see being easily used to basically do this: "Detect +13% lean fuel trim, correct XYZ to fix that lean condition". It can't be that hard. The other half of the challenge is figuring out the EcoBoost tuning strategy.

I haven't had to change fueling yet, just started increasing boost and timing but from the Cobb tutorial posted earlier, if I remember correctly it stated to change slope to modify fueling to match fuel trims. Someone can correct me if I'm remembering wrong but I believe it said to increase slope to correct for negative trims and decrease for positive ones.

[GapRider]

I know, right!

The theoretical discussion on how the ECU handles the data is nice, but there's nothing that I can see being easily used to basically do this: "Detect +13% lean fuel trim, correct XYZ to fix that lean condition". It can't be that hard. The other half of the challenge is figuring out the EcoBoost tuning strategy.

I'm hoping that if we learn the theory and then be able to plug parameters into a spreadsheet, that we will be able to tune just as you describe. That's what all the theory discussion is aimed at. If I was smarter we'd be done already : )

[GapRider]

I'm stuck on the [term*1.0 -14]. I'm assuming the Q and R results are mass and would atleast be to the same power of ten (same decimal range)?
For that assumption to be true, the [term*1.0 -14] would have to be in the 3.53831E-08 range. I haven't been able to get there.

[GapRider]

Yeah, I posted that to get your feet in the water.

I can write up on the basis of the new system when I get some time.

I do a lot of EcoBoost tuning and have it pretty down pat.

OK, my feet are soaking wet, I'm ready for the quadratic explanation : )
I can't plug numbers into ax^2 + bx +c and get anything that looks believable

[GapRider]

I plotted a graph of one set of speed density parameters (OP I think). I had to multiple the result by 1E-16 to get it in a range that looks believable.
Using this:


Am I totally off the path?

[murfie]

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

"To avoid certain problems with a conventional quadratic polynomial fit, the ‘orthogonal polynomial’ method was used to make the quadratic term independent of the linear slope and intercept terms. Thus carryover calibrations could be used with the quadratic term set to zero or the quad term could be added for increased accuracy without changing the linear term. Thus, in one embodiment, the quadratic term is utilized, while in another it is set to zero."

The link is an example problem in which orthogonal polynomials are used on a set of data. very math heavy but understandable. you can even relate it to the data that we have access to.

[PST]

Wouldnt this all be easier to figure out if we were to measure actual manifold pressure with a map sensor? Then we could solve the equation with an actual number for that term. Ive been using a standalone MAP sensor forever, only ever used it to see boost pressure on setups that didnt have a gauge or didnt have sufficient MAP sensor for the boost range.

I suppose I could set it up to log on a supercharged Coyote Im working on.. But the resolution will not be that great, as I only have an AEM 3.5bar map to use and the car only makes like 14-15psi with the current pulley.

[GapRider]

Right now I feel like that would be a step in the forward direction.

[PST]

Ok, Ill get it done.

I actually have some nGauge logs with a map sensor if youre interested.. HPT wont open the logs (they're .csv files), and I doubt I was logging all the parameters you need but it may be worth a shot. Otherwise, give it a week tops and Ill have this one up and running again and will send you any logs you want!

GapRider, Murfie, if you want the tune and logs send me a pm with your email addresses.

[GapRider]

Ok, Ill get it done.

I actually have some nGauge logs with a map sensor if youre interested.. HPT wont open the logs (they're .csv files), and I doubt I was logging all the parameters you need but it may be worth a shot. Otherwise, give it a week tops and Ill have this one up and running again and will send you any logs you want!

GapRider, Murfie, if you want the tune and logs send me a pm with your email addresses.

Thanks man

[GapRider]

I noticed when trying to calculate and plot OP speed density graphs that there seem to be peaks at 1000,2000,3250,4500 and 5000 RPM.

I thought the speed density values were experimentally obtained by the manuf based on logging actual data on the engine on some kind of super dyno where they measure everything imaginable.
Anybody know why I see those "peaks" for slope and quadratic values at those particular RPM's? Car is 13GT Coyote.
OP speed density mass.JPG

[Higgs Boson]

I noticed when trying to calculate and plot OP speed density graphs that there seem to be peaks at 1000,2000,3250,4500 and 5000 RPM.

I thought the speed density values were experimentally obtained by the manuf based on logging actual data on the engine on some kind of super dyno where they measure everything imaginable.
Anybody know why I see those "peaks" for slope and quadratic values at those particular RPM's? Car is 13GT Coyote.
OP speed density mass.JPG

ever seen a GM VE table? looks like new mexico.